Need some help understanding this heater control circuit

[WatchfulEye]

I had a slightly malfunctioning OCXO, so I took it apart to see what was inside.

I found the attached heater circuit, which surprised me with its complexity, and in particular, the unusual feedback structure. I was expectings something rather simpler, like a wheatstone bridge sensor and low pass/integrator filtered control loop. Instead, there is a seperate current sense feedback with that amplifier having an unusual negative feedback arrangement.

C5-C8 are all small (0603 grey ceramic so probably C0G), but I couldn't practically measure them, and the resulting time constants don't really make sense for the thermal model of an OCXO.

[BillyO]

It must have been a fairly old OCXO.  That circuit is an analog Proportional Integral Derivative (PID) controller.  It's needed because of the thermal inertia of the physical components being heated.

More modern OCXOs are now using microcontrollers to do the same job.  Some things seems a little odd.  Are you 100% sure of your reverse engineering?

[WatchfulEye]

I am 100% certain that the circuit is correct. I did make a simplification only showing 1 of the 2 heater transistors, although only 1 has current feedback.

The reason for my confusion was what was the reason for the various parts, when there is no derivative or meaningful integral inputs to the control loop. If it was just a proportional controller, why use such a complicated scheme?


However, having had a chance to play with the circuit in LT spice (attached), I think I understand it better. Can anyone tell me if I am right?

I think that U2/Q5 are a level shifter and 128 Hz LPF with gain of 0.1. They shift the high side current measurement to low side, and attenuate the voltage by 10. So, in effect replicating the effect of a Q3 collector-side sense resistor of 120 mOhm.

The current sense feedback for the main loop is just to ensure that heater power is proportional to the error. This could have been simplified to just taking the output from the op-amp, as the emitter degenerated output transistor is quite linear in power anyway. There is a small amount of LPF, but cutoff is in the kHz region, so has minimal contribution to loop stability.

Current limiting transistor Q1 doesn't do anything. It has a trip current of 1 A, but the MJD210 output transistors don't have enough gain to deliver 1A with a 1k base resistor.  R8 and C9 are a snubber for Q1, and are also superfluous.

The main contribution to loop stability seems to be the power proportional constant (set by R10/R12 and the current sens resistor).

[BillyO]

Part of U2 is definitely configured as an integrator.

The differentiator is another story and they may be using Q5 for that, but it's a bit of a mindbender.  It looks like the combination of U2 and Q5 are handling both functions.