Put together a test flow-graph today that is just a UHD-source and an
1024-bin FFT sink at 10frames/sec.
I configured the source for 50Msps, with 8-bit samples. My AMD Phenom
II X6 1090T was able to keep up, but if I inserted
even a single multiplier block between the source, and the FFT, I
started getting nearly-continuous ‘O’.
This suggests to me that actually “doing anything” with data at 50Msps
is going to be a challenge. The way the FFT sink works
is that it decimates the input quite aggressively (in vectors of
FFT-length) before doing anything with the data–it throws most
of it away in order to support the configured frame-rate.
–
Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium