ATX power supply analyzer

[edgaras006]

Hello everybody,

I would like to hear your opinion about a project idea. I am thinking about building proper ATX power supply analyzer (checking full Intel ATX specification, well not full but everything what can be tested from the outputs) which would permanently sit in 5.25" optical drive slot. I can check specification with active load and oscilloscope, but that is very inconvenient. I could skip transient testing and check only ripple, but I still need an oscilloscope. Which is better, but still inconvenient if I have to take my oscilloscope to a relative where a potentially sick power supply is.

I thought that low cost ATX power supply analyzer should exist on the market, but they start from $800. There are power supply testers which checks DC levels but this information is almost worthless.

So I thought that I could make one and maybe even try to sell few units. What do you think? Is it worth a time?

[janoc]

Don't think this is going to be worth the effort. Given that an ATX PSU costs between 20 to 80 euro, depending on quality and power rating, they are effectively a consumable that is rarely worth repairing. And if you want to diagnose a faulty one, it is much simpler to just have a known good one and temporarily swap them - if the fault disappears, the supply has a problem, if not, it is something else. If it was the supply then you can cart the bad one out to your bench where you are much better equipped.

Also such analyzer will need a fast ADC to be able to measure ripple, the same for measuring any transient testing, together with a bunch of control electronics. And in order to be able to measure the full load capability, you need something to dump all that current into - either large resistors or a bank of very beefy mosfets, along with the required cooling. That's going to be big and expensive. Without testing the load a marginal supply with bad capacitors could still look fine if you only measure the ripple ...

I don't see how you want to integrate a tester like this to fit into a 5.25" drive way with all those cooling requirements and having to unplug the original load (motherboard, drives, etc.) in order to do any load testing on it anyway.

 

[edgaras006]

I was thinking about a convenient device which would be like pass through. It could monitor ripple using peak and dip detectors which could be interfaced with much slower ADC, something like 500K or 1M sample. It is still not an easy job, but way easier than oscilloscope front end to 200M samples ADC interface.

About transient testing, I do not have to run it 24/7. Something like 5-10A bursts for sub 1s would do just fine. If runs were 0.5s for each rail, and tests would go every 5s average power would be just 12W which is OK.

After all, I do not have to test transient response all day long. The most important thing is continuous ripple monitoring and alarm when it gets close to the limit or over the limit. I believe that continuous monitoring would show more than transient, because PSU would be under real load - actual PC.

Another important note is that PC is way more expensive than power supply itself and sometimes data is the most important. I am looking at a device like this as insurance for my PC, so I could know when it is time for PSU repair.

Initially, I thought about making few for my self and my relatives, made some circuits, but later I thought about maybe selling few.

I almost made an LCD for before deciding to ask people whether that device would be in demand.

Here is how it looks (I know, numbers should be centered).

[janoc]

I think you are taking it from the wrong end. If your data and PC are important, then make backups and buy good quality power supplies. If you are really paranoid, you could buy a case with redundant supplies and built-in failover, like many servers use.

But even large expensive servers don't monitor their supplies in the way you are proposing - it would be completely pointless because the supply could fail in many ways that such a monitor wouldn't catch. Your monitor essentially handles only the situation of dry/bulging capacitors, which is what causes excessive ripple (assuming the supply is otherwise designed in a sound way). That's easily remedied by buying a good quality supply.

However, if the supply suddenly decides to go BOOM (and fry the MB or your disks in the process, or even starting a fire) because of a power surge or because one of the switching transistors committing suicide, you are back to square one, regardless of whether you have a complicated monitor in place or not. The same if the fan in the supply dies.  Etc.

Basically, buy a good quality supply and you won't need this type of monitor.

OTOH, if you decide to sell this, I am sure that a lot of people will buy it - plenty of folks put various pointless blinkenlights, fan controllers and what not on the machines to make it look important ...

[David Hess]

Also such analyzer will need a fast ADC to be able to measure ripple, the same for measuring any transient testing, together with a bunch of control electronics.

Ripple may be completely characterized through subsampling.  The only limitation is ADC input bandwidth and how long you want to wait for a result.

Ripple is one of the more interesting measurements and will reveal output capacitor health however as the ATX testing standard shows, differential measurement is required for consistency so the ripple test should start with differential probing which is actually not to great a requirement given the 20 MHz bandwidth requirement.  For an informal test, I would use a lower bandwidth like 1 MHz anyway.  A lower bandwidth also allows the possibility of making an analog AC to DC ripple measurement.

[David Hess]

But even large expensive servers don't monitor their supplies in the way you are proposing - it would be completely pointless because the supply could fail in many ways that such a monitor wouldn't catch.

They do not for ATX type power supplies but large servers do not use them.  The current and future generations of "digital power" type power supplies have self monitoring built in.

[edgaras006]

Thank you all for the responses. I know that there are many failure modes for the power supply, but I have never seen a power supply which literally blows up without any warnings. Even high-quality power supplies die over time. My experience shows that most of the time caps die first. I have many PCs which runs around ~100 hours a week each. This is how my experience looks like.
Zalman ZM-600HP recapped twice since 2005 still running fine.
Corsair VX430 recapped once from 2011 still running.
Nexus I do not remember the model, but I had to recap it once since 2011.
Corsair CX430 recapped once since 2013, still running too.
Seasonic S12II-430 bought last year, already out of spec. I am not sure if I recap it or attempt to get a warranty.
Recently I just bought Seasonic Focus Platinum 550. We shall see how long it will last.

Every system is loaded from 100W to 350W. If I had had that device it would have allowed to plan repairs.

And yes, for ripple I am making AC coupled DC measurements.