Project: PSU Tester (A Real One)

[hitokiri_808]

What awg wire are you using on the MOSFET banks? I like the housing you got, I think it could use more ventilation though. CFM on those AC fans?

 

[hokiealumnus]

Not sure of CFM, they're second hand (as is most of it). I'll look at the specs when I'm at home and at least get you an amperage rating.

MOSFET banks have 18AWG wire for the collectors & emitters. They will be soldered into 10AWG to collect them and then go into bus bars mounted to the side.

 

[hokiealumnus]

Ok, go specs on the fans. They are 115vAC, one is .20A and the other is .24A. There is a third I'm going to try and fit as well. Airflow shouldn't be too much of a problem. I'm leaving the back panel off as pictured above and I'll be cutting some additional vents in the front panel provided there is room with the switching and instruments.

 

[WonderingSoul]

Status?

 

[Deathhorse]

Status?

yeah what he said i read the whole thing an it really interesting then it just stops. hope all is well

 

[hokiealumnus]

Heh, all is well. We've been out of town a lot and our one year old got sick....making us sick too (baby germs are evil) on the only weekend I had to work on it. Sorry for the lack of updates. Life has gotten in the way. It WILL get done, just not as soon as I would like. All of the parts except for a shunt or two are physically at my house. There just isn't enough time to work on it.

On the plus side, my father-in-law is coming down for Thanksgiving; so I'm motivated to get it as far along as possible in the event I need his help.

 

[hokiealumnus]

BUMP! Heh, bet you thought this project was dead, didn't you? Well, I'm very happy to say it's not. As planned, my father-in-law was in town (along with everyone else in our family) for thanksgiving. Not a lot of time right now to explain, but I can post some photos for you to puzzle over. Hoping against hope to be able to post a decent schematic for this thing for your perusal.

Right now, I’m just happy to report we have one of the load circuits completely operable and working nominally. It dissipated 50A at +12v without even getting hot to the touch. The other three circuits are just a (not insignificant) matter of replicating the control connections. Then comes the monitoring fun. Here are some photos of our progress. Sorry for the mess on the workbench.

 

[hokiealumnus]

Well, fter looking at the schematic I did at first, the only real change is that the center battery lead (purple wire) had to go to the emitters instead of the collectors on the transistors. As you can see, there are also only four transistors on the heat sinks instead of five. There are six of them, two each for both of the 12v load circuits and one each for the 5v and 3.3v. Nothing else is really different.

 

[noxqzs]

This is a type of project that I could stay stay up half the night without realizing what time it is. You know when you get so involved in a something that time just flies by.

One question on the monitoring side. Have you had a chance to mess around with the labjack? I have been debating on getting a similar device that I can interface with, and this one seems to be pretty versatile for programming.

 

[hokiealumnus]

Not yet. I've installed the software but that's it. The monitoring PC I was planning on using (PIII 750) appears to like to randomly restart itself, so I've placed it on hold for the time being. I will say that it is one versatile piece of equipment. The user's manual is substantial. Just the getting started basic manual has a ton of information.

Heh, if it's as good as I hope it is (and it's a big if), I won't even have to use an oscilloscope for ripple. Depending on how fast it can sample data, I may be able to use one of the lab jack inputs for it. Did I mention it's a huge if?

This is the model I got (the HV version). Check out the quick-start guide & user's manual to give you an idea of what this little thing is capable of, which is a LOT relative to its price.

 

[hokiealumnus]

Ode to a PSU Tester



Let me tell you a story. It's a story of the PSU tester that was to be.

When my wife and I took our son to visit his grandparents over Christmas, we took the tester with us. It took lots of tetris to get it to fit, but we managed it. Over our time there, I spent a lot of time with my father in law finishing this puppy up.

We did it. It worked beautifully. We put 10-turn pots on because the MOSFETs were more sensitive than we had earlier considered. Every rail was pulling its weight. The design was perfect.

Tested both 12v rails and the 5v rail. The only problem with any of those in short testing was that the 5v rail was having trouble pulling more than 30A - the low voltage combined with too much resistance (even with all of those resistors paralleled down to something like .2ohms), so that would have needed a bit more designing.

Then we got to the 3.3v rail. You see, this rail was designed without any load resistors. The wattage going through these wasn't enough to justify needing any load resistors to dump heat. We tested it at 12v to make for a quick & dirty test to ensure it worked properly...that turned out to be a good thing.

