my LM324 electronic load is oscillating

[ledtester]

I've been testing my build of the "150W 10A 72V LM324 Electronic Load Constant Current Tester Kit TL431 Reference" that is prevalent on ebay/aliexpress/etc. and I'm encountering weird behavior especially at lower load voltages.

I've made a couple of videos demonstrating the issues:

- Part 1:
- Part 2:

The video descriptions have all of the details of the set up.

Basically when the load voltage is 5V I'm seeing oscillations start at a load current of 2.6A; at 10V the oscillations begin at 5.1A; at 15V the load is stable up through 6A (the limit of my power supply.)

Part 2 shows other weird behavior when I sweep the power supply's output voltage from 5 through 30V at various load currents.

The power supply (a Topward 6303A) is second hand. How should I go about debugging and fixing these issues?

Schematic of the LM324 load kit attached.

[ledtester]

FWIW, here is a video of the sweep test using the "XHDZ-FZ-2G" electronic load kit:



Schematic: https://goo.gl/Olr6k2

[Kleinstein]

The resistors R5,R8,R13,R16 in front of the gate are relatively large. This makes the control relatively slow an may reduce the stability. Normally some 100-220 Ohms should be sufficient - it's just to isolate the OPs output from the capacitive load.

The Resistors R10,R15 ... are odd: they couple the FET control signal to the measured current. This would reduce the accuracy a little, but should not cause oscillations - its more like helping a little. The resistors are quite large (The R21 value is probably a mistake), so the effect may not be that bad.

There could be a possible problem with the known cross over distortion of the LM324 - this may sole down the OPs to make them prone to oscillation. One could try loading down the OPs a little (e.g. some 1-5 K to Ground)
For the initial tests it could be a good idea to use only 1 of the 4 channels. This makes changes easier and the currents a re lower.
It may be a good idea to have some kind of switch to turn of 3 channels (e.g. the control signal from the pot). This gives an extra lower current range.

Normally an electronic load would need some RC damping element like 10 Ohms an 1 µF at the output (e.g. across the load terminals) to avoid high frequency problems.

[xavier60]

The  feedback resistors should be removed. The feedback capacitors might need to be larger.
But the main problem might be because there are 4 loops controlling one output. They could be trying to correct each other.

[iMo]

The R6/R18 (etc) create an inverting amplifier with gain 220. The shunt voltage is amplified and closing the fet. The oscillation is around 70kHz from the video. You may experiment with the 1nF feedback capacitors values.

Also I would remove the 10uF capacitor at the TL431's output as the TL431 may oscillate when loaded with certain capacitor and ESR values.

ESR>2ohm (10uF) is recommended for I_431 around 1mA and 2.5V output.

[StillTrying]

I'd try decoupling the pot's wiper, 100n to 10uF.

[Kleinstein]

The 4 parts are not interacting much. They all do a current regulation on there own. So the current in one path changing would not effect the others very much.

For the high frequency range, where the voltage change does have quite some effect (e.g. via the drain - gate capacitance), there should be the RC combination at the output to dampen possible oscillations.

[toeeks]

I know this thread has been dormant for a while, but as the kits are still available for purchase, here's how one can fix the oscillations:

I've noticed the labels on the PCB ask for 75N75 FETs (4000pF input capacitance, 80 amps continuous), but my kit shipped with 110N8F6 (a whooping 9130pF on the gate, 110 amps continuous). Swapping these monsters out for a set of IRFZ24N (370pF, 17 amps) removed the oscillations entirely. As there's four FETs in parallel, they could still dissipate more than 50 amps, even though this PCB doesn't look like it's made to carry that much current for an extended time...

I also tried replacing the LM324 by a TLC274CN before I replaced the FETs. That had no effect.

[Kleinstein]

The limiting factor with the FETs is usually the SOA - that is how much power they can handle without local overheating on the chip. This is often relatively poor for lower voltage FETs like IRFZ24.
There is not DC SOA spec for the IRFZ24 - chances are it could be relatively poor - like a maximum of some 0.1 A at 40 V, only a small fraction of the rated power dissipation. The smaller Fet can be OK for low voltage, like 12 V. To be on the safe side with a small FET this would be more like an IRF510 or IRF620, so defenitely higher voltage types.
So there is a good reason for using relatively large fets - this is not because of the current, but because of the power rating.

With the same resistor values for the gate and fedback path, the gate should have a capacitance lower than the capacitor in the feedback. So either use small FETs or adjust the resistors / caps. 

[toeeks]

Good catch. Let me still remark that I haven't had issues with the IRFZ24N FETs so far, which I have mounted on a PR138/94/M3 heat sink with two 45mm SUNON ME45101V1-000U-G99 fans.

When fully turned up, the pot limits the maximum current to 8 amps in total (2 amps per FET). The heat sink gets hot after a while when it's dissipating more than 40 watts, so that's the limiting factor for my rig. Running it at 5V/8A, 10V/4A, 20V/2A and 40V/1A for a couple of minutes each didn't lead to anything blowing up, but of course your mileage may vary.

[Kleinstein]

For low voltage, like 5 V or 12 V the low votlage FETs is usually OK. The critical case in the list is 40 V/1A - same power as 5 V and 8 ampls, but much more critical for the FET. How much individual FETs can withstand can vary and one can be luky to get relatively good samples that can withstand quite some power. The problem however is that not all samples would survive it - some may just fail within fractions or a second, and usually with a short. It is not so much about how long they stand, the point is the risc for them to fail early one.

If used without reliable SOA data, it would be a good idea to test the unit at slightly more severe conditions - after that one can be reasonable confident all is OK, even without a data-sheet to confirm. Testing at 12 V and 4 A does not guaratee it would stand 40 V and 1 A - more like a test at higher voltage can be used for the same power also for lower votlage.

A problem with SOA data is that for many FETs there is no DC curve, no data. Even worse some data-sheets give a SOA curve, that is plain wrong and promisses more than the FETs can deliver. This happens if the curve is derived from transient thermal resistance.  A SOA curve with just a simple power limit for all voltages is suspect, if not extra mentioned in the text.
The power specs for many MOSFETs is also exaggerated, ignoring any thermal resistance from the case to the heat sink - more like a theretical best case estimate, but not realistic (e.g. ignoring that the thermal resistance can go up with temperature).

[Vovk_Z]

The resistors R5,R8,R13,R16 in front of the gate are relatively large. This makes the control relatively slow an may reduce the stability. Normally some 100-220 Ohms should be sufficient - it's just to isolate the OPs output from the capacitive load.

The Resistors R10,R15 ... are odd:

+1. I think TS should lower gate resistors to 22-100 R, get rid of R10, R15, R18, R21. And increase a bit C3-C6 to 2.2-4.7 nF.