[Frank, Tara]

We checked the input referred noise of the UPDH and found it to be 31nV/rtHz, so we bypassing U6, and the input noise is reduced to 3.3nV/rtHz. Most of the excess noise was introduced at the adjustable gain, U6.

        Before the modification, the noise level at TP13 was 170 nV/rtHz. The gain at U1 was 26.4dB, and -6dB at U6 (0 on the knob). Thus the total amplification was 14.8 (a factor of 5.5). The input referred noise was then 170/5.5 ~ 31 nV/rtHz.   After we scraped out the trace from U6 Vout to TP13, and short TP3 and TP13, the noise at TP13 level was 37nV/rtHz. This level is reasonable (cf OP27 datasheet). The amplification was changed to 20.8 dB (a factor of 11). The input referred noise now is then 37/11 ~ 3.3 nV/rtHz. This input referred noise can be convert to frequency noise by the frequency discriminator gain (latest value was 31kHz/V). It is only 0.1 mHz/rtHz which is  low and will not be the limiting source. This input referred noise will be added to the noise budget. The cavities can be locked properly after the modification. 

fig1: part of the schematic of UPDH after modification on R14 and R12, and U6 is bypassed. The whole updated schematic can be found on svn

Right now the gain is not adjustable, we will solve this problem later.