Warning: this is a rant.
I use a UPS (Uninterruptible Power Supply) to protect the Great Beast of Malvern from power outages and lightning strikes. Every once in a while I have to buy a replacement UPS and am reminded of how horribly this entire product category sucks. Consumer-grade UPSes suck, SOHO UPSs suck, and I am reliably informed by my friends who run datacenters that no, you cannot ascend into a blissful upland of winnitude by shelling out for expensive “enterprise-grade” UPSes – they all suck too.
The lossage is extra annoying because designing a UPS that doesn’t suck would be neither difficult nor expensive. These are not complicated devices – they’re way simpler than, say, printers or scanners. This whole category begs to be disrupted by an open-hardware design that could be assembled cheaply in a makerspace from off-the-shelf components, an Arduino-class microcontroller, and a PROM.
How badly do UPSes suck? Let me count the ways…
I know people who hook up car batteries to salvaged UPS electronics and get 10 years of life out of the rig. UPSes could be designed with the kind of deep-cycle gel batteries used for marine applications like trolling motors to last even longer and be even more reliable. But noooooo. Buy a UPS and the vendor (even one of the relatively good ones) will sell you the crappiest, lowest-cost power cell that might, with a squint and a following tailwind, possibly achieve the dwell time printed on the packaging; it will in fact be a piece of shite with so little deep-cycle endurance that it will crap out in usually less than three years.
(This isn’t just needlessly expensive, it’s bad for the environment. Dead batteries are nasty things to put in the waste stream. Doing that as seldom as possible is real care for the ecosphere, not mere idiot virtue signaling like, say, ‘recycling’ paper and plastic.)
Yeah, and that dwell time will always be at least what Mark Twain called a “stretcher” and from less scrupulous vendors an outright lie. Vendors commonly measure it with tiny monitors and half the other peripherals slept; you’ll be lucky to get 50% of the rated value on a system in real use.
Next: automobiles nowadays are are equipped with intelligent battery-current sensors that measure not just output voltage but discharge current and battery temperature. This is enough to do accurate state-of-charge and state-of-battery calculations, so you (a) know how much dwell time to expect during an outage, and (b) get warning your battery has entered the bad end of its bathtub curve well before it craps out.
But are these intelligent current sensors deployed in UPSes? Why, no! That might add a couple of cents to the BOM and of course we can’t have that. Far better to inflict unexpected battery death on the customers who, you know, were paying money exactly so that sort of thing wouldn’t happen to them.
Now we get to what actually triggered today’s rant: the terrible user experience produced by the vendors’ grim determination to pump out least-possible-cost designs that ignore what users actually need. I was awakened from a sound sleep at oh dark thirty yesterday morning by the alarm on my UPS. Upon examining it, I was greeted by a flashing idiot light.
What’s missing from this picture? A text error-message display, like you’d see even on a rather low-end printer or scanner, to tell me “Scheduled periodic dwell test failed – your battery is dead”. To get a clue that this is what that particular alarm tone and flash pattern meant I would have had to lay hands on the device documentation, which of course you can always do instantly when awakened to deal with an emergency alarm in the middle of the fscking night.
It gets better. The tech support drones at my vendor’s call center couldn’t tell me what that alarm behavior meant either. Eventually they issued an RMA for a new battery anyway, but I couldn’t reconstruct what had actually gone down until discussing the whole sorry mess with AD&D regular John D. Bell the next day – he runs a university data center and has seen incidents like this before much more often than I have.
(It took some struggle to get the vendor to issue an RMA, because although I bought the device just three months ago a serial number lookup reveals that it was manufactured five years ago – the battery spent almost all of its service life sitting in a succession of warehouses. To be fair, this part is not the vendor’s fault – it’s the kind of risk you take when you buy from an electronics retailer that likes to stack other peoples’s overstock goods in the front of the store with an eye-catching discount on the pricetag.)
Back to what is absolutely the vendor’s fault: all but the lowest-end UPSes have monitoring ports (usually USB these days) that, in theory, should let you get useful status information from the device – perhaps even hook it up to a monitoring daemon on your computer that can do a clean shutdown when the device has been on battery for more than a transient brownout’s worth of duration.
In theory. In practice, this sort of thing is such a pain in the ass to set up that despite having contributed to two different UPS-monitoring daemons (NUT and apcsupsd) I gave up on this capability years ago. The problem divides into at least two parts:
1. The wire protocols these things speak are generally undocumented. The vendors think it’s sufficient to provide a Windows binary blob that gives you a fixed-function monitor GUI. Which is almost invariably so badly designed and poorly documented that it might look kinda purty but you can get little actual use out of it.
2. That first problem might be surmountable if you could watch the port yourself and reverse-engineer the protocol by watching what datagrams come up the wire. And if it had been designed by anyone with a clue about how to do application protocols right they’d be in some self-describing metaprotocol like JSON or XML or at least NMEA0183-like text sentences.
But no. This never happens. UPS protocols are invariably cryptic, half-assed crap designed by an EE in a hurry who thinks every “unnecessary” byte transmission is a sin against nature. Fields have no names. Numbers, if they’re not binary-encoded in unspecified endianness, have no units. Opaque status codes abound. The protocol grammar is full of defect-attracting corner cases. The device never IDs itself or provides a protocol version. Discoverability: what’s that?
If this sounds like the same sort of mess that afflicts GPS reporting protocols, that’s because it is. Actually it’s worse here, because there’s no equivalent of NMEA0183 to even begin to address the discoverability problem. Relentless vendor cheapness is at the root of both messes – given a choice between spending NRE on a decent design and shipping inadequate crap that piles hidden long-term costs on customers, UPS vendors unfailingly opt for the latter.
This whole product category begs to be disrupted by a maker design – because it’s possible, and otherwise the incentives on the vendors won’t change. Without disruption, the whole category could stay trapped in a nasty crab-bucket equilibrium that never rewards the risk of spending a few more pennies than ones’s competitors to field a decent design.
Let’s invert the above gripe list to specify what a decent design would look like:
0. Open hardware design, open-source firmware design, open-source device-monitor code.
1. Designed to be used with deep-cycle marine gel batteries that will last next to forever, for minimum long-term cost and least environmental impact.
2. Uses EV-style intelligent battery-current sensors to enable accurate projection of battery performance.
3. Has a textual alert status display in addition to alarms.
4. Has a USB monitoring port that speaks a decently-designed and fully documented wire protocol, probably JSON datagrams a la GPSD.
I could write the firmware, but I don’t have the chops to design the hardware. Anybody game?
Sounds like a good Kickstarter project. Go for it.
I would absolutely back this on kickstarter, even knowing that hardware kickstarters are risky. (I have been burned on 2 hardwere kickstarters that arrived years late and were notably different then advertised).
The reason is that the mere existence of such a thing will set a floor on the quality of saleable USBs. If there is a free alternative, then commercial units cant afford to be too much worse then that, unless they have some sort of lock in.
This is a case where a stealable licence might work faster then a viral GPL-like. Vendors can make professional assembly, casing and pretty-skinning into their value-added, and get units into stores where grandma can buy them.
>The reason is that the mere existence of such a thing will set a floor on the quality of saleable USBs. […] Vendors can make professional assembly, casing and pretty-skinning into their value-added, and get units into stores where grandma can buy them.
You’re basically describing my sinister master plan, here.
You could claim to be perfect, except that you have to work on your tendency to be too humble. It should read “.. my sinister master plan, here. BWA HA-HA-HA-HA!”
>It should read “.. my sinister master plan, here. BWA HA-HA-HA-HA!”
Correction accepted.
A few problems with the assertions made here:
(1) UPS systems *all* have deep cycle batteries.
(2) Marine batteries are less capable of deep cycle operation than “deep cycle” batteries used in UPS systems.
(3) Car batteries won’t last a usable 10 years
(4) Car batteries are not good for deep cycle applications and will be destroyed by a UPS or any other deep cycle application.
(5) Most moderate quality UPS system have replaceable batteries. All my APC and Tripp Lite systems have a way to replace the battery.
(6) UPS systems use standard batteries that can be bought on-line at places like Apex Battery.
(7) The life of a deep cycle battery is a well understood science, its 3-5 years depending on environment and usage.
I will grant the user interface and serial monitoring sucks, but most of your rant seems to come from buying a 3 year old UPS from a disreputable vendor.
>I will grant the user interface and serial monitoring sucks, but most of your rant seems to come from buying a 3 year old UPS from a disreputable vendor.
No. It was APC. :-)
Being APC means nothing.
Was disassembling dead battery packs from APC Symmetra, some of them had CSB batteries inside, some had really crappy non-name batteries. Both types of packs had APC QC slip and were bought from reputable reseller.
Non name batteries tended to swell quite a bit, to the point where we had to really use force to pull packs from the UPS since battery swelling expanded metal casing of the pack as well.
They also had tendency to die much faster than CSB loaded packs.
A few problems with _your_ assertions:
I really doubt that marine “Gel” batteries are less capable of deep cycle operation than UPS batteries. We use gel batteries in solar power installations, and many of them are marketed originally for the marine market. 3-5 years? You have got to be kidding me. I lived off-grid for 16 years. I got the whole 16 years off the lead-acid batteries powering my well system (they _did_ fail catastrophically at that point, right after I’d sold the property, but before the deal closed!). On my main power system (also lead-acid), installed later, it’s 15 years and still counting. Those batteries were guaranteed for 10 years, so they’re clearly not expected to have a 3-5 year life; and they’ve been very deep-cycled.
And esr never suggested using car batteries, just that people have. I’d be surprised if they _haven’t_ had ten years out of them, though it’s probably not normal.
You’re comparing longevity of a proper battery (probably something around 100Ah minimum) with smallish 7.2Ah, when both see probably same discharge current?
500VA UPS will pull some 30A on full load from single 12V battery.
You really believe they will both have comparable longevity? It is impossible.
Also, Puekert’s law (higher the current, lower the perceived capacity).
Also, as batteries get older, internal resistance increases, and for specific amperage pulled from the battery, it’s voltage will be lower, and UPSes of course measure battery voltage and based on it decide whether battery has gone bad.
Had few examples where UPS cuts off at 10.5V, and battery drops pretty fast from fully charged voltage to 10V and then sits there for ages and actually puts out usable current for much longer than battery that seems healthy. While such battery is clearly not 100% good, it will still work for lower loads, or paralleled and properly balanced with other batteries, or with inverter that doesn’t try to be too inteligent.
I think it would be a major achievement to get any mainstream UPS vendor to sell the batteries separately. That solves the “5 years in a warehouse” problem, and gives some tiny incentive to build the rest of the electronics with a design aesthetic above “disposable.” On the other hand, standardizing the batteries probably ends all UPS battery innovation, which we probably do still want for a while.
Also, who puts a monitor on their UPS? Dude, that thing is a power-sucking boat-anchor that will cut your runtime in half. ;)
>Also, who puts a monitor on their UPS? Dude, that thing is a power-sucking boat-anchor that will cut your runtime in half. ;)
No, the kind of LCD strip display I have in mind has really small power draw.
Plus, surge protection. Don’t want the monitor to get blown out.
Most UPSes feature surge-suppression ports that aren’t powered by the battery. The monitor can (maybe should, if only to keep the same voltages relative to ground) live there.
An eInk display would be fine for that context and the draw rounds to zero!
>An eInk display would be fine for that context and the draw rounds to zero!
Good point!
I’d probably go with ePaper. Much easier to interface and close to zero power draw, and a more responsive display
> >Also, who puts a monitor on their UPS? Dude, that thing is a power-sucking boat-anchor that will cut your runtime in half. ;)
> No, the kind of LCD strip display I have in mind has really small power draw.
I think Zygo was talking about powering your computer’s monitor from your UPS. And yeah, the old CRT-based ones would be a sure battery killer.
On the other hand, a modern flat-screen is only (stops and looks at the back of his Dell 20″ display) – huh – this thing claims it draws 1.5A, so over 150W. Worse than I thought.
On the gripping hand, a computer without a display is nigh on to useless, yes?
It can, if properly configured, throw itself into suspend or hibernate.
This can save lots of work.
How much power would it require for the UPS to speak JSON to an app on your phone via wifi? Everyone’s already manic about keeping their phone batteries up…
I’m imagining an app that would not only tell you time remaining, but would remind you during work hours that your UPS won’t make it until the next working day.
(I suspect most UPSes were designed with the expectation that if main power went down, they should wake the admin up, right the hell now, no exceptions, the redcoats are coming…)
So long as that’s not the *only* interface. I want something wired as an option, if only so that I don’t have to worry about security compliance for yet another wireless gidget. But the option would be nice to have.
Why put the WiFi in the device itself? A UPS is a solidly stationary device which is generally expected to be easy to get wires to. Just give it ethernet, USB, and a basic on-device display. If you want WiFi, add a WiFi AP to the ethernet network it’s plugged in to.
WiFi is for mobile devices and low-bandwidth “appliance” type devices which aren’t likely to be easily wired. A UPS is pretty much the furthest from either of these things.
Because a SOHO installation might not HAVE wired ethernet, or at least not convenient enough to run a cable from the base station.
I have two LED displays that draw over 100W each (high quality DACs, big panels, and really, really bright). I can push the desktop behind them to use a similar amount of power if I run all the benchmarks at once, but normally it’s 80-100W. So somewhere between 50% and 70% of the power goes into the monitor.
It’s a lot less power to run a network/WiFi AP on the UPS (5W-10W), ssh into the machine from a nearby laptop or phone, and do whatever it is that needs to be done (hit save & quit?), than it is to try to do that with the desktop’s usual monitor.
It depends on the larger power-loss strategic goals. One strategy is to try to keep running without service degradation and without power as long as possible–essentially making the desktop a heavy laptop. You can do that if you have unlimited UPS battery money, but if you really need continuous user interaction during a power failure, a laptop will do it better (maybe that’s the only thing laptops do better, given your very specific CPU requirements).
The other strategy is to use the UPS strictly to ride out short power bumps (like up to one minute), then keep running just long enough do a suspend-to-disk (or long enough to start a backup generator and feed the UPS from it) for anything longer. If you can take a break while the first part happens, and the second part happens automatically, then the monitors don’t need power during that time.
I have a big NEC PA-301W LCD monitor that puts out a surprisingly large amount of heat. Not as much as a CRT but it’s quite warm. It’s also surprisingly heavy – 24kg.
The batteries absolutely need to be a separate module from the electronics module. Batteries age much faster than solid state electronics,and your electronics feature upgrade cycle should be completely uncoupled from your battery purchase and replacement cycle.
I have a simple emergency power system that is two 12v 35 amp hour AGM batteries, a couple of solar charge controllers, and wiring with standard power pole connectors. Add a couple of inverters for 120 v AC. Completely modular.
Something like this approach would work for UPS design.
Almost ALL but the lowest priced consumer crap has a way to replace the battery.
Come to think of it, cordless power tools already have fairly high-current batteries, in sort-of-standard from factors (they at least have to support a line of tool products), and they’re available at many hardware stores. Would these make good UPS batteries?
>Would [cordless power tool batteries] make good UPS batteries?
They’re probably the same kind of deep-cycle gel battery I was pointing at – given the electrical requirements and low weight they’d almost have to be.
Cordless power tools are either NiCd (if they’re old-style batteries; for example, DeWalt 18V) or some sort of Lithium chemistry (e.g., DeWalt 20V MAX or flex volt batteries). Compared to a marine battery their big disadvantage is rather rapid self-discharge and very poor tolerance for deep discharge. (Lithium controllers correct for this by dropping the battery out at low voltages, NiCd tools normally don’t have controllers.) You also cannot get anything like the total energy storage of a marine battery in a single unit. I don’t know about lifetime on a float current; I suspect the NiCds won’t age well that way, at least.
I think (but am not sure) the cordless power tools, at least the consumer-grade ones, all use some kind of lithium-ion cells with proprietary interface and packaging. I wonder if batteries intended for small off-grid solar installations would be a better bet?
The constraints for tool batteries are different to those for UPS — namely, tool batteries need to be light enough that the tool is usable, while UPS batteries can be any mass that doesn’t cause the server rack to plunge through the floor.
Tools, notebook computers, etc. use battery technologies that are optimized for stored joules per kilogram (and hang the expense, more or less).
I believe lead-acid is still the champ in terms of stored joules *per dollar*. For a UPS, joules stored wins over portability.
>I believe lead-acid is still the champ in terms of stored joules *per dollar*.
But maybe not if you factor in expected lifetime. Marine-grade gel batteries are at least supposed to last longer and handle deep discharge better. (I have not tested this myself, mind you.)
Gel cells are lead-acid. They just use a gelled electrolyte, making them less likely to spill acid everywhere.
@esr “But maybe not if you factor in expected lifetime. Marine-grade gel batteries are at least supposed to last longer and handle deep discharge better. (I have not tested this myself, mind you.)”
The radio tower that allows me to get Internet out here in the boonies has 2 auto-parts-store marine batteries in series for 24V. It uses solar panels and a cheap charge controller to manage the battery state. The system was first put in service May 2009. The first battery replacement was winter 2014-15 (5.5 years).
Mind you the batteries sit out in in the hot sun in the summer (in a battery box) and in the freezing temps in the winter and are subject to long periods of discharge during cloudy low sun periods in the winter.
Based on this and other experiences, I’d say 5 years is about what you’d expect in a well designed system with garden-variety lead-acid batteries.
Other battery technologies (e.g. like those used in off-grid solar installations) could go 10+ years easily, but at much greater cost, and probably not maintenance free.
In this application, those batteries will rarely be deeply discharged. That radio equipment, compared to the car starter those batteries are designed for, is like a garden hose compared to Niagra Falls. If you removed the solar panels, the thing would run for a couple of weeks.
Yes. Correct. I do believe the 5ish year mark is a largely unavoidable process of simple materials degradation due to thermal cycles, electro-chemical cycles, and simple age. Note that these are pretty-much the cheapest of such batteries available.
In addition to cordless power tool batteries, I’d look at the power sources used for studio strobes. Something like this one. They have been used for keeping laptops charged, as well as their intended purpose of powering strobes.
I don’t know what your datacenter-dwelling friends consider an “enterprise grade UPS” but the ones we use in REAL data centers do most of what you’re describing. They also cost more than a house and take almost as much space as one. We’re talking Liebert grade stuff here, not APC grade. The batteries that go in them resemble car batteries more than they resemble the consumer grade gel-cells that are found in most low end units. The charge levels are constantly being measured, and the onboard computer can tell us with a good degree of accuracy how long we can expect the current load level to be able to run on battery power.
Obviously, however, we only need our UPS banks to run long enough to keep the data center running until the generators are running … for a single computer you probably want to run it for a while and/or shut it down cleanly. And what you’re saying is true: there are no “pro-sumer” grade units available. I’d totally prefer to see a “batteries not included” design that requires you to go over to Sears for a couple of car batteries to back it.
I second this. Love to be able to swap batteries in and out; the value is in the electronics and the human/computer intermediation.
The UPSs we had in Iraq would run long enough for us to get servers powered down.
Well, they’d probably last longer than that, but when the power dropped the chillers went offline, and usually didn’t come back. I don’t think we had UPSs on them.
And we went from 75 ambient to 105 ambient in about 5 minutes.
But they always did keep the power up.
And we went from 75 ambient to 105 ambient in about 5 minutes.
/me dies just thinking about the temperature shock.
This was Iraq.
Temperature shock was not a primary concern.
We left some equipment up to melt in place if necessary because it /was/ a matter of life and death.
Cisco switches and routers (at least the ones we had) can run for a while with room temperatures at about 110.
Oh, very much understood. Doesn’t change the reaction I have to the mere thought of it.
Well, if it’s any comfort, the chips wouldn’t walk out of hteir sockets because of the weight of the dust.
> We’re talking Liebert grade stuff here, not APC grade.
