I've always noticed that when I start up osmocom_fft at 250kSamp/s with those RTL dongles plugged in, the spectrum looks like a bunch of samples are FFT'd, and then the application waits for another batch of samples and the FFT GUI doesn't update, then some more samples come in and get FFT'd, and this process goes on. I up the rate to something like 1M or higher and go about my business.

I've been working with an application that can use lower sample rates, and in my investigations, sample rates below 901kSamp/s exhibit this "burst of samples, then nothing" repeating pattern. It seems to me that samples are being dropped, i.e., the sampling is not continuous. With osmocom_fft -F (fosphor) running, if I set the sampling rate to 901k vs. 900k, I see a marked difference in the spectrum, and how often it is updated in the drawing. My suspicions about the samples not arriving / being dropped comes from doing a simple FM radio demod, which sounds great at 901kSamp/s, but plays continuous, stops and underflows, and repeats that again and again when things are run at 900kSamp/s (GNURadio flowgraph attached). I've also seen other really weird stuff, like if you ask for 215k sampling rate.

Has anyone else seen these issues? Can anyone replicate? Anyone have solutions or information about what sample rates are properly supported by the hardware? Any documentation to the RTL2832u that might explain this behavior?

Here is my setup:

Ubuntu 14.04, gnuradio built from tagged release, osmosdr / fosphor built from git master some day in the past

--