High power transistor drive suggestions:

I am improving a 400W power load . (target is about 400W per module, more of these are paralelled)
The load driven by a current sensor and properly scaled,smoothed analog current feed made from a 12-bit PWM.

Schematic = see attachment.

Today: T1 is a 2N3055, it’s collector (A) is directly connected to (BC) T2,and T3 is MJ11012 (2x30A, 200W) darlingtons.

I am not afraid of thermal runaway as T1,T2,T3 is mounted on same heatsink - and the current on T1’s basis is controlled by measured current.

I plan to replace T1 with TIP122 (darlington hFE > 1000), to be able to drive more loads from one PWM pin.

Questions:
-is there any reason to connect a resistor between A and BC , or two resistors between A-B and A-C ?

Is there any other issues that should be adressed ?
Maybe T2 and T3 should have it’s own RL, (I think it’s wrong).

Please remember: the circuit is designed to be just bad, it’s all about providing a powerful resistive load.
RL is multiple paralelled 100 resistors , and T1,2,3 are on another heatsink. Both heatsinks are colid copper and using 8 heat-pipes each to transfer heat to cooling fins where a fan ensures good flow.

load.png

I am not afraid of thermal runaway as T1,T2,T3 is mounted on same heatsink - and the current on T1's basis is controlled by measured current.

Great if it removed your fear of it. Unfortunately it will not remove the reality.

-is there any reason to connect a resistor between A and BC , or two resistors between A-B and A-C ?

You might want to limit the base current to stop the transistor being destroyed.

You should be OK on the heat as long as most of that 400W is going into the resistors, and you're talking about those monster CPU heatsinks - the good ones can dissipate 100W without even breaking a sweat (I mean, look at the TDP spec of a high end part - that's before the overclockers who buy those heatsinks jack up the heat generation). Make sure you're making a good thermal contact.

Either a resistor between A and BC, or between A and B, A and C will limit the base current.

I'd be more comfortable with 2 resistors, so that if one of the transistors doesn't behave quite the same as the other, it would be less likely to destroy itself (imagine one transistor that conducted a little more readily than the other - it could be passing much more current, and you wouldn't know it until the part failed, resulting in all the current going to the other transistor and trashing that one too).

Don't understand why you are doing this...

A single medium sized mosfet on a very small heatsink, driven with PWM, could have done the same job of accurately heating up resistors.

If you want current limiting in you original schematic, add a resistor in the emitter lead of the driver T1.
e.g. if you are driving the T1 base with 5volt max, the emitter will be 0.6v less.
There will be a ~4.4volt drop over it's emitter resistor.
With ohms law, you can work out the value.
Darlingtons with a DC gain of 1000 won't need a lot of base current. It can even kill them the way you are driving them.
I suppose ~100mA base current is more than enough. 47ohm. 100mA can be done with e.g. a BD135.
Then you still have the saturation voltage over the darlingtons, so the resistors are never fed with the full 20volt.
Back to the simple 1x1cm 75Amp mosfet.....
Leo..

Grumpy_Mike : Funny comment about reality vs believing :slight_smile: I calculated that 2x30A darlingtons, each able to dissipate 200W , should never get close to be over loaded if RL is dimensioned for the planned 10-27V , even with quite uneven load. - and the cooling is very good, the casing (TO-3) is barely ~50degree.

DrAzzy: I were unsure about the need to limit BE current (or if slightly diferent resistors would made any hFE difference worse) - while T2,T3 have a max VBE of 5V, 300mA @30A , I guess the PID loop would never allow it to go that high in the first place because the control current is regulated by sensed current in the circuit.

Wawa: The reason I moved to anlaog circuit, instead of the digital PWM I had first (IR2125 driven IRFP064N) - is that the PWM way was very difficult to filter perfectly, and even with some filtering, the fast hall-effect current sensor needed to be sampled many times and smoothed out digitally, (31khz PWM) , also, doing a analog RC filtering to smooth it out, proved difficylt whithout sacrificing some precision/response-time.
So I decided to make a fully linear load instead of PWM -which was easier - but not the filtering of such currents..

The reason I am mentioning using a TIP122 darlington as driver, is that perfect smoothing of the PWM is easily done in one step is I only need very little current.
Thank you all.

