Would combining double-conversion UPS and ATX power supply make any sense?

[Whimsy]

Hey all, utter newbie out of his depth here.

I'm thinking about a project to force me to learn more about electronics in general and thought I could use an UPS for my server and workstation (in home use, obviously). To make it more interesting, I wonder if I could combine double-conversion (aka online) UPS and two ATX PSUs by stripping DC-AC circuit from the UPS and primary transformer and rectifier from the PSUs.

Besides learning, motivation for this would be avoiding conversion losses online UPS normally brings while still providing good quality power to the machines. I'm planning to buy a old-ish UPS with depleted batteries which I'd replace with bunch of gelled lead-acid motorcycle batteries in parellel. I don't know make or model of the UPS yet, PSU's to be butchered would be Corsair's AX 850*2.

Bypassing transformers in the UPS and two PSUs would remove the isolation between the three pieces of equipment and it'd essentially be a single battery backed PSU with two 70 amp 12v rails +minor rails. Would PSUs cause interference to one-another? Possibly from switching noise?

I'm not sure what's the voltage after transformer in the PSUs but I'll likely need a regulator. Noise here shouldn't be an issue since PSUs are designed to work on noisy 230v line. If I assume that input to PSUs varies between 14.5v when the system is on-line and then declining to 11.5v as batteries discharge (after which system would issue suspend/shutdown PCs plugged in until power is restored). Current draw would be high tho, up to 140A from input, should I expect to run into problems there?

But most importantly, does this even make any sense? While I'd be doing this to learn, I'd perfer to have a working system in the end. Preferably one that won't fry the PCs connected or worse, be a fire hazard. High current draw makes me a little nervous. 140A is a lot of juice.

I should note that I'm not expecting to hit anything near that 140A on live system, peak draw would hardly go over 70A and median should be around 35A, maybe less. But if I'm using two 850w PSUs (which I have already), designing the system to be able to provide that sounds like the correct choice.

Insight would be appreciated. I realize that this all sounds like I'm oblivious to my ineptitude but I prefer to think I'm simply overreaching

[Galenbo]

I'm thinking about a project to force me to learn more about electronics ...

... PSU's to be butchered would be Corsair's AX 850*2. ... Would PSUs cause interference to one-another? Possibly from switching noise?

I'm not sure what's the voltage after transformer in the PSUs but I'll likely need a regulator. Noise here shouldn't be an issue since PSUs are designed to work on noisy 230v line. If I assume that input to PSUs varies between 14.5v when the system is on-line and then declining to 11.5v as batteries discharge (after which system would issue suspend/shutdown PCs plugged in until power is restored). Current draw would be high tho, up to 140A from input, should I expect to run into problems there?

You can do this. Get the schematic of every part you use, put them in the simulator and look at the results.
Don't forget to be specific. There's no such thing as 'a' schematic of 'a' stabilized voltage source.

[dtnicholls]

You've got several problems here.

Firstly, you're getting rid of one of the number one benefits of an online UPS: Isolation.

Next, you'll have a heck of a time hacking an atx power supply to take a low voltage DC input. A good part of your problem is going to be getting a 12V regulated output from 12V-13V input, there's just not much overhead there. As far as I know (not a lot to be fair) most ATX supplies use a multi-tap transformer to produce the various voltages that are then regulated to produce the supply output. Hacking a low voltage DC input into that would be near impossible, short of redesigning most of the supply. There are DC supplies you can get but they're usually pretty low power affairs for running car based PCs.

Hacking the UPS should be pretty easy though, just grab the output of the battery and go from there... Should be able to leave all the circuitry as it is, you could disconnect the output side (pull the switching fet) so that you're not producing any AC to save a little power if you wanted to, don't know what that will do as far as the effect (if any) output sensing has on the charging side though.

At the end of the day the efficiency gains had from not converting to the higher voltage and back is likely to be negligible and in terms of cost, materials for achieving it are likely to be many times more expensive than what you'd save on your power bill anyway.

Next thing worth addressing is your choice of battery. An UPS needs deep cycle batteries. AGM batteries from a motorcycle don't like being over discharged, as is likely to happen if you put them in an UPS, even proper deep cycle batteries don't fare too well if you are mean to them. Motorcycle batteries are also likely to be just as expensive, if not more expensive, than deep cycle ones made for the task.

Sorry to burst your bubble there mate, but I reckon it might be worth picking a new project and just using the UPS in the usual fashion. Getting an old one that doesn't work (assuming it's more than just dead batteries) should prove enough of a challenge to get working again though!

Hopefully someone more knowledgeable than I can be more encouraging...

[macboy]

Hey all, utter newbie out of his depth here.

I'm thinking about a project to force me to learn more about electronics in general and thought I could use an UPS for my server and workstation (in home use, obviously). To make it more interesting, I wonder if I could combine double-conversion (aka online) UPS and two ATX PSUs by stripping DC-AC circuit from the UPS and primary transformer and rectifier from the PSUs.
...
I'm not sure what's the voltage after transformer in the PSUs but I'll likely need a regulator.
...

You do not understand how an ATX power supply works.
First, the AC is rectified and filtered to provide a DC voltage, usually 340 VDC (from the rectified 240 VAC or voltage doubled 120 VAC). This high DC voltage is then used in a switch-mode DC-DC converter (push pull type) to provide several voltages, mostly 5 and 12 VDC. Those two voltages are not rectified separately; instead the turns ratio on the DC-DC converter transformer and the common filter inductor determines the 5-to-12 relationship, and an overall feedback provides voltage regulation.  This is very different from the classic linear power supply design with a transformer, then rectifier and filter caps, and finally a regulator. Notably, this DC-DC converter uses a transformer to isolate the primary and secondary.

