I'm doing a quick and dirty project to build a resistive DC load to test some ATX12V SMPS. I've got access to a bunch of 5 ohm, 50W power resistors at a bargain price, and I have some aluminum and copper bar stock for heatsinks. I need ~800W total load, about ~700W @ 12V, ~75W @ 5V, and ~30W @ 3.3V.

I'd like to use a rotary switch to vary the +12V load by 2.4A at a time. A local shop as some 6 position switches. So one of those would allow me to vary the +12V load between 0A and 12A. Get 5 of them and I have 0A to 60A, or 720W. Great! Do the same thing with 3x for the +5V and 3x for the +3.3V and bam.

Here's the problem though. That means I need 55 power resistors. That's a couple hundred bucks. Ok. I was expecting that. But it also means about fifty bucks in IRF740s with my current design, and any transistor I can find that will do the job is around the same price or more. I also need some diodes. And some resistors, but I have access to a ton of 5% 0603s so no biggy.

How can I condense my design to use fewer semiconductors? The power resistors I already have sourced and they're my best deal, but I feel like I can use fewer transistors and diodes, but it just ain't coming to me right now. It's been a rough day; I need a beer.

Any advice?

 

IRF740 is 10A 400V MOSFET. Take a look at IRFZ44N which is 55V 49A which would cost almost the same.

Put 2x or 3x 5Ω 50W resistor in parallel for each MOSFET and you can save more than half the cost on transistors and switches.

I doubt the 5V DC to the gates of MOSFET is enough to make the mosfet conduct fully. I guess you'd need 12V there as the diodes would drop 0.6V each on the gate voltage.

Allen

 

It looks like you are wanting about 25 steps in your adjustment range. How about making it 31 steps and have them be binary weighted?

So make it 2A per step (giving you 62A max, or 744W). You then need six branches with {2,4,8,16,32}A respectively. Each branch it turned one or off with a simple two position switch (though you can get rotary switches that have 32 positions with 6 binary encoded outputs). For each branch, just put the right number or resistors in parallel and control them with a single suitable solid state relay or power MOSFET.

Given the basic topology you have, you will probably end up with something between 2A and 2.4A per step. There is one problem, which may or may not be significant in your case, and that is that the voltage across the power MOSFETs are not going to be the same because they aren't carrying the same current. Thus you will have some non-linearity in your data, but it is pretty easy to calibrate and compensate for when you plot and analyze the data.

 

The IRFZ44N looks like a good pick, thanks mate. So is the idea of doubling up resistors, which is so obvious I actually had it in an earlier design, but for some reason forgot about it later. That makes it a bit cheaper, since I can have two sets with 4.8A steps and one "fine tuning" knob that does 2.4A steps.

The diode voltage drop was a concern for me. However, I don't want to use the PSU's +12V output to drive the mosfets if I can help it. If the power supply fails under testing the +12V might deliver some pretty nasty transients, and I don't know what that'll do to the power mosfets if delivered across the source and gate simultaneously, but I suspect the result might be pan fried silicon.

The +5VSB in ATX SMPSs shares the EMI filter and bridge rectifier, but from that point on is an isolated circuit, so it's isolated from the most common failure points: PFC circuit, PWM circuit, and secondary rectifiers. It also stays present even when the PSU is off so I can use it for aux circuitry like if I want a voltage pass/fail circuit (fun with comparators! Or I could cheat and use an ADC ).

Calibration will be a pain, but shouldn't be too difficult. Besides, if I needed real dead-on precision I'd just mortgage my soul and buy a Chroma testing rack.

 

Why not just build a constant current circuit, using a big light bulb as your load, and set the current by changing the reference voltage of the constant-current controller? You need only one big resistor as a current shunt, one op-amp to do the controlling, and a bank of MOSFETs sized to handle your worst-case current. I think worst case is when the MOSFETs and the load are dissipating the same power.