One of the other people who had been in the chat with Eric and me about UPSes, himself also being a pro systems administrator, specifically cited Lieberts as being not good. His complaint – “Insane bespoke firmware.”
Having worked with some of them in the past, I can’t disagree.
Agreed here, but I _can_ tell you that with the network card (at least on the older unit I used to run), you could collect a number of stats including draw (input, output, and bypass) and estimated runtimes via SNMP, which wasn’t terrible.
The panel instructions were cryptic as hell, the manual was almost unreadable, but I’d give a couple of small body parts just to be able to have a cheap (APC-class) ups that failed to by-pass rather than failing off. That would at least fix the annoying problem that has lead us to simply ride the lighting on our network at my current gig, which is that the stupid things fail off and then you have to go to the site to restore the network at random times (and usually when driving conditions are at their worst). Not good in a 24/7 manufacturing shop.
As a side note, the UPS I ran (which we bought used) was originally purchased by Industrial Light and Magic for their render farm, so was probably part of the server farm used to render the Prequals. Kinda cool.
At the manufacturing shop I worked at we had process control relays to lift backup power to primary power so that it wasn’t a situation of primary power, fails, switch to UPS, then switch to backup generator power. Instead we switched the backup and primary power feeds so that the UPS was the backup of last resort for the computer room power.
Again, it goes to show you how a lot of these things, like your mentioned ‘fail to backup’ just aren’t easy to configure with UPSes.
You don’t BUY a UPS like that, each is a custom installation specific to the building it will support.
I would buy instantly 2 ups which would be over-engineered. Go for it ;)
I feel your pain, and I’d put $100 into a Kickstarter to get this done. I bought the $200 unit from Costco, and the experience was bad. I plugged it in. It is rated for 800 Watts. My computers pull less than 100 Watts. After a few hours what happens? Yep. An ugly screeching sound, and the UPS cuts out the power. I try again. Happens again a few hours later. Finally I just leave it overnight, and now it seems to work ok. The UPS actually caused power outages! I guess I should have given it some time to charge first. But… that makes no sense. The UPS is rated at 800 watts of throughput, and I was using a fraction of that; surely it didn’t MATTER that the battery wasn’t charged up? Surely it could charge itself in the background while acting as an expensive surge protector in the meantime?
There you go, using logic again.
Disappointed that there isn’t already a link to the opensource project to solve this problem.
Come on Shenzen rockstars, design us a UPS control board with a decent charge controller, battery condition monitor, and inverter. Sell the 3d printed case and the control blinky parts on http://www.seeedstudio.com and let us source our own deep cycle cell battery. Be a hardware hacking hero.
Cordless tool batteries are a good idea. The electric bike battery industry may also provide another option.
A more DIY solution would be banks of 16850 Lithium Ion batteries. These are commonly used in E-Cigarettes, (and Tesla Roadsters). You can wire them up into to battery packs as large as needed, and they have a fairly good cycle life. Plus, you can swap out old cells as their performance degrades.
18650 batteries are awesome. My weapon light mounted to my bedside AR-15 uses one, and pumps out 1000 lumen with some pretty great runtime. I’d love a UPS that used a bank of them with a decent charge controller.
APC recently came out with a new line of UPS products based on removable lithium battery banks.
Honestly, I don’t trust APC that much.
Shite products and expensive. I’ve had nothing but problems with them over the years. They worked fine as long as I planned on not using them for more than a few minutes at a time and replacing them every couple of years.
So, yes please! Is there some hardware guru who could design this? I’d happily fund a kickstarter for this.
This. Tesla has an incentive to get batteries that maximize consistent performance, so if that’s what they’re using, it’s what you want. Maybe you just want a Tesla Powerwall.
Rather than making a UPS, make a charge-controller unit that plugs into AC and two or more discrete banks of these batteries [such that you can be replacing the batteries in one bank while the other(s) keep your system running, which has both Ethernet and USB interfaces for info.
> Open hardware design, open-source firmware design, open-source device-monitor code.
Add to this (even if it’s assumed by the above, it needs to be explicit):
Open device-monitor protocol. People who want to write their own monitor code can do so without reference to the open-source code (that might, if they’ve seen it, “virally infect” the code they write) so long as they adhere to the protocol spec. The open-source implementation can be the reference implementation, but that’s no substitute for getting an open protocol.
I was going to suggest this project be named “OpenUPS”, but a quick search shows that’s taken by something that can only put out 24V.
>Maybe you just want a Tesla Powerwall.
I just reported on G+ that I looked seriously into the Tesla Powerwall when it was announced. Really liked the concept but concluded that I just couldn’t justify spending that many kilodollars for the amount of dwell wattage it offers.
At present and given my power usage profile, flashlights and individually UPSing must-stay-up devices remains far more cost-effective – like by an order of magnitude. I think that gap would need to close a lot before whole-home power cells seriously threaten UPses.
It also depends on where you live. PA has a very different amount of sun reaching the ground than does CO.
Here it would make more sense.
But I’ve got a friend who’s working for a company that expects to radical change that industry if they can get to production.
Big IF after Solandra.
I also wonder if part of the problem is we have the wrong component boundaries here. For your use case, at least, I suspect it would be simpler and more efficient to build the battery into the computers’s own power supply (which is mostly just a AC/DC converter), maybe with a internal data port of some kind… makes this project a lot more complicated, though.
They’ve had those since the early ’90s, that I know of. Plenty of room in an old PC power supply case. There’s not a whole lot of room even in a full size ATX power supply, though. In “compact” and pizza-box machines, the power supply box is mostly full of power supply; no room for batteries.
When I lived in Australia I had a few 60hz thingies that really didn’t want to run on 50hz.
The 220->110 conversion was fairly trivial, but the 50hz to 60hz was EXPENSIVE.
What I almost did was take a 220/50hz battery charger and attach it to two 6 volt batteries in series and then to one of these https://www.amazon.com/Sunforce-11240-Inverter-Remote-Control/dp/B000WGNNUQ Not really a “UPS”, but on the other end you’d get good, clean power.
Might have to replace the batteries more often though. And I expect some loss.
> I suspect it would be simpler and more efficient to build the battery into the computer’s own power supply
For a laptop at relatively low power consumption, yeah.
For the Great Beast of Malvern, which has a 1000W (output) power supply – nope, nope, nope. You want all that heat and bulk and noxious chemistry in its own box.
One aspect I feel as absolutely unacceptable is what APC does (or did in the past) with the DSUB-9 ports on their devices.
Childishly assuming you could interact with such a device like a network router or so (connect it to a computer, run $ cu -l /dev/ttyX, get a shell, query status, run commands) results in a power outage of everything that is connected to said UPS!
For some reason unknown to me they decided to use RS232 as a protocol but to use a non-standard (non RS232) pinout. Connecting a normal RS232 DTE closes a circuit (or so) that instructs the UPS to shut down and cut power from the south end.
Why would anybody do something so stupid?
Of course that mistake can be avoided by consulting the documentation before connecting something to the device, but in the situation of a power outage a technician has to act quickly and inactivity of usage is the major property an UPS must have in such a moment. Sadly they don’t.
>For some reason unknown to me they decided to use RS232 as a protocol but to use a non-standard (non RS232) pinout. Connecting a normal RS232 DTE closes a circuit (or so) that instructs the UPS to shut down and cut power from the south end. Why would anybody do something so stupid?
Ah. I think I might actually know the answer to this.
Before UPSes speaking badly-designed wire protocols over USB monitoring ports there were ancestors speaking badly designed wire protocols over RS232 ports. Those, in turn, descended from a mercifully now extinct species called a “contact-closure” UPS that didn’t speak any protocol at all, but could only signal the host machine that line power was down by changing the state of some RS232 handshake line.
The botch you dealt with probably up-gunned a contact-closure design so it would speak text on the other RS232 pins. This sort of shenanigans was pretty common about a decade ago; one of the UPS monitor daemons I contributed to carried a lot of info in its documentation tree about how to work around – or emulate – various crazy nonstandard pin assignments on the monitoring port’s RS232-like D-shell connector.
I’m guessing you, poor Grumpy, rolled snake eyes and got a unit where pulling DTE to active interacted fatally with the implementation of the legacy contact-closure feature. Yes, very stupid indeed, but I’ve seen stupider in the past. As bad as today’s UPSes are, their ancestors used to be worse.
Jeezus.
These people should be hooked to the Taser of Enlightenment by nipple clips while being beaten with a cluestick.
>These people should be hooked to the Taser of Enlightenment by nipple clips while being beaten with a cluestick.
Yeah. There’s a reason I stopped working on that kind of software and gave up configuring my own systems to use it. The vendor suckage just got too depressing.
I use apcupsd. It sucks less. Thank you.
We actually had that thing, where if you plugged the UPS into the wrong serial port on the back of the Sun box (the one *not* under thrall of the APC supplied software) it would cut all power. Took a few tries to figure it out. About 20 years ago in my dev lab.
Sweet Ghu, William, that made me laugh so hard the tears are running down my cheeks. My poor kitty retreated from my lap to the fireplace and is looking at me like I’ve sprouted another head.
In a previous life I had two UPSs, one in the form of a laptop, the other in the form of a small commercial product. The latter ran my router and the smallest infrastructure machine, the one tht did the failover for DNS and the external web site (1 page).
The laptop was elderly, but I could buy a new battery. The UPS was new, and contained “no user-servicable parts”. After the latter failed, and its replacement failed, we said to heck with it. We bought a new laptop and a high-bandwidth ethernet dongely-thing and made it the failover for the router, the DNS and the website.
That disruption is a “Solved problem in computer science”, just looking for a biz-boy to make it a product.
I don’t forsee quality consumer-grad UPS’s existing for much longer, let alone being ripe for disruption. The main reasons:
1) everyone is switching to laptops and tablets. Desktops are a niche device
2) whole-house or partial-house battery backups will be come more prevalent
#2 is the interesting part. Off-grid homesteaders and preppers lead the way for a while, but now with the advent of affordable solar power, this is becoming more mainstream. Consider integrated products like the Tesla Powerwall. Or more prepper-oriented solutions like Renogy and Go Power that make charge controllers that you can hook up to arrays of car batteries. The Tesla solution is more self contained, and more expensive, but pretty darn friendly. You need to know a bit about electronics and home wiring to make the DIY systems work, but you can swap out individual car batteries every 7-10 years or so as they degrade in power.
>1) everyone is switching to laptops and tablets. Desktops are a niche device
Aaron, you’re usually a pretty smart guy. I’m surprised to hear you repeating dim-bulb conventional “wisdom” like this.
For lots of desktop use cases, only a full-sized keyboard and a monitor too heavy to hinge will do the job. Forget about the size of the CPU case, it’s the ergonomics of these peripherals that will keep “desktop”-class systems with mix’n’match parts alive, even if the processor module the peripherals plug into is as physically small as a smartphone (and might in fact be your smartphone, as I’ve projected before).
Full size monitors, in particular, imply a power draw for the whole system that will need a UPS if you live in an area where power outages are common during the storm season. Again, it doesn’t matter that your CPU module might be a thing with an inboard battery (smartphone or laptop); your monitor won’t have a battery of its own. This in itself implies a continuing market niche for UPSes.
As a semi-separate issue, I think we’ve already reached peak tablet. If they were going to take out desktops they would do in laptops first – the ergonomic spread between what you buy with these two device classes is less. And this clearly is not happening.
Instead, tablets occupy a niche as Internet consumption devices and gamer toys, basically anything where you only rarely need a keyboard. I think consumers have already found the limit here; more people carry laptops than tablets and this seems unlikely to change.
If you stop thinking of the computing-device market as being stratified by CPU power or physical size, and start thinking of it as being stratified by ergonomic task requirements, I think you will predict better. Major stratifiers are questions like “do I need a physical keyboard?” and “does my productivity increase with every square foot of pixels on my display(s)?” and another is “can I cope with the way a chiclet keyboard is going to fuck up my hands if I use it constantly?”.
The Surface Pro line (at least the SP4 I own and the Surface Book I’ve pawed) are laptop-killer tablets (though the SPB seriously blurs the line). The iPad I have from work would be suitable for light-duty for everything I do for work if it had a full-up browser and I could attach it to the same VPN I can my work laptop. Higher than “light-duty” if I could attach a “real” KVM to it when I’m not mobile. Because all my paid-for work is done in a browser or an SSH client, either connected to a remote device.
The only thing I can’t do on my SP4 that I use a full-up “laptop” for is graphically-intensive gaming (and for that I’ve got Steam Connect). My next machine will be going back to a desktop because of the ease of intermittent partial upgrades to the hardware, but I might just end up running the SP4 as my “remote console” to it for a while.
Now, the SP4 runs a full-up desktop-grade OS (in my case Windows 10, though I have a linux VM squirreled away on the SSD someplace). If I want I can (and have) attach a full-size KB and monitor to it
The Surface Pro line (at least the SP4 I own and the Surface Book I’ve pawed) are laptop-killer tablets (though the SPB seriously blurs the line).
But it’s not the tablet part of the device that’s a laptop-killer, it’s the plug-it-into-a-KVM-and-it’s-a-desktop part.
Really, once we start seeing that sort of functionality properly implemented in phones (the hardware is there already, and has been for years, but the industry is taking its sweet time doing the software side right), I expect both tablets and laptops to disappear, leaving just phones and KVM switches (and a few desktops for when modularity matters and mobility doesn’t).
You may be right, but it depends on how big the phone form factor gets, and how acceptable wireless headsets get (and how much more room for improvement in voice command there is).
Phone screens are too damn small for reading on, and the on-screen keyboards all suck.
> For lots of desktop use cases, only a full-sized keyboard and a monitor too heavy to hinge will do the job.
I’ve got one of these:
https://www3.lenovo.com/us/en/laptops/thinkpad/x-series/x140e/
Hooked up to https://www.ikbckeyboard.com/product-page/kd87 , http://www1.la.dell.com/sv/en/corp/peripherals/monitor-dell-e2311h/pd.aspx?refid=monitor-dell-e2311h&s=corp. and a trackball (the keyboard and trackball through a USB switch). It’s got 4 really small cores, 16G of ram and will run up to 2550xwhatever (I have a higher resolution monitor, but that’s hanging off my work lapdog using the the same keyboard and trackball, which is also hooked to my iMac.).
I can get the 22 inch monitor, one of the laptops, power cords, monitor stand, keyboard and trackball plus other cables in a Pelican 1600 for transport almost anywhere in the world.
That method of analysis works for me. I copy edit for a living. It made a big improvement in my work experience and productivity when I invested in a screen that could display two pages side by side. I’ve used my wife’s laptop and the virtual keyboard on the household iPad, and neither would be endurable for a day of typing and cut/pasting. The iPad is a marvelous substitute for a backpack full of game books, and it makes for quite a pleasant two-person video experience, but processing text on it just doesn’t work; even browsing is less satisfactory than on my desktop. Give me big display area every time!
> I copy edit for a living. […] Give me big display area every time!
Yes. Users like you, who are content creators rather than mainly consumers, are the reason the “desktops are disappearing” talk is just nonsense. Tower cases may become as obsolete as the Sun pizza-box form factor is today but the affordances of human eyes and hands pin desktop ergonomics in place.
It’s a more stable technology cluster than most people realize – 44 years old this year, dated from the Xerox Alto that pioneered the style in 1973. That’s more than half the entire lifetime of the stored-program electronic computer.
Tower cases may become as obsolete as the Sun pizza-box form factor is today but the affordances of human eyes and hands pin desktop ergonomics in place.
First of all, the young whippersnapper wonders why the pizza box form factor is obsolete. Being the type to keep two desktop machines on the same desk, it would be nice if I could stack them.
Secondly, even if the tower becomes obsolete as a single-machine form factor, I can see it living on as a server rack form factor for raspberry-pi-sized machines.
It’s obsolete because if you’re not running 5 hard drives, running more than 2 monitors or replacing stuff every couple months as a hobby your needs can be met with a laptop (using an external monitor, mouse and keyboard) or something like this: http://www8.hp.com/us/en/products/desktops/product-detail.html?oid=15292277#!tab=specs.
Up to 64G of ram, Latest generation I7 processor, two DVI ports, ethernet, VGA (WTF?) and a REAL SERIAL PORT.
This is essentially the pizza box formfactor for the 2010s.
Or something like this: https://system76.com/laptops/galago
Designed to run lInux, up to 32G of ram, HDMI and MIniDP (dunno if you can use both at once), nice screen, plenty of USB ports.
I’m seriously considering that laptop for my next computer.
My choices are constrained by what Apple offers, because I’m a user rather than a programmer, and because, having run both the Mac OS and Windows on my current desktop system, I really don’t want to suffer with the Windows interface unless I’m being paid to do so. And the Mac Mini is just way cheaper than any Apple laptop, and takes up less desk space.
Funnily enough, for me the Windows interface is not great, but tolerable, and it’s the Mac interface I don’t really want to deal with without payment. That said, I grew up on Windows.
I certainly don’t have a tower case. My desk holds a Mac Mini and a 1TB external hard drive. Among its other virtues, it will cost less than half what C paid for her laptop when I replace it with the current model. Laptops strike me as luxury goods for consumers rather than tools for working people.
Of course, the nearby UC campus is covered with students carrying laptops everywhere, but in today’s university environment, I think a lot of them may have to be counted as consumers.
It depends on your ego and your needs. I use this laptop (see reference above) for “work” occasionally. One of my client’s clients needs basic Linux administration, and with a 22 inch display I can run 3-4 terminal windows (cygwin) and a web browser.
I can also take it with me when I travel–which is harder to do with a desktop.
Also it was pretty cheap. I think I paid ~600 for the base laptop, 16 G of ram (moved the 4 gig stick to daughters lapdog) and the 250G SSD (to replace the spinny drive).
“Laptops strike me as luxury goods for consumers rather than tools for working people.”
It depends on the user and the business case. I have been issued a laptop, not a desktop, in my last three office jobs. My primary usage is a combination of web browser-based tools and Microsoft Office. This is probably the norm for 90% of the people in the HQ buildings.
The laptop can be used from my desk, from the table in a conference room, from the airport, or from my home office. Wherever I go, I have my files, my applications, Wi-Fi capability, etc.
However, at both my office desk and my home office desk, I have an external keyboard and one or two external monitors. The laptop gets plugged into these much of the time, which makes it much easier to use.
William’s single-location usage is just one possible pattern. It’s not universal.
At my current job, I have a laptop that plugs into a “dock” to provide the extra monitors and keyboard you describe. It doesn’t even have a spinning HD in it; it’s all SSD. I store everything of value on a network share because that SSD is not backed up, and the NASes are. It happens that I’m one of the handful of people who also has a desktop machine, but that’s because my job has some requirements that are best met by letting me handle some things there. So one of those external screens runs an RDP session to the desktop, while the other screens are left for Outlook and web browsers.
So I agree that the desktop isn’t “dead” for everyone, but for the overwhelming majority of people who had them, say, five years ago, they are.
What’s actually happening is that tablets are morphing into laptop-like things, and vice-versa, a bit like Kipling’s “The Beginning of the Armadilloes” — for reasons you state, that there are many tasks for which you just need a damn keyboard and nothing else will do; and yet the handiness of a device that you can tuck in the crook of your arm and poke at with a finger or stylus — or, if you’re creative, do precision things like draw with that stylus directly on the screen — cannot be overstated.
Microsoft really kickstarted this with the Surface, but I’m typing this on a Samsung Chromebook that bears many of the same ergonomic features. (It’s one of those things you can fold backwards, automatically disabling the keyboard so you can just poke at the screen.) Laptop-tablet hybrids may be the future of low-end computing devices.
Recently I switched to a laptop only world, and it works even for serious stuff. The laptop (2017 Macbook Pro) is hooked to a 49″ 4K Samsung TV/monitor, and via BT to Mac overpriced keyboard and trackpad. USB connects to backup and external drives.
