Hardware Secrets Power Supply Test Methodology

[Olle P]

Welcome!

That's a very interesting observation you make there.

Originally Posted by jolphil

Of course the consumer could elect for an expensive one but that may not be a consideration for the average desktop user.

I sincerely doubt that the "average desktop user" bothers to invest in an UPS in the first place.

In my mind there are three separate reasons for using an UPS:
* You live in an area where brownouts or complete power loss are frequent.
* You use the computer for a living and do highly sensitive stuff that must be saved to prevent a huge loss of money.
* You run a computer that absolutely must be available 24/7. (Possibly connected to a network that's also supported by UPS.)

Cheers
Olle

[jolphil]

I sincerely doubt that the "average desktop user" bothers to invest in an UPS in the first place.

That may, be but I can tell you that all of my family uses a UPS..I guess mostly because of my urging..

I realize that some mfgrs. do specify a "holding" of anywhere between 12 to 17 MS and thats good, but it does not tell the whole picture of whether the Power supply will start after power is interrupted ..I did a small test on my Budget UPS to show what I mean about the waveform..First picture on the right is a waveform with the power connected to the UPS.Reasonably nice sine wave..(note the amplitude is incorrect as my scope could be damaged by over voltage.A restive divider was used)..The middle picture shows the waveform with no UPS load and the AC power removed..The left picture is with a very light load to the UPS again without AC power..
This shows it is not a square wave but a pulse waveform and as noted from picture two and three, the duration and amplitude of the pulse varies with the load on the UPS..I did not run a heavy load test but my guess it that the waveform gets closer to a real square wave..
The reason I brought this up was that reading some of the reviews from a very popular power supply company and the Company's tech stated that his supply needed a sine wave to operate..
Heres the link to show the waveforms..(I hope you can get them)

http://my.imageshack.us/v_images.php

Jolphil

[Olle P]

Originally Posted by jolphil

I realize that some mfgrs. do specify a "holding" of anywhere between 12 to 17 MS and thats good, but it does not tell the whole picture of whether the Power supply will start after power is interrupted .

Very true, but still a non-issue if you "never" get any such interruptions.
Where I live we have such interruptions on the power grid roughly half a dozen times per year, which means that the probability it will happen when I'm using the computer is extremely slim.

As noted in my previous post: Frequent power drop-outs is a reason for using UPS.

Originally Posted by jolphil

Heres the link to show the waveforms..(I hope you can get them)
http://my.imageshack.us/v_images.php

That's the login page, not where the image is.

Cheers
Olle

[jolphil]

First Olle..Thank you for your welcome..
Second I apologize for not giving you the proper link..Try this one..
http://profile.imageshack.us/user/jolphil]jolphil[/URL] at 2009-06-03
I guess your power grid is much better than in the US as there are a lot of UPS devices being sold here..I know I live in the Lightning capital of the world in central Florida..That being said we seriously need them..
Thanks again for your welcome..
jolphil

EDIT: Oh Oh!!! still not the correct URL for the images..As you can tell by now I am struggling with the correct
method of how to share images..I will get back if and when I learn how to do it..

[jolphil]

[IMG][/IMG]

I post more if this works..
Jolphil

Yippee!! Heres the second and third..
[IMG][/IMG]

[IMG][/IMG]

[Olle P]

I've come up with a new idea...

When testing high power PSUs you often run into the problem that your power meter can't handle as much output as you want. Not on the 12V lines and not in total.

A simple, yet accurate and pretty cheap, solution would be to attach a precision resistor or two (like this one) in parallel with the power meter. Each 2ohm resistor would draw about 6A/72W, modified by the exact voltage.
Just slap the resistor to a sizeable heatsink, connect it to a Molex connector and you're ready to go!

This way you can test with more power drawn from the 12V line(s).

__________________________________________________ __________

Some comments regarding the thermal measurements:

- What's the purpose of measuring the temperature of the PSU housing?

