I may build my own Desktop PC UPS

[rebelrider.mike]

I am a bit aggravated with my store-bought UPS, and UPSs in general. The theoretical issue I have is that they are not truly uninterruptable. And you wouldn't want them to be because the inverter sucks. At least for the ones I can afford. Anyway, they operate by using a relay that switches the supply from grid to battery super fast in the event of a power outage or some other dirty power event.

My real life issue is that I have frequent power interruptions lasting only for a second or two. Usually the UPS switches over fast enough for my PC to stay running. But once or twice a month, it's not quite fast enough for whatever reason, and my computer reboots without warning. Every time this happens I get a little more aggravated. To the point where I'm thinking of building my own UPS where the battery and inverter are always in use so there is no switchover.

Maybe this is overkill? I don't know. Can't hurt to design one and see if I'm willing to buy all the parts vs. maybe upgrading to a UPS with a faster switchover time. Is there such a thing? I've never researched it. The UPS I have now is an APC 750. I don't know the exact model, but it has a THD rating of 45%. Good enough to run things for a few minutes while I save stuff and shut down the computer, but I wouldn't want to run off of it continuously.

In fact, remember that time I made my own THD meter with a laptop and a battery charger? The APC did pretty good for it's rating between 38.5% and 44.6%. Not particularly important, but at least we know that APC is being honest about their THD values. I've fount that other companies are not so truthful.

Anyway, here's what I've got so far:


I'm thinking I'd keep the APC, because it has some nifty features built-in that I like. High/Low voltage alarms, an app for the PC where I can see things like power disruption instances and battery health. (It tests its own battery periodically.) An internal surge protector. It's also it's own power strip. I can leave the monitors plugged into it since a brief power disruption won't bother them much. Worst case is they will reset which takes hardly any time.

I'd use a regular power supply plugged into the APC to get DC to a regulator, which will feed the battery and/or inverter. The battery would have a compatible voltage range as the inverter, and the inverter would be pure sine (less than 3% THD) to provide clean power to the computer. I'll have to put my PC on a power monitor to see what kinds of Amps are needed to run things.

The battery will need to be able to handle both output current and input current for multiple conditions.

• No grid power: the battery will have to provide all the current to run the inverter.
• Grid power, full battery: the power supply, regulator, and battery will have to provide the current to run the inverter.
• Grid power, low battery: the power supply, and regulator will have to provide all the current to charge the battery and run the inverter. Also, the battery will have to be able to accept the current provided by the power supply and regulator. If the PC is off, the battery will receive all the current available from the regulator, so a maximum voltage and current will have to be set so as not to trip the BMS. But enough current needs to be available to run the computer and charge the battery if the computer is running. So a battery will be needed that can accept high enough current for both cases.

Does that make sense? I think it will once I figure out the actual power (and therefore current) needed to run the PC. Although the computer's power supply is rated for 750W, I already know that it is unlikely to need more than 300W because the APC app lets me monitor power usage in real time. But that's with both monitors also plugged in, and I'd like to use a second device in case the APC is not accurate. As I write this, my PC is using about 55W. I need to run some kind of stress test though to see what the maximum might be.

I'm using a Poniie PN1500 Electricity Monitor. I've seen a couple YouTubers using them and they're pretty low cost so I thought I'd get one. I've been very happy with it. I'll report back when I have more current data. (See what I did there?)

 

[Korishan]

To the point where I'm thinking of building my own UPS where the battery and inverter are always in use so there is no switchover.

Maybe this is overkill? I don't know.

What you're talking about is an "In-Line UPS". Another term is "Active UPS". Basically, Grid power is fed through a rectifier and then passed back through the inverter. The only difference is that if the grid goes out, it just switches to the battery, which is seamless as it's the same source as far as the inverter part is concerned.

The UPS I have now is an APC 750

Is this the one the computer is connected to? If so, this is probably a big reason, it's too small. The 750VA is only rated for about 560Watts. You derate the VA by about 75% to get the wattage it can handle. Your computer with monitor(s) is probably close to that limit. Even being within 10% is enough to make it not be fast enough.