From a UPS standpoint – the computer won’t go down on a power outage. Display will, of course, but that is less tragic. Disks will be unreachable but most frequent storage access is to the internal flash drive.
So, yeah, in my case, the UPS may no longer be needed – supporting the idea that laptops are eating into that space.
How do those external drives handle a power outage? IMO, they either need to be on a UPS or run a filesystem that absolutely can’t be borked by an untimely loss of power.
For keyboards I use separate keyboard connected through usb – no issue there. Can do the same with display (I didn’t bother so far as I have no need for bigger display right now, my only issue is that many apps don’t work correctly with extremely high dpi displays (my laptop does 3840×2160 on 15 inch display)). And it is quite powerful so no issues there as well.
Oh, they’ll live on, but they’re becoming more and more niche. I don’t think the absolute number exist to make a huge consumer revolution. If there is something better, it’ll come from the DIY community, not some scrappy startup that seeks to dethrone APC and its ilk, because the market just isn’t large enough.
I’m perfectly happy with a laptop for most things, but yeah, external monitor, keyboard, and mouse is absolutely necessary for any real work. That being said, the only thing I have on a UPS is my router and my NAS. The rest just on surge protectors, because again, if my monitor goes out, I can still un-dock my laptop and shut down.
Most people who aren’t engineering types I know rely entirely on cloud storage (whether proper backups or just Dropbox like services).
Again, I’m not arguing that laptops and mobile are “better”, but more and more people are moving in that direction, because for their use cases (which is increasingly and depressingly Gmail, Facebook, Amazon, and Netflix) it’s simply good enough, and takes up much less space.
@esr: Instead, tablets occupy a niche as Internet consumption devices and gamer toys, basically anything where you only rarely need a keyboard. I think consumers have already found the limit here; more people carry laptops than tablets and this seems unlikely to change.
As it happens, I am configuring a 10″ Android tablet, whose intent is to replace a laptop as travel device, with something smaller and lighter. Plug in an external keyboard (or a USB hub and plug KB and mouse into that) and I’m largely good to go. I’m not actually on the machine that much when traveling, so use cases are check email, some browsing, the occasional doc or spreadsheet file, and a game or two. Anything beyond that can wait till I’m at the desktop at home. In landscape more, the display is large enough (1280×800), and the Logitech portable KB is reasonable. (A happy surprise is that the one I have has a full sized USB port, so I don’t need to mess with an OTG connector to plug in the keyboard. I *can’t* get external storage plugged into that port recognized, but don’t really need it.)
The major annoyance is that I’ve been unable to root it – 6.01 Marshmallow apparently patched the flaws previous rooting exploits used, and everything I’ve tried bounces off. That’s not a deal breaker, as I can do what I want without rooting it, but it is annoying.
>Dennis
That utter lack of control over the device and Google the primary reasons I don’t own a tablet (other than my two cellphones, which are just small tablets these days).
@William O. B’Livion
I have reasonable control over the device. Rooting would give more, but most of that would be the ability to remove instead of just disable vendor installed bloatware. I have enough control for most needs.
It’s primarily used in local mode, has various protections when its online, and I basically don’t care about Google.
>Dennis
“6.01 Marshmallow apparently patched the flaws previous rooting exploits used, and everything I’ve tried bounces off.”
I did a quick search on duckduckgo and quite a few hits come up on “6.01 marshmallow root”. I don’t know if any of these actually work, but at least some users are claiming it can be done.
@Cathy:
I’ve looked too. On prior devices, Kingo Root did the job. I do it by connecting to a Windows host via USB and the Kingo desktop version pushes the exploits. Kingo also has an apk that can attempt to root *on* device without the host connection. (A friend who is all Linux was able to use the apk to root an older model of the tablet I have.) Kingo bounced off, in host based and on device versions.
King Root, a similar product with a similar name but different manufacturer also bounced off in host based and apk versions.
Framaroot, which appears to be apk only, also failed.
My device does have a developer option to unlock the bootloader, so it looks like I could flash a custom ROM and get root that way, but that’s something of a last resort.
I don’t actually require root to use the device as intended – I’m simply accustomed to having it. It’s like an itch I can’t scratch.
But I’m not in a hurry to scratch it, and can wait till a known to work on my device solution appears. It’s like Linux distros – I want to spend my time using the device, not hacking to make it usable.
>Dennis
Since I am usually travelling, I am usually on a laptop. And when using the laptop I seldom use more than four gig of ram and 200 gig of SSD drive.
But when I get back home to my real computer, it is like night and day. Going from my laptop to my real computer is like going from my phone to my laptop.
And, when my UPS fails, and I find myself forced off my main computer to my travel laptop, I resent it mightily.
“But when I get back home to my real computer, it is like night and day. Going from my laptop to my real computer is like going from my phone to my laptop.”
Assuming you use an external monitor, external keyboard, and external mouse with your laptop when you are not traveling, I’m not clear on why you perceive so much difference between laptop and desktop.
If you aren’t using these peripherals, you are missing out.
That said, I am typing this on my desktop with two external monitors — but this is my gaming machine, with a good (for its day, now somewhat long in the tooth) video card.
Careful what you wish for. Before the open source community can get its shit in one sock, the UPS industry will in fact be disrupted — by Apple.
The PowerPod will be a large, white, glowing dome that sits on your desktop or the floor. It will sport USB-C and Thunderbolt jacks only; no mains-power outlets (though a special adapter ($130 retail from the Apple Store) will let you plug your Mac Pro or iMac in). Its firmware will be a special build of iOS, and it will project icons and other UI elements onto the inner surface of the dome, which is touch sensitive. Not that you’d have much occasion to use it, as a typical use case would be:
1) plug PowerPod into wall
2) plug devices into PowerPod
3) profit!
The PowerPod would automatically push notifications to every connected Mac, iPhone, iPod Touch, or iPad, as the power jacks double as monitoring ports. With minimal extra work it could push notifications to your iPhone even remotely, through iCloud. Macs could be configured to start a Time Machine job when power gets transferred to the PowerPod’s batteries, and to shutdown cleanly when the batteries are low, with just a couple of changes in the Settings app. The PowerPod would even be able to accept voice commands. (“Hey Siri, how much charge is left on my PowerPod?”)
The advertised dwell time would be enormous for a consumer-grade UPS, and the PowerPod could be expected to uphold that dwell time throughout its entire service lifetime (five or six years). However, after about a year, in order to meet that dwell time the actual voltage supplied to the PowerPod’s ports will drop significantly.
No other devices will be able to speak the PowerPod’s monitoring protocol. Some enterprising hacker will crack the protocol and put proof-of-concept code up on GitHub, but Apple will have it DMCA’d within the month.
Regardless, everyone on Hackernews will own one, and wonder among themselves how Linux and Windows people ever get along without such an elegantly designed, easy-to-use device.
>to shutdown cleanly
“Shutdown” is not a verb.
It is now.
No, it isn’t.
“Shut down” is a “phrasal verb”, but “shutdown” is not a verb. Follow the link.
English is three languages in a trenchcoat, with irregular and unsavory habits.
I perhaps should have emphasized “now.”
None of which is the joining of a verb and a particle into a single word in some conjugations, but not in others.
It never fails to amaze me when people of above-average intelligence and computer literacy can be so damned careless with natural languages.
Because, at universities, they tend to be drawn more towards the linguistics department, which scientifically studies how languages actually do work, rather than the English department, which comes up with arbitrary prescriptions about how they think English should work.
I don’t think that’s really relevant. Linguistics departments tend to study primarily spoken languages. But the difference between “shut down,” “shut-down,” and “shutdown” is almost entirely a feature of the written language, which is governed by convention. I make my living as a paid prescriptive linguist, a job that includes knowing those conventions. In this case, the convention is that when a preposition is used as a particle to modify the sense of a verb, the two are spaced; when the same words are used as a noun, they are typically closed up; if they are used as an adjectival modifier, they may be hyphenated (“The problem was the shut-down power supply”). Linguistics is certainly relevant to identifying which of those three cases you’re dealing with, but the typographic expression is a matter of custom.
On the other hand, it’s useful to have a typographic distinction that helps you tell them apart, as it makes the written sentences easier to parse.
I’m right with you, being a professional copy editor (among a few other things) as well. But then there’s the question of context, and there’s very little need to be pedantic about someone else’s misuse of a noncrucial word in the comment section of a blog.
It should be an eponymous law — copy-editing the Internet is and always will be a futile endeavor.
I used to be very careful with language, but let me tell you, it can be very freeing to vomit something up through your keyboard and project an, uhm, urban vibe. And annoying people like you who care about proper language usage is the closest gamma males like me will ever come to the juicy juicy feeling of having power over other people.
You need to start doing bench presses and playing with real weapons in a martial context.
MUCH better for you.
>You need to start doing bench presses and playing with real weapons in a martial context.
I second this, Emanuel. There’s no reason anyone as smart as you can’t upgrade. Lift weights, get some hand-to-hand combat training, learn how to shoot. Unless you’re very unlucky the latter two will not be skills you need to actually use, but you’ll walk taller and your T level will go up. Women notice these things, oh yes they do.
An empirical test for whether “shutdown” has become a verb yet might be whether strikers have started chanting, “On strike! Shutdown it!”
(I’m thinking of Tok Pisin where, I have read, “bugarup” is a common verb and can be used in such statements as “You burgaup im good!”)
No, it’s a command.
SHUTDOWN(8) BSD System Manager’s Manual SHUTDOWN(8)
NAME
shutdown — close down the system at a given time
SYNOPSIS
shutdown [-] [-h [-u] | -r | -s | -k] [-o [-n]] time [warning-message …]
DESCRIPTION
The shutdown utility provides an automated shutdown procedure for super-users to nicely notify
users when the system is shutting down, saving them from system administrators, hackers, and
gurus, who would otherwise not bother with such niceties.
And shell command lines are not English sentences.
They are now.
*laugh*
thanks for that.
cat reply | sed -i ‘s/are now/have been for some time/’
Like all the best satire, I’m only 99% sure that this is satire and not a prophecy or a message from a post apocalyptic time traveler.
There is no need for usb anymore in a world of ubiquitous wifi.
you could start with something like this: https://en.wikipedia.org/wiki/ESP32
and program the needed stuff in lua.
I like the idea of using an e-ink display, but (in the dark) a backlight would be helpful.
Thing is, a whole new generation of very smart battery controllers now exist for all sizes of lithium batteries and it would be nice to leverage that – but I have no idea if any of the manufacturers of such have made their code public. certainly tesla made their patents public….
I don’t think wifi is quite right for this application–you’d have to use some sort of SNMP like trapping mechanism and that could get mucky.
Unless you just want to poll for status and don’t care about sending the hosts a “shutdown gracefully” signal.
> I like the idea of using an e-ink display, but (in the dark) a backlight would be helpful.
Back it with luminescent paint of the type that glows for hours after being “charged” by exposure to light. Not optimal for readability, but then you’re probably only going to need a couple of lines of text. You won’t be reading novels on it. :-)
This, and include LEDs around the perimeter that “charge” the paint while AC is functioning nominally. The fact that they’re lit is part of the UI that way. Provide a button someone can push to energize the LEDs while depressed even when they’re otherwise off if they’d like.
And maybe a button to turn on the monitor. IMHO, there should be a monitor power jack on the UPS, but it should not power the monitor when the UPS runs out of power from the mains. However, if someone pushes a button, the monitor should be powered on for a couple minutes – just enough to make sure the attached device has properly shut down.
Why would I want a remotely-hackable gizmo that will never get OS updates when I could get something that connects over USB?
’cause it would be open source and get OS updates unlike anything “smart” that runs over usb?
Why should it be remotely hackable? Put a “read only” switch (hardware switch) on the OS firmware side, store persistent variables in a itty bitty little piece of non-volatile ram and you’re off.
This isn’t a general purpose computer that does lots of things, it’s a freaken smart battery.
Last time I checked, anything capable of WiFi comes with serious amounts of firmware and a neverending list of security problems (with new ones affecting “all devices released so far” every year or two).
Now you might be thinking a ESP8266 serial-to-WiFi module puts the WiFi on the other side of a serial interface, complete with AT commands (Hayes just won’t die), so it’s harder to hack the UPS controller…except that anyone who owns the ESP8266 now has a repeater into the WiFi network, and can spoof the UPS control protocol too (unless you use SSL which is another security treadmill).
A Bluetooth interface might be OK just because nobody does anything important on Bluetooth, but you’d have problems if there was a power failure and there are more than 7 UPSes in the same building.
That said, there are YOLO customers out there who would want this capability. Maybe leave a header socket on the board so they can plug in a WiFi module if they really want this.
I gave up on UPSs when I noticed that my last 3 power incidents were due to a dying UPS, and I could not remember when I last had a blackout in my area.
I replaced it with the same sort of high end consumer-surge protector that I have on my television, etc. Sure my computer is going down if the power goes out, but I save my documents and most applications these days are more resilient to unexpected power-cycle.
>I could not remember when I last had a blackout in my area.
Lucky you. In a good year we get three or four of those here in south-eastern Pennsylvania (and I’m talking a wealthy exurb that the power company cares about, not a rural backwater). In a bad year we might have a dozen.
Summer thunderstorms and ice on the wires in winter are our big problems. Elsewhere, like in California, they have rolling brownouts for purely political reasons (and that’s going to get more common).
When I lived a few minute walk from Kendall Sq, Tech Sq, AI Lab, Draper, etc etc etc we had power events DAILY.
Our local substation killed people. Great infrastructure.
I don’t get blackouts often in my area, but I do get transient spikes and other power issues now and then that would be nice to have a good quality UPS to filter out.
I would definitely contribute to a kick starter for a UPS that could operate off an array of Deep Cycle SLA batteries. Though, it might be nice to have a whole-house UPS like the Tesla Powerwall that ran everything but the high power equipment (leave off the electric stove, microwave, and dryer, but power the lights and other things.)
My dead UPS story is that I was sitting talking to my mother one morning and heard something fall off my desk behind me. I went to see what had fallen and couldn’t find anything on the floor. At that point my mother (who was sitting opposite me as we were talking) commented that she had seen a flash from that direction. At that point I looked at the power LED on my desktop: dark. There was a brief moment of panic while I wondered if it was the computer’s power supply or the UPS that had failed, and when I determined that it was the UPS, whether the failure had propagated to any connected equipment. Fortunately, nothing but the UPS was dead, but isn’t a failing UPS supposed to give a warning tone apart from “bang” when it dies, given that there was no indication (brightening/dimming/flickering in the room lighting, failure of equipment elsewhere in the house, etc) that there had been any kind of transient on the mains that would have killed it suddenly?
Low-end UPSes don’t give warnings, they just fail. Sometimes they even fail during their bi-weekly battery self tests, leading to questions from ops like “why does this server reboot exactly every 13.7 days?”
A low-end UPS is typically unconditionally worse than a surge protector. They might be worthwhile for a few months if you have daily outages, but after a year or two, the dead battery will be the least of the problems.
I feel compelled to note that plugging my APC Back-UPS Pro 1500 into my Mac Pro’s USB instantly did the Right Thing: the Mac knows what the charge state on the UPS is and whether it is running on battery, and will shut down cleanly when there’s only a few minutes power left.
It’s possible to not suck, but the Linux world apparently hasn’t achieved it yet.
And since when are you warning us you’re about to rant?
> It’s possible to not suck, but the Linux world apparently hasn’t achieved it yet.
I’ve been a Apple guy since 1989, and a a Linux guy since 1993.
When my (currently 9 year old) 27″ iMac dies, it’s VERY likely that it will be the last Mac I will have ever bought.
Apple is behaving worse than Microsoft these days.
It is not, as noted above that Apple plays nice with APC, it’s that APC plays nice with Apple.
I’m kind of surprised RHEL doesn’t.
Agreed that it would be nice. My wish would be for more devices with switchable inputs from 110 volt 60 cycle to 12 volt DC or equal. It offends me to go from DC to quality sine wave AC so I can go back to DC again. That’s not going to happen either.
My observation is that as mentioned the market for a consumer grade UPS has been driven so low by the rise of phones, tablets and laptops – combined with the low performance hence lack of benefits – that the market incentive is insufficient for a really superior UPS with a battery that itself is a couple hundred dollars. I paid far more for my last top of the line car battery from Sears than I did for my first car. Looking at Boeing’s experience with the 787 I’d sure be leery of fancy lithium battery power supplies that target consumers.
Time was any power user had at least one UPS now few bother. Off hand I may know more people with solar cells and off grid ability than anything close to a meaningful UPS. Might as well lament that Windows 10 doesn’t feature a GUI interface hex editor prominently.
Agreed that I’d buy a bring your own batteries really smart device. As it is my main machine is a laptop with a high end i7 that doesn’t lose any data during frequent power failures. Everything else depends on operator switching. Sometimes reliable fail safe automation is more trouble than it’s worth.
>Time was any power user had at least one UPS now few bother.
Can’t be true. If it were, MicroCenter wouldn’t bother keeping stacks of UPSes on the floor – power users are exactly their main market.
The other tell is that there is still enough product-development money going into UPSes that I see different models every time I have to buy one. It’s just sad that the money isn’t being spent on anything a user can’t see on the box, leading to the nasty crab-bucket equilibrium I described in the OP.
As you mentioned, the recreational marine industry has battery monitoring kit that makes data available on the NMEA 2000 bus (similar to, but more proprietary than NMEA 0183) but they’re stupid expensive – Maretron’s is $500 per battery. If you do find someone to go ahead with this project they might want to look at the marine units from Xantrex and Victron as a starting point. There are similar units for RVs; maybe some have data output.
What now? It’s not like it was 15 years ago. In SF at least we haven’t had power shortages for many years. No blackouts except for some short ones restricted to certain neighborhoods. Just the usual glitches.
>In SF at least we haven’t had power shortages for many years.
That’s in SF. Talk to somebody in rural Central Valley about this. Makes a difference which areas have money and clout.
The reason I say it’s going to get worse is that California utilities aren’t building out to meet even current loads, let alone future ones. They’re caught in a regulatory/environmental/NIMBYism straitjacket – it’s a lot like the California housing crisis, and the looming water crisis. (New England has similar though less severe issues.)
Not to worry, though, metro SF (along with the wealthy parts of LA and Silly Valley) will be the last affected. The utility bosses know their asses will be toast if rich people start feeling the effects.
California’s problems date back to 1996, when the utility companies engineered a deregulation bill, which passed both houses without one dissenting vote, then got signed by the governor. The utilities then sold their generating plants a electricity rates tripled… Essentially the problem was a bill with insufficient consumer protection getting signed into law… and everything has gotten worse since.
https://www.pbs.org/wgbh/pages/frontline/shows/blackout/california/timeline.html
People complain when their ox gets gored in a bill aimed at a particular race, gender, or sexual choice, but the worst damage is frequently done by obscure bills which change obscure regulations – and those regulations are frequently “obscure” because they are working properly.
In political terms, the average person is watching “professional wrestling” while some mousy little guy they’d ever notice picks their pocket.
> California’s problems date back to 1996
California’s problems date back to the state legislature being taken over by Marxist envirowhackos. That happened in 1970.
Funny how not allowing any new power plants to be built for 50 or so years eventually catches up with you.
Good summary of what went wrong in California:
https://www.nae.edu/Publications/Bridge/OurEnergyFuture/TheCaliforniaElectricityCrisisLessonsfortheFuture.aspx
This is what happens when regulation and consumer-protection laws go away, and what regulations remain serve the powerful special interests. It all boils down to the whole “by OECD-nation standards, the USA is a banana republic” thing. See also: telecommunications and transport, throughout the country.
“and what regulations remain serve the powerful special interests.”
To a very close first approximation, *all* regulations wind up serving the powerful special interests.