We watched the ammeter raise & lower as expected (on all rails). It was a miracle...the design worked exactly as it should with no complications. It was GREAT.

Then the amperage started creeping up...then going faster. I turned the knob back down...to no avail. I turned the switch off, with no effect. Solder popped out of the MOSFETs. Two of them then literally burst into flames.

It's a good thing we put the industrial strength quick disconnects in there (that are designed for currents far in excess of what we're doing). Grabbed that rope and yanked hard, then blew out the fires.

Crap.

Pulled out that HS and started removing the MOSFETs to reinstall new ones. Perhaps at 3.3v there wouldn't be that problem. It wasn't really designed for what we were doing with it.

Hooked one of the 12v rails back up to test pulling current for a longer time and see how hot the load resistors and MOSFET heat sinks would get. As we toyed around at 35'ish amps, it was barely getting warm. This was phenomenal.

Then the amperage started creeping up...and running away again. POP...more solder shot out of the MOSFETs (they're not loud, but not quiet in their death cry). Quickly pulled the disconnects before these caught on fire too.

Damnit.

So that's that. This design is great. It does exactly what it is supposed to do with gusto. Unfortunately, the components were not up to the task. Turns out MOSFETs (at least the ones we were working with) very much prefer to be on or off and are not happy with operating in the linear (meaning in between on & off) range.

We're not giving up though! My father-in-law kept the unit up there and is completely redesigning it. He had to procure parts he has access to, so it would have done little good here. He expects to have it fully functional with a simpler, but less novel design by the spring. I'll update when I know more.

RIP original design, you were a great experiment. Hope your replacement doesn't end up on fire like you did!

 

[hokiealumnus]

Another bump from the depths!

Over the Easter weekend at my in-laws' place, my father-in-law and I radically changed the design. I wanted to cry disassembling a perfectly good piece of hardware that just couldn't do what we needed it to, but such is life.

The new design is going to hinge on real transistors (not MOSFETS) that are fine operating in the linear range (in between full on & full off) and switched resistors.

For each rail, there will be a series of switches. each one will put a resistor inline on its assigned rail. They won't all the the same, but we're aiming for roughly in the neighborhood of 10 amps per switch (i.e. a series of 1ohm resistors, which in theory pull 12A at 12v, but with line loss, you're looking more like 10A). Once we get close to the PSU's threshold (say, within 20A), we'll start bringing up the transistor bank assigned to that rail for the fine tuning closer to the OCP point.

No ETA sadly. He has it up there and will work on it as he's able. But this design is much less risky and more of a 'sure thing' rather than an experiment with MOSFETs, which is really what the first design was.

On a side note, since I've left OCT, when it's completed, this tester will be purposed for reviewing power supplies for Overclockers.com!

Oh, and re: Soldering station, the one I ended up getting for "Christmas" (meaning Easter, heh) was a BlackJack SolderWerks BK3000LF, which is straight up AWESOME. It heats to an adjustable set temperature, monitors its temperature to keep it there and has a 30-min no activity shut off. It also takes virtually no time to heat up. This thing is super sweet. Now I need to find something to solder!

 

[hokiealumnus]

Bumpity bump bump. Posted an update in another thread, thought I'd put it here too:

Whenever my father-in-law is done with it. He has made progress (saw it over a weekend in August) and has all the parts (he was waiting on some shunts), it will just take time for him to complete it.

I asked him to take it and work on it, but it would take him more time to draw out his plans on paper than to just build the thing, both of which would happen in his spare time. Unfortunately that spare time isn't very plentiful...he still works at 68 and loves what he does (engineer). Loving what he does means 9-5 days are never and weekend work is frequent.

I can't tell you how happy I am he decided to go with real shunts (100A=100mV, if anyone wants the spec). Previously he thought we'd feel out a shunt in an area of wire, calibrating with an ammeter. Not only is that tedious, it makes maintenance a pain...if something happens to that wire, you have to 'make' another shunt. To my great surprise and relief, he ordered and got real shunts to use!

 

[kayson]

Very cool project you've got going. I spent the summer working at a company in their power mosfet division so I got to work with some $1k+ electronic loads Do you have a schematic for the current design? Seems like you've switched up some stuff.

