I am trying to use the parts I have on hand, to build an active load for testing the output of DC power supplies. The idea is based on this design: https://www.deeptronic.com/electronic-circuit-design/variable-dummy-load-for-power-supply-testing/

But uses MJ11016 NPN darlington transistors in parallel for higher power https://www.onsemi.com/pub/Collateral/MJ11012-D.PDF

I have attached a drawing of the schematic. Please let me know if I've done something drastically wrong here! I have some giant diodes and will be inserting one for polarity protection, but I really want to know if the basic circuit is all wrong.

If you have an idea for improvement please let me know that as well.

 

Over what voltage and current range do you want to test?

 

I am hoping to be able to cover from zero or just a couple of volts up to 100V, and as much current as can be sunk. Obviously this is going to blow off lots of heat. These are on big heatsinks with a high-powered blower to keep them cool, and I can always get more heatsink for them if necessary. In the most-extreme case I have a giant water-cooled heatsink, along with the necessary radiators/fans/pumps. I think the temperature of the transistors is not going to be my main problem. The maximum limit seems to me to be defined by the safe operating area.

Here is the SOA graph from the datasheet, with the scale on the right showing the original and the scale on the left showing 4x the original current levels.

 

as they are BJT the SOA rating is at Tcase = 25 deg C, if the heatsink gets to 75 deg C say, you will be lucky to get 30W out of each xtor before they hit 2nd breakdown ... also current sharing at 100V and low current will be an issue ...

 

That's the sort of "second set of eyes on a problem" thinking I came here for!

The 1.5 ohms resistors, I *think* should be ok for current sharing at high voltage. What else can I reasonably do, to ensure good sharing between transistors under these conditions?

Medium amounts of math follow, tl;dr: I think I can get 90W out of these per device in the expected use case, vs. 200W.

Understanding that I am talking about absolute maximum ratings...

The transistors are rated to have a junction to case resistance of 0.87ºC/W. It is may be impossible to find out exactly for the insulators I have, but I found one source that says 0.009ºC/W and I found one source that says I could expect 0.36ºC/W with mica washers. So call it 1.2ºC/W from junction to heatsink.

Total dissipation per device is rated at 200W. These have a rated junction temperature of 200ºC. 200/0.87 gives about 230W, but then you add an ambient temperature on there and the 200W rating makes sense. I have a thing about running heatsinks actually-hot, so for grins I'll postulate I let them get up to 50º with forced convection. Then I have 150º to work with on a warm day. If my thinking is right, that gives me 120W per device on an infinite heatsink at 50º.

These are on about 6" each of heatsinks that look exactly like Boyd/Aavid's 61820 profile, which they say is 1.94ºC/W/3" with Perimeter=23.64. With a correction factor of about 0.75 for about 5.5" length each that's 1.455ºC/W. Then correct it again for a 30º rise over ambient (1.257) and it's back up to 1.83ºC/W. Multiply by 0.6 for forced convection and it's about 1ºC/W.

Profile specs from https://www.boydcorp.com/thermal/air-cooling/aluminum-extrusion-profiles.html
Looks like the same correction factors on a clearer print with (to me) more useful explanations: https://www.mouser.com/catalog/supplier/library/pdf/Aavidselectionguide.pdf

So aiming for a 50º heatsink with forced air cooling it looks like I can call it 1ºC/W. Plus 1.2 for the joints and I'm at 2.2ºC/W for the combined heatsink/interface cooling solution. For 200W that's 440º which is juuuuust slightly ouver the maximum temperature. So it looks like I can reasonably expect to be able to get 91W out of these, each.

 

On further review, I think the 1.5 ohms resistors may not cut the mustard. At high voltage, I don't need to be able to sink so much current thanks to the SOA limitations, and these do not generate much voltage to discourage curent hogging at low current. At lower voltage/higher current, the resistors can be set to exceed their maximum power ratings pretty easily. They have a 5x for 5 seconds over-power rating but I would like to keep everybody under their maximums if possible. I might need to have at least two sets of resistors, one for higher and one for lower current operation.

 

The comment about the emitter resistors was a very good lead! Having done some modeling (having FINALLY figured out how to put a new model into LTspice) I figured out some stuff.

At high voltage, I want to use a relatively giant resistance, currently thinking 10 ohms, on the emitters. This gives almost a volt difference at high current, about 10mV/mA so I think there's plenty of wiggle room, even for driving a darlington transistor, even at low voltage. By comparison, that's 7x the vbe degeneration as the previous 1.5 ohms. Somebody please correct me if I'm still off here!

I also dropped the lower resistance resistors to 0.75 ohms for lower dissipation/higher current capacity

I bumped up the base resistors from 2k to 3k because it gave me an easier time adjusting the sunk current at high voltage and didn't hurt anything at low voltage, except that it cuts out a bit of current sinking ability at the very lowest voltages.

I've added another pair of transistors so there's 6 circuits. Here are some model results at various voltages. Shown are 4 circuits, so there can be visibility of high and low voltage emitter resistors with currents and voltage each, without having to change anything but the operating point.

 

Just to close the loop on this: I built it and it works well! I have used this to test the outputs of a few different power supplies now, up to 60V and about 8A using the 10 ohms emitter resistors and 18V 30A using the 0.75 ohms resistors. The heat sinks didn't get warm with their fans blowing, and the transistors didn't get very hot either. Nice.

I ended up running everything through a 0.01 ohms resistor on the + input so I could use a volt meter to sense current, and I have digital voltage and current meters which is super convenient. I made a maximum current-at-voltage table and stuck it on the base of my load so I can easily compare voltage/current limits and not go too far. I still may go back and fit a fat diode to give some reverse polarity protection, but the power cables are labeled and it's only me using this, so maybe not.

At high voltage/current the current likes to creep up a little bit, but a careful hand on the fine adjustment knob keeps that in check.