Strange, subtle problem in IQ demod

Just about everything about Red Pitaya
Post Reply
tjrob
Posts: 6
Joined: Wed Jun 09, 2021 3:17 am

Strange, subtle problem in IQ demod

Post by tjrob » Tue May 24, 2022 5:42 am

We use two Red Pitayas with pyrpl to lock two semiconductor lasers to adjacent fringes of a small Michelson interferometer (FSR ~ 258MHz). We have a photodetector looking at the heterodyne between the lasers, with a 1.8GHz frequency counter that samples at 1kHz. This works reasonably well, but when we look at the FFT of the heterodyne frequency, we see a "frequency comb" that limits performance (first attachment) -- that corresponds to an impulse repeating every 17.2 seconds. Remarkably, we found a smoking gun inside the Red Pitaya:

I did a loopback test of the IQ demodulator, with a short cable connecting its modulation output (out2) to its input (in2), and also put a PID on its quadrature output (P=0, I=100Hz). In this test there is just a single Red Pitaya, and the lasers are not involved; the only connection is a short cable from out2 to in2. The scope looks at the IQ output and at the PID output (which also goes to out1 that is disconnected). I setup the IQ: modulation frequency=37.5E6, amplitude=0.01, phase=305.5, quadrature_factor=1. By carefully tuning the phase I can get the PID output to remain at zero within 1LSB for a few minutes, except for isolated pulses. With the scope running at maximum decimation of 65536, the scope auto-triggers every 9 seconds. Four successive traces at 0.002V/div (attached) show a clear 3mV impulse every ~17 seconds (the scope is averaging the decimations; the dotted blue line is zero; the LSB is clearly visible). This pattern repeated for many minutes, with the impulse moving left by about one second every other trigger, with intervening triggers at zero; the impulse varied in shape and amplitude. Of course the sharpness of the FFT comb implies a very stable interval between impulses. Note this is excellent stability for a system running open-loop with high gain.

I discovered that varying the Modulation Frequency made significant changes in both the impulse interval and its size. We of course need two Red Pitayas with different modulation frequencies (two lasers locked to one optical cavity), and the best pair we found so far has improved the performance (Allan Deviation) by more than an order of magnitude -- this effect is not small, even though it is only a few millivolt impulse at intervals measured in seconds. Note that it is still present, and we can see that eliminating it would improve performance significantly.

As it is so sensitive to the modulation frequency, I believe the problem is somewhere inside the IQ demod, but I suppose it is possible that it is in the ADC, DAC, or PID. Note it is happening at intervals of 645 million cycles of the modulation (!).

Has anyone else seen this? Has anyone even looked?
Any suggestions how to fix it?

Note: I tried to reproduce this using the pyrpl GUI, but that is far too cumbersome. We have our own GUI that puts all the control parameters in one window and the scope in another window, so I can vary the IQ phase and immediately see the result in the IQ demod output (sweeping ~ 10ms/div) -- it is tricky to set the phase precisely enough to get stability. If you want to investigate this, reply to this post and I'll make a small program to let you play with it.

Note: "Sorry, the board attachment quota has been reached." so my attachments are lost.

tjrob
Posts: 6
Joined: Wed Jun 09, 2021 3:17 am

Re: Strange, subtle problem in IQ demod

Post by tjrob » Thu May 26, 2022 10:23 pm

I made a test version of our GUI that connects the IQ demod input to out2, which is its modulation output. So this is the same loopback test, but without the DAC cable, or ADC -- it's all digital. I still have PID0 reading the IQ demod quadrature output, P=0, I=100Hz, and use the internal scope to look at both the IQ demod out and PID0 out.

It is more difficult to hold PID0 output at zero, but it can be done by carefully setting the IQ phase. The impulse is now periodic at 17 seconds, but it is about 8 seconds wide and 10 millivolts tall.

While playing with this we found that for modulation frequencies of 31.25E6 and 42.5E6 the PID0 output stays right at zero, without any impulses. 35.5E6 is almost as good. The first is 125MHz/4, and the second is close to 125MHz/3, but neither 125MHz/3 nor 125MHz/5 work as well.

This digital loopback shows that the problem is definitely in the IQ demodulator. It remains both strange and subtle.

Post Reply
jadalnie klasyczne ekskluzywne meble wypoczynkowe do salonu ekskluzywne meble tapicerowane ekskluzywne meble do sypialni ekskluzywne meble włoskie

Who is online

Users browsing this forum: No registered users and 2 guests