Regulatory capture. It’s a thing.
Spoken like someone who hasn’t spent time in a country where public infrastructure, government services, regulation, and consumer protection laws actually work.
I recommend Germany, Sweden, the Netherlands, or New Zealand. Hell, Canada would be an improvement.
Once again, American government suffers from deep systemic flaws which should not be confused with flaws of government in general.
“I recommend Germany, Sweden, the Netherlands, or New Zealand. Hell, Canada would be an improvement.”
Average price of electricity in Germany: US$ 0.35/kWh.
Average price of electricity in United States: US$ 0.12/kWh
So, costing close to three times as much is your definition of “actually working”? Whatever, dude.
>So, costing close to three times as much is your definition of “actually working”? Whatever, dude.
But…but…they virtue-signal with such passion and commitment!
Compared to that, what does it matter that European standards of living are so low in terms of actual, measurable wealth and options that an American has to live in Alabama (47th of 50 in income ranking) before purchasing power falls enough to be comparable?
(And that comparison was done by Swedes, who have it good by European standards.)
the base cost of electricity in the Netherlands is around a few cents per kWh, but it is the Taxes on energy (use) that increase that number to the 0.35 figure. That IS an incentive for CONSUMERS to look for alternatives, i.e. solar panels. And yes the surplus energy they generate may be delivered back into the grid and is payed back for!
You do not mention the ‘special interests’ which want the what becomes the straitjacket because they have no idea of the Unintended Consequences (a side plug for John Ross’ book of that name: at fee.org).
New York is presently suffering from a heating crisis. The governor (? someone?) recently called for everyone to turn down their thermostat to preserve heating old and gas. Why? Because greeny marxist special interests have banned fracking and banned construction of gas pipelines. Meanwhile northern and western Penn state is flooded with fracked gas depressing the market price.
Yes, and Oklahoma is flooded with much higher than usual seismic activity because as it turns out, fracking causes earthquakes.
@esr: The reason I say it’s going to get worse is that California utilities aren’t building out to meet even current loads, let alone future ones. They’re caught in a regulatory/environmental/NIMBYism straitjacket – it’s a lot like the California housing crisis, and the looming water crisis. (New England has similar though less severe issues.)
I’m not sure anyone is really building out. I’m in NYC, and the last thing ConEd wants to do here is build more centralized generating capacity. They are pushing energy efficiency and conservation for all they’re worth.
The first problem is the cost of new capacity, which will be round billions. To get the money, they’d have to do a bond issue, with various regulators tossing roadblocks. Once they have the financing, they have the issue of where to put it, and NIMBY comes into play.
And when they build it, what fires it? The likely solution is natural gas, though environmentalists will scream about fracking. The alternatives are oil, coal or nuclear, which most folks will consider worse.
And once they get the thing built and putting power into the grid, they face the issue of raising rates to get the money to pay for the financing when the bonds must be redeemed, with politicians and regulators.throwing more blocks into the path.
They don’t want to go there.
>Dennis
>I’m not sure anyone is really building out. I’m in NYC, and the last thing ConEd wants to do here is build more centralized generating capacity. They are pushing energy efficiency and conservation for all they’re worth.
Which is a nice idea and makes everyone feel virtuous, but is insufficient when baseload demand is rising faster than you can capture additional efficiency. Which is California’s situation and New York’s. And it’s a good thing – you don’t get economic growth and increasing standards of living without that rise.
The real problem both places have is sclerotic Blue-state politics. The gentry-liberal elites that run both jurisdictions are fooling themselves that conservation and renewables will be enough. They’re wrong, and – hm, I was going to say they’ll find out the hard way, but being rich gives you insulation against bad policy choices. It’s the poor, people who can’t afford PowerWalls, that are really screwed by high electricity prices and rolling blackouts.
@esr: The real problem both places have is sclerotic Blue-state politics.
That’s one of the problems. The biggest, I think, is the sheer cost. Paying for the new capacity will involve raising rates. That will be a political football anywhere, because nobody wants their rates to go up, and folks will will look for ways to push the burden onto others.
> That will be a political football anywhere, because nobody wants their rates to go up
Given a choice between electricity at a higher price, and no electricity, I would pay for it. But I’m lucky – I’ve got more than enough income to afford to do so.
For the poor? Maybe tired rates – you get the first N kwh a month for $X, and subsequent ones cost more. Then at least you can make a rational decision on what you can afford, as opposed to being subject to arbitrary loss of service.
(Nope, can’t afford to run the AC all day – but we’ll run it at night to sleep comfortably. And we won’t lose power and wreck all the food in the ‘fridge.)
Why, necessarily? More usage means more billing at the same rate–electric customers pay per kWh, not a flat charge (and even then, there’d be more customers).
Hmm, sounds like a project I could participate in.
I can probably do system engineering, prototype production and integration myself, and bring on and finance somebody to do board design and layout.
/me thinking of an overbooked schedule …
>/me thinking of overbooked schedule …
You could get me to do the wire protocol for sure – not many people I’d trust to get that right.
Lets brainstorm potential collaboration scenarios over email?
If you’re going to make this happen, it has to be completely open like the keyboardio keyboard; no spyware capability!
What I mean is, the design makes it easy to open up and visually inspect, and you can reflash the firmware or even swap out the arduino board. So, it is easy to verify that there is no spyware on board.
I note that CyberPower actually makes UPSes with nice, reasonably-informative LCD displays, rather than just beeping. They seem to otherwise be the same crap as other UPSes I’ve tried, but at least you don’t have to guess what the beep codes mean.
I wonder how much the limits and trade-offs of various battery chemistries, and the requirements of consumer safety standards come into play here. All consumer UPSes that I’ve seen have used sealed lead-acid batteries. SLA batteries have properties that make them useful in a consumer setting: They can be oriented in any direction, so there is reduced risk of spilling dangerous chemicals if the UPS were knocked over or dropped. They vent less hydrogen gas, which reduces fire/explosion risk. Their construction aids charging time. These useful properties however come at a cost. Their sealed nature means you can’t replenish or maintain the electrolyte, even if you wanted to. Thus they are less tolerant of overcharging, and their useful life is less than that of a well-maintained flooded battery.
In bigger settings like datacenters, the considerations get flipped. Now you want as much capacity as you can fit into the space available, and at the amount of money you’re going to spend on capacity, it makes sense to buy maintenance service as well, so that capacity you paid good money for will last as long as is practical. The hazards above can be mitigated by putting the UPS behind a access-controlled door away from those who don’t know what they’re doing.
Lithium ion batters pose a different set of hazards and trade-offs. They have much greater risk of thermal runaway and thus fire/explosion hazard. This requires smart circuitry that prevents under/over-charging/discharging from turning the battery into a bomb in addition to reducing it’s useful life. And some cases of manufacturer error[1] or case breach[2] mean you get a fire/explosion hazard anyway. But they can use more of their capacity than a lead-acid battery can, and are lighter to boot.
I’m not going to dispute the firmware situation. I’ve never touched it, only been on the receiving end of the functionality offered. You are right about the little Windows blob that can shut down the computer once the UPS reports that less than a preset estimated runtime remains. The undocumented and undiscoverable hell of the monitoring port protocol itself doesn’t surprise me terribly. It’s just good enough to get the blob to work. And good enough seems to be the order of the day with consumer goods.
It seems to me there’s a gulf in product quality and fuctionality between (usually cheap) consumer goods, and goods intended for professional/business/enterprise use, not just in UPSes, but goods in general. If you need/want something better than cheap consumer goods, but don’t need or can’t afford the whole enterprise-grade offering, well, good luck with that.
1. The 2006 Sony battery recall
2. Pretty much any case breach which causes a short that the control circuitry can’t stop.
I used to have a UPS and got rid because it was so crap that not having one resulted in fewer power outages.
If a good one existed, I’d be in the market.
I think this is an awesome idea. Please continue the regularly scheduled disruptions!
If someone will build a UPS like esr describes, I’ll be one of your first customers. I’ve needed several of these for YEARS now.
As they say … just take my money.
(But beyond that. I’ve written firmware for large battery maintainers and could contribute ideas.)
Amen. Me too.
Eric,
While I do not completely disagree about the “lowest possible cost” aspect of mid-size UPS units, I respectfully disagree that mid price small business grade UPSes are as bad as you state. The unit in question failed after 2.5 years and was replaced under warranty. Batteries are easily available in the secondary market and cone in standardized sizes. The SX1000 allows up to 4 additional banks of batteries for extended run times, although this does limit the accuracy of the runtime display.
have a APC SX1000. It monitors current and voltage of the cells and offers a fairly accurate prediction of remaining battery life. The unit also allows load shedding at certain battery levels, allowing more critical loads to be kept up longer. The unit has a 2-line LCD that shows details of the last 10 errors detected and the last 10 self test results and other status information such as current load, input voltage, etc.
I have not tried the serial comms stuff at all and it would be no surprise that the protocol would be difficult to reverse engineer. However, there is a simpler answer – just tap into the contact closure circuit that drives the beeper (or use a microphone) to determine that an alert is sounding. Once that information is known, physical intervention is likely required anyway.
Server auto-power down would be handy, but for many users running Linux, the system will be just fine even if it shuts down uncleanly.
Designing an AC connected relatively high power circuit is not of the faint of heart. UL, TUV and CE approvals are all requirements for any serious design. The amount of energy stored in the batteries is somewhat significant. I would be hesitant to buy any unit did not have all of these regulatory approvals.
As I said, your comments about overall build quality are right on target, but some available UPSes implement a reasonable feature set, absent serial communications.
You have a good point about the approvals for such a device. After talking with Mr. Dave Stetzer on the topic, each new device costs between $50,000 to $100,000 to get licensed and approved in the USA. So the kickstarter has to raise $100k just for government approvals and paperwork. All other costs are on top of that.
http://www.stetzerelectric.com/
So sell parts & pieces that can be connected together to make a whole. We could refer to it as being “open” or somesuch. Or just sell a kit.
Does that help?
I’m all for sensible safety regulations. I’m not for safety regulations that represent a barrier to entry reducing competition and innovation.
One of the ways that computing steps forward is to separate the control plane from the data plane. In the case of the UPS the data plane is the batteries and power circuitry and the control plane is (doh) all the sensors to detect the health of the batteries, input voltage etc. as well as the communications with the devices connected to said UPS.
It seems to me that there’s a clear use for a product that takes standard but generic batteries (marine batteries, car batteries, Lion batteries, other rechargeables) plus some appropriate middleware – i.e. connectors and chargers etc. – that can be controlled from an arbitrary raspberry pi or similar. Said controller can then spit out sensible error messages via web, usb, screen etc.
Doesn’t HAVE to be a pi because in many ways its overkill but OTOH a pizeroW would provide lots of connection flexibility as well as a powerful dev environment. It would be sensible to have the pi for the initial prototyping and then have a cheaper dedicated device for the mass production variant.
Arduino?
I’ve built line-switching circuits with both classes of control device. The failure modes are terrifying with that amount of energy involved.
We want the parts of the system that are touching line voltage sealed away, locked in a closed physical box with non-trivially-modifiable firmware that says “no” when software asks for the kind of stupid that will damage things and hurt people (like rapidly toggling a relay to keep it constantly arcing, or connecting main input and inverter output to the load at the same time). There are quite a few electrical rules that should not be violated no matter how open we’d like the software to be.
This is kind of like the PowerSwitch Tail approach–a box of line-voltage relay full of UL compliance with an optoisolated 3-5V TTL input that hobbyists can connect to their protoboards with impunity.
Same problem with the battery side. If there are multiple battery types supported, the firmware needs to know exactly which one is connected. This is why battery packs tend to include that firmware, especially those that tend to slowly turn into bombs when they are misconfigured slightly.
So we are left with the host computer interface, the component that runs the display and alarms, and the component that decides whether and when to power off the load when the host isn’t connected (assuming the battery firmware doesn’t override that by withdrawing battery power), that are reasonably programmable without extensive power systems experience and religious data sheet reading.
That’s three microcontrollers to provide the required safety isolation. Or, if you’re a UPS vendor, you use one, write all three parts of the firmware yourself, use a proprietary battery interface and sell the batteries so nobody will connect the wrong kind of battery, lock it all down against user modification, and handwave your way through the safety audits.
>We want the parts of the system that are touching line voltage sealed away, locked in a closed physical box with non-trivially-modifiable firmware
I had already had the thought that the PCB for this thing needs to be in 2 parts just so we don’t need to cycle through UL approval yet again every time we tweak the firmware or the low-power logic.
The microcontroller I’m playing with is under $2.50 in quantity–and it’s a Cortex M4F with onboard Flash, built-in BLE, and all sorts of other goodies. Purpose-suited MCUs would probably add about $3 to the BOM for the project–enough for a Big Ol’ Manufacturer to combine, certainly, but hardly a killer amount for the kind of project under consideration (especially if the innards end up reusable enough to slash NRE).
So, in the event that somebody uses your punch list to make one of these things, please add “Optionally fails to bypass” (the “option” can be a simple mechanical switch that connects/disconnects the bypass circuit upstream of the ATS) to that list, so when the UPS dies, the equipment will come back on when power is restored. Absolutely my biggest irk on the small UPSs. Ideally, the fail-over should be < 200ms, which is short enough that a switching power supply probably will carry the load over the hiccup, in the event that the device fails while the power is still applied.
It may have to be a mechanical switch (possibly flipped by a solenoid, but not a traditional spring-loaded relay that resets to a specific state when power is removed). Implementing the option in software seems hard, especially if this UPS is supposed to protect against intermittent voltages and overvoltages too (i.e. cases where bypass is the wrong response and there’s no software running due to lack of power).
…or put a dedicated hardware overvoltage sensor/switch module close to the AC input, so that it cuts off the input AC source from everything, including the UPS’s battery charger (which now no longer has to work at 450VAC) and bypassed outlets. Overvoltage then just looks like power failure, so the UPS responds with the inverter if it has charge, or switches off power if it doesn’t.
I note that “fail to bypass” and “withhold load power until a reserve battery charge is available” are diametrically opposite goals.
UPS designs that don’t fail to bypass are not failing to achieve a design goal–they’re trying to prevent rapid power cycles and out-of-spec voltages on downstream equipment. They’re assuming you’re swapping out UPS batteries on schedule so the UPS continues to work, and assuming “off” is better than “roller-coaster of voltage.”
>I note that “fail to bypass” and “withhold load power until a reserve battery charge is available” are diametrically opposite goals.
Very good point.
“I note that “fail to bypass” and “withhold load power until a reserve battery charge is available” are diametrically opposite goals.”
I wouldn’t consider UPS shutdown due to low battery a “failure”, I’d consider it proper behavior. If we are shutting down for that situation, then by all means kill everything, but if the UPS fails _unexpectedly_ , e.g. the charging circuit dies, a filter goes out, the control board takes a dump, then the device should simply cut-out and (if the bypass circuit is charged), drops the load to mains-power via the bypass.
In the big UPSs, the cutout is handled with controlled breakers, e.g. a breaker with a solenoid in it that when you hit it with a voltage, it disconnects. A dual-pole version of this would be what you would need for a bypass circuit. The one on the 150kVA I ran was about $600 when I had to replace it.
Have the monitor speak JSON to a MQTT server. An esp8622 would fit the bill for $3 in low quantities. Supply the source, and Robert’s your mothers brother.
I have been struggling with ups failure for a very long time.
Every laptop has a UI about its battery state. Ideally, when you plug in your ups, your mains computer will just display the same UI with no further fuss
The Distributed Management Task force already has a standard which is supposed to provide a uniform interface to laptop batteries and external UPS, which supposedly every laptop battery obeys, and every standard ups completely ignores.
Ubuntu already has code for hooking to the standard, thus it should not be that hard to connect up to the ubuntu battery display to an external battery.
Ideally what would happen is that you plug a USB cable between the battery controller and the computer, ubuntu asks what kind of USB device this is?
Oh, battery controller. Activate the battery display.
>The Distributed Management Task force already has a standard which is supposed to provide a uniform interface to laptop batteries and external UPS, which supposedly every laptop battery obeys, and every standard ups completely ignores.
DMTF conformance goes on the feature list. then.
>Ubuntu already has code for hooking to the standard
OK, there’s your answer, Jay. It’s not Linux that’s sucking, it’s the UPS vendors. You happen to have got one of the models that conforms.
My old UPS was an APC (I think), and Ubuntu treated it just like a laptop battery. When it failed with a bang, I decided I didn’t want to buy that brand again, and my new UPS doesn’t have anywhere near the ability/will to talk to my system that my old one did.
>My old UPS was an APC (I think), and Ubuntu treated it just like a laptop battery.
Was it some service daemon separate from the window manager, I hope? I sometimes miss out on hotplug device discovery because I run i3 – CDs and thumb drives don’t automount and pop up a directory window.
It looks like the DMTF USB GPS management standard may have obsolesced the whole concept of an explicit report and command protocol over serial emulation while I wasn’t looking.
Too bad about your APC experience. They used to be about the least bad of the major vendors. One guesses that unwise cost-cutting may have taken a toll.
IIRC, there was a package in the repositories with some APC-specific command line stuff, and I think a daemon. The graphical interface was the standard MATE battery interface that I knew from my laptop, I’ve not researched if that hooks into a system power management daemon independent of MATE or interfaces with the APC tools directly.
I would really like one (actually would purchase many) w a network port (not wifi!) that would allow shutdown of x number of servers monitoring w NUT or equivalent.
It has already been disrupted by the major cloud providers.
I’ve wanted a respectable UPS for a long time, but none meet my needs.
I need a UPS that will effectively slam the door on ground loops for my DAW – total isolation.
Add that to your spec ;)
>I need a UPS that will effectively slam the door on ground loops for my DAW – total isolation.
I don’t know what this means.
Searching for “DAW” on Wikipedia, it looks like he’s talking about a Digital Audio Workstation.
So my guess is that slight differences in potentials for different grounding conductors in his setup are causing current flow where there shouldn’t be any that’s messing with audio capture, and he wants a UPS that won’t have that issue.
Wikipedia also talks about ground-loops[1]. You’re essentially correct — different conductors supposed to be ground reference end up having different potentials. It’s the story of the Magic / More Magic switch. If Dan wants a UPS that makes sure all its outlets are at the same ground potential, that sounds reasonable. But if he wants a UPS that disconnects the ground from a port when a ground loop is detected — hell no. that ends up leaving a whole piece of gear ungrounded. Ground lift[2] should be done at the shields of the audio cables, not the chassis.
1. https://en.wikipedia.org/wiki/Ground_loop_%28electricity%29
2. https://en.wikipedia.org/wiki/Ground_lift
DAW = Digital Audio Workstation
I had a *lovely* experience where my audio interface (connected via USB3 to the DAW) generated a glorious hum whenever using my RF headphones.
I eventually solved that problem (after an embarrassingly long 24hr thunk time ;) with a $10 isolator placed inline with my monitor output from the interface to the headphones base station.
Products like HumX are expensive for what they are, and I always thought a decent UPS should be able to provide the same protection. Alas no.
I think Jeremy has the right idea for a safe solution.
I should perhaps add that my DAW is a Fedora-based system using JACK and Ardour.
If any budding open-source musicians wish to understand how to setup a pro-audio installation like this, I would be happy to offer advice.
My friend Phil Salkie puts in a plug for lithium-iron-phosphate batteries. He says they don’t quite have the energy density of Li-Co but have a longer service life and no thermal runaway issues.
Seconded. Was going to say this (and describe certification problems) once I got done reading the thread…
If your goal is to get a hardware/firmware/software reference design that can then be manufactured at reasonable cost (or used as the core of a product) then does the type of battery much *up front*?
We’re all going to have different needs–for example I have a UPS on my DSL router that is as useless as tits on a boar hog because (apparently) the upstream DSLAM goes down like a democrat intern every time the power starts to think about dipping…but I digress.
