Slashdot Mirror
ESR's Newest Project: An Open Hardware/Open Source UPS (ibiblio.org)
An anonymous reader writes:
Last month Eric S. Raymond complained about his choices for a UPS (Uninterruptible Power Supply), adding that "This whole category begs to be disrupted by an open-hardware [and open-source] design that could be assembled cheaply in a makerspace from off-the-shelf components, an Arduino-class microcontroller, and a PROM...because it's possible, and otherwise the incentives on the vendors won't change." It could be designed to work with longer-lasting and more environmentally friendly batteries, using "EV-style intelligent battery-current sensors to enable accurate projection of battery performance" (along with a text-based alert system and a USB monitoring port).
Calling the response "astonishing," Raymond noted the emergence within a week of "the outlines of a coherent design," and in an update on GitLab reported that "The response on my blog and G+ was intense, almost overwhelming. It seems many UPS users are unhappy with what the vendors are pushing" -- and thus, the UPSide project was launched. "We welcome contributors: people with interest in UPSes who have expertise in battery technology, power-switching electronics, writing device-control firmware, relevant standards such as USB and the DMTF battery-management profile. We also welcome participation from established UPS and electronics vendors. We know that consumer electronics is a cutthroat low-margin business in which it's tough to support a real R&D team or make possibly-risky product bets. Help us, and then let us help you!"
There's already a Wiki with design documents -- plus a process document -- and Raymond says the project now even has a hardware lead with 30 years experience as a power and signals engineer, plus "a really sharp dev group. Half a dozen experts have shown up to help spec this thing, critique the design docs, and explain EE things to ignorant me." And he's already touting "industry participation! We have a friendly observer who's the lead software architect for one of the major UPS vendors." Earlier Raymond identified his role as "basically, product manager -- keeper of the requirements list and recruiter of talent" -- though he admits on his blog that he's already used a "cute hack" to create a state/action diagram for the system, "by writing a DSL to generate code in another DSL and provably correct equivalent C application logic."
He adds to readers of the blog that if that seems weird to you, "you must be new here."
Calling the response "astonishing," Raymond noted the emergence within a week of "the outlines of a coherent design," and in an update on GitLab reported that "The response on my blog and G+ was intense, almost overwhelming. It seems many UPS users are unhappy with what the vendors are pushing" -- and thus, the UPSide project was launched. "We welcome contributors: people with interest in UPSes who have expertise in battery technology, power-switching electronics, writing device-control firmware, relevant standards such as USB and the DMTF battery-management profile. We also welcome participation from established UPS and electronics vendors. We know that consumer electronics is a cutthroat low-margin business in which it's tough to support a real R&D team or make possibly-risky product bets. Help us, and then let us help you!"
There's already a Wiki with design documents -- plus a process document -- and Raymond says the project now even has a hardware lead with 30 years experience as a power and signals engineer, plus "a really sharp dev group. Half a dozen experts have shown up to help spec this thing, critique the design docs, and explain EE things to ignorant me." And he's already touting "industry participation! We have a friendly observer who's the lead software architect for one of the major UPS vendors." Earlier Raymond identified his role as "basically, product manager -- keeper of the requirements list and recruiter of talent" -- though he admits on his blog that he's already used a "cute hack" to create a state/action diagram for the system, "by writing a DSL to generate code in another DSL and provably correct equivalent C application logic."
He adds to readers of the blog that if that seems weird to you, "you must be new here."
If a UPS is going to burn down my house, I want it to be a UL listed device. The insurance company is not going to give a shit if I was super careful in putting it together. They're kinda pissy about homebuilt electrical stuff that burns down your house.
This is one of a plethora of currently closed-source products that could stand some competition! Ignore the nay-sayers, of which there plenty, and continue to push the boundaries. We need guys like you, whether we know it or not...
There is probably a lot of potential for open electronics instruments as well. Multimeters, oscilloscopes, low end audio and RF spectrum analyzers and such. Dave Jones has had a very well received (AU$ 644,674) Kickstarter project with the 121GW multimeter. It's not entirely open (the firmware is proprietary,) but the hardware is open (schematics, components details, etc.,) the MCU is an easy to deal with STM32 and the programming headers are deliberately easy to get at, so ultimately open source firmware will emerge.
Maw! Fire up the karma burner!
Or replace the replaceable batteries, like you are SUPPOSED TO DO.
Sigh.
Shouldn't we be working on an open hardware/open source FedEx instead?!
#DeleteFacebook
It's all about the batteries. You can get a <$100 UPS which will last 2-3 years and keep a PC going until you can do a graceful shutdown, assuming you're in front of it. Then spend $40 for a new battery or buy a new one to get a few more years.
Or, you can get a full-blown generator or solar system which will keep the whole house up for possibly days.
