Dummy Load for ATX PS testing.

[SuperHerman]

Hi,
i am new to this forum (at least new to not just reading) and am working on a Dummy load for ATX PSU testing.
My goal is to build 2 Dummy loads, one for the 12V rail with about 70A current capability and one for the 5V rail with 50A capability.
For quite some time now i have a bunch of old Darlington power Modules lying around without any use for them, so i thought it might be worth a try to use those for the job.
I am aware, that there are many Threads on this topic (although not on using darlingtons) and have read most of them.
The Darlington module i am going to use for the 5V rail is a Toshiba MG100G2YL1 which is a 400V 100A 700W@25°C unit with 2 Darlington Transistors inside.
For the 12V Rail i want to use QM200DY-2H, a 200A 1000V 1560W @Tc=25°C Module, again with two Transistors in one case.
As always with these higher power approaches the cooling will be a big part of the Problem, i am very much aware of that.

I have already build a first prototype which uses a big old CPU Cooler which is good for about 200W continuous.
I have not drawn a schematic of the circuit, but it is quite simple.
The current is mesaured by 6 parallel 0.01 Ohm Resistors, that current singal is amplified by a factor of 30 so that 100A produces 5V at the output.
Then comes the erroramplifier, which compares a 0-5V reference (10-Turn Pot) to the currentsense signal and drives the basis of an npn transistor.
The collector-current of that transistor is mirrored into each basis of the darlington module.
After compensation of the erroramp that worked ok, but the voltage was not distributed evenly across the 2 Modules, so i added a servo amplifier which compared the middlepoint voltage of the module to a reference voltage derived from the complete voltage across the module and deflected some of the current to the basis of one module.
Now it works quite good and the power dissipation is shared across the modules evenly.
One problem with this circuit is, that everything is referenced to ground, which is connected to the emitter of the lower darlington transistor in the module, so the max. voltage i can connect the load to is about 3 Volts above the midpoint voltage. This is not an issue when testing 12V or 5V, but it limits its use to somewhere below 24V with a 15V Power supply for the control circuit.
I think a better way would be to use the collector of the upper power transistor as the reference ground and use a floating supply for the control circuit.
If anyone has some thoughts about this and maybe about a better way to implement that voltage-sharing servo, i would appreciate it.
Attached are some images of the ugly prototype and of the powersupply i used for testing, so far i only testet it up to 50A without any problems besides temperature rise of the Heatsink.
I will soon draw a schematic of the thing and post it here too.

Greetings from Germany
Ps.:
Sorry for not reformating the pictures, simply forgot it.

[SuperHerman]

Hi,
i just created a schematic diagram of the circuit i build, the values are not identical because i tried to simulate and did not have the correct models for the parts i used, so to make it work in simulation i had to change a few things.
Over the last weak i repeatadly used the Load to test ATX PSU's with good results (meaning no failures of the Load).
So maybe that circuit can be useful for someone who wants to give a purpouse to his old slow Darlington modules as heating elements.
If anyone has any suggestions  or sees obvious flaws in the circuit, please don't hesitate to tell me so, i am still at the start of learning electronics.

Greetings