We all have different needs in different places–I’d like the one holding up my DSL router (pretty low power) to last 20 or 30 minutes with that load–which is 7.5 watts.
OTOH, my Dell (2000s era Precision workstation 690 with 4 dual core Xeons) probably uses that much in sleep mode.
So you have a design for the guts that does the work–part of which is a programmable battery charge monitor that can load different profiles based on the type of battery attached (I bet someone smart could look at how the battery was charging and tell what was attached). Then you have another *set* of reference designs for various sizes of UPS, ranging from a rechargable Li battery for keeping access points/DSL modems up for a couple hours to a 4 marine battery setup designed to keep a RV running for a week of cloudy weather.
Or am I misunderstanding something fundamental?
>Or am I misunderstanding something fundamental?
I don’t think so, but I think you’re proposing complications beyond what we should try to handle for a first design.
I think we should get one battery pack working first, then test with other packs that have the same specs, and only then start thinking about multiple profiles.
Of course there are way to factor the control software that are either more or less friendly to changing the battery-type parameters, but since I’m a competent software architect I would make sure that dependency is properly isolated in one data structure even if I were never planning to support more types at all.
I have a total absence of practical knowledge of the relevant technology and the kind of free-wheeling mindset that is good during initial design brainstorming but should be shot on sight during requirements gathering. Since that’s roughly where we are, you have been warned.
I envision three modules, with very small interfaces between them. The first is the battery unit, with D.C. in and out, and a small set of control and status lines. It’s purpose in life is to provide a steady D.C. voltage and an indication of how long it will continue to be able to do so. When not under load, it feeds back charge state and such things.
With luck, this will not require UL or equivalent approval, so building various versions for the various battery technologies can be done without having to get each variant certified.
Module two is the line-voltage unit: AC in, AC out, DC in, and status/control lines to the other two modules. As little policy as possible is done here, since any change probably leads to a recertification cycle. It is where the “safety” rules are enforced, like fail-to-bypass, only one source feeding the equipment, etc.
Module three is the controller proper, again hopefully able to avoid needing to be recertified whenever it changes. Ideally it wouldn’t need to care about the battery technology because the battery module hides it.
Your module 3 is the Pi.
This sounds a lot like the three-controller design Zygo was muttering about upthread.
It’s exactly the result of his comments bouncing around in my head long enough to buff off the serial numbers. I think I made the interfaces and responsibilities a little more explicit, and I think putting voltage regulation into the battery module might be mine.
Module 3 is definitely a candidate for a Pi, or a repurposed home router motherboard, etc.
The battery module as specified needs at least a microcontroller, or maybe something a bit beefier, but definitely sub-Pi level complexity.
I don’t know enough to decide if the high-voltage module needs that much smarts or if it’s better to go low-tech to ease regulatory approval.
The line-voltage module 1 should get most of its safety from hardware design, e.g. set up the relay wiring with SPDT relays, put a circuit breaker on the inverter output, so that software can’t make some mistakes and nobody gets hurt (assuming nobody is hurt by the load not being powered) if it makes other mistakes. If UL auditors are anything like the safety auditors I work with, they will insist on this.
The battery-voltage module 2 needs something with super-Pi robustness and a sub-Pi low-current sleep mode (module 1 doesn’t need low-current sleep mode because module 2 can just kill its feed from the battery). I’m not sure this module escapes the need for UL approval–the voltages are low, but the currents are not, and there’s also a charger in the design somewhere that will be consuming line-voltage AC input. If a fully-charged battery module breaks something or hurts someone, it can continue doing that for quite a while before it runs out of juice.
Module 3 is a small Linux host computer plugged into a dumb UPS with a DC output. Everything else gets turned on or off by module 3 using software AC outlet switch and inverter output controls exported from module 1 and/or 2.
I used to have a UPS, made in 1980, that used a standard wheelchair battery. When I got it, it was about 14 years old and the original battery was kaput. (Dated, so I’m sure it was original.) Replaced the battery for $80 and got another 14 years out of it before what little electronics it had finally got zapped and died, along with the battery. It was zero’d out multiple times, and could support an ordinary PC with CRT for about 2 hours. Dumb as dirt (all it knew was On vs Off) but couldn’t beat the lifespan.
I had good luck with my original APC and my first two Tripp-Lites — they lasted 7-8 years each, and were finally killed by lightning strikes. Since then, tho, I’ve had multiple Tripp-Lites die at around 2 years, and now it looks like the CyberPower units are doing the same — one that’s a bit over 2 years old is down to about 30 seconds capacity; the other is a little newer and still okay. (These are all midrange consumer units.) Well, now I know why!!
I don’t know if it’s relevant, but the current models have only been completely drained once. The ones that lasted several years got zero’d out several times.
At any rate — consider wheelchair batteries as an alternative to designed-to-fail-one-week-out-of-warranty. In fact, now I’m wondering if they could be adapted to an existing unit.
The APC Back-UPS Pro 1500 that Troy Hunt chosen in https://www.troyhunt.com/specing-choosing-and-testing-ups-for/ does have an LCD screen that shows, among others, expected remaining time (last photo in that blog post).
It just cost an order of magnitude more than cheapo almost-UPS…
@esr:
Half of what you want is here: a battery charger/inverter. You can attach almost any type of battery to thing thing and it provides AC power output.
On another note, I have a request for the device design: The ability to take multiple power inputs. Imagine a device which could power eg. your furnace in the event of an extended blackout. You want A/C charging/use for most short-term blackouts. But you (me) want to be able to hook up a solar panel/thermoelectric generator/something to be able to charge with whatever meager power I can produce, spitting it out in bursts to keep the house warm.
>he ability to take multiple power inputs. Imagine a device which could power eg. your furnace in the event of an extended blackout. You want A/C charging/use for most short-term blackouts. But you (me) want to be able to hook up a solar panel/thermoelectric generator/something to be able to charge with whatever meager power I can produce, spitting it out in bursts to keep the house warm.
This is underspecified. What voltage and wattage ranges do you want to be suppoorted as input? What physical connector(s)?
“Laptops strike me as luxury goods for consumers rather than tools for working people.”
My current observation is that working people are more likely to have laptops as their desktop. The critical tool is a Linux cluster or equivalent while the personal computer has become a combination smart terminal and Powerpoint creator.
I used to envy a friend who had a pair of Sun machines so he wouldn’t have to wait for one to finish before starting on the next. Now I can easily have more power than he has locally while all he does locally is R.
The real distinction is having as much or more money in monitors than the laptop, pizza box or tower case. Developers who compile locally with local power are a special case in a small corner of the enterprise.
To the extent this is a serious discussion about backup power I’d suggest not wasting time specifying a really nice practically universal commercial product that nobody will bring to market. That assumes nobody will bring it to market.
For my purposes at least it has to be the brains of a bring your own battery device. Else I can continue to power down when the line power goes down and shift focus. The real need for productivity is to keep the big monitors going
>The real need for productivity is to keep the big monitors going.
True. that’s what makes “just use a laptop” not viable.
Chipping in the business perspective rather than a tech one:
Does the “backup” function then call for a 7-inch tablet display on a long retractable power cable to kick in when the main power goes out? Leave the wide screen monitor powered off, I mean. In many of my hopes for a UPS. I want just enough PC i/o to see what’s running, save current status, and shut down cleanly. Salvage my existing productivity, not try to extend it during the outage. I don’t need a 30 inch monitor for that. Not completely sure I need a keyboard and mouse.
I do like the idea of a kindle-style e-ink display for the UPS status indicators. Also, just for me and mine, the “idiot” lights: green is main power on and battery full; yellow is mains on and battery drawing a fresh charge, red is mains out and battery discharging, and no lights at all meaning UPS battery exhausted. Maybe arrange the idiot light cluster (as above) vertically and add a horizontal row next to the yellow indicator, four blue lights. Four is a battery of 80% or more charge remaining, three is 60-70%, two is 40-to 59, one steady is 20-39%, and one blinking is “you had better fire up the outside generator really darned soon. “
>Salvage my existing productivity, not try to extend it during the outage.
I’ve written up three different user stories describing UPS cases. This one corresponds to SOFTLANDER; then there’s SUMMERTIDE (powering through short outages and brownouts) and WINTERMUTE (powering through long-duration outages).
Each of these scenarios has different requirements. SOFTLANDER is the only one where you might get away with not powering your normal monitor. The far end of WINTERMUTE is the data center case and not addressable without huge amounts of battery mass.
Consumer-grade and SOHO UPSes are really designed for SUMMERTIDE. They don’t have the host-shutdown capabilty required for SOFTLANDER, not unless you augment them with something equivalent to NUT or apcupsd. This needs to be kept in mind when contemplating the incoming flood of “Arrgghh! Yes, UPSes suck!” triggered by the OP. If you don’t address SUMMERTIDE you’re not addressing the problem most people buy UPses to solve.
@esr: I’ve written up three different user stories …
Did I miss it? Have you published this?
>Did I miss it? Have you published this?
I will soon. I’m waiting on a mailing list approval from freelists.
My use-case goes like this:
0) Handle spikes and brief outages in the 0-10s range that are common during storms, by switching the load to the battery transparently, then switching back as soon as the line is back in spec.
1) If the battery power drops below a certain threshold, initiate a save-to-disk hibernation. When the computer has shut down (as detected by power-consumption dropping to the trickle that keeps the hardware able to monitor the “power” switch) remove power entirely.
2) When line power is back in spec, check battery power to be sure it’s above a higher threshold before restoring power to the computer to let it boot up.
I see this as “SOFTLANDER” with some overlap into “SUMMERTIDE”, but I’m not sure how much. In fact, I see “SUMMERTIDE” and “WINTERMUTE” as a spectrum, defined by how long the duration of the outage is (and therefore how long before transitioning from 0 to 1 in my points above). The far-left end of that spectrum might be considered pure “SOFTLANDER”.
> In fact, I see “SUMMERTIDE” and “WINTERMUTE” as a spectrum, defined by how long the duration of the outage is
That is true. In my research I have learned that 15 minutes is the 95th percentile of the distribution of outage times. (I expect the distribution is Poisson but none of my sources address that.)
I have therefore set 15 minutes as the upper bound of SUMMERTIDE and the target dwell time for the UPSide-1 under a typical load which includes a host and a large monitor. That’s going to be in the near neighborhood of 300W and I think the way the math works out the battery needs a capacity of 1600mah. This is achievable and not even very expensive.
I apologize that I missed seeing ESR’s “use case” discussion before posting my own ideas about the differences in problems between Grandma’s and Corporal O’Reilly’s.
The problems of a military user (not the service host for a military data center or communication hub — those are still another big ball of stinkin’ wax) tend to lie at the extreme edge of a general range of cases in ESR’s SUMMERTIDE; not all of which arise from the problems of war. In Australia, the news makes the AEMO electric grid in Melbourne and Victoria sound about as reliable as a politician’s campaign promise. Anything that can keep the corporal’s reports making his suspense would be fully adequate for the rest of us.
So that in mind I’d suggest an upper bound measured by hour rather than minute.
In the alternative, a “UPS” with an alternate power “input” connector. That should allow a user to add on a 300 to 750 W inverter, powered with jumper cables to a car battery/electrical system, and which expects the UPS features to condition whatever powerline noise comes out of such a jury rig to a standard safe for the PC.
>So that in mind I’d suggest an upper bound measured by hour rather than minute.
Because of the range of milliampere-hour ratings available in battery packs, above 20 minutes gets sharply more expensive. Upper bound in hours would sure be nice but is very unlikely to be practically achievable.
I really don’t get why you keep insisting the monitor be powered. Provided that the UPS can reliably initiate suspend-to-disk (with a hook for user-defined operations like “sync my local data store to the cloud now” [assuming your router and broadband “modem” are also on the UPS]) , you don’t lose anything but the time the power’s out, and you massively reduce the power requirement for the UPS, making it that much cheaper to build.
What do you think you need your monitor for?
He wants to keep working for 12 of the 15 minutes his power supply will last, and use the last 3 to suspend to disk.
Since 12 minutes gets you (guessing here) through 90 percent of the outages, and the remaining 10 percent rapidly get into hours, that’s a reasonable use case.
Part of this is “server” v.s. “workstation” thinking. Until recently my use case was my servers. Now those are SEP.
Oh, and on my iMac and my wife’s desktop the monitor is inseparable from the computer–you can’t turn it off without turning off the other one.
>He wants to keep working for 12 of the 15 minutes his power supply will last, and use the last 3 to suspend to disk.
That’s right.
>Since 12 minutes gets you (guessing here) through 90 percent of the outages, and the remaining 10 percent rapidly get into hours, that’s a reasonable use case.
10 minutes is 90%. 15 is 95%. That’s according to Eaton’s capacity-planning guide.
If SOFTLANDER were the central use case UPS vendors sell into, you’d see USB ports universally on even the lowest end units, and a lot of vendor-supplied software analogous to NUT and apcupsd shipping with them. On the other hand, you can’t do WINTERMUTE in consumer/SOHO-grade UPSes at all. I have a much clearer grasp of the costs than I did a week ago, I’ve been doing homework. So we’re left with SUMMERTIDE – that’s the only use case the portable UPS units on the market actually fit.
The only respect in which my use case is unusual is that because repo conversions take a damned long time the cost of having them interrupted is really quite painful. This makes my load more datacenter-like…but I’m not planning the UPSide-1 around that. Gets too expensive if I try.
> If SOFTLANDER were the central use case UPS vendors sell into, you’d see USB ports universally on even the lowest end units, and a lot of vendor-supplied software analogous to NUT and apcupsd shipping with them.
Maybe you and I have been buying different UPSes, but every UPS I have ever owned has had a data port on the back, and come with media in the box that contained software for talking to the UPS and enabling it to trigger a safe shutdown or hibernate. Hell, I’m pretty sure that even the tiny UPS that powers my home’s router and cable modem has a monitoring port on it.
Granted, the vendor-included software has been Windows and Mac only, but if I am honest, I have to acknowledge using Linux as my primary workstation puts me into a small niche that probably isn’t worth the investment for the vendor to accommodate. That said, it doesn’t excuse the terribleness of the wire protocol.
I think UPS vendors are going for SOFTLANDER, but only the majority of cases, not the corner ones.
>Maybe you and I have been buying different UPSes, but every UPS I have ever owned has had a data port on the back, and come with media in the box that contained software for talking to the UPS and enabling it to trigger a safe shutdown or hibernate.
Our sample clearly differs. Maybe you’re used to buying further up-market than me.
My present UPS (which admittedly is pretty low end) was no port. I don’t generally see ports on the inexpensive ones I end up buying at MicroCenter because a battery went kaput, much less shutdown software.
In fact in a couple decades of dealing with these things I’ve seen lots of ports but I don’t think I’ve *ever* seen one with vendor-issued shutdown software. That’s why I got involved with NUT and apcupd a decade ago. I mean, I’ve heard some vendors do ship such a thing, it’s just never intersected with my experience.
Not disagreeing, just questioning.
Where are you seeing the data on duration of outages, and does it also show the incidence or frequency of such outages?
I’m wondering about the bad weather cases where the power may go out for 10 to 20 minutes, come back on for an hour, then go back down. (Continuing my military analogies, when the power is out while the #1 diesel generator s refueled, comes back up, then goes down again as #2 gets fueled.)
Does the data indicate the UPS batteries usually recover the necessary 20 minute charge in the interval between outages?
>Where are you seeing the data on duration of outages, and does it also show the incidence or frequency of such outages?
https://powerquality.eaton.com/thoughtleadership/choosing-ups/design-considerations.asp
>Does the data indicate the UPS batteries usually recover the necessary 20 minute charge in the interval between outages?
I don’t know. I’d love to find actual data on this.
I stopped buying the UPSes with no data ports years ago because they’re toxic waste I’d have to pay to throw away (i.e. lead batteries) with exciting random power-interruption features (i.e. they don’t fscking work) and undesirable failure modes (i.e. they turn off one day and never turn back on again, ever, so they don’t even work as heavy bulky power bars). You shouldn’t buy them either. I keep waiting for the day when nobody buys them any more, and they stop being made (though that day could be five years ago and they would still be on sale today).
A decade or two ago, Costco started selling APC units with an apcupsd-supported USB protocol and built-in display with menus and runtime estimation, at affordable prices. I’ve been using those ever since.
>A decade or two ago, Costco started selling APC units with an apcupsd-supported USB protocol and built-in display with menus and runtime estimation, at affordable prices. I’ve been using those ever since.
What vendor & model is this? And given these exist, what are you looking for in UPside?
A few minutes gives enough time for the computer to save state and cleanly shut down. However, the programmer also needs time to save state and cleanly put away his work that he can more easily pick it back up when the power returns. Keeping the monitor functional is kind of necessary for that.
Suspend to Disk saves state cleanly. When the system resumes, all user-space processes are restored to the state in which they were when the Suspend operation began.
> That’s going to be in the near neighborhood of 300W and I think the way the math works out the battery needs a capacity of 1600mah.
Um, I’d check your math –
300 W * 0.25 Hr [[15 min]] / 65% efficiency => 115 WHr
115 WHr / 24 V [[nominal pack voltage ]] => 4.8 Ah == 4800 mAh
Not impossible, but larger than you calculated. The 65% efficiency figure is a worst-case figure-of-merit for nominal stored energy to usable output at the AC sockets on the back panel; we should be able to do better than that (with corresponding reduction in the necessary capacity of the battery pack).
>Um, I’d check your math –
I did mention that I don’t actually know electronics, didn’t I?
> WINTERMUTE
And they say cyberpunk is dead…
@Christopher Smith: “Paying for the new capacity will involve raising rates.”
Why, necessarily? More usage means more billing at the same rate–electric customers pay per kWh, not a flat charge (and even then, there’d be more customers).
You’re an electric utility. You spent a couple of $billion to build a new generating plant. You got the money by floating a bond issue. When you sell bonds, you are promising the buyers that when the bonds come due, you will pay the face value of the bonds plus accumulated interest.The bonds have come due and must be redeemed.
That’s a lot of money, and you are highly unlikely to be able to fund it out of ongoing revenues if you haven’t raised your rates. Yes, usage will expand, but not that much.
And being able to float the bonds in the first place requires confidence among the buyers that you will redeem them when they come due, which will involve confidence that you’ll be able to raise your rates to have the money to redeem the bonds. The less confident the buyers are, the harder it will be to sell bonds, and the more you will have to pay in interest when the bonds come due.
It’s like taking out a loan. The first question the lender will ask is “Are you good for it? If they make the loan, will you repay?” Bonds are a different form of financial instrument than loans, but the same questions apply.
>Dennis
An open hardware UPS would be a pretty cool undergraduate senior project. Does anyone have any resources I could look into to gauge whether this would be out of my depth or not?
>An open hardware UPS would be a pretty cool undergraduate senior project. Does anyone have any resources I could look into to gauge whether this would be out of my depth or not?
Watch this space. I’m initializing a wiki you might find interesting.
Any UPS is channeling a lot of energy. I had 2@ trace 4KW inverters/chargers on a bunch of (Canadian) submarine batteries and solar panels, but I had lots of lightning arresters, and other protective circuits. We lost about 1/3 of the solar panels and our roof in a hurricane, but we were the only source of ice for our whole neighborhood for several weeks.
So it is all the probability of the use case. Sunny+hurricane=lots of solar and batteries = is not the same as northern winter = > need stored hydrocarbons like oil or gas.
I guess I am saying go cheap and take your chances, or go way more expensive, and live out a hurricane or other massive disaster.
So, it looks like the latest-and-greatest capacitors on the market might be the way to go. At 500 Farad and 16 volts, that’s about 64 kJ stored. Obviously, you can’t drain it to 0, and there will be some loss from the voltage regulator, but even at 50% efficiency, that’s enough to run a 150W setup for 3-4 minutes, and that’s with about $200 worth of capacitors. To work well, it would probably need to replace the main PSU, since stepping up to 120V, just to drop back to 5V is terribly wasteful.
