We opened one up and soldered a co-ax cable to one of the +6v power contacts at the output of the filter board.

Connecting this to a 'scope we observed a level of +/- 5mv noise with no FEE connected.

Connecting one FEE made little difference.

Re-connected the FEE to the power supply in the cabinet and wired a co-ax to a spare FEE power plug. 

Ran the system with two FEEs on the bench top and two in the cabinet. The benchtop FEEs were connected to the NIM bin in the cabinet with copper braid.

Using a PB4 in the NIM bin and my test input cards on the FEE mezzanine connector, the following readings were taken using the 'integrate' function in the spectrum browser.

Shaping 8us, negative pulse.

aida02, ASIC1, ch0 Peak width => 16.58 channels ( full 16 bit spectra )

aida03, ASIC1, ch0 Peak width => 18.48 channels ( full 16 bit spectra )

aida04, ASIC1, ch0 Peak width => 18.05 channels ( full 16 bit spectra )

These could be better but the output of the PB4 was not perfect. I think ASIC 1 Ch0 is normally the worst on the mezzanine.

Attachments 1 to 6 are the waveforms and some zoomed in.

The co-ax from the power supply was connected to a 'scope and showed a noise level of +/-50mv. 

Fourier analysis of this noise showed two peaks 109kHz and 218kHz. ( Attachments 7 & 8 )