The cost and unreliability of UPSs are the batteries. He suggests "deep-cycle marine gel batteries that will last next to forever," which is naive. Marine use is very different than UPS use - cycle life vs. chronological lifetime. No battery lasts forever, and those who want long-term backup for occasional outages (see: voice PBX systems) use flooded lead acid batteries..And that's just for hours - there's a generator involved longer term. Gel cells are better suited for lower current draws over a longer time, and are significantly more expensive than wet cels, per capacity.
What he seems concerned about is getting some warning about impending battery failure. You can fix that by doing regular deep cycle tests, but that shortens battery lifetime and makes for periods where the batteries are drained and you're taking a risk should the power actually fail.
Myself, I have a years-old enterprise class UPS from eBay, assumed the cost of getting new batteries, and will again in a few years.
"National Security is the chief cause of national insecurity." - Celine's First Law
It's about time we moved past shitty lead-acid batteries for UPS systems.
Where are my Li-ion UPS systems? Where are my server/workstation 1800 watts sustained over 30 minute battery options? I want my damn server to have enough time to shut everything down. No more of this "suspend to ram/disk", I want it actually shut down, and if something hangs, not sit there at the screen waiting (I'm looking at you windows) for someone to close whatever is running/hung.
and god damn windows et al software updates that want to run when you shut down.
Bulk of the consumer grade UPS are based off sealed lead acid - fairly idiot proof but only moderate charge capacity, limited peak draw capacity and limited life (normally the battery capacity has seriously degraded after two years of continuous service).
Making the battery user replaceable would be a big step in the right direction.
Shifting to Lithium Ion cells (preferably the iron phosphate variant) and applying a little de-rating will significantly improve the situation (requires a charge management circuit but these can be easily be sourced). The manufacturers' data sheets state to charge lithium ion cells to 4.2 volts per cell, keep the charge limited to 4.1 or even 4.0 volts per cell will dramatically increase the life of the cells (leave this as an exercise for the reader to research).
If you want to make the equipment "industrial rated" then component ratings need to be reviewed. Capacitors are available in various voltage ratings however if you want the equipment to last 10 years in +100F/+40C conditions then the voltage rating needs to be nearly double the normal operating voltage (there is a study done/sponsored by one of the US military branches on longevity and component ratings).
There is some fairly heavy engineering required on the power side of stuff and I don't believe ESR has yet to wrap his head around all the issues. Perhaps that is why he has put the call out for assistance.
Or replace the replaceable batteries, like you are SUPPOSED TO DO.
Sigh.
Yep. That's what I do every 3 years. Replace the batteries with higher Ah batteries, if possible. I then re-use the old batteries by hooking them up to solar panels at the camp to run 12v items like led lights. I put them in parallel to increase the amperage and, when the voltage drops to 7V or so, put those batteries in series to provide 12-14v.
Why would you want batteries that have less capacity per dollar? You don't need a portable UPS for a server so using lighter batteries doesn't make any sense. There's a reason other than for a few boutique UPSs that they all use real lead acid batteries.
There's already a product that does this, it's been around for years. Actually there are lots and lots of things like this out there.
The killer isn't designing the thing, that's relatively straightforward ("relatively" meaning you need a couple of experienced EEs, 6-12 months, and $50K or so to get the kinks worked out). What's not straightforward is getting it manufactured in quantity and UL rated. Open source works OK for software where the "manufacturing" cost is zero and there's no need for any safety certification, but less well for anything else.
And as such, instead of a single commercial pressure on the existing UPS makers to up their game, we will end up with a "background noise" of ever-changing, starting-then-fading, projects that become increasingly incompatible. And therefore instead of a united front that competes with the UPS makers, there will be loads of minor players that the big guys can safely ignore.
A better way to approach this would have been to introduce a mature product, build a user base and offer off-the-shelf solutions. Much like with RPi or Arduino (though admittedly, both of those suffer from fork-ism). Rather than to hail the coming of vapourware and getting bogged down in design arguments.
politicians are like babies' nappies: they should both be changed regularly and for the same reasons
A real 1,8kW, not a nominal 1,8kW? At 1,8kW for 30 minutes (after DoD limits, a kWh or more, versus a several dozen to a couple hundred watt hours in a traditional consumer UPS battery), and such a output power need, you're straddling a sort of awkward middle ground between a UPS and a home backup system (such as a Tesla Powerwall). Which is indeed a UPS (plus a grid-tied inverter and other things), although too pricey for your average UPS user's needs since the smallest you can get (if you can find one) is Powerwall 1, at 6,4 kWh / 2kW ($3k). Powerwall 2 is easier to get, but is even further out of spec for you (13,5 kWh / 7kW peak / 5kW continuous, $5,9k). #1 would run your computer for nearly 3 1/2 hours, while #2 would run it for over 7 hours.
