KeriProductions ATX338 PSU Load Tester

[KeriJane]

Hello.

I noticed something.
There is a shortage of test equipment designed to accurately test ATX Power Supplies.
At the low end, there's the ULTRA brand go/no-go voltage tester that's pretty useless. At the high end, there's SunMoon and higher.
In the middle, well, there's nothing! Well, at least I couldn't find anything.
Until now.

Introducing the KeriProductions ATX338 PSU Load Tester. This is a two-channel load tester, allowing scope output and voltage display of two different circuits at once. The load resistors are switched via a bank of toggle switches. The ATX ON switch is interlocked to the POWER/FAN ON switch so that the fans must be turned on before the PSU can be activated.

It is based upon industrial wirewound resistors of 3Ω/50W for the 12V+, 1Ω/25W for the 5V+, .5Ω/25W (20w for now) for the 3.3V, a 10Ω/25W for the 5V standby and a 30Ω/25W for the -12V. The total wattage comes out to around 338. Which should be high enough for testing low to mid power PSUs. Extra space remains for two more resistors. They are mounted to the case with simple Carriage Bolts and have some wavy washers to allow for expansion.

The resistors are rated for temps up to 350C but derate rapidly over 50C The two fans appear to keep the resistors under 60C. (with the help of a borrowed infrared thermometer)

Load testing was done tonight with the help of a recapped TruePower 380. I have not yet wired the Voltmeters or scope outputs, but did verify that the fans run, a fair bit of heat is generated, the indicators for Power, ATX Power Good Signal and the rudimentary -12V load indicator all function.

Configuring the Velleman LED displays was an adventure, but I got them set for the 0-20V range with voltage dividing resistors despite the instructions written for people a lot smarter than me.

The casing was just a case that I picked up at Fry's. After drilling a GAZILLION holes in it and grinding out the rectangular holes with a Dremel tool and a file, I painted it with high temperature paint. The Label was created with OpenOffice DRAW.


The next step is to wire the selectors, voltmeters and scope outputs. As I understand it, decoupling caps should be attached to the scope terminals inside the case.

I should have it finished up this weekend.


Any comments or suggestions are welcome.

Thanks for all the help,
Keri

[dood]

Re: KeriProductions ATX338 PSU Load Tester

Looks GREAT! You're going to build mine next, right?

[Oklahoma Wolf]

Re: KeriProductions ATX338 PSU Load Tester

Looking good

[PCBONEZ]

Re: KeriProductions ATX338 PSU Load Tester

I LIKE IT!

Putting together something like that has been on my 'to do' list for oh,, 5+ years.

[KeriJane]

Re: KeriProductions ATX338 PSU Load Tester

Hello.

OK!
I got the Load Tester 99.9% complete. The .1% is that I "forgot" to connect the decoupling caps on Channel 2. Oh, and 1 of the 5V resistors isn't here yet.

So...... Off to some load testing. It works! The panel meters conveniently display voltage, the scope outputs provide signal to my scope, the fans make a reassuring drone, hot air comes out of the left side, .... in short, it does everything I expected.

Wow, do those decoupling caps make a difference!
On CH1 the readings are as expected, with a sawtooth ripple pattern. It gets more ripple as load is increased. The 10uf electrolytic and .1uf ceramic caps are in parallel right behind the BNC connector.

On CH2 without the caps, there's those giant gnarly spikes all over the place. Even on the 5VSB with the PSU running. With the PSU off, the 5VSB is pretty clean, but not as clean as CH1.

So here's my questions:

The decoupling caps affect the readings a lot. Aren't they distorting the signal?

I have the scope set for a 1x probe. Is this correct for this application? The scope signal is taken from the back of a rotary selector switch that is exposed directly to voltage taken from the ATX connectors, before the load switches.
The case (and BNC connectors) is grounded to the PSU ground wires. The power for the meters and fans is provided from a 9v wall adaptor and is completely isolated from the ATX circuits and grounds.

If I put a "Warranty Void if label is removed or damaged" sticker on this thing will Oklahoma Wolf come snooping around with an insatiable desire to void the Warranty?

