Yesterday, I pulled out the AI board for the PRM/BS SUSs. (After the investigation it was restored)
Contrary to our expectation, the board D000186 was not Rev. A but Rev. B.
According to Jay's note in D000186 (for Rev.D), the differences of the Revs are as follows
Rev.A: Initial Release (Analog Biquad version, 4dB 4th order elliptic with notches)
Rev.B: Filter implemented by Freq Devices chip
Rev.C: Differential input version with better RF filtering
Rev.D: 3rd order 0.5dB ripple Cheby with notches at 16K&32K, DB25 input version
I went to the WB EE shop and found bunch of AI filter modules. At least I found one Rev.A and six Rev.D.
I found at least one Rev. C.
I took Rev.A and Rev. D to see the difference of the transfer functions.
Rev.A has more ripple but steeper roll-off. Rev. D is flater at the pass band with slower roll-off.
Rev.D has more phase lag, but it will be fine once the entire frequency response is shifted to x4 high frequency.
The notch frequency of the Rev. D looked right.
I made the empirical pole/zero modeling of the transfer functions.
The LISO models are attached as the ZIP file.
I faced an unexplainable phase behavior at around one the notches for Rev.A.
This may suggest there could have been internal saturation is the stage during the sweep.
More importantly, Rev. D has differential inputs although the connector formfactor is different from the current 40pin IDC.
In fact we should not use Rev.A or Rev.B as they have single end inputs.
Currently the inputs of the AI's for the SUSs are single ended while the DACs are differential.
This means that
1) We waste a half of the DAC range.
2) The negative outputs of the DACs are short-circuited. OMG
3) The ground level fluctuation between the DAC and the SUS rack fluctuates the actual actuation voltage.
Now I am looking at the noise performance of the filters as well as the DAC output noise and range.
I hope we can use Rev.D by replacing the connector heads as this will remove many of the problems we currently have.