Bricked my PSU, Was it my UPS/AST PDU?

Hardware Hub
, ,
1

I recently got an AST PDU and a UPS to switch my servers over to battery backup in case of power outage. Now, both the “source a” for the PDU and the UPS in was connected to a Furman surge protector.

BUT, I also had my 2x 3090 machine learning box connected to that same power strip.

Somehow during me setting up the cable for the PDU/UPS (and me testing the failover functionality), the PSU on my machine learning system has DIED.

Chatgpt is saying it had something to do with line interference or something. Like a voltage sag.

But given how AI is wrong alot, I thought I would ask here.

Is the PSU really dead for something so simple? Its an EVGA 1600watter. I need to make sure this is likely the reason.

Is it truly bricked? If so, what happened?

2

Servers? And one of them was making use of a 1600W PSU?

Add up the total max load (in watts) on that single PDU and report back.

3

The ML threadripper system was using the 1600w psu. But it WASNT plugged into the PDU. Thats the confusing part. It was plugged into a different power strip.

Wall → Surge protector power strip

Then all of these plugged into the power strip:
<PDU source a
<UPS (which is also source b for PDU)
<threadripper system

4

Let’s explore this.

I believe you are saying PDU Source A is wall power (via the surge protector) and PDU Source B is the same wall power, by way of an inline UPS (via the same surge protector).

Is this correct?

5

Yes. But if source A goes down, the UPS battery would be source B. So I thought it wouldnt matter.

6

OK. Where do Source A and Source B derive? Are they separate power circuits? I do not believe this is the case as you inform they both come from the same surge protector. Correct?

You mentioned servers, plural. Was anything else attached to the Furman surge protector at the time of failure?

7

Other things attached to the surge protector were an Xbox and monitor. Both of which work just fine now.

The other server was a dell, it was the only thing connected to the PDU. I was using it to test the PDU failover.

Source A = direct into the surge protector.
Source B = into UPS (which in turn is connected to the surge protector)

8

Another piece of info that might be helpful: The threadripper system did some weird power stuff. Or at least seemed to. I could have sworn I had turned it OFF. Then later it was back on. So then I turned it off again, and after that it hasnt turned on no matter what I do.

9

Start simpel. 1600 watt Psu. How many watts was it powering. P=u X I. Something shorted? Measured fase to 0 and fase to earth. Did you have eyes on the set-upall the time. Not that someone Plughed in there 250 watt power brick for their empty laptop. Open the Psu short the condensator so It won’t kill you when you accidently touch it. Measure the fuse. And look for burnwd wireing.

10

Im not near skilled enough to do all of that. I just need to know if it is possible something on the surge protector killed the PSU and how to prevent it next time. In other words, what happened?

11

Inside the Psu is a fuse. They call it a fast fuse because it will blow fast(and thus protect) the other components in the Psu. What you can do is buy a fuse which you can install before the Psu which act even quicker. I assume there was a power surge which blew the fuse inside the Psu. If you had to much watts behind the 1600 watt Psu. Then the voltage will drop and amps will go up. That’s bad that can burn wirering (and a building) down. Do you maybe have a simpele drawing on how you connected everything.

12

Tell us the model of PDU and UPS used for testing. Be as specific as possible.

13

CyberPower PDU44001 Switched ATS PDU
Eaton 9PX1500RT 9PX 1500VA 1350W 120V Online Double-Conversion UPS

14

This UPS is not compatible with the selected PDU (at least, NOT in your as-tested configuration).

Let’s break down why this is true.

CyberPower PDU44001 Description

Under Product Features, the following is provided:

  • Automatic Transfer Switch
    Dual power cords (A and B) provide redundant power sources to devices with single power supplies and can be plugged into separate compatible sources* to ensure continuous access. Users select the preferred source as the primary input. CyberPower Smart App Online and Sinewave series UPS are highly recommended for use as the power source. *Note: Only sine wave AC power sources are compatible with this ATS PDU. Simulated sine wave AC power sources are NOT compatible.

Here we can see that ONLY pure sine wave AC power sources are compatible with this ATS PDU.

Eaton 9PX(1500RT) Product Datasheet

Reference Specifications table (page 2), Output:

Crest Factor: 3:1

Output AC Waveform (Battery Mode): Pure Sine wave

At first glance, one might be inclined to believe this UPS is a pure sine wave UPS… and they would be wrong. The UPS provides a pure sine wave output ONLY in Battery Mode (when the output load is supported by the battery).

You will note that nowhere in the product literature does Eaton go so far as to label this UPS as a “pure sine wave” UPS, even though they do specify that the output AC waveform is a “Pure Sine wave” in Battery Mode.

The (dirty) trick is hidden in an understanding of the meaning of Crest Factor.

Crest Factor (CF) is a mathematical relationship used to analyze waveforms.

CF = [Peak Value of Waveform] / [RMS Average Value of Waveform]

In AC power systems, voltage is produced as a variable output, a sine wave, where the magnitude of the voltage is changing in time.

RMS, or Root Mean Square, is used to describe the average value of a time-varying signal. For pure sine waves, the RMS value is [Peak Value / square root (2)] and so if we plug this expression into the above expression we find:

CF = [Peak Value of Waveform] / [Peak Value / square root(2)]

The peak values cancel out and we are left with CF = square root(2) or approximately 1.414.

This means that if the sine wave is a perfect sinusoid (here “perfect” means pure), then the CF, regardless of peak value, should be 1.414, and cannot go above 1.414.

But the Crest Factor for this UPS is 3:1, not 1.414:1. Ergo, when NOT operating in Battery Mode, the output is NOT a pure sine wave, even though the input from the wall IS a pure sine wave.

This is a low-end UPS and has most likely resulted in the failure of your Threadripper system PSU due to incompatibility with the paired PDU in this configuration (I cannot stress this enough).

I suspect these components may have caused an unspecified variation in line voltage during the switching process (from Source A to Source B, or vice versa) resulting in an out-of-specification low line voltage, magnifying the in-rush current into your TR PSU during turn-on (lower voltage, higher current) which has blown an input safety fuse in your PSU.

It’s dead, Jim.

I would caution you on opening the PSU and replacing the fuse yourself unless you understand proper electrical safety consideration and are experienced in such operations. Just because the PSU is unplugged does it mean those big input caps are fully discharged.

Your mistake was putting the UPS in front of the PDS (as a source). You should have put the UPS on the output of the PDS. There was no point in having the UPS as an input as both the direct input from the wall (via the surge protector – Source A) and the input from the UPS (via the surge protector - Source B) both ultimately derived from the same source (the wall). This does not provide for redundancy as Source A is Source B. You can label them “A” and “B” all you want but they were really “A” and “A” or “B” and “B”, certainly not “A” and “B”.

There really is no point in switching from a non-battery backup source to a backup-source as 1) if the wall power goes out, you’ve lost input to “both” sources, and 2) if the wall power goes out, the battery picks up the load (so what’s the point in switching over to a dead wall outlet?).

Redundancy is only achieved if “Source A” and “Source B” really truly are different sources (i.e., different circuits). And then only are you protecting for a failed circuit or a failure of the power line from the wall to the input of the PDU (which is a pretty damn unlikely scenario). I do not get the impression you gave this too much thought.

Because the restriction on the use of the PDU references the sources (and not the outputs), placing the UPS on the output would have been fine.

Now you know (and knowing is half the battle).

1 Like
15

As a stretch (and to cover all bases), I would encourage you to go to your breaker box and check to make sure the breaker for the affected circuit is still closed.

It may have tripped on overload during the switching process due to the mismatched AC waveforms going into the PDU.

It’s possible you think your TR system PSU has failed because the outlet at the wall is unpowered.

16

Thanks for this write up. But this is a $2,000 UPS. It’s not low end. I dont think?

" Eaton go so far as to label this UPS as a “pure sine wave” UPS, even though they do specify that the output AC waveform is a “Pure Sine wave” in Battery Mode."

Dont all UPS’s operate on AC mode until the power goes off?

Also, the point of my plan was to run on AC power unless a power outage then switch to the UPS. This is to avoid having to depend on the UPS (wearing out the battery?) and to avoid firmware errors which I have experienced with my other Eaton.

17

The give-away (for those in the know) that the Eaton UPS is NOT a pure sine wave UPS is the use of the term “Double-Conversion” in the description:

Eaton 9PX 1500VA UPS

Here “double-conversion” mean that the AC input waveform from the wall is rectified to a DC voltage and then inverted back into an AC waveform on the output, hence “double-conversion” (AC → DC → AC).

This is a dead give-away the output is NOT going to be a pure sine wave as the ONLY way to create a PURE sine wave is with reciprocating generators (where an armature spins around a stator… like a turbine generator or motor generator). The fact that it is a SINE wave derives from the fact the generating impetus is a rotation.

There are no moving components in a UPS so no way to make a PERFECT sine wave.

“Double-Conversion” is advertised as it sounds cool. Ohhhh, DOUBLE conversion. Malarky. Cheap… NOT pure sine wave.

When I say “cheap,” I don’t mean you might not have paid a lot. I mean it’s a cheap design. Personally, I would never touch such a unit… it’s not pure sine wave, which is critical for sensitive equipment like computers.

18

I dont know what to believe! Chatgpt says:

Good news:
The Eaton 9PX is not a “line-interactive” unit like cheaper UPSes.
It’s an online / double-conversion UPS — meaning:

It always converts incoming AC → DC → AC through an inverter, even when wall power is present.

That inverter’s output is a pure sine wave 24/7, not only when the battery is discharging.

So the commenter’s claim that it’s “only pure sine in battery mode” is wrong — they probably confused it with Eaton’s 5S or 5P (line-interactive) series.

19

Also you are forgetting the the PDU DID switch over to the UPS. Something it did not do with my cheaper UPS. So this one IS compatible with it.

20

Go back and read what I wrote. I said NOWHERE does Eaton go so far as to label this UPS… i.e., they never say it’s a pure sine wave unit, and it’s not.

Yes. There’ no DC anywhere until an internal PSU rectifies AC to DC and then regulates down to the various line voltages for distribution to various internal components. Everything from the wall to your PSU is AC, including everything into and out of a PDS, UPS, etc. Here in the US we don’t distribute DC, we distribute AC.

There is no load on the battery until the input voltage falls below whatever may be the lower input threshold (not necessarily 0V). Running off of a UPS does not wear the battery. If you’ve had this problem in the past is because all batteries have a limited life, whether in use or not. They don’t last forever. Use will accelerate aging, but it’s not the only factor. Keeping your battery at lower temperature is the number one consideration you have control over.

It doesn’t sound like you needed to purchase a PDU. Run the UPS from the wall (you don’t need a surge suppressor as that functionality is built-in) to the PSU and be done with it. If the power goes out it will automatically switch to Battery Mode until the AC is restored, at which point it will re-charge the battery and resume normal AC operation from the wall.

Power Distribution Units (PSU) with ATS are for making use of redundant circuits in case of circuit failure, not complete power failure (ALL circuits go down)… unless of course you have a home generator powering either some or all circuits in your home. Not sure why you decided to buy an ATS PDU based on what you’ve shared thus far.