I am having a very similar problem attempting to use the SCXI-1125 module with the SCXI-1327 terminal block to measure the ac rms voltages from an alternator. I am ultra new to NI (as in I had never touched it prior to this past Wednesday), so I am still feeling my way around quite a bit and trying not to blow things up in the process. Similar to what the OP stated in his problem, my voltage amplitude also rises with frequency with measurements taken with MAX and LabView falling increasingly short of the actual values at the top of the speed range. I get decent measurements at the lower frequencies, but as I near the top of my speed range (roughly 2.1 kHz), my measurement accuracy drops off drastically. At full speed (2.1 kHz), I measure 36 Vrms on the Yokogawa meter (which I trust) but only 30 Vrms in LabView. I did a little tinkering and found that I could recreate this phenomenon with a function generator by generating a sine wave, keeping the amplitude constant, and increasing the frequency. Definitely looked like an issue of unwanted low-pass filtering. After rooting around on the NI site for a while, I ran across this document...
http://www.ni.com/pdf/products/us/4scxisc296_ETC_196.pdf
In the table on page 299, it lists the dynamic input characteristics of SCXI-1120, SCXI-1120D, and SCXI-1125. According to this table, the bandwidth of the SCXI-1125 is only 2.6 kHz (even with the filtering set to 10 kHz) when paired with the SCXI-1327. My alternator waveforms have a good deal of harmonic distortion (almost squarewaveish), so this looked like the source of my measurement problems, as the harmonics of 2.1 kHz in a square wave would be 6.3 kHz, 10.5 kHz, etc., etc., and would most certainly get killed by a 2.6 kHz bandwidth. (I also saw much worse accuracy when supplying a square wave from the function generator, so I thought I was on the right track.) I switched over to the SCXI-1120D (whose bandwidth is listed in the table as 22.5 kHz) expecting this to solve my problems. No dice. In fact, this combination behaves even more peculiarly. If I supply a pure sine wave, I get decent results up to about 2 kHz. Around this point, I start getting HIGHER measurements than I should. Not what I would expect from a low-pass filter. Around 4 kHz the accuracy is getting pretty bad. By the time I get up to 14 kHz, I'm measuring almost double the actual voltage that I'm putting in. (I see this on the MAX test panel as well as in the LabView interface, so it's not an issue with how I'm processing the information in LabView.) Then past 30 or 40 kHz the low-pass filtering kicks in and I get next to nothing as expected. It seems as if the frequency response of the filter in the SCXI-1120D has a hump (gain greater than unity) before it starts to drop off and attenuate the signal. That is not all that peculiar for a multi-order low-pass filter. The strange part is how low in frequency this hump starts to show up. With an advertised input signal bandwidth of 22.5 kHz, I would expect to be able to measure a 2 kHz square wave without much problem. Maybe a little inaccuracy due to attenuation of the higher frequencies and the resonance of the filter near the cutoff frequency, but not the drastic overmeasurement that I'm seeing. Any thoughts or suggestions would be appreciated. I'm pretty stumped.
Oh, by the way. I set the gain and filter jumpers in the SCXI-1120D according to this document....
http://www.ni.com/pdf/manuals/320425d.pdf
...on pages 2-10 through 2-13. I notice that the filter jumper settings for the second stage filter are different for the SCXI-1120 and SCXI-1120D. The SCXI-1120 requires the even jumpers from W25-W40 for high bandwidth while the SCXI-1120D requires the odd jumpers. That arose a bit of suspicion in me. I also wonder if maybe this is an issue of pairing the SCXI-1120D with the SCXI-1327 like it was for the SCXI-1125. I didn't see anything about that in any of the other datasheets, but the document that I linked was the only one that I saw that even mentioned it for the SCXI-1125.