- The most important temperature is the inlet air temperature, somewhat represented by the "room" sensor. (Sensor could possibly be mounted at the air intake, unless the flow messes with the accuracy.)

- Then come component temperatures inside the PSU. These are for all practical reasons impossible to measure directly. Some indication could be gained by measuring the outlet temperature, since outlet air < heatsinks < components.

Cheers
Olle

[Gabriel Torres]

Originally Posted by jolphil

Hello,
I have just been reading how you test power supplies and think so far, You are
doing more "real world"testing than any article I have read..That being said, I failed to see, or missed, anything regarding PSU'S with active PFC how they behave with typical UPS units..Some "average" UPS units supply a square wave to the computer/PSU rather than a sine wave.. I Read On Some Newegg reviews, the Tech response to that is that they need a sine wave UPS to operate after loss of AC power..Overall, many units can slip by this critical feature and even cause the computer harm..
At the very least, the owner would be completely surprised that his computer
shut off incorrectly after a loss of AC power with a UPS being used..Of course the consumer could elect for an expensive one but that may not be a consideration for the average desktop user..I would like to see some load results test where the AC is switched off and on several times at say 90,115, and 120 volts or some test parameters you choose..I realize this is NOT an ATX specification but it is a real world condition for many..After all this does happen..
Overall you folks are doing a great job to help us with deciding which power unit is best for me and others..
Thank You,
jolphil

Hello Jolphil,

Thanks for your nice words about our reviews. What you ask is on my list of ideas for a long time now. I want to do a review on what you ask, problem is finding time to do so. Thanks also for the oscilloscope pictures, it shall help other users to understand how is voltage output on UPS units.

Cheers,
Gabriel.

[Gabriel Torres]

Originally Posted by Olle P

When testing high power PSUs you often run into the problem that your power meter can't handle as much output as you want. Not on the 12V lines and not in total.

This problem you describe we face with our load tester, not with our power meter. The solution you are giving has several drawbacks and we can't use it:

1. It is a passive solution; we prefer using an active solution (resistors are passive components; our load tester uses transistors to generate load).

2. It can't be easily reproduced. We always prefer to use ready-to-use solutions.

3. Added components generate parasite components and/or noise that can and will affect the results.

The real solution is upgrading our machine with a more powerful one...

Originally Posted by Olle P

What's the purpose of measuring the temperature of the PSU housing?

So users can have an idea how much heat the power supply will introduce inside the case and also how efficient the PSU cooling system is. Good power supplies will have its temperature only 1º C to 5º C higher than enviroment (air) temperature, but power supplies with poor cooling solutions the difference can be of 10º C or more.

Originally Posted by Olle P

The most important temperature is the inlet air temperature, somewhat represented by the "room" sensor. (Sensor could possibly be mounted at the air intake, unless the flow messes with the accuracy.)

It is an option. However we decided not to do this way because the heat on the air coming out of the power supply is not only generated by the power supply, as the power supply gets air from inside our thermal box (which is already very hot).

Originally Posted by Olle P

Then come component temperatures inside the PSU. These are for all practical reasons impossible to measure directly.

We could put a thermometer probe on the secondary heatsink, for example. Not complicated to do. I only do that when we face overheating issues.

Cheers,
Gabriel.

[Gabriel Torres]

Hello all,

We updated our methodology today. Here are the main differences:

1. We bought a Instek GPM-8212R power meter power meter, which has an amazing 0.2% precision.
   
2. We bought a Fluke 52 II thermometer, which has an amazing 0.05% precision.
   
3. We finally built our "hot box" and retired our old cardboard box. See the pictures!
   
4. Added pictures from our lab.
   
5. Added a list with the complete breakdown of how much all equipments have costed to us so far. This was done for two reasons. First, to show you that we are very serious and commited with our power supply reviews. And second, to help other publications willing to build their own lab.

Enjoy!

Cheers,
Gabriel.

PS: I will try to redo all our recent power supply reviews in order to fix efficiency numbers using our new power meter.

[PM650]

the link to the sunmoon load tester appears to contain a worm...google seems to agree.