I'm sure that there's some "middle ground" li-ion backup products out there, though. I assume you want li-ion for the longer lifespan and higher peak output of a properly managed li-ion pack?
"Lock and load, Brides of Christ!"
1) skip the fucking usb. Go with network. Or at least allow a usb/network adapter. 2) it is long past to have a direct dc line from battery to going past computer power supply. Converting from DC to AC and back is so inefficient. Instead have 48V connector and let power supply take it down. 3) ideally, rack mountable.
I prefer the "u" in honour as it seems to be missing these days.
normally the battery capacity has seriously degraded after two years of continuous service
As someone whose job it is to monitor industrial UPSes at a major hazard facility, let me say: WTF ARE YOU DOING! You should be easily able to get a UPS to run for 5 years unless you're horribly abusing it environmentally or electrically.
Unless you're defining "seriously degraded" as below 90% or something silly like that. Or listening to the vendor's sales guy, that's another expensive mistake.
Already exists, it's called OpenUPS and NUC-UPS. In fact OpenUPS is pretty close to what ekr is proposing, only it already exists as a finished product. Guess he didn't do much Googling before he came up with his one...
The problem is the batteries. Lithium batteries have a distinct disadvantage in that they can die of old age - so even if you never experience a power failure, after a couple of years, the batteries have degraded, just by being 2 years older.
Lead-acids do not have this problem, but they do have problems when they're not treated right, which shortens their lifespan. Basically a lead-acid's lifespan is determined by its use - the more deeply you discharge them, the shorter their lifespan. That's a big difference between the ISL batteries in your car (ISL - ignition, starting and lighting) and deep-cycle batteries. ISL batteries are designed to offer high currents for short periods of time (i.e., ignition and starting), but cannot take much discharge. Even in the deep of winter, a heavy start of a modern computer-controlled car will take less than 1% of its capacity, delivered over the 30 seconds it takes to crank. But don't you dare run it below 50% or you will run into problems.
Deep cycle batteries are designed for lower peak currents over a longer period of time and can tolerate deeper discharges. (But none can take to 0%).
Unfortunately, a UPS demands both high peak currents and deep discharges, so either kind of battery is poorly suited.
Lithiums are great, but with their limited lifespan, are really not suited for standby applications - they really hate sitting on charge continually. There's a reason why lead-acids are still around, and still used in a lot of standby applications. A float charger is dead simple, they tolerate that well, and if you design it right, the standby load can be low that it really doesn't tax them.
The last UPS battery I had, I realized I had changed 10 years ago before failing last year.
Do you mean the OpenUPS? It's DC only, and one of the hardest parts of ESR's UPS is the AC part. He needs it to run things like monitors that typically are not available in DC versions.
The AC part can be designed a number of ways, but from ESR's requirements it sounds like line interactive is required. The simplest way to do that is to take AC mains and convert it to DC. Use the DC to feed a DC to AC converter that produces the output AC voltage, and have a switch over mechanism in case the AC input goes away. Big caps to cover the gap during the switch over period. The DC also charges the battery when required.
As well as providing very good isolation of the output from the input, this method also makes it easier to handle various AC voltages and frequencies. The down side is that it is less efficient and much more complex than a simple switch-over system.
Designing such a system that can safely handle 1000W peak and say 500W continuous is not trivial. Getting it certified as safe is very expensive. Even the equipment to test such a design seriously will run into many tens of thousands of Euros/Dollars.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
OpenUPS2 and NUC-UPS can only provide about 60W, maybe 100W peak for short periods. And they are DC only, so you can't run your workstation monitor from them making saving your work and shutting down a bit tricky. ESR specifically says he wants to run a 150W monitor from it.
Those little things are great for low power DC servers, but useless for workstations.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Am I the only one with a $120 APC UPS with $50 replacement batteries every 2 years that does exactly what OP was looking for? And I get 35+ minutes of action time for my network gear, primary computer and one flat panel, complete with usb connectivity that shows up and is addressable in Linux.
Nevermind, just open source it. That solves the problem.
I have an old APC Smart-UPS 1500 (the black version that Dell sold, bought at a blowout price from TigerDirect back in the day), and one thing I found was that the default hair-trigger response was murder on my batteries, due to a daily power grid switching transient that would unnecessarily trigger the unit for a few seconds. Setting the sensitivity to low made a huge difference in battery life, and another thing that helped was to switch to monthly self-tests instead of weekly. I do a manual battery calibration once a year.
The 1500 is a bit overkill-ish for my setup, but it has served me well.
Oh, no! You have walked into the slavering fangs of a lurking grue!
I put them in parallel to increase the amper
Is that how that works?? ;)
Yes, actually. With older lead-acids, the internal resistance increases, leading to reduce efficiency and peak current capacity. Paralleling them up lets you get a bit more life from them.