Edit: on a side note, if you're getting a lot of mosfets or other discrete parts, you can usually just order samples off the manufacturer website and get them totally free

 

[hokiealumnus]

Oh yea, it has changed considerably. I haven't had an opportunity to make a diagram, but will make a rough one when I have time. No promises on time frame for it though. Since the schematic is in his head, it will be more of a rough guess than an actual schematic.

We've dropped the MOSFET idea; they couldn't hold the loads in a linear fashion. We're going with mostly resistive load and using transistors (like this one) to fine-tune it.

So I guess the diagram hasn't changed all that much, it's just heavier on the resistance side and lighter on the transistor side. The resistance is going to be "active" - we will use switches to load at roughly 10A/switch until the load is near where we want it, then we'll use the transistor bank the rest of the way with pots.

 

[kayson]

Oh yea, it has changed considerably. I haven't had an opportunity to make a diagram, but will make a rough one when I have time. No promises on time frame for it though. Since the schematic is in his head, it will be more of a rough guess than an actual schematic.

We've dropped the MOSFET idea; they couldn't hold the loads in a linear fashion. We're going with mostly resistive load and using transistors (like this one) to fine-tune it.

So I guess the diagram hasn't changed all that much, it's just heavier on the resistance side and lighter on the transistor side. The resistance is going to be "active" - we will use switches to load at roughly 10A/switch until the load is near where we want it, then we'll use the transistor bank the rest of the way with pots.

Yeah makes sense. Power mosfets aren't quite designed to be used as variable resistors. What kind of current are you loading? More interestingly, what are you using for your power supply?

 

[hokiealumnus]

For the tester's meters (there just to get a ballpark during operation; real values will be measured with the LabJack) we're using a 5V wall wart and for the transistors we'll have a small power supply mounted inside. I think 12V, but I'm not positive on what voltage the transistors need for operation; haven't asked.

If you meant the power supply to test, we're burn-testing on an old Thermaltake 420W I left up there with him. Once we're satisfied it won't blow up again (or won't blow up the PSU), I've got a Cooler Master 850W unit ready to plug in.

As far as current loading, we're designing it to take 100A total on two +12V rails. It would have been one, but Oklahoma Wolf helpfully reminded me that some PSUs (such as the Corsair HX1000) actually use two discrete power supplies in one box. The other two will have capacity for 35A each on +3.3V and +5V. We're not going to bother with the -12V rail.

Ironically, the hardest one to get loaded as high as we need is the +3.3V. You have to get resistance so low (and still retain the ability to dissipate heat), it's quite a challenge. V=IR, or for this purpose of course I=V/R. So if you have large resistors that are difficult to get < 1ohm, 3.3V / 1ohm = a mere 3.3A. We've got a bunch of resistors parralleled up using a (relatively) small space and are up to ~20A so far. It's going to take some thinking to get it the rest of the way and still maintain some semblance of reasonable size.

 

[kayson]

For the tester's meters (there just to get a ballpark during operation; real values will be measured with the LabJack) we're using a 5V wall wart and for the transistors we'll have a small power supply mounted inside. I think 12V, but I'm not positive on what voltage the transistors need for operation; haven't asked.

If you meant the power supply to test, we're burn-testing on an old Thermaltake 420W I left up there with him. Once we're satisfied it won't blow up again (or won't blow up the PSU), I've got a Cooler Master 850W unit ready to plug in.

As far as current loading, we're designing it to take 100A total on two +12V rails. It would have been one, but Oklahoma Wolf helpfully reminded me that some PSUs (such as the Corsair HX1000) actually use two discrete power supplies in one box. The other two will have capacity for 35A each on +3.3V and +5V. We're not going to bother with the -12V rail.

Ironically, the hardest one to get loaded as high as we need is the +3.3V. You have to get resistance so low (and still retain the ability to dissipate heat), it's quite a challenge. V=IR, or for this purpose of course I=V/R. So if you have large resistors that are difficult to get < 1ohm, 3.3V / 1ohm = a mere 3.3A. We've got a bunch of resistors parralleled up using a (relatively) small space and are up to ~20A so far. It's going to take some thinking to get it the rest of the way and still maintain some semblance of reasonable size.

For the 3.3V rail you might want to consider using power mosfets. It would be hard to get an exact load, but you can get power mosfets that are on the range of 0.010 ohms. You wouldn't have quite as much control over the current with those though. You could use that to get up into the higher amps and then discrete resistors to fine tune the load.