Yuheng, Michael

We continued looking at the CC PLL bandwidth using the summing amplifier transfer function measurement.

I saw an old post by Marco Vardaro which says that the nominal bandwidth of the PLL is 40-50 kHz, though maybe the settings were much different back then. Analog devices datasheet says the loop bandwidth is about 20 kHz (fig 1). In this case it was just inspected from the width of the spectral peak, so not particularly accurate.

We took open loop transfer function by checking the coherence in FFT mode (fig 2). In this case, the frequency response is achieved by injecting white noise and taking Ch2/Ch1. We start with a small noise value and increase it until the coherence is approximately 1 over the frequency band of interest. For the CC OLTF the maximum noise excitation is about 30 mV. Also for whatever reason the spectrum analyzer initially gave me frequency response magnitude in dB/rtHz which was a bit odd, I don't know what setting was doing that. It went back to just dB after I switched to swept sine and then back. The unity gain frequency is about 11 kHz and the unity gain phase is +50 degrees (fig 3), which is a bit strange. Judging by this as well as the earlier PLL data it seems there is room to increase the bandwidth. We also checked in swept sine mode for which the optimal excitation amplitude was 15 mV. It gave almost the same result 11.2 kHz UGF.

Afterwards I noticed that this was actually measuring both the fast and slow loop. We repeated the measurement again with only the fast loop. In this case the optimal swept sine amplitude dropped to 10 mV and the UGF very slightly decreased to 10.8 kHz (fig 4, 5).

We used the ppol PLL as a reference check. Looking at the fast loop we see a unity gain frequency of 9.6 kHz with phase 50 degrees. In this case the optimal noise excitation was 150 mV and optimal swept sine was 100 mV (fig 6, 7). Even though the unity gain frequency is the same, it seems the ppol PLL is more robust against unlock. In this case the possible issue might not be control bandwidth but rather the dynamic range of actuation. 

While searching for elogs about PLL bandwidth I came across a previous post by myself and Yuhang. At that time it seems like the correction signal for the CC PLL was too small. This was when we were having small glitch issues and not the major instability. In that case it seems that when the CC PLL was left floating 2 MHz off the setpoint, the fast correction signal was only 200 mV, versus the CC laser PZT tuning coefficient 1 V/MHz. At the time the ppol PLL fast correction signal gave the correct value. When I tried it today, CC detuning of about 7 MHz and ppol detuning of 3 MHz maxed out the fast correction signal.  This is a bit strange so we should check again in detail tomorrow.