Odd circuit for LED light

I've just finished modifying an LED light that had a rather odd circuit. I haven't got any questions here and I'm posting because what I found might be of interest.

The LED light had significant flicker at about 100 Hz when I got it. I'm sensitive to that so I had to change it.

The light runs from a regulated 24 V supply, and when I measured the LED voltage, it was being switched at around 50 kHz and the 100 Hz, variable duty cycle PWM was enabling that. I worked out the circuit, which is "before.pdf". There is sufficient gain in the op-amp and the two transistor that the circuit oscillates at about 50 kHz, so the op-amp is controlling the average current, not an instantaneous current. The current produced a voltage drop across the 1 Ohm resistor, and that was compared with a reference voltage divided down from 5 V

I don't know the values of the unlabelled capacitors, but and I don't understand what they are doing, but without the between the op-amp input the oscillation frequency is near 1 MHz.

The PWM was achieved by turning on and off the supply to the op-amp, so the op-amp was powered from the I/O pin of a microcontroller.

I moved the op-amp supply to the 5 V, and connected the current reference to a smoothed version of the PWM signal. That worked quite well but the lights never turned off. I added the 1k and 1M resistors to add small offset to make the op-amp turn off completely at 0% PWM. The circuit is in "after.pdf".

The were a couple of other things that I don't understand about the circuit. Firstly, it was all through-board components, and looked like it could have been made in the 1990s, although the LEDs are on a separate aluminium board, and they are surface mount. Secondly, there are two independently controlled banks of lights, and the arrangement of supplying the op-amps with the PWM signals meant that two separately packaged op-amps had to be used. They had used dual op-amps, LM358s, so their odd arrangement cost an extra op amp.

This looks like it has been tweaked to stop it oscillating as opposed to necessarily being designed that way. The engineer probably built the circuit, found that it oscillated and started putting capacitors in to quieten it down not knowing what he was doing. The capacitor between the 2 op amp inputs is doing this - limiting the frequency response of the circuit and trying to stop it oscillating. However, with this capacitor comes a phase shift, and from what you say, this is sufficient to keep it oscillating.

You would be better putting a resistor in series with the inverting input of the op amp and a capacitor from op amp output back to the inverting input. This will provide high gain at DC, but lower the gain at AC, thus giving good DC accuracy (the voltage at the top of the 1 Ohm resistor equalling the voltage at the non inverting terminal of the op amp), but rolling off the gain at AC to prevent it from oscillating. So move your 1k resistor to be in series with the inverting terminal and move that capacitor between op amp output and inverting input. Remove the 1M. You might have to fine tune the circuit but hopefully that will work.

Well the 1 M resistor is needed to make the circuit turn off completely when the PWM is at 0%. I added that and the 1k resistor to slightly offset the input voltage of the op-amp and to force the output off in that situation.

I had realised that putting the capacitor between the output and the -ve input would help stop oscillation.

I really don't know if the circuit was supposed to run as a linear circuit, and I'm not going to worry about it or go back to it unless the switching causes electrical noise. I would expect to have to limit the gain of the transistor circuits as well as the op-amp if I wanted to make it run linearly. I would put a resistor in series with the emitter of the NPN transistor if I were doing that.