I'll report back when I have more current data. (See what I did there?)

no, no I didn't


Aside from the learning experience, I'd say either go with an in-line up, or go full on DC power to the devices. Monitors are pretty easy to convert to DC as the power is rectified anyways. You would either need to input 120VDC to them, or you could bypass the initial circuit and go straight 12V/5V internally. The computer is a little more tricky. There are DC PSU's on the market that take 12 to 48VDC input and break out to the various loads in the system.
If you wanted, you could even just get a bunch of those buck converters that can do 10+Amps. Not sure how "exact" the voltage needs to be or if the voltage can fluctuate a few mV up/down w/o causing issues. Altho, always could put some capacitors on the output side to smooth out the rippling.
This is the kind I was thinking of:

You can get them to output 5V as well. Altho, you'd need another buck converter to get the 3V3 lines needed for the mobo.

 

[OffGridInTheCity]

My off-grid AIO MPP Solar 3048LV is set to 'battery' -> AC priority so it's basically running on battery all the time (as I understand it) and uses the incoming AC for charging the battery. Thus there's when the AC drops, there's no switch over / acts like a UPS. Maybe buy a small, used AIO and use it in place of the UPS?

Heck, one could coble together a DIY battery -> inverter + AC -> charger -> battery.

 

[rebelrider.mike]

Found it: APC Back-UPS ES 750G. The app says I can put up to 450W on it, though I've rarely seen it use more than 280W even when doing graphics intensive stuff like gaming or video rendering. The PC is currently using about 56W, and the app says that 85W are being used. The two monitors together then must be using 29W. So 15W each I guess. I'll measure them directly later.

I found a stress-test program called OCCT that ran my CPU and GPU at 100%. For nearly 7 minutes my PC became a space heater. I've never heard the fans work so hard, and I could feel the heat coming out the back. The test stopped when the CPU got to 90C. The power reached 409W, and my APC app announced that I had too much load for it to reliably back up. I've never gotten that message before. So that was fun. The GPU never got above 87C.

I was able to get some numbers today:

• Off: 2.55W
• Sleep Mode: 2.45W
• Idle: 60W
• Video Streaming: 90W
• Video Gaming: 250W
• Video Rendering: 170W
• Stress Test: 409W

These numbers are all variable. The meter never sits at one value. It's always changing. So these are approximate maximums that I've observed. My spontaneous reboots typically occur when the computer is idle, or in sleep mode. But there are lots more times when the UPS switches over just fine and the computer never notices. That's what has me a little bit baffled. Why is the computer only rebooting during some of the power interruptions? Why not all? Could it be that the UPS is switching fast enough but some weirdness in the power disruption is causing the RAM to reset?

I've begun looking into inline UPSs. I'll keep looking, but so far I'm not liking what I find. Either they're way too expensive or you have to ask for an estimate (which amounts to the same thing.) Also, the ones I'm finding still have switchover times in the description. The inline part is supposed to eliminate that isn't it? There is an actual company out there called Active Power UPS which really screws up searching for active UPSs. And they don't even have active UPSs from what I saw. But like I said, I'll keep looking.

I've thought about converting my PC and monitors to DC before. It seems silly to take AC and convert it to DC for the battery, then back to AC for the power supplies which convert it to DC again. But pure sine inverters are pretty darn efficient, and I'd have to do a lot of testing to find every 12V, 5V, and 3.3V component and then see how many amps each one wants. I also looked into ready made DC power supplies for PCs. Too pricy for me. At least they were at the time.

 

[Korishan]

My spontaneous reboots typically occur when the computer is idle, or in sleep mode. But there are lots more times when the UPS switches over just fine and the computer never notices. That's what has me a little bit baffled

How old is the UPS?? That might be part of it. Perhaps the FETs are degraded and don't fully switch when they are supposed to. Or could be the caps not discharging as fast as they use to; this is more likely I think. Or, the caps don't hold as much capacity as they use to as they degraded. This happens quite often for Electrolytic Capacitors. You could see about replacing them with identical or large uF capacity ones as a side quest.

so far I'm not liking what I find. Either they're way too expensive or you have to ask for an estimate (which amounts to the same thing.)

