Andrew,
Thank you for a useful and informative explanation. Yes unexpected experimental results can be startling, but also informative. I wish you luck in your quest.
As I said, I am an electrical engineer who worked for a General Motors advanced research facility for 37 years. In the last 15 years of my tenure, I was sent all over the corporation to solve electronic problems, many of which involved frequencies and interference like those under question.
Sometimes, problem solutions are very obscure. A case in point was a piece of military electronics on a light armored vehicle (LAV) that malfunctioned sporadically and randomly when the 25mm cannon was fired. I studied that problem for many months, attending firing exercises all over North America.
I finally found the problem was direct sunshine illuminating a laser sensor created a photodiode bias voltage, which was blocked by a coupling capacitor, so the bias voltage was blocked into further electronics. It turned out that shock from the cannon firing compressed the glass-epoxy pc board enough to change its capacitance and create a voltage: I = d/dt(CV). This displacement current created a voltage in the associated circuitry. Changing and thickening the pc board material solved the problem.
Yes, I would expect the Peltier module control switching frequency is significantly lower than the CMOS sensor scan frequency. With pulse width modulation (square wave) used to control the switching regulator for the Peltier cooler drive wires, high frequency harmonics are present, albeit somewhat filtered. As you know, inductors have capacitance, and they only function as inductors below their self resonant frequency. Short high frequency current paths are also very important.
I would agree that the Peltier PWM switching frequency is probably many orders-of-magnitude below the video clock frequencies. You also have the opposite orders-of-magnitude difference between the switching regulator waveforms and the video output signals (before on-chip A/D conversion).
Since the banding is somewhat suppressed, i.e, only affects lower level signals, even more attenuation of interfering signals to the video signals may be needed.
When I see a potentially problematic electromechanical design, like the Peltier drive wires in a big open loop directly above low level video signals, I get concerned.
There are simple tests that can be done to prove or disprove this conjecture. The simplest of which is to raise the Peltier drive wires away from the video flat cable and press them together to minimize coupling area. Compare the banding in before and after photos.
Subsequently, add a shield over the video flat cable area. Such a shield can be a small piece of single-sided 0.31" PC board material. The ground connection should be as short as possible. Retest, as before.
If there is an improvement, then the above conjecture is valid.
You have a tiger by the tail, judging from the blogs complaining about the problem, i.e., disappointed customers who bought an expensive camera.
I hope you find a solution soon.
Regards,