Pics to follow,
Have Fun,
Keri

[Oklahoma Wolf]

Re: KeriProductions ATX338 PSU Load Tester

The decoupling caps affect the readings a lot. Aren't they distorting the signal?

Yeah. They're supposed to simulate motherboard filtering.

If I put a "Warranty Void if label is removed or damaged" sticker on this thing will Oklahoma Wolf come snooping around with an insatiable desire to void the Warranty?

Give it a try and we'll see

[Per Hansson]

Re: KeriProductions ATX338 PSU Load Tester

Wow, that really looks great
If you want to make a couple of them I'll offer to laser cut the chassi for you in stainless steel, granted I get one of course

Get the production going!

[PCBONEZ]

Re: KeriProductions ATX338 PSU Load Tester

How are the caps installed?

I think you are using bypass caps and calling them decoupling caps.

Decoupling caps would not simulate motherboard filter caps.
Or rather. I don't see how they could.

.

[Bob Parker]

Re: KeriProductions ATX338 PSU Load Tester

The decoupling caps affect the readings a lot. Aren't they distorting the signal?

Yes.

[KeriJane]

Re: KeriProductions ATX338 PSU Load Tester

Hello.

I was wondering.....

Would it be a good idea to build another load tester with analog meters instead of the digital ones?

Or, put another way, does anyone else have a preference for watching a moving needle rather than lit up numbers?

It would be sightly cheaper and simpler to construct and could even run the cooling fans from the PSU power.

The big question is:
Can a 0-15VDC analog meter compete with a 0-20VDC 3.5 digit digital one?

Sometimes watching a needle tells more than a number. In particular if the sampling rate is low on the digital meter, a needle might react faster to fluctuations.

Or maybe I could build one with a digital meter on one channel and an analog meter on the other.
But that would look asymmetric and off-balance. Ick.

Oh! PCBONEZ, Yes, they would appear to be Bypass caps, not Decoupling ones. I read "Decoupling caps" somewhere relating to this application and adopted it without researching what "Decoupling" means in this instance. Thank You for that correction.

So very much to learn,
Keri

PS: I'm gonna get some more pics of this thing! honest!
And some results of testing stuff.

[KeriJane]

Re: KeriProductions ATX338 PSU Load Tester

I don't know you KeriJane but only a couple of hours ago I was reading about your home-brew ATX PSU load tester - looked pretty good! You should post more details about it when you get chance.

OK, here's more details and a schematic:

The KeriProductions ATX338 PSU tester is a simple set of switched power resistors cooled by two fans.

The resistors are all connected to the ATX, P4 and Molex 4-pin connectors through a bank of switches at one end, and a common Ground bus at the other end. Two resistors for -12v and 5vsb are considerably smaller than the others and are merely glued to the housing to dissipate heat.

The ON switch is actually DPDT (not shown in the schematic) with one pole running the fans and voltmeters while the other one is wired in series with the ATX ON switch to reduce the chance of running the PSU without the fans running. A fan speed, airflow or temperature sensor should really be added to stop the PSU as well.

The ATX ON switch pulls the ATX signal from the PSU to ground, thus starting the PSU.

The three indicators are "12v and 5v" LED types with built-in resistors.
One runs off of the 9v for the meters and fans, one is driven by the PSU's PWR GOOD signal, and the third is wired in parallel with the -12v resistor just to see if there is -12v present.

Two voltmeters with BNC outputs have their own rotary selector switch.
This allows testing of two circuits at once while using a dual-channel oscilloscope.
The voltmeters require 9vdc on a separate circuit to operate which is provided by a wall adaptor that also runs the two 12v cooling fans at a somewhat reduced speed.

The BNC outputs have two small capacitors each, a 10uf electrolytic and a 0.1 ceramic disc that emulate being connected to a motherboard and seem to be necessary. Thanks to Oklahoma Wolf at JonnyGuru for that tip.
And, Oops! the 2nd one isn't shown on the schematic.

