Introduction Initial LIGO accelerometers were sampled at 2048 Hz.  This rate gives accelerometer coverage over about 1/8 of the DARM band. This coverage is better than it may seem because displacement, and the effect on DARM, tend to drop rapidly with frequency. Nevertheless, complete coverage over the detection band may help build the confidence of a detection paper reviewer, allowing us to say that there was no increase over ambient motion during the event.  For this reason I have been pushing to increase the bandwidth. Peter F. has tentatively approved an increase in bandwidth to 4096, and I mentioned to him that I might want to redistribute that, having some accelerometers sampled at lower rates so that some could be sampled at higher rates. Hence this investigation.

How high in frequency do the accelerometers sense the vibrational background?The accelerometer specs claim +/- 3dB out to 1300 Hz with a resonance at 2.4 kHz. But, for us, the data is useful beyond the flat region if the accelerometer senses the ambient background (and is thus able to detect any changes). Figure 1 shows that accelerometers on a test mass chamber see background up to at least 7200 Hz.

How well are high frequency vibrations sensed by the 16k microphones?   To test whether 16k microphones offered comparable coverage to accelerometers for high frequency vibration of the vacuum enclosure, I used a piezo shaker to inject 3000 Hz vibrations onto the GV-1 gate valve seat (which is visible to the test mass, has been identified as a potential glint site, and could conceivably be excited by a failing annulus ion pump or actuator), and compare the responses of nearby accelerometers and microphones. Figure 2 shows that accelerometers in the vicinity tended to have at least an order of magnitude greater SNR than microphones. This suggests that microphones are not a very good substitute for accelerometers for high frequency vacuum enclosure vibrations. The Y-manifold and other more distant accelerometers also show the signal with fairly high SNR, suggesting that the signals travel pretty far through the vacuum enclosure and so a smaller number of accelerometers could provide fairly good high-frequency coverage.

So I propose we increase the rate of a subset of accelerometers to 16384 Hz, and maintain the total data rate by getting rid of all quad sum channels and reducing all other accelerometers to 2048 Hz.

Robert

Increase the following channel rates to 16k, reduce all others to 2048 and eliminate accelerometer quad sums:

H1:PEM-CS_ACC_PSL_PERISCOPE_X

H1:PEM-CS_ACC_BEAMTUBE_MCTUBE_Y

H1:PEM-CS_ACC_BSC1_ITMY_Y

H1:PEM-CS_ACC_BSC3_ITMX_X

H1:PEM-EY_ACC_BSC10_ETMY_X

H1:PEM-EX_ACC_BSC9_ETMX_Y

H1:PEM-CS_ACC_HAM6_OMC_Z

L1:PEM-CS_ACC_PSL_PERISCOPE_X

L1:PEM-CS_ACC_HAM6_OMC_Z

L1:PEM-EY_ACC_BSC5_ETMY_X

L1:PEM-EX_ACC_BSC4_ETMX_Y

L1:PEM-CS_ACC_BSC3_ITMX_X

L1:PEM-CS_ACC_BSC1_ITMY_Y

L1:PEM-CS_ACC_HAM2_PRM_Z