>To work well, it would probably need to replace the main PSU, since stepping up to 120V, just to drop back to 5V is terribly wasteful.
This keeps coming up as a problem. The reason we’re stuck with the conversion losses is big monitors, really. 90% of the real use cases for a UPS involve powering monitors. Until every monitor has a power jack that takes a standard DC voltage over a standard connector, eliminating the PSU in a desktop setup won’t get us to a place where we don’t need AC.
@esr: The reason we’re stuck with the conversion losses is big monitors, really. 90% of the real use cases for a UPS involve powering monitors.
Your use case, perhaps. Consider data centers, with rack upon rack of servers in a “lights out” configuration, administered remotely via SSH. What big monitors? UPSes there have two purposes – power conditioning, including momentary blips, and providing enough backup power to let you gracefully shut things down and avoid potential data loss in a blackout until power comes back.
Big monitors are the least of your worries, because the monitor has to have a working system to connect to.
(I did once encounter a case where the monitor was the problem. The systems house I worked for had sold a Unix box to a customer, along with a console terminal and big UPS. Users connected via dumb Wyse75 terminals.
The customer complained that if they turned on the console terminal, the system rebooted. And it did. The issue was that the UPS had the capacity to power both server and console once running, but that particular terminal drew a lot of current on power on. The amount it drew made the UPS think a power event had occured, and kicked in the UPS’s safeguards so it shut down the server to protect it. The solution was trivial – don’t plug the console into the UPS. Plug it into a wall outlet.)
>Dennis
>Your use case, perhaps.
No, it’s not just me. I’ve been paying pretty careful attention to the many comments on this thread from people complaining about having struggled with bad UPSes for years.
>Consider data centers,
That’s a different case with different requirements than anything this crew designs collaboratively can hope to address. Cleverness doesn’t help when the core requirement implies literally tons of mass sunk into battery cells.
>So, it looks like the latest-and-greatest capacitors on the market might be the way to go.
I looked into this and nope. The discharge curve on supercaps is all wrong for a UPS-like deployment, where you really want dwell time of 15-20 minutes. Where a battery delivers close to constant voltage until it’s nearly drained, a supercap has steep linear falloff from the word go.
>To work well, it would probably need to replace the main PSU, since stepping up to 120V, just to drop back to 5V is terribly wasteful.
Replacing the PSU is also a showstopper. See my earlier reply to LP – monitors are the blocker.
The conversion losses are a shame, but a unit that can’t deliver AC has wandered outside of UPS territory and the normal UPS use cases.
I will make the comment here that the power infrastructure for modern electronics is rather screwy. We have whole bunches of things that really need low voltage DC, but use mains power because of infrastructural inertia. Ideally, we’d have computers, monitors, etc. that had plugs for 12V or 5V DC and no internal power supply, and then we’d have external power supplies, one per room or per desk (or per house, with DC wall jacks? I don’t know enough about the design of electric power systems to know the optimal distance between point of conversion and point of use for low voltage DC) that would power a whole bunch of things together, provide charge points for portable electronics, etc.
>Ideally, we’d have computers, monitors, etc. that had plugs for 12V or 5V DC
I agree. Alas, it’s not the world we live in.
The world we live in is moving away from DC power supplies, to boot.
In my day job I “own” around a dozen boxes that run on 48 Volt DC power supplied (which is then stepped down for the COTS bits inside the boxes). Those power supplies are basically impossible to find from the vendor, and at any rate the rooms they live in are likely going to get converted to standard AC power supply as soon as practical. (Which will involve me moving my gear out to an AC-powered area).
On the contrary, more and more gear is using low power DC (5V @ <1.5A), but it's all disguised as "USB power". The current (no pun intended) limit is around 10W (5V @1-2A, depending on some vendor-specific crap). USB-C is slowly infiltrating, though, with a max power spec of 100W (20V @ 5A)(!) so there's a defacto DC power interface – it's just shaped like a data interface (USB).
Good point. I’ve just transitioned to using USB-C for my work laptop and my primary phone. The number of adapters I haul around today is annoying, but I will pay that transition price to get USB- C everywhere.
When USB-C can supply the power to the device as well as the “Suspend to disk and then power off” signal on a single cable, we’re golden.
Already can – USB-C carries data, after all. It’s a USB connection after all.
The top end of USB current ratings use the data cables to carry some of the power. you can be a power cable with the higher current rating, or a power+data cable with a lower current rating, but not both (or at least not both at the same time).
Not sure if that applies to USB C, but it does to every USB standard before (and even if C dedicates some wires to data, someone will eventually decide they really need that extra 5W per wire…).
USB-C carries the mongo power (5V @ 100W, whee!) on a separate set of pins from the data and the “low” power, per wikipedia.
I don’t see a power supply deciding to add a couple more watts (at a lower voltage). I’m sure someone somewhere will do it, but not a purpose-designed UPS.
(FWIW, the 100W maximum is at 20V.)
That’s what I get for going on memory from a Wikipedia article, especially when someone above stated the correct information well before I posted.
I worded that poorly. USB-C can do that. The question is whether we’ll have device and OS support to do the suspend-to-disk-and-power-off, and USB-C-aware charging ports.
Those are both software problems. Consider that there already has to be negotiation from both ends of the cable for devices that can either sink or supply power depending on what the user wants to do.
> I don’t know enough about the design of electric power systems to know the optimal distance between point of conversion and point of use for low voltage DC
Long story short, it’s much more efficient to transmit AC power over long distances than DC (because you can do it at higher voltages, implying lower currents, resulting in lower resistive losses. As usual, there are complications on both sides).
Thomas Edison and George Westinghouse did this great debate at the end of the 19th century, as widespread electrification was catching on. Start with “The War of The Currents“, or ask any EE with a strong background in power distribution how this goes.
Arguably, having a medium-voltage (someplace between 12 and 48V) DC “bus” at the room level and a standardized way to plug desktop electronics of all sorts into it would be more efficient. But think of trying to get every manufacturer to agree on the standards, etc. (In fact, your local landline telco central office did just this with giant racks of 48V batteries.)
Long story short, it’s much more efficient to transmit AC power over long distances than DC (because you can do it at higher voltages, implying lower currents, resulting in lower resistive losses. As usual, there are complications on both sides).
I know *that* much, I just don’t know the specifics well enough to know whether you could have a household DC supply at a given voltage or whether it would need to be room by room, etc.
You could probably do my house (2600 sq feet stacked mostly vertically).
You might not be able to do Bill Gate’s house.
Also, some users occasionally power non-computer things with a UPS, like printers. “Little power bumps” suck at $5/interrupted page, or $100 for a ruined ink cartridge.
@Pouncer:>Does the “backup” function then call for a 7-inch tablet display on a long retractable power cable to kick in when the main power goes out? Leave the wide screen monitor powered off, I mean. In many of my hopes for a UPS. I want just enough PC i/o to see what’s running, save current status, and shut down cleanly. Don’t need a 30 inch monitor for that. Not completely sure I need a keyboard and mouse.
You’ve described the normal usage of UPSes – enough power to smooth out blips and keep stuff running through brief outages, and enough power to let you shut down gracefully if there is an extended outage. You may not need keyboard or mouse.
The UPS may be designed to communicate with the system and tell it to shut down without human intervention. The question becomes configuration, and whether the UPS is correct that shutdown is required. Your intervention may be “Get the alert, check the status, and determine whether or not to let the UPS shut things down.”
I do like the idea of a kindle-style e-ink display for the UPS status indicators.
I don’t. The advantage to eInk is two-fold. The first is low power consumption. Once a page has been rendered on an eInk display, no power is required to maintain it, unlike LCDs which require a constant trickle to keep the screen painted. Users of dedicated eBook viewers using eInk like Kindles report going weeks without needing to charge, because the screen is the biggest power draw. The second is that some folks report finding an eInk display simply easier to read for extended periods, and complain of eye strain trying to do so on LCD displays. eInk can also be read outdoors in direct sunlight, which is problematic with LCDs.
The disadvantage to eInk that lets it out here is that it’s monochrome. If you need color, eInk isn’t what you get. I do need color, so my eBook viewing device is a 7″ android tablet with LCD screen.
For the use case here, I don’t see the advantage to eInk. Yes, it takes power, but the amount of power is negligible, and how do you read it in the dark if the power does fail? And you aren’t going to be reading it for extended periods. You are unlikely to be looking at it at all unless there is a power event. For that matter, the display doesn’t even need to be active unless there is a power event. You want to be able to connect to the UPS over the network and monitor history and status from your workstation, and not need to look at a dedicated display.
And using eInk adds to cost. What is being discussed here needs to be cheap and modular, with a controller that monitors line conditions and can kick in UPS functions, and replaceable batteries it connects to to provide emergency power when needed. You want function and reliability, and you spend what you need to get that, but most of this should be off the shelf kit.
>Dennis
>how do you read [e-ink] in the dark if the power does fail?
Others have already proposed backing it with phosphorescent paint.
But I’ve been developing the opinion that this is the kind of gimmicky, over-clever solution that is best avoided. It is not something we can buy COTS and expecting a makerspace build to handle phosphorescent paint seems a bit much.
@esr: But I’ve been developing the opinion that this is the kind of gimmicky, over-clever solution that is best avoided.
Yes, since it’s solving a non-existent problem. eInk requires less power, but the power required to drive the UPS display is trivial. You add complexity for no net save.
>Dennis
The other side of the problem is what happens when it fails 10 hours into an 11 hour night cycle. Phosphorescent paint is usually crap and in my experience tends to decay over time.
Might as well put a tritium vial in there. When you can’t read the display any more it’s time to replace the unit :)
Has the advantage of making the device illegal in Berkeley since it’s a nuclear free zone.
E-Ink is an effectively opaque surface due to the solid white and black particals that fill each micro capsule, the lighting has to be in front of it. Therefore a phosphorescent paint backing is wasted.
Kindles used side mounted LEDs and a diffusion layer to light their E-Ink screens when ambient light is to low.
So then tritium vials would be PERFECT.
:)
> Yes, it takes power, but the amount of power is negligible, and how do you read it in the dark if the power does fail?
https://www.amazon.com/Streamlight-73001-Miniature-Keychain-Flashlight/
https://www.amazon.com/Pack-Super-Bright-Aluminum-Flashlights
https://www.amazon.com/Finware-Keychain-Flashlight-Batteries-Included
I’d suggest the latter because they’re less unpleasant to hold in your mouth while you use both hands.
Edited to add: Don’t get the red ones. If you’re trying to preserve your night vision, get the green–that way you can see blood.
Frankly if you DON’T have some sort of battery powered lighting device near to hand almost all the time you’re deep into a failure mode already.
>Frankly if you DON’T have some sort of battery powered lighting device near to hand almost all the time you’re deep into a failure mode already.
There’s an app for that.
Try holding that in your mouth while you get a tourniquet in place.
Note that DX or Aliexpress will happily sell you off-the-shelf eInk modules, available also with red or yellow one-color support.
eInk isn’t something only big customers like Amazon can implement these days.
(I mean, it’s gonna cost a little more, and you’re right about the relative power consumption being irrelevant.
But the design doesn’t need to be minimum-cost, and frankly it’s nice to have higher resolution vs., say, a 1982-level character display, which is what a lot of cheap LCDs seem to be.
I’d pay a few bucks more for a nice display with a frontlight, and since they can both use i2c comms, it’s not like there’d be a serious problem with having the option?)
>I’d pay a few bucks more for a nice display with a frontlight, and since they can both use i2c comms, it’s not like there’d be a serious problem with having the option?)
I’m going to try to avoid options in the initial design. They’re too likely to add more complexity than value.
Yale College used to and for all I know still does have residential colleges served with DC from the steam heating and generating plant. There have been other DC survivals around New England from early adapters.
I like the Surefire 2211 wristwatch/light for an always light. Like a handgun to fight to your rifle it’s a handy guide to something with longer life.
I do wonder how the idea of having subset of sockets with one designed as master – other sockets in the group can be configured to be powered only if the master takes current. This means that if this feature is turned on, other sockets presumably to peripherals such as monitors will have power to them killed if the PC goes to sleep or is turned off.
Also, what should be presented on the IPS display? Battery charge and remaining time is a must, but I think that having current power consumption would be nice to have.
Big button to turn off the alarm is also nice.
Also, what should be presented on the IPS display?
– Battery life/degradation/capacity.
– Peak power consumption.
– Power factor (ok, this might be a luxury item).
– Status of connected host.
– Estimated battery runtime at current consumption.
Big button to turn off the alarm is also nice.
This.
And a switch/setting to turn off the audible alarm permanently.
> Big button to turn off the alarm is also nice.
…and that button should not be confusable with the button that turns off load power, and certainly should not be the button that turns off load power.
Yes, I’ve had UPSes designed with one button that turns off the alarm (one press) and the load (two presses…or was it three or one, and is it a different number when mains power is off?), and no, I haven’t managed to murder the designer of that UPS yet.
Zygo — you might mean “execute” not murder. Justifiable …
Ill considered opinion,
In neighborhoods where the power is usually reliable (always on, always in spec) a UPS only kicks in for an unusual event. And only needs cover an outage of a few minutes (shut everything down safely) to a few hours (keep it going until the utility company patches the downed line, after the unusually windy storm.) My idiot lights, as previously described, work fine for me and Grandma. Salvage productivity.
In neighborhoods where the power is usually UNreliable (only on for part of a day, as the government imposes “rolling blackouts”; only available from diesel generators, on days after the fuel trucks have safely convoyed to the base; on, but with crappy fluctuations in voltage or frequency, etc.) THEN I would want a display on the UPS/ surge protector / line conditioning device. I want to continue advancing productivity even with most of the other office equipment dark. I would want to augment idiot lights to show how many hours of the past 24 the power has been on at all, and what the freqs and volts have been, and if the battery idiot light is yellow with two blues is it charging, discharging, or bouncing back and forth? This might be the use case for Corporal O’Reilly at a 2nd echelon base, for instance. And in such a case, every watt diverted to the display is one not available to staying productive. So, e-ink in the luxury model UPS?
@The Monster: I really don’t get why you keep insisting the monitor be powered. Provided that the UPS can reliably initiate suspend-to-disk (with a hook for user-defined operations like “sync my local data store to the cloud now” [assuming your router and broadband “modem” are also on the UPS]) , you don’t lose anything but the time the power’s out, and you massively reduce the power requirement for the UPS, making it that much cheaper to build.
What do you think you need your monitor for?
I think Eric’s concern is SUMMERTIDE – providing enough power to run the Great Beast and his big monitors so he can continue to work through a temporary outage, which is what he tends to see. WINTERFALL, where power is out for an extended period beyond what the UPS can cover is a separate issue.
I see the occasional power loss, when something overloads a circuit to my place and I must reset a breaker in the closet in the hall. My desktop goes down, but I largely don’t care. It boots off an SSD, comes back up quickly, and I don’t lose data because I’m normally in Firefox which updates its profile on SSD, and when I re-invoke it, I’m back where I was. I’m not doing stuff where data loss is a concern.
Eric does care because he is – the last thing he needs is the Beast going down in the middle of a repository conversion or a software build.
>Dennis
>I think Eric’s concern is SUMMERTIDE
That is correct. Furthermore, the grades of UPS you can buy in a store (as opposed to having custom-designed for your data center) are designed as though the vendors think SUMMERTIDE is the central use case of their customers.
@esr: Furthermore, the grades of UPS you can buy in a store are designed as though the vendors think SUMMERTIDE is the central use case of their customers.
And I think that’s a correct assumption. Most folks have power reliable enough that temporary outages are what the customer sees.
The last time I had to deal with WINTERFALL was the great NE blackout years back. I was systems/network/telecom admin for my shop. My UPSes allowed me to gracefully shut down servers before the lights went completely out. And my then Verizon landline had its own telco supplied power, so I could keep in touch with the rest of my company, monitor status and know when power was back at my office, and wander in and start bringing things back up. (Living in walking distance of my office was a boon.)
It was a “something I’ve done once and don’t want to do again” exercise. But it was also heartening in a way – the attitude of folks in NYC seemed to be “We’re all in this together and need to help each other. We can get back to our religious/sectarian/political arguments when the power is back.”
>Dennis
Instead of a full text display, one could provide decent error messages for the cost of an LED. Just hook the LED I/O port up to one of those tiny speakers. Then, you can drive the I/O port with a voice file. This produces decently intelligible speech. The voice can say, obviously, what is wrong.
>Then, you can drive the I/O port with a voice file
That’s an awful lot of complexity to pay to avoid buying a $15 part. No; violates KISS.
Imagine a rack of 30 identical UPS units in a tiny wiring closet, all simultaneously–but intelligibly–saying “THPTMTPPTDMMOTF OOPPWDOCWORRRWSFW DWPSPTWSHPCTPHFPHP LTLLSOSUCTPTPDEPTPUP CTSPUTPTEPEEEDCRRRR DSTMAUIAIPAIMAMMMCMT MMDININGURE.”
Even so, I admit I wouldn’t mind an audio countdown timer (“FIVE. Minutes remaining. … FOUR. Minutes remaining.”). But the load computer (there is a load computer in this plan, right?) could do that easily, and probably better. Eric is right–don’t build things into the UPS hardware design that could be just as well done from the load computer with a standard discovery and information protocol.
Maybe leave some empty space in the box to swap out the shipped control board with a roughly Raspberry Pi-sized device, and make sure all the important signals are available on a header somewhere. That should suffice for most geeky custom mod requirements, and the engineering overhead should be minimal (if anything, it skips a round of shrinkage between prototype and production enclosure designs).
>Maybe leave some empty space in the box to swap out the shipped control board with a roughly Raspberry Pi-sized device, and make sure all the important signals are available on a header somewhere
Interesting alternative: Have the inboard computer run OpenWRT. Control operations are ioctls in a custom driver. The management logic is a daemon started at boot time. Hey, why not?
I’d keep the base model simple and cheap–use a $2.50 microcontroller with a USB interface, and put all the complex behavior in the load host…
…but make it easy for anyone who wants to run OpenWRT on a $25 ARM Linux board powered from the USB battery (i.e. it can stay on while the load is turned off). Less up front cost for you, and you don’t have to personally support 20 different ARM boards because everyone wants different things.
> Imagine a rack of 30 identical UPS units in a tiny wiring closet,
I don’t think that’s really the target market for this, and if you did something like that to save a few bucks, well, choices have consequences.
>>A decade or two ago, Costco started selling APC units with an apcupsd-supported USB protocol and built-in display with menus and runtime estimation, at affordable prices. I’ve been using those ever since.
> What vendor & model is this? And given these exist, what are you looking for in UPside?
apcaccess says the current model is a Back-UPS XS 1300G with a battery from 2014, in service since 2015. Before that it was whatever model Costco was selling at the time (and before that, it was 200…6? A few address changes ago, at least) but only the last model has the display.
My favorite feature would probably be to not power up the load after power is restored until the battery was charged up to a configured minimum runtime–and a display telling me this, so I’m not trying to figure out why it won’t start after power is restored. There’s no point in turning the load computer on until there’s enough reserve to complete a boot and shutdown cycle if the power goes away again.
Another feature I’d like is to have the load host control socket switching, so the load computer can shed loads it isn’t using. Some UPS models have this feature as a load sensor on one of the outlets, but it’s unreliable and stupid when the load host has a USB connection and can control the UPS switching itself. This is also handy for other reasons, like rebooting devices remotely (it’s why I currently own so many Raspberry Pis, Power Switch Tails, and little boards full of relays). That adds a bunch of BoM cost though, and only people who like to remotely reboot devices would pay extra for it. APC charges a lot for this privilege.