No Ammeter is provided or needed. The resistors are of fixed values and adequate wattage. So long as the cooling fans keep the resistors under 50-70° C, (verified with an infrared thermometer), current will remain fairly constant as determined by voltage / resistance. (Ohm's law) For my reference, I printed voltage and amperage at each switch.

My theory of PSU testing is that the PSU need only be partially loaded to get an adequate measurement of voltage, ripple and noise.
It's not as dramatic as JG's SunMoon maxing out the latest PSU until it screams or explodes but I just want to test under normal loads and conditions.

Few computers fully load even a modest 300w PSU which this tester can (and has) overload, and a reasonable assessment of condition can be had at half power on most systems.

As my tester is completely manual, one must be careful in choosing how to load a PSU. Overdoing the 12v on a cheap or ancient PSU that favors 5v and 3.3v can kill it right there as can overloading the 3.3v or 5v on more modern, primarily 12v models.

Never depend on overload protection to work!
It's no fun at all to explain to someone why their PSU died during testing.
Fortunately, I haven't had to do that yet except for myself.

Have Fun!
Keri

Created and posted entirely on a Mac Pro with OODraw

[c_hegge]

Re: KeriProductions ATX338 PSU Load Tester

YAY! Great to have you back KeriJane!

FWIW, I built myself a similar load tester (only, it looks more amature-made and goes up to 1kW). To cool it, I have 3 120mm Fans, two Delta Mega Fast 250CFM ones and an NMB out of a Dell (with the the speed controller disabled). As you can imagine, it sounds like a Boeing 727 taking off.

[kaboom]

Re: KeriProductions ATX338 PSU Load Tester

Welcome back! I was a 'board reader' before I officially joined and enjoyed your posts even then.


Nice tester. I'd rather have resistors than huge current sinks/sources, the least of the reason behind that is (to a point) resistors can take the heat. You have to derate semis as their temp goes up... If you could manage the feat of keeping a TO-247 device at room temp, maybe it's better than a resistor- I don't know..

I'll use my dummy loads and 50W 12V bulbs! And for those real pigs, I connect a 12V to 120V inverter up and 'play.'

I'm not too picky- all I'd probably do is solder some pigtails to molex connectors to make connecting things easier...

Again welcome back!
-Paul

[ipman]

Re: KeriProductions ATX338 PSU Load Tester

There's nothing better than a pure resistive load. And an inverter is no such thing.

[kaboom]

Re: KeriProductions ATX338 PSU Load Tester

What else is there at DC? You certainly can't have a reactive load with DC.

With an inductor, after 5 IR constants, current will be limited by the DCR.
With a cap, after 5 RC constants, current will be limited by the (usually) very high leakage resistance of the cap.

Even though the first stage of the inverter boosts a nominal 12V DC to 170V DC with MOSFETs, a transformer, rectifiers and caps, to the power supply feeding it, it certainly appears resistive.

The DC power supply knows or cares not whether it's a fraction of an ohm resistor or a bunch of power semis that it's feeding.

-Paul

[KeriJane]

Re: KeriProductions ATX338 PSU Load Tester

Thanks for the WB's

I considered using the far cheaper bulbs instead of resistors. Well, OK, they're not really cheaper, just easier to find.

Bulbs have a couple of characteristics that I didn't like:

1- they get really hot,

2- there is a brief period of time while they warm up. During this time, the resistance is very low and can overload the PSU. A PSU is sensitive to extremely rapid current fluctuations and a bulb warming up is one of them.

3- Bulbs are pretty big. A car headlamp is maybe 60w and having 5 of them in one place is a bit much.

I chose resistors because I'm not smart enough to figure out a more elegant solution using switching transistors and really big heatsinks.

Resistor derate with temperature too. The ones I used are fairly stable up to about 60-70° C. Above that, resistance goes up a lot and limits current. The max temperature rating is something crazy high but they probably would never get there even with the fans off due to the increased resistance.
Turning the fans off is not something I'd care to try because even if the resistors are OK at 300C, the nearby voltmeters and my floor might not like it so much.
Because they're wire wound, there's bound to be a small reactive force though I believe they were designed to minimize reactivity.


