We are using the digitizer in a wideband amplifier test station, where the time variables are signal frequency and cycle resolution (samples per cycle - this is a constant throughout a test). This way the signal analysis is mathmatically identical regardless of signal frequency. We chose 500 samples per cycle, and because our amplifiers typically do not operate beyond 100kHz, we did not see this effect until experimenting with higher frequencies. In our range, the large decimation factor (n) kept our sample rate very near 35MHz. Originally we kept the sample rate above 40MHz, but found that at the high end of n=2, the sample rate would enter the 70MHz region. We moved the target frequency to 35MHz, so instead of a n=2 fs=72MHz, the program would request n=1, fs=36MHz. Therefore the only problem we have now, is when the requested signal frequency and cycle resolution amount to 72-79 MHz. Because this puts a giant hole in our test range, for all practiacal purposes it caps our sample rate to 70MHz.
I have received an email from an applications engineer at NI, Jesse Ormston, who said he also recreated the problem, but was able to bypass it by not using the PXI star, and instead routing the signal through coax on the front panels.