I calculated that 2x30A darlingtons, each able to dissipate 200W , should never get close to be over loaded if RL is dimensioned for the planned 10-27V , even with quite uneven load.

I think you miss my point. Transistors even on the same heat sink will not share currently evenly. I have been on many projects that attempt to do this ( lead by senior engineers ) and they have all, without exception failed.

Still confused.
Why do you need a current sensor with PWM power control.
You know the exact resistor values I presume.
And the exact supply voltage if you are using a regulated switchmode supply.
Then the power in the resisors is calculated by the "ON" time of the PWM signal.
I don't see why the normal ~500hz/8bit PWM out from an Arduino can't do the job.
That will give you a better than 0.5% power resolution.
Maybe better/easier than sampling current AND voltage (and calculating power) from your analogue circuit.
Leo..

P.S.
Small emitter resistors for T2/T3 will solve the imbalance.
Two separate wires from each emitter to the positive supply will do.

Its trivial - drive with PWM, put the low pass filter on the current sensor output.

[ Actually that's a rather glib statement - with 400W the power supply has to be
able to absorb the PWM current pulses without strain ]

If you wanted smooth current you still wouldn't use linear current regulation,
you'd ideally use a switch-mode regulator, which means some tricky circuit design and
big inductors.

Grumpy_Mike:
I think you miss my point. Transistors even on the same heat sink will not share currently evenly. I have been on many projects that attempt to do this ( lead by senior engineers ) and they have all, without exception failed.

This is a powerful battery tester, - 400W per module on even the lowest voltage(10V) , would mean 40A , unless the transistors (30A) are so very far apart, as one pulling over 30A while the other conducts under 10A - then I don't think it's a real issue. (?) - they would need to be very, very different for that to happen.
So far, this have not been a problem.

Wawa:
Still confused.
Why do you need a current sensor with PWM power control.

Small emitter resistors for T2/T3 will solve the imbalance.
Two separate wires from each emitter to the positive supply will do.

I need PWM to be able to load battery packs in ~10…~40V with constant current or constant effect (to simulate actual load in UAS)

  • will add 47ohm resistors “just in case” :slight_smile:

MarkT:
Its trivial - drive with PWM, put the low pass filter on the current sensor output.

The problem is : with narrow PWM (like if I wish 10W load) - It's big/expensive to smooth it perfectly (big coils etc), or else I will get 31khz ripple on current measurement - smoothing out hall effect sensor is a tradeoff between accuracy and speed.

Regarding PWM resolution, I don't mind small changes in load a set 200W may vary between 198 and 202W many times a second, as long I can measure it accurately enough and calculate mAh/Wh properly. But if I dropped big/expensive filtering, and were using PWM all the way, - the current sensor AD sampling is not able to keep up with the ripple and accuracy is bad.

No, low-pass filter the current sensor in hardware of course, multipole filter if necessary.

I think with the last posts we have a better understanding of what you want to achieve.
Personally I would go back to the first drawing you made. Analogue load, but with an adjustable current, done with PWM.
And A/D reading of the battery voltage and current, so you can adjust current and work out total power and log that over time.
I made a crude drawing, so please add your comments.
As said, those darlingtons (I hope you mean the MJ11013 PNP) only need a few miliamps to drive.
Just a small 1A power transistor is more than enough.
The 47ohm resistor limits current at just below 100mA, and makes PWM control smoother.
I have added 2x20A emiter fuses. They also act as balancing resistors.
I have drawn a 0.1ohm load. Low, because of the 10volt battery requirement.
That means that the transistors do a lot of the heavy lifting.
Study the SOA graph on the datasheet. You might have to add a third one…
The PWM out is controlled, in software, by the current feedback.
Leo…

Whanks Wava - I'll add fuses, good idea, just in case.
I'm off to the "vestisen" to fly UAV for a few weeks now, but will follow up when back.

Thanks for the drawing, the circuit have been operational for some time now, - and yes, I will stick to the analog solution after testing the digital one.
I am using a 50A hall-effect sensor for current, and a voltage divider for voltage. Both calibrated against Fluke instruments for precision.

Everything low-current and the 32u4 ""leonardo" is on an small PCB on the backside of the display (see attached picture)