One possible idea is to use a high voltage bulb as a resistor. One that won't glow at all at 12v. This would get around the whole warming up issue. It would probably take several in parallel.

[kc8adu]

Re: KeriProductions ATX338 PSU Load Tester

i found out bulbs are not good for psu testing.had a mess of 1157's rigged up as t battery test load.about 30 amps worth.cold start resistance too low and tripping the astron rs35m i was testing.ran fine on a carbon pile load at full rating.

[kaboom]

Re: KeriProductions ATX338 PSU Load Tester

2- there is a brief period of time while they warm up. During this time, the resistance is very low and can overload the PSU. A PSU is sensitive to extremely rapid current fluctuations and a bulb warming up is one of them.

I hear ya on that. The OCP can (will?) trip when bringing up the bank of bulbs. If I bring three of them up at once, most of the time the supply will trip, just a twitch of the fan before the primary shuts down. Two is not a problem with the single-transistor-forward Bestecs.

It's more interesting with the older half-bridges. Two bulbs usually trips them. No cycle-by-cycle, then lockout, current limiting as with a 3842- of course! When the two 'back-bias' caps in the base circuits of the two power xsistors start going bad, the switching may become unbalanced. Makes the transformer -very- unhappy during those high-current starts. Which then trips the 'derived' OCP via 'ratio-derived feedback' via the driver tx (in supplies so equipped.)

The Hipro I rebuilt into a 'big' supply did this. Two bulbs would not allow it to 'bootstrap.' After recapping, it would almost bring up two bulbs, and starting with one seemed to get easier. Less 'chatter' from the main TX upon power-up. Then I wound a bigger TX and went 'overkill' with it. That thread is way down there...


A trick from these new microwave boards is to energize the load with an NTC, which is bypassed after ~200ms, in series. Maybe something like that could be used to control the 'inrush' of these bulbs.

I stagger-start the bulbs if there's a chance of their inrush tripping the OCP. I've also been known to submerge low-value 10-25W WW resistors in small cups of water. As long as they are submerged, they're not getting much hotter than the water. Of course, the element will be hotter, as it's not in direct contact with the water.

And several old hard drives during the load test are nice, too.

[ratdude747]

Re: KeriProductions ATX338 PSU Load Tester

Keri-

As for the digital vs analog thing- you could integrate what I put in my PSU tester... What I am suggesting (for your unit) is to use analog displays and have bannana jack output and a switch to connect a DMM to (if you needed a digital readout).

.......................................................................

all my psu tester is is a little device that has an atx plug, 2 6 position rotary switches, a power switch, 2 banana plugs, and a probe. what it does is it can be used to start a psu and once external load is added (if desired), you use the 2 rotary switches to set the testing voltage which is sent to the v+ banana plug. the common banana plug is always connected to ground.

settings:

rotor 1:

1. 12v+
2. 5v+
3. 3.3v+
4. Rotor 2
5. Probe
6. Ground (both plugs to ground

Rotor 2:

1. 5vsb
2. PWR good
3.12v-
4. 5v-
5. unused
6. unused

the test socket fits 20 or 24 pin sockets. anything not on the 20 pin part (4 additional ins, sata, molex, p4, pci-e, etc) you use the probe to test.

to go with it I may make individual load units... they would plug into molex/p4/pci-e plugs and be used to give the amount of load that is wanted tested and how much on each rail (one molex unit would have just 12v, another would have just 5v).

[boggystudios]

Re: KeriProductions ATX338 PSU Load Tester

i found out bulbs are not good for psu testing.had a mess of 1157's rigged up as t battery test load.about 30 amps worth.cold start resistance too low and tripping the astron rs35m i was testing.ran fine on a carbon pile load at full rating.

I just ordered a bunch of parts to make an ATX PSU load tester. I think I will use a carbon pile for the loads but can only find them sold as battery load testers. Guess I will have to rob a few.