Jenne, Evan
We looked again at RFAM on the 9 MHz REFL readout.
The attachment shows REFL9I (and REFL LF) for four different PD powers. Additionally, at the fourth power we locked the ISS outer loop (dc coupled with boosts) just to check that the RFAM does not change. The REFL centering loops were on the whole time.
At no power were we able to see anything that looked shot-noise limited, so we were not able to independently check the transimpedance and demod gain for REFL9I. However, the limit placed by the high-frequency noise already seems to indicate that we are missing some gain in the signal chain. Thus far I have been using 2900 V/W for the PD transimpedance plus demod TF, but this would produce spectra that are below the expected shot noise level.
I double-checked the schematics for the REFL LF path and found that the current digital calbration from counts back into watts seems OK. The PD powers given in the attachment are based on this LF calibration.
Finally, note that the dark noise at several kilohertz is larger than the noise when the PD has power on it. No clue about why.
The attachment shows REFL9 and REFL45 signals (in ct/rtHz), with the colors corresponding to the powers given in the previous attachment.
REFL9Q appears to be shot noise limited above 100 Hz, since the noise grows with the square root of the power.
In comparison, REFL9I looks fishy. Its dark noise is higher than REFL9Q, and above 500 Hz shows a noise that is lower than the REFL9Q shot noise.
We are tempted to say that the analog Q channel of the REFL 9 demod board is broken (recall that analog Q corresponds to digital I, and vice versa), or the corresponding whitening channel is broken. (However, since analog Q is teed off and hooked up to the summing node board, it's possible that there's some strange interaction with the SNB.)
Anyway, if we use REFL9Q to calibrate the rf gain of REFL9, this implies a total gain of 2.88e6 ct/W (= 4.4e-4 ct/rtHz / sqrt(4*h*nu*P0), with P0 = 29 mW). We can refer this back to the demod board output by undoing the digital gains (0.18 ct/ct), the ADC conversion (2^16 ct / 40 V), and the whitening gain (12 dB) to arrive at an rf gain of 2600 V/W, which is not too far off from the old value of 2900 V/W.