Yup, this is part of the reason why they aren't as common for households. What you would look at in this case is for "used" server UPS's. They are usually faster switching, have higher switching capacity, and generally are quieter and more resilient. Plus have higher battery capacity, even if using LeadAcids. I was able to get 2 APC SUA3000 units for about $100 each. Fortunately they were close enough by that I could drive and get them, saved $150 in shipping, had them immediately, and got to take a scenic drive up to Georgia

I'd have to do a lot of testing to find every 12V, 5V, and 3.3V component and then see how many amps each one wants. I also looked into ready made DC power supplies for PCs

Yeah, that's the biggest issue right there. The time to track everything down
As far as the cost, that's why I suggested the Buck converters and not go straight to the DC PSU's

The good thing is that those bucks I mentioned all share a common ground, so you wouldn't need to do any weird isolations or conversions to make them grounded. Which means you can put them in a single box with a common ground output with several voltage outputs.

 

[rebelrider.mike]

For now, I think what I'll do is add up the hardware I'd need to build my own, and compare it with the other solutions you've all suggested.
Another possibility is that I could try to resurrect my old APC 1300. The one with the dead battery. See if it works any better.

 

[rebelrider.mike]

I set this project aside for a bit, but looking at the components I'd need to build my own online UPS, I'd need about $116. That's for the power supply, DC regulator, and pure sine inverter. And also a diode.

The cheapest used one I've found so far is $150 on eBay. An EMERSON Liebert GXT3-1500RT120. I did some looking and found this diagram for the GXT3:


Similar to mine but with extra goodies. The differences that stand out to me are that they have a direct connection from the power in, through the power supply (they call it a rectifier) to the inverter, to the output. The battery charger, battery, and DC converter are in parallel. There must be a good reason for that, but I don't know what it is.

My design wouldn't need a DC regulator between the battery and inverter because the inverter can use the full voltage range of the battery. Not sure what the benefit is there.

There's also a "Dynamic Bypass". Maybe for replacing the battery without shutting off the unit? The manual probably says what it's for, but I can't be bothered to read it unless I actually buy the thing. (Edit: this is a joke by the way. I'll read it, just later.)

Inside the "Rectifier", there seems to be a Power Factor Correction circuit. Mine would definitely not have that because I don't know how to make one. I'm curious though. If there is power factor correction, then why are the Watts different that the VA? If it was corrected, I'd think they would be the same.

I was planning to keep my existing UPS for the surge suppressor, and it probably has the filters built in too. I'm guessing it doesn't have a PFC circuit, but I'll go check.

 

[Mark_T]

rebelrider.mike,
Google suggested to check some settings. Did you check it?
Check BIOS Settings:
Review and adjust power-related settings, ensuring that auto-restart options are disabled for unexpected shutdowns.
BIOS/UEFI Settings:
Incorrect power settings in the BIOS (like USB Wake Support or Wake on AC) or an oversensitive "Auto On Time" can cause the computer to power cycle when power is restored.

 

[Korishan]

rebelrider.mike,
Google suggested to check some settings. Did you check it?
Check BIOS Settings:
Review and adjust power-related settings, ensuring that auto-restart options are disabled for unexpected shutdowns.
BIOS/UEFI Settings:
Incorrect power settings in the BIOS (like USB Wake Support or Wake on AC) or an oversensitive "Auto On Time" can cause the computer to power cycle when power is restored.

This isn't a problem with his computer, it's a problem with the UPS not keeping power supply steady during a grid blip.
If it were the computer it would happen more often or he would have to reset/restore the BIOS settings often. I don't think this is the problem regardless what AI says it is

 

[Mark_T]

it's a problem with the UPS not keeping power supply steady during a grid blip.

I have no doubt about that too.

 

[Mark_T]

I made a discovery!
In sleep mode PC will shut down power supply.. only 5V for stb stays ON.
i connect +5V from laboratory power supply to violet wire (stb power 5V)
and disconnect power from PC.. waited like 30 sec and connected back and pc
resumes from sleep mode! repeat it 2 times.
Current consumption on those 5V was 0.13A
Power supply is Liteon ps 6351-2 rated 100-127 volt/ 1A,
but in sleep mode stays ON down to 20V(AC)

 

[rebelrider.mike]

I had a 5 second outage today, followed by a big boom. Likely one of those transformer fuses blowing up, and the power being rerouted. This time my UPS decided to behave properly, and my PC and monitors just kept going like nothing happened. I still don't get why the UPS works 9 out of 10 times, and then sometimes doesn't. Maybe the switchover relay is just getting old?

 

[Korishan]

Maybe the switchover relay is just getting old?

Possible. Depending on the age it could be actual relays, or it could be FETs. The FETs would be controlled with capacitors which could be weakened.