Speaking of Raspberry Pis…If the UPS can be monitored and controlled via IP (wifi or ethernet) instead of USB, I’d be able to get rid of the raspberry pi’s that I currently have glued onto the sides of my APC UPSes.
>My favorite feature would probably be to not power up the load after power is restored until the battery was charged up to a configured minimum runtime […]Another feature I’d like is to have the load host control socket switching,
Added to starting points.
>Speaking of Raspberry Pis…If the UPS can be monitored and controlled via IP (wifi or ethernet) instead of USB, I’d be able to get rid of the raspberry pi’s that I currently have glued onto the sides of my APC UPSes.
I’m now thinking it might be smart to use a Pi as the onboard computer in the UPside.
Yes, more expensive than the Arduino or a $25 ARM board. On the other hand, it’s a development environment that’s really well understood. Hook the control switches to the GPIO, add a few A2D converters, and suddenly your UPS management is a boot-time daemon in the userspace of a Linux. We trade some hardware cost for tremendously reducing the complexity of the software engineering, and get stuff like an Ethernet port for free.
A Pi is a logical choice for prototyping. A Pi Zero is a $5 ARM board with no Ethernet, but it’s pin-compatible with bigger Pi boards that do have Ethernet (although a USB-to-Ethernet dongle might be cheaper than the marginal cost of a Pi model with integrated Ethernet).
One gotcha to watch out for if you are manufacturing at any scale is availability. Pis come and go in waves–just look at any contemporaneous blog posts about how hard it is to buy any recently released Pi model–and the Raspberry Pi foundation will probably not guarantee your supply (UPSide is a little off-mission for them). You might need to cultivate a relationship with a second ARM board vendor who can build you something with a compatible connector and software stack (assuming there isn’t one out there already).
If you think in terms of “3.3V Linux ARM board with 8 GPIO pins, 2 I2C interfaces, one SPI interface” then you don’t have a supply chain crisis when one popular hobbyist product is superseded by another–you just map UPS hardware interface signals to IO pins on the new hotness, and build a cheap PCB connector (or start shipping a different one) to match its pinout.
>You might need to cultivate a relationship with a second ARM board vendor who can build you something with a compatible connector and software stack (assuming there isn’t one out there already).
Banana Pi. Cheaper and in some ways better; 4 USB ports is overkill for this deployment, 1 would be right.
>If you think in terms of “3.3V Linux ARM board with 8 GPIO pins, 2 I2C interfaces, one SPI interface” then you don’t have a supply chain crisis when one popular hobbyist product is superseded by another
The situation is a little better than you apparently know. There are multiple manufacturers supporting the Pi HAT (Hardware On Top) pinout – I found this out when I was researching hardware for the NTPsec micro-server HOWTO. Orange Pi is another one.
“My favorite feature would probably be to not power up the load after power is restored until the battery was charged up to a configured minimum runtime–and a display telling me this, so I’m not trying to figure out why it won’t start after power is restored. There’s no point in turning the load computer on until there’s enough reserve to complete a boot and shutdown cycle if the power goes away again.”
THIS!
@esr: I’m now thinking it might be smart to use a Pi as the onboard computer in the UPside.
Yes, more expensive than the Arduino. On the other hand, it’s a development environment that’s really well understood. Hook the control switches to the GPIO, add a few A2D converters, and suddenly your UPS management is a boot-time daemon in the userspace of a Linux. We trade some hardware cost for tremendously reducing the complexity of the software engineering, and get stuff like an Ethernet port for free.
Yes. This is not a place to shave pennies. In the context of the total cost of the system, the difference between Arduino and Pi is negligible. The easier development and greater control more than compensate for the cost. How much would something like this be worth to you? If you need it, you need it badly enough to pay a decent price for it. The problems discussed here ultimately stem from trying to do it cheap.
>Dennis
> The problems discussed here ultimately stem from trying to do it cheap.
Cheap and dumb. I now think building it around an RPi is going to be cheap and smart – yeah, you pay $20 more than for an Arduino or ARM SBC, on the other hand now the low-power electronics and firmware don’t need a custom PCB and you’ve already got both USB and Ethernet ports on board. The BOM for the whole system might actually drop.
Vcc for the thing is going to be a pain.
>Vcc for the thing is going to be a pain.
Obviously Vcc is short for something, but I am only an ignorant software guy. What are you talking about and why is it going to be a pain?
Vcc is the voltage feeding the Pi. Grossly oversimplified, a regulator hooked up to the battery terminals (so it gets power from either the battery charger or the battery) should suffice; however, there are assorted problems to solve:
The inverter might spike the battery if the load is too high or the battery is too weak, which will crash the Pi. UPS control systems are generally set up to default to load-off, battery-on-trickle-charge when Vcc hits zero. This creates a problem if we want to be on load-bypass, battery-on-trickle-charge in some but not all instances when that happens (i.e. load-off or load-bypass is a configuration option).
The Pi has a non-trivial boot time, during which all the relays have to be in a safe and sane position. It adds latency to the “restore power NOW” path, and creates a possible safety issue if the Pi crashes during or before an overvoltage event.
Little microcontrollers don’t have these problems–they’ll run on as little as 1V, they don’t need a big capacitor to keep running during voltage spikes, and they boot (or reboot) in a fraction of a second.
Maybe a hybrid design, where there’s a little micro that directly controls the hardware, charges the battery enough to run the Pi, and runs the startup sequence until the Pi is up, then takes its orders from the Pi thereafter. The task of the little micro is to use the UPS electronics to maintain a stable DC power voltage for the Pi under startup conditions. The Pi then provides monitoring and controls load relays and battery charging by sending commands to the little micro.
There are also subtle requirements on the Pi’s Linux distro, e.g. it must operate with a mostly read-only filesystem because its power will fail while running, and if you think SSDs are bad when writing during power failures, wait until you see what SD cards do.
Also what happens to the UPS during a dist-upgrade or kernel upgrade on the Pi? The Pi would need to tell the little micro to hold the fort during a maintenance reboot.
It’s probably best to let the little micro handle switching to battery during power failures itself anyway–better latency guarantees than you get from a Linux host that might need to page in the monitoring executable before it can respond.
A Raspberry Pi makes an awesome Internet-connected UPS monitoring host. Whether it makes sense to be the UPS is a very different question.
>A Raspberry Pi makes an awesome Internet-connected UPS monitoring host. Whether it makes sense to be the UPS is a very different question.
In my visualization of the Cosmic All, the Pi is doing policy – like “what do I do on a button press?”, driving the status display, and things like battery modeling based on the sensor data from the pack.
We were probably going to need a separate custom PCB with its own tiny microcontroller for the high-power electronics, anyway, if only to minimize the parts of the design that have to undergo compliance testing when they change.
Agreed, but here’s another point anyway just for completeness:
When the batteries run too low for the inverter, we don’t want to be using them to feed a 100mA+-when-it’s-completely-idle Pi, or we’ll damage them. Better to shut the Pi off then, and let a microcontroller that uses 0.1mA wait out the rest of the power failure. At that rate it can run on “dead” batteries for months.
OK, this triggered a grey cell. Something that may have already been stated … I haven’t read the whole thread at this point, but there’s two general operating modes for UPSs. The small ones (APCs, etc) run on bypass, then switch to battery on power fail. The big guys run “through” the batteries, meaning mains power is rectified and cleaned, then passed onto a DC bus where it is split between the charge controller and inverter, which in tern powers the equipment. On these systems, the maintenance bypass circuit cuts the entire UPS (rectifier, inverter, charge controller, etc.) out of the loop. I wish I was still working at the data center so I could show you the block diagram on the display, it makes that clear… sadly, I don’t have any photos I took of the silly thing when I wrote docs for it years ago, and I wasn’t able to find a clear example online, sorry!
>the small ones (APCs, etc) run on bypass, then switch to battery on power fail. The big guys run “through” the batteries
This used to be the case (I know it was during my last close look at the tech about a decade back). Nowadays I believe it’s all double-conversion (through the batteries) all the time, even on low-end units. A way you can tell is that nobody quotes standby-to-battery transition times any more.
Ironically, the older style (standby power system, SPS) would work better today than it used to. Most of the loads UPSes drive now have switching power supplies, which (unlike the older inductive style still in wide use a decade back) can hold voltage for a small amount of time at the beginning of a dropout and can thus ride out some transition time.
Another consequence of switching power supplies: they don’t really give even half a crap about pure sinusoidal power, so UPSes made to drive computers (as opposed to, say, the refrigerator on your RV) barely even gesture in that direction any more. It’s now typical to ship a fairly coarse stairstep approximation of sinusoidal AC.
UPDATE: Turns out a few very low-end devices still do SPS – like, $40 boxes from CyberPower. I wouldn’t touch ’em with a 10-foot pole.
I wrote:
>Nowadays I believe it’s all double-conversion (through the batteries) all the time
I was wrong. It is true that SPSes are almost extinct but units below 1KVA aren’t double-conversion, they’re what’s called “line interactive” – automatic voltage regulation of mains power with fast failover to the battery if mains voltage looks like it’s going to wander far enough out of spec that the AVR can’t cope.
Mea culpa. I thought everybody had gone to full double conversion because nobody competitively quotes mains-to-battery cutover time any more, but it looks like what actually happened is the failover time of line-interactives got so short that it’s less than a switching power supply’s voltage-hold time.
The bigger problem is that while older PSUs were dumb and had no problem with line-interactive UPSes, the newer ones are smarter and can detect the power failure before the UPS intervenes. What happens next depends on the PSU design — if it starts drawing full power to keep its internal caps charged, it can overload a smaller UPS. A bigger UPS that could handle this isn’t an option because the customer is cost sensitive and doesn’t think they need it.
Vcc is mostly an anachronism at this point.
Transistors have three legs. In a bipolar junction (traditional) transistor, one of the legs is the collector. In an integrated circuit, the Vcc pin is connected more-or-less directly to the collector legs of the output transistors. Vee is for negative-going transistors, on the emitter leg, but sometimes (very rarely) used to mean ground when the power supply is ground and positive only (unipolar power).
But unless you are using recycled parts from the 70s, your chips are almost certainly built with field effect gates internally, which don’t have collectors, emitters or bases, but instead have sources (Vss), drains (Vdd) and gates.
In actual usage, Vcc and Vss mean either “I am the positive terminal of the power supply” or “Please connect me to the positive leg of the power supply”.
I think using a microcontroller for the device interface is going to be mandatory, even with a Raspberry Pi acting as liaison to the human interface and the data interface. The Pi boots from a SD card, and tends to corrupt that card when it is being accessed while the pi loses power. Something in there needs to be smart enough not to let the Pi boot until the batteries can ensure enough time for the pi to boot up and shut down cleanly, otherwise you’ll eventually brick your UPS until you can physically remove the SD card and reflash it.
“accessed” or “written to”?
With random SD cards, who knows? Think “bad SSD firmware x 1000”. I’ve had SD cards randomly die after a few years, even when powered continuously and never written to.
Usually when people build embedded devices with embedded MMC flash storage, they boot the device off of chips that were sourced and soldered onto the board, so that they can avoid having to deal with every possible SD firmware bug at boot time (they might still have an SD slot for data, but at least they can pop up an error window if those fail, and they can work around all the bugs that happen to be in the chips they’ve soldered). BeagleBoneBlacks are built like that. Raspberry Pis are not.
On the other hand, all that would be required to fix a bricked UPS is to open the lid, swap in a fresh SD card with a good image on it, close, and…restart? Turn it off and on again? Is there a reset button here?
Maybe the non-Pi components implement a heartbeat protocol with the Pi, and if no heartbeats are detected, they fall back to dumb UPS behavior (OK, which dumb UPS behavior?)? Or maybe there’s an idiot light that says “replace Pi”?
>Or maybe there’s an idiot light that says “replace Pi”?
More precisely, “Replace Pi SSD image.” That’s the simple way to handle it.
This strikes me as a good reason to use one of the Pi clones with onboard NVRAM.
…and a reset button, because original Pis don’t have those either.
Or maybe the exact user instruction is “remove Pi, replace SD card image, reinsert Pi.” That could be tricky to do on a UPS with a live load.
A packetized serial protocol with ECC for the downstream micros would prevent spurious commands from being sent during Pi hotswap. A 5V switch should be available to turn off power to the connector as the Pi is inserted or removed.
Maybe bolt the Pi into a custom box with PCB and connector that is designed for hotswap. The box would interact with mechanical components to turn off power first and disconnect ground last. No need for a separate manual switch then, and various issues ranging from noise to ESD are avoided.
Accessed, but writing makes it much more likely. I now automatically split my Pi images into boot and root with a SD card and a USB stick. That reduced my corrupt SD card images by 99%, but not 100%.
@esr: The BOM for the whole system might actually drop.
It might indeed. But even if it doesn’t, it’s still a net win for the reasons discussed.
And since more than one hand will be working on the software for this, a widely used, well understood hardware platform with a full toolchain available makes it much easier for those other hands to get development systems they can test code on that can be verified and merged into the master repository.
I’d be startled if various folks in this thread didn’t already have Pis they were poking at for other reasons. The questions is probably “What Pi configuration is required to perform the functions under discussion?”
>Dennis
My daughter wants a fishtank. With fish.
I don’t want 20 gallons of water in a glass bucket in her room.
But I’ve got a couple raspberry pis and a spare LCD around.
Now if I could just get xfishtank to run properly under a modern screen saver.
I’ve had a bunch of UPSs die. Batteries, sure, they’re consumables. But some went into frenzies of beeping that couldn’t be turned off. Others simply quit charging. Last week, one that apparently started developing some kind of electronic seizures; the two machines plugged into it would be off on some mornings, came back up with no informative error messages, until one day a tiny LED, not mentioned in the manual, came on. I replaced the UPS and the machines haven’t crashed again.
More than once I’ve wondered if I could lash up something with a car battery and one of those solar charger/inverters…
Please, make also 12/24V version of it. Most of ups-protected devices doesn`t need double conversion from 24 to 127/220 and back to 12/5 v with all associated losses and HV protection.
>Please, make also 12/24V version of it. Most of ups-protected devices doesn`t need double conversion from 24 to 127/220 and back to 12/5 v with all associated losses and HV protection.
You say “most”. What devices are you thinking of that can take 12/24V direct?
Also, see the first item under Rejected Ideas.
High end network switches and routers take DC direct or AC. With optional DC output, you get telecoms as customers.
That said, I would like to see something as simple and as soon as possible. Can always add features for version 2.
Every time someone says “12V output please” I think of one of these. That’s a standard 12V output colloquially known as a cigarette lighter. UPSide could do that, easy.
DC power supplies are not created equal. They have different behavior at startup and different responses to variations in load over time. Often the powered device is never tested with anything but the power supply it ships with. If the slope of the voltage output at startup is too steep, it could end up frying a low-spec component with surge current; too shallow, and the device browns out at startup; too much wire or fanout between devices and the power supply, and devices brown out during operation. Then there are the arcs when devices are hotplugged, surges when a heavy load is disconnected, minimum load issues, capacitance, feedback, etc. Cheap low-voltage devices aren’t designed to solve these problems because they never arise when the power supply is an AC wall-wart feeding only one device.
There may be a market for 48V DC outputs for telecom, and presumably that market knows how to interface to a shared third-party DC supply and designs their equipment accordingly. I’ve seen the brochures but I’ve never been there.
I have only two routers I’d like to keep powered. One has power-over-ethernet and the other has some unspecified DC voltage (most likely 12V). The rest are just laptops, tablets and phones so no need there. Having a USB port to charge a phone/tablet would be a nice bonus. So no AC power required.
Can there be enough people like me to make a separate use case?
>Can there be enough people like me to make a separate use case?
It is remotely possible we might want to support PoE, as that is a pretty well defined standard.
Zygo’s remarks make me terrified of trying to support any form of direct DC other than very well specified and standard forms such as USB 5v, PoE, or telco 48v. I don’t want to go anywhere bear the possibility of damaging equipment that was designed around the idiosyncracies of one wall wart.
So, er, not just “no”, but flees the idea in fear no.
I think the input power connection spec is too low, C13 cables are easy to get but normally limited by IEC standards to 10A, 15A if heavy-duty conductor cables are used. US 20A circuits should only be loaded to 16 A (a mere 1760VA) which may be under the desired output of the reference design. At IEC limits we are looking at 2300VA for high voltage operation, still possibly too low to allow for battery charging and the target connected loads.
Better to go with a C19 connection as input, that way you can go from a US NEMA 5-20 connection all the way up to a NEMA L6-30 (4,992VA at 208VAC) depending on design output.
>Better to go with a C19 connection as input
If we do that, are we excluding using the kind of 3-prong grounded jack found in residences? That would be a crash landing.
No, C19 is a three-prong connection with a 16A IEC design limit. C19 <5-15P and C19<5-20 cables are available from TrippLite so we aren't talking about exotic parts..
>No, C19 is a three-prong connection with a 16A IEC design limit.
Roger. So we have this reasoning on the record, please open an issue on the tracker requesting this change. I’ll close it and do the edit.
Done
Another design question: Should the solution allow parallel-able battery+conversion modules to scale max load and/or run-time?
Cost: more complex design, both for the AC side of the converter and the host chassis.
Benefit: allows scaling run time by adding Power Modules instead of futzing with adding battery packs (either parallel or series connected) with all the impact on the battery monitoring subsystem.
>Another design question: Should the solution allow parallel-able battery+conversion modules to scale max load and/or run-time?
This is already under Rejected Ideas for the UPSide-1. Best to keep it simple for the first iteration, but this would be a plausible scaleup for a future version.
@esr: This is already under Rejected Ideas for the UPSide-1. Best to keep it simple for the first iteration, but this would be a plausible scaleup for a future version.
The question that occurs to me is precisely what the first iteration end result will be.
As I understand it, there are three separate parts involved:
The battery packs that store the power needed to handle transients, and keep systems up long enough to be shutdown gracefully in an extended outage beyond the capacity of the batteries.
Controller circuitry that interfaces with mains power and determines when battery power is required and switches to it, then switches back when battery power isn’t needed, and monitors the health of the batteries.
A smart interface that sits between the UPS controller and the systems being protected, and provides a friendlier and more powerful interface for the user to monitor power conditions, store historical data for tracking trends, and give a way for the user to issue commands to the UPS. That’s the Pi. It connects over the local network, and precisely where it resides may be a detail.
Current UPSes combine one and two in one package, and lack of something like three was your initial complaint.
Because the design is essentially modular, the end product might not need to be all three, and certainly won’t be all in the same package. If this gains enough traction, for example, current UPS vendors might see the value to supporting something like three as an external, and provide a way to interface it with their product. In that case, all the user may want is the Pi. They already have battery packs and controller circuitry.
Because of this, some things discussed I don’t see as issues, like the problems flash based file systems can have if being written to when the power goes out. Yes, they can, but from our viewpoint, the Pi is simply one more thing fed by the UPS, and can be told to shut itself down before the power goes out completely.
Another thing I don’t see as a separate use case is Zbyn?k Winkler’s desire for PoE. You can already buy power strips that include USB ports as well as three prong plugs for standard power cables. I don’t see the UPS needing to supply this directly. The dual-format power strip is just another thing the UPS powers.
The more interesting problems will be the UPS controller and the connection to the battery packs. Since a design goal seems to be user serviceable and replaceable batteries, it sounds like the controller also needs to be an external unit, capable of interfacing with more than one sort of battery.
This design sounds like the sort of thing that could be a DIY project for the users. Buy the parts and put them together. How feasible this is may depend upon the regulatory environment, with Underwriters Laboratory certification the tip of that iceberg.
The last question becomes how many of these things need to be made and sold at what price to be a viable commercial product, where the user is buying one, two, and three, and exactly who will make them. But we can’t reasonably ask that question till we have something like a final first iteration design. It’s simply something to note for the future.
