I tried getting the relative gain between the AC and DC paths of the New Focus 1811, essentially repeating the measurement in elog #1152 but (a) taking measurements for both the DC and AC paths and (b) taking measurements in the region 10 kHz to 100 kHz, where the DC and AC bands overlap. According to the manual, the DC path is meant to be used up to 50 kHz, and the AC path is meant to be used down to 25 kHz.

I again had roughly 1 mW of light on the PD. I took spectra with the Agilent 4395A, which has 50 Ω input impedance. For the AC path, I took the spectra in W/Hz and multiplied by 50 Ω, giving a spectrum in V/rtHz referenced to the input of the spectrum analyzer. For the DC path, I took the spectra in W/Hz, multiplied by 50 Ω, took the square root, and then multiplied by 40/(2*0.82) = 24. The factor of 40 is the value given in the manual for the relative gain between the AC and DC paths, the factor of 2 accounts for the 50 Ω output impedance of the AC path, and the factor of 0.82 accounts for the apparent 11 Ω output impedance of the DC path. (The DC voltage out of the PD was 0.66 V when fed directly into the 1 MΩ input of the scope, but dropped to 0.54 V after I inserted a 50 Ω feedthrough; the quotient of these two numbers is 0.82 and this implies that the DC path has an 11 Ω output impedance). If the relative gain value (40) as given in the manual is correct, the DC and AC spectra should overlap so long as they are dominated by intensity noise rather than detector noise.

The first plot shows the AC and DC spectra, both light and dark. The intensity noise is not as dominant over the detector noise as I had hoped. Still, in the second plot I've quadrature subtracted the dark spectra from the light spectra to produce what are in principle the spectra of the photocurrent alone. Even with the caveat that certain bins in these spectra are dominated by detector noise, it appears that the spectra don't even have the same shape. My guess is that we may already be seeing the action of the high-pass rolloff on the AC path.

It might be worth having another go at this measurement with a higher photocurrent, but I'm guessing that this isn't a great way to get the relative gain calibration.