bonjour,
attached are 6 measurements of the transfer function of your cic 
filter. it was done by injecting 9 successive synthesizer frequencies 
into your card, then FT and magnitude calculation (the broad feet of 
the lines are an artefact of this magnitude calculation). the 
bell-shaped curves are the expected response, fitted to the center 
peak.
all data are for a decimation factor of 8. all vertical scales are 
(essentially) the same, because the data on the card have been 
appropriately bit-shifted.
these experiments are at 85 MHz central frequency, 10 MHz above the 
ADC clock rate, because i wanted to simulate the expected results for 
70 MHz input frequency, and a 60 MHz ADC rate, as we will have in the 
future.
(for 70 MHz/75 MHz the results are less nice, i have not yet prepared 
the corresponding .pdf files).
of course my printed output format is that for a regular nmr 
experiment, so many of the printed-out variables have no meaning in 
this case. the only important variable is "bandwidth". the first 
digit is the number of stages in the filter, the second digit is the 
m-factor, and the rest is the bitshift. note however that a second 
digit of "2" corresponds to typing that value into your pb_cic_setup, 
but clearly the m-value of the resulting filter is 1, as shown by the 
bell-shaped curves (but luckily typing "1" sets a value of 2). note 
also that here is given a case where the 15-bit shift works, whereas 
it does not for certain other filter settings.
note the important spurious signals for "bandwidth"= 1210 and 1111. 
these have in common that it is a one-stage filter. contrary to what 
you suggest in your filter manual, this is not a recommendable 
choice. the wide band filter 2213 is already an acceptable choice 
(two stages, "real" m=1).
i also found that it is important that the input signal is not (much) 
larger than 1 V pp. apparently, the input stage + ADC of the card 
"saturates" rather than "clips" for larger values; just looking at 
the shape of the signal it is not immediately obvious that this 
saturation is happening. nothing wrong with this behaviour, of 
course, but the user should beware.
regards,