Am I missing something obvious, or is that an accurate summation?
>Dennis
>The question that occurs to me is precisely what the first iteration end result will be.
See Context and economics.
>Am I missing something obvious, or is that an accurate summation?
Pretty accurate. You might want to look at Three-controller design and some of the pages near it.
Nobody has talked about one of the most important design parameters for me.
A UPS that fails when it is supposed to work is worse than no UPS at all. It creates an expectancy and then fails to perform. It can happen in several ways. It might fail directly on incoming power failure, or it may run for part of the expected time, and then suddenly fail. Almost all of the time, these problems are due to the batteries ageing. If you very seldom have outages, the likelihood that the UPS will fail is very high.
For this reason, you want individual monitoring of each battery in the UPS. You want to measure voltage over the battery, current through the battery and temperature on the surface of the battery. You may also want to measure tension on the surface of the battery, as some kinds of battery cells expand when they age. By gathering the data over time, you should be able to determine pretty accurately when your UPS is no longer up to your required specifications and needs new batteries. This may require gathering data sets from the community, but would increase the reliability of UPSes enormously. Preventive maintenance is seriously underrated.
>For this reason, you want individual monitoring of each battery in the UPS. You want to measure voltage over the battery, current through the battery and temperature on the surface of the battery. You may also want to measure tension on the surface of the battery
Oh, believe me, detailed battery-state modeling for battery-lifetime projection is in the plan. Go look at our design wiki.
The one thing on your list we are not going to get to do is surface-tension measurement. We’re not going to build our own batteries – the hazard issues around that are nasty. No, we’re going to use nice, safe, sealed merchant-market batteries.
Which don’t have integral surface-tension sensors, and if you think we going to go poking holes in them to install sensors you are out of your fscking mind.
:-)
Tension sensors are glued on the outside of the battery. The nice thing about them are that they are cheap, very precise and easy to work with. Every single battery pack my sysadmins have replaced in our UPS has been visibly deformed (5 mm bulge on each side), and tension sensors will give a readout long before anything is visibly happening.
Now, you may get information that is good enough through the other readings, but this is one way to get reliable data.
I did some work with tension sensors for electronic scales many years ago. In those applications I measured the elastic deformation of thick steel rods to find out how much weight was put on top of the rod.
You know about LiFePO4 batteries, right?
https://en.wikipedia.org/wiki/Lithium_iron_phosphate_battery
energy density isn’t as good as most other LiIO, but
longer life, more cycles
safer (non-explosive, even when ruthlessly overcharged)
high discharge current
same voltage as lead-acid, so drop-in replacement is possible
you can get e.g. 18650 cells & make your own packs, but that’s not really the point (as e.g. LiCo has >2x density). pre-made packs exist that are the same shapes as common lead-acid batteries (including UPS batteries) so they’re a drop-in replacement. Or an upgrade…
http://www.batteryspace.com/12-8v-lifepo4-battery-packs-from-19.8ah-to-200ah.aspx
wow, I hadn’t seen the new LiFeMnPO4 prismatic cells!
http://www.batteryspace.com/lifepo4-prismatic-module-3-2v-60-ah-10c-rate-192-wh—–un38-3-passed-dgr.aspx
$0.50/Wh is <1/2 the price of standard LiFePO4
single modules exist from 64 to 640 Wh
>You know about LiFePO4 batteries, right?
Yes, we’re investigating those.
I just want a block with batteries in it. All my stuff runs on 12v, my Raid is a DS214 and it has a 12v power brick, so does my PC.
Why step up a 12v battery to mains voltages, and step it down again? All that loss for no reason. When all I want is a little box with alkaline disposable batteries in it, that sits after the power brick and before the Synology raid and cuts in when the 12v PSU dies. I’ll use the same on the PC. Everything else… well they don’t need UPSs.
I would like to see a USB connector too so it can tell the Raid to shut down, but as long as it lasts past a short power cut I’m happy.
I can always run the batteries through a normal battery tester. They’ll just be 8 AA batteries (or 16 or some larger ones, or whatever is needed for the power draw). No special complexity here. I can just use a battery tester every few years.
>Why step up a 12v battery to mains voltages, and step it down again? All that loss for no reason.
Monitors.
Until monitors have a standardized DC input jack, this will remain true.
ATX power supplies using DC-DC converters rather than an inverter are better than UPSes.
We don’t need to run the inverter!
Desktop motherboards and or power supplies should have a lithium ion battery port.. like a laptop..
There could be a few standard battery pack port types… 1000watt ones would be pretty expensive, but really with proper power management the cpu and gpu could go into low power mode.. My laptop works pretty well on battery.
Why not have a 5.25inch port on a desktop house a battery?
I’d love to see something like this as well.
The base electronics consisting of a modified sine wave generator running at a common denominator frequency to support both 60Hz and 50Hz operation (some high multiple of 120Hz and 100Hz as you only have to generate a half cycle and then invert) controlled by an Arduino with the basics of current and voltage monitoring is pretty straight forward. An H-bridge constructed with a stack of MosFET’s is standard operating stuff in the land of DC motor control so that aspect of off the shelf is pretty straight forward.
The power transformer could be tricky depending on sizing requirements (for off the shelf), but you can stack those as well. The primary issue will be that most off the shelf power transformers are rated using the RMS voltage (secondary) whereas the voltage we need is the peak to peak voltage as that is what the batteries will be sending. This might be “modulatable” by altering the pulse width of the generated sine wave but doing that reduces the total output power. Off the self components will necessitate compromise.
Here is what I would like to see in the open design of a UPS:
1. Separate power electronics and battery housings, easily stackable.
– i.e. A battery box configured like a paper tray accepting the various sizes of standard gel cell batteries. The stack grows by putting the battery box ON TOP of the inverter; battery boxes can be stacked.
2. Charging electronics inside the battery box allows for flexible power sources.
3. Integral temperature monitoring of the batteries, adjusting charge current appropriately and looking for battery boil temps (battery boiling is the number one killer of UPS’s in my experience).
4. Simple user interface (basic LCD 2 line x 40 will do). Some way of forcing the user to replace the battery when the UPS tells you to replace the battery :) I can’t tell you how many times I’ve replaced a UPS that had the “replace battery” LED lit for a year before the charging electronics finally destroyed the battery and everything blew up the next time the power went out. I also can’t tell you how many times a customer threw out the UPS because “the little red light was on” instead of just replacing the batteries. Crazy.
5. Both a well documented open format data reporting capability via the USB port as well as SNMP monitoring via a standard Ethernet port.
As an aside, I have most of the electronics for the inverter side sitting on my work bench beside me now. I hadn’t really thought about building a UPS so much as just experimenting with different modified sine wave approaches and a simplified method for pulling pulse widths out of a PROM to create either a 60Hz or 50Hz sine wave but I’d love to work on this project with its’ expanded scope.
J
Awsome idea, having experience with several consumer grade UPSes (5Kw and 10Kw).
Could you please consider some of the next ideas ?
0) Please consider desining it when the input frequency is unstable, a 10% to 15% frequency jump happen several times each day at my place (there is nothing that can be done with that). APC just refused to operate in such conditions and I had to put an AVR before the UPS to make it work.
1) Please consider designing the device to work with over 45C , when power goes does even if it is only 30C outside the internel building can get up to 45C very fast (because the A/C stop working)
2) Please consider designing the to operate AC based devices, I use my UPS to run a 2kw air conditioner and few servers. so pure sine would be a HUGE plus.
3) Make it possible for your UPS to be u[gradable above the 3000W limit most consumer grade UPS have.
4) Please consider allowing internel DC input – many building installation already have the 12Vdc, 24Vdc and 48Vdc – You choose what socket type is the best for you, the users will get an adjuster from thier existing DC socket to your socket.
5) Please consider allowing plugging an existing power bank to either DC input of the UPS (like above ) or an addition to the existing batteries (paralel). I for example have a ~30Kw power bank (5x 500 ah boxes for 12v) with it’s own cycle (3 step ) charger – being able to reuse a power bank and not needing to replace it with your battery of choice would be a great benifit. Such power banks are not uncommon in rural areas where power goes down every once in a while (it is usually a special room or a basement).
>Could you please consider some of the next ideas ?
DON’T POST IT HERE!
https://gitlab.com/esr/upside/issues/38
Really, I’m overwhelmed.
Some things I would consider making standard on this:
0. Network connectivity (gigabit friendly would be nice but arduino controllers only have 100Mbit devices so that will have to do.)
1. Simple hitachi based display, let’s make this thing so easy to program even a novice can goof around with it.
2. Basic environment monitoring a DHT 22 that can be mounted to the device or remotely mounted would make this something that could wind up in a server room.
3. Serial isn’t dead (RS-232 monitoring)
4. SNTP syncronizing
5. Linux / FOSS daemon – first (windows last)
I have some firmware skills and some Hardware skill. I would be able to help.
>Some things I would consider making standard on this:
DON’T POST FEATURE REQUESTS HERE!
They belong on the General Feature request thread on the project issue tracker.
>I have some firmware skills and some Hardware skill. I would be able to help.
Where are you in meatspace?
There is a reason why the UPS space is such a mess.
There is a strong tendency to low “first cost”; once you’ve owned your first UPS you are going to be skeptical about any claims how long the next one will last. Thus you are not going to believe that you can spend a little more money and get something better.
Second there is no set of inputs and outputs that will make everyone happy. I just bought an APC UPS to protect my cordless phone base station and a DSL modem that has WiFi built in. With those two things I have 21st century connectivity to the many battery powered laptops, tablets, game consoles, etc.
Both of those are powered by wall warts and in theory I could use something that supplies 12V DC, probably having to make cables for the two devices. Ideally I would like a runtime of days for that small load; as it is, UPS systems seem to be sized to deliver a larger load for a shorter period of time. Thus the inverter is scaled such that It appears to consume as much power in waste as those two devices consume. At least APC sells this model with an external battery pack that lets you increase runtime.
With a better inverter you could get better efficiency and better runtime. But who wants to pay for an inverter which will sit idle 99% of the time? Who wants to pay for a better battery charger? Do you want an inverter which makes “good enough” electricity for certain uses or do you need a much more expensive “sine wave inverter?” Even if the inverter is ideal, the power supply it feeds is probably not so you lose in terms of runtime there.
Trimming the fat would likely make a much smaller and cheaper device that handles my specific case (keeping two pieces of telecom equipment running with long runtime) but other people have different needs; they all agree they don’t want to spend too much, but that gets us stuck in the place where we get a bad product that still costs too much.
As for monitoring, I’d much rather see an Ethernet plug on the machine that would plug into that DSL modem, but that is probably asking too much. The APC protocol to communicate with it over USB is acceptable to me, but I don’t have any computers with a USB port that are close to the “telecom closet”
>As for monitoring, I’d much rather see an Ethernet plug
The SBC has one. That’s just software.
Maybe you could use some of this cheap Class-D amplifier boards (tas5630) with an transformer in behind.
I had a reclaimed APC 2000 watt UPS that i hooked up to 4 RV batteries. That setup lasted for 7 years without a problem and we had frequent power interruptions. I did have an external SNMP box (APC branded) that kept track of things. Of course, with that much capacity, it seldom drained the batteries more than about 50%. The system failed miserably by blowing one of the batteries into shards. I was glad I installed the batteries in an outdoor enclosure.
But I am not disagreeing with most of your rant. I can usually replace one of the “consumer” grade units for less money than the battery.
I’m optimistically hoping that this project will bear fruit, but a review of the GitLab repo suggests no activity in some time. Is the project stalled or dead?
>I’m optimistically hoping that this project will bear fruit, but a review of the GitLab repo suggests no activity in some time. Is the project stalled or dead?
Stalled. My hardware lead is building a farm in Indiana. Expects to be able to put some serious time in over the winter – I just talked with him.
An idea for fast prototyping: What about hacking a standard UPS that features <your favourite battery type> and putting in a new brain? Take a suitable APC model, kick out the serial port and replace with a RasPi Zero plus a MCP3202 A/D converter and some optocouplers. Then add a “f_upc” driver to the Linux USB OTG “gadget” architecture to get USB plug-and-play.
No hassle with certifications etc. I’m sure all these “makers” will love to do an actually useful project… Just setup a Web to collect recipes and experiences.
>An idea for fast prototyping: What about hacking a standard UPS that features
One good reason is, nobody is really doing the LiFePO battery we want in a consumer-grade device yet.
The other problem is that it would be really tough to get the smart monitoring for charge state out of a device without analog circuitry designed for that.
Thought I’d check in and ask if there are any more updates since the last query 4 months ago. Prompted by one of my UPSes which is about to die. :-)
>Thought I’d check in and ask if there are any more updates since the last query 4 months ago. Prompted by one of my UPSes which is about to die. :-)
As it happens, Eric Baskin called me last week to say he expected to be able to resume work soon. His near-term goal is to design suitable high-power circuitry that (a) doesn’t need a heavy custom-built inductor, and (b) isn’t covered by blocking patents. Not trivial.
Cool, thanks for the update!
Hi Eric
This is a really important topic and it’s fantastic you are doing something about it by creating a free and open-source hardware UPS that the users can control.
I’ve had similar problems with UPS. Further, it’s seems also impossible to repair your own hardware as the manufacturers don’t release service manuals or schematics :-( I’ve searched online and found many cases of people trying to repair their UPS, but don’t get far due to the lack of help from the manufacturers who only want you to upgrade….
For the development of the design and hardware of the UPS, have you had a look at KiCad (https://kicad-pcb.org)? It’s being used for the development of some very complex hardware, for example the Librem-5 linux based phone (https://puri.sm/posts/how-we-designed-the-librem-5-dev-kit-with-100-free-software/).
So you may get more help and support. May I suggest you promote this awesome project on forums and groups that would appreciate the goals of you and your teams efforts. Such as:
EEVblog
https://www.eevblog.com
Free and Open Source hardware section at EEVblog forum
https://www.eevblog.com/forum/oshw/
Hack-a-Day
https://hackaday.com/
Hackster.io
https://blog.hackster.io/
EE-Web
https://www.eeweb.com/
FixIt
https://ifixit.org/
Or if you are really keen, you could go all out and start a crowd funding project.
Crowd-supply
https://www.crowdsupply.com
I suggest CrowdSupply because they are more ethical than Kickstarter as they are focused on user rights and sustainability. They’ve also deliver 100% of their backed projects, unlike Kickstarter…
Best wishes
PS. I’m not affiliated with any of the suggested projects/websites above.
>For the development of the design and hardware of the UPS, have you had a look at KiCad
The good news is our power engineer, Eric Baskin, knows about KiCad and I think he’s using it.
Bad news is the project has been stalled for months because he was busy with growing season on a farm. But I heard from him last week and he’s working on a hardware prototype again.
Hi Eric,
I just noticed there has been updates at the gitlab repo…. happy to see.
Great to hear you are considering KiCad. If Eric Baskin or the project uses KiCad, I just found you can promote the project on kicad’s Made-with-Kicad page at http://kicad-pcb.org/made-with-kicad/
Free and open source projects like these need all the support and promotion they richly deserve.
It sucks to live in the USA, now doesn’t it? In my country all hydro is underground, I can’t even remember any power outage in the last 30 or 40 years or so. Hydro poles like you have all over the country don’t exist in my country for almost 60 years. Get with the times foolish yankees, problem solved.
>Get with the times foolish yankees, problem solved.
I’m going to guess that your country is probably geographically quite small, heavily urbanized, and wealthy. Matter of fact I’ll lay odds it’s smaller than the individual state of Pennsylvania, where I live.
Without all those traits burying all power lines is (much as I like the idea) impractical. Americans aren’t actually stupid about those things, but we do have a geographic-scale problem that Europeans don’t grasp.
There’s tradeoffs in burying power (and other utility) lines – they’re more expensive to install, maintain, repair, and upgrade.
I genuinely wonder if this is the case. Power lines are buried in my area, and I strongly suspect that is why no one around here seems to bat an eye while other counties are reporting massive outages whenever a major storm hits. Same with cable outages.
I don’t know what their expense sheets look like, but I can readily accept that it’s noticeably more expensive to dig a trench, lay a line, and fill it back in, then to dig several smaller holes, put wooden poles in, and string line from that. And that it’s more expensive to dig up a broken line, patch it, and re-bury it, than to repair it when it’s hanging from a pole. But does that continue to be the case when one has to repair the hanging line several times per breakage of its buried counterpart?
@Paul Brinkley I don’t know what their expense sheets look like, but I can readily accept that it’s noticeably more expensive to dig a trench, lay a line, and fill it back in, then to dig several smaller holes, put wooden poles in, and string line from that. And that it’s more expensive to dig up a broken line, patch it, and re-bury it, than to repair it when it’s hanging from a poleBut does that continue to be the case when one has to repair the hanging line several times per breakage of its buried counterpart?
I think the issue is complicated by the business nature of the power companies. They are regulated local monopolies, with pricing largely set by the regulators. Many are also explicitly non-profit. The regulators will see their mandate as protecting the customers, which devolves to keeping a lid on pricing.
Doing things like entrenching power lines is a capital expenditure, and higher capital expenditures require higher rates to fund them. Digging holes and planting poles in them, then stringing power lines from the poles will be cheaper, and depending on the area and how spread out the population served by the power company is, possibly much cheaper.
Fixing the outages caused by storms can be paid for out of present revenues, and that anticipated cost is likely built into rate schedules the regulators approve.
The notion of spending more up front for lower costs over product life is likely not something that will be considered in a case like this.
>Dennis
>The notion of spending more up front for lower costs over product life is likely not something that will be considered in a case like this.
That’s a straight-up function of the utility’s planning horizon. If it’s long enough, the fixed cost of burying cable plus the expected cost over lifetime of servicing it will underrun the fixed cost of aboveground installation plus service, with the difference going negative due to storm outages. The fixed cost of buried line amortizes away, the downstream service costs don’t.
It’s not just the cost of break-fix, but also regular maintenance, including maintaining the access points (manholes) and conduits, in places where the cables are needed to be accessible; or the cost to maintain the tools needed for the purely buried ones. Obviously, this analysis isn’t applicable any more, or we wouldn’t have buried cable utilities, but it’s a factor.
And of course it’s path dependent – to bury existing above-ground utilities would require buying an all-new grid. The existing grid in my area (Central NJ) was originally built out in a period where things were simply less reliable and access was considered more important than armoring. Consider that when electrification occurred, they HAD underground utilities for water supply and removal, and for town gas etc. And they still chose to put electricity and phones in above-ground. And that even after the devastation of Sandy, they repaired rather move underground.
In the 10 years I’ve been in this house, I’ve lost power for significant lengths of time once (Sandy) and that was because of problems with the substation (flooded and subsequently rebuilt) as much as with the above ground lines; and my in-laws with buried power were out of power longer. Small Sample Size, and data is not the plural of anecdote, but that’s my experience.
(I got by with a short drive to a friend’s house in a neighboring town who still had power, bringing a crate full of UPSes and other batteries to charge overnight)
@esr: That’s a straight-up function of the utility’s planning horizon.
Unfortunately, it’s also a straight up function of the regulatory environment the utility exists in. I’ve seen stats elsewhere that electric power utilities that are explicitly for-profit in places where the regulatory environment is not actively opposed to that notion have better service records and cheaper prices to the customers. Where I am is not one of those places.
You commented upthread about this: The real problem both places have is sclerotic Blue-state politics. The gentry-liberal elites that run both jurisdictions are fooling themselves that conservation and renewables will be enough. They’re wrong, and – hm, I was going to say they’ll find out the hard way, but being rich gives you insulation against bad policy choices. It’s the poor, people who can’t afford PowerWalls, that are really screwed by high electricity prices and rolling blackouts.
I don’t define things in exactly that fashion, but it’s the sort of issue I’m talking about.
>Dennis