You could ditch the AC-to-DC part of the SMPS and just feed the appropriate DC voltage directly. Since the DC-DC converter has an isolating transformer, each power supply would in fact be independent, more so than the are currently, since the outputs of ATX supplies are always grounded (the black wires are always connected to mains earth).

The trick is the DC voltage required: 340 VDC is nothing to play around with, and a battery bank to provide this would be a nightmare. Even if it will work at half that, you are still looking at a dozen batteries in series, and of course, you need some way to keep those charged, balanced, and maintained.

Just using an offline UPS will give good results for a home user. ATX power supplies are very tolerant of voltage changes; they are usually specified to run from any voltage from 90 VAC to 250 VAC, and will usually tolerate an even broader range. They will (read:should) also happily provide stable output during the short blip in the AC before a normal UPS kicks in. Using an online UPS to feed a SMPS is not so useful.

[Whimsy]

You do not understand how an ATX power supply works.

Well, you can say that again. I was under the impression that PSU converts the voltage to 12V DC and minor rails are drawn from that with step down converters. Thank you for the explanation and clearing this up. With my level of expertise, this is more than I can chew.

Getting an old one that doesn't work (assuming it's more than just dead batteries) should prove enough of a challenge to get working again though!

That's a very good idea and less likely to end up on fire or abandonment out of frustration.

[richard.cs]

If I were going down this route I would go for a high voltage battery bank that the ATX supply can run from directly. If you have a PSU with a switchable doubler (even if it's auto switching) then the ratings plate will say something like "100-130V / 200-250V" and it will internally use a DC bus of around 300V, though it may work a bit lower. If you use a PSU with active power factor correction then it will take a continuous range of voltages (often 90-260V AC) and you may be able to feed it 100V DC and have it generate its 300V DC bus from there.

Your problem then reduces to designing a battery bank at perhaps 120V nominal with a reliable charging circuit that can still supply the ATX PSU at well as charging the battery. You could seperate the functions and diode-OR a 150V supply with your battery bank but then you loose the advantages of a duel conversion supply in that your battery no longer absorbs/clamps crud on the incoming mains.

[rs20]

First, the AC is rectified and filtered to provide a DC voltage, usually 340 VDC (from the rectified 240 VAC or voltage doubled 120 VAC).

In a modern, power-factor corrected power supply, it's more than just rectified and filtered -- it's essentially (rectified and) boost-converted up to that several hundred volts DC. That's how the PFC works, you need to suck current out of the mains even when the cap is at several hundred volts and the rectified AC supply is down close to zero volts. You need to boost the voltage. A normal rectifier would just suck bursts of current at the peaks of the AC waveform, which is a bad thing.

Here's a pretty crazy idea (because the circuit is not designed with this in mind at all, and boosting voltage by 20-30x from 12V to several hundred is a tough ask): there is already a boost converter inside the ATX supply. Look into adapting it for boosting 12V? Much bigger transistor & inductor, different controller (normal DC boost controller as opposed to a fancy sine-wave tracking PFC boost converter).

[eas]

You might also look at the PicoPSU and clones, which are made to provide ATX power off of low-voltage DC.

[Whimsy]

If I were going down this route I would go for a high voltage battery bank that the ATX supply can run from directly. If you have a PSU with a switchable doubler (even if it's auto switching) then the ratings plate will say something like "100-130V / 200-250V" and it will internally use a DC bus of around 300V, though it may work a bit lower. If you use a PSU with active power factor correction then it will take a continuous range of voltages (often 90-260V AC) and you may be able to feed it 100V DC and have it generate its 300V DC bus from there.

Your problem then reduces to designing a battery bank at perhaps 120V nominal with a reliable charging circuit that can still supply the ATX PSU at well as charging the battery. You could seperate the functions and diode-OR a 150V supply with your battery bank but then you loose the advantages of a duel conversion supply in that your battery no longer absorbs/clamps crud on the incoming mains.

PSU's do have active PFC's. I'm queasy about modifying battery pack to high voltage, tho. High discharge current combined to highish voltage is a fire waiting to happen considering my overall experience level. I'll consider it in any case.

Here's a pretty crazy idea (because the circuit is not designed with this in mind at all, and boosting voltage by 20-30x from 12V to several hundred is a tough ask): there is already a boost converter inside the ATX supply. Look into adapting it for boosting 12V? Much bigger transistor & inductor, different controller (normal DC boost controller as opposed to a fancy sine-wave tracking PFC boost converter).

Sounds like a brilliant hack if doable. It'd certainly take the points for elegant sollution. Out of my current reach (due to noobishness) but I'll look into it in any case. I mean, how hard can it be Thank you for the suggestion!

You might also look at the PicoPSU and clones, which are made to provide ATX power off of low-voltage DC.

I haven't seen the name, I'll look into it. Thank you!

[wraper]

BTW online UPS unlikely to be powered with just one 12V battery. Likely there would be at least a few of them is series, therefore voltage 12V * count of batteries. Higher power online UPS can easily have 400V from batteries.

[Cliff Matthews]

Most new ATX supplies also have a separate standby transformer (the smallest of the 3, typically between the heat sinks). This is called VSB and runs the soft turn-on circuitry and aux devices (like Ethernet NIC's waiting for "magic wake-on-LAN packets" and USB ports).

Newer ATX supplies will provide VSB of 5v@>3amps (good for USB charging) and older ATX supplies were <2amps (the schematic is back from the 90's - I shaded the transformers and the "second supply" was really gutless @ 500ma). This is yet another matter to address.