We knew we might have to match the BCM bandwidths to that of the detectors. Des pointed this out long ago and the PREx guys got bit by it last year. Unfortunately, since the MD preamp bandwidth is 25 KHz and the Lumi style is 15 KHz, it's not a well posed problem. We'll have to study the machine noise spectrum above 10 KHz and come up with a solution if merited.
In the short term, there is a much bigger fish to fry (big like a whale shark). Hypothesis formed with Kent this morning: The BCM1 and BCM2 bandwidths are not even close to being matched (probably because the BCM2 card has a much lower cutoff frequency left over from polarized target running), and it matters because the current jitter spectrum sometimes has important (.geq. 100 ppm) contributions from above the BCM2 cutoff frequency. That would mean Aq1-Aq2 is not a robust estimator of the BCM1/2 random noise, and BCM2 is not yet reliable for normalization.
Today I plan to quietly review some 10 year old notes on the rms-to-DC boards. Then I have to pull the BCM2 rms-to-DC card, and if the hypothesis is confirmed, spend 2 hours modifying the filter with a soldering iron. There's a 10% chance the board won't come back. But if I decide to pick up a soldering iron, it means the BCM2 chain is useless to us now anyway.
Notes from Wednesday nite attached. A more careful estimate than I did yesterday suggests that any BCM noise at the O(100) ppm level is safely above the normal ADC noise floor of 1 microV/sqrt(Hz) (which corresponds to about 15 ppm on a 1V signal for 1/960 second quads). Also, if the BCM noise is white and meaningful when the beam off, then one can estimate the noise level seen by the parity daq with the digital scope "measure rms" function and a little algebra.
check of BCM signals for noise levels
24November2010BCMRackCheck.pdf