Skip to content

Fix sample rate mismatch on FSK flow-graph OGG output

Mark Jessop requested to merge darksidelemm/satnogs-flowgraphs:fsk_ogg_output_fix into master

The OGG file recordings from all FSK observations since the latest satnogs-flowgraph update (1.2.1) are ~1/3 the length of the observation (at least for 9k6 FSK).

Examples: https://network.satnogs.org/observations/2361693/ (CubeBel-1, on my station #232) https://network.satnogs.org/observations/2352818/ (MCUBED-2, on station #49)

The reason for this is there appears to have been a change to the decimation prior to the Quadrature demodulation bits of the flowgraph. The input sample rate (57.6 kHz for 9k6baud FSK) is decimated by (decimation//2). The decimation for 9k6 baud is 6, so the resultant sample rate is 57.6/3 = 19.2kHz.

In the path that leads to the OGG output, there is a resampler, which resamples by audio_samp_rate / (baudrate * decimation). This ends up reducing the sample rate even further, to 16 kHz (for the 9600 baud case), before it heads to the OGG output. If the OGG files are interpreted as this sample rate, they are usable for decoding.

For example, I was able to decode 12 packets from the CubeBel-1 example above using the UZ7HO soundmodem by replaying the file at 16 kHz. (The FSK flowgraph did not decode any of these packets, presumably due to this ongoing issue: gr-satnogs#193 (closed) )

One solution is to just set the OGG output sample rate to the actual sample rate. However, to maintain consistency with all the other flow-graphs that use a 48 kHz sample rate, the final resampling step should be ( audio_samp_rate / (2*baudrate) ) to correct for the earlier decimation.

Testing has been conducted on station #232. Observation https://network.satnogs.org/observations/2362433/ shows the correct length audio output for the observation. Inspection of the sample rates in the flow-graph for different baud rates shows the sample rate conversion is correct. Further observations are scheduled on station #232 for further confirmation.

UPDATE: I have added a separate non-decimating low-pass filter for the OGG output. This allows the FM demodulation and the ogg output to be generated from a higher sample rate source, which has the potential to improve the decode performance of external FSK decoders that use the OGG output (not performance tested, but removing an un-necessary decimation and interpolation seems reasonable?).

UPDATE2: Nevermind about the above update.

Edited by Mark Jessop

Merge request reports