cu8 loader produces incomplete readouts

[n6ham]

A cu8 loader produces incomplete results depending on a presence of -f 915M argument, despite the presence of 915M in the file name.

In one case, lack of -f 915M results in missing one out of two signals

 rtl_433 -r g001_915M_1000k.cu8 -F json

{"time" : "@0.139808s", "model" : "ThermoPro-TempSpikeXR", "id" : 138, "color" : "black", "temperature_int_C" : 31.900, "temperature_amb_C" : 32.300, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}

rtl_433 -f 915M -r g001_915M_1000k.cu8 -F json

{"time" : "@0.139808s", "model" : "ThermoPro-TempSpikeXR", "id" : 138, "color" : "black", "temperature_int_C" : 31.900, "temperature_amb_C" : 32.300, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}
{"time" : "@0.161949s", "model" : "ThermoPro-TempSpikeXR", "id" : 138, "color" : "black", "temperature_int_C" : 31.900, "temperature_amb_C" : 32.300, "is_booster" : 1, "booster_battery" : 3, "mic" : "CRC"}

In other case, presence of -f 915M results in missing one out of three signals

rtl_433 -r g004_915M_1000k.cu8 -F json

{"time" : "@0.048643s", "model" : "ThermoPro-TempSpikeXR", "id" : 156, "color" : "white", "is_docked" : 1, "temperature_int_C" : 26.500, "temperature_amb_C" : 32.400, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}
{"time" : "@0.105515s", "model" : "ThermoPro-TempSpikeXR", "id" : 156, "color" : "white", "is_docked" : 1, "temperature_int_C" : 26.500, "temperature_amb_C" : 32.400, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}
{"time" : "@0.162386s", "model" : "ThermoPro-TempSpikeXR", "id" : 156, "color" : "white", "is_docked" : 1, "temperature_int_C" : 26.500, "temperature_amb_C" : 32.400, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}

rtl_433 -f 915M -r g004_915M_1000k.cu8 -F json

{"time" : "@0.048643s", "model" : "ThermoPro-TempSpikeXR", "id" : 156, "color" : "white", "is_docked" : 1, "temperature_int_C" : 26.500, "temperature_amb_C" : 32.400, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}
{"time" : "@0.162386s", "model" : "ThermoPro-TempSpikeXR", "id" : 156, "color" : "white", "is_docked" : 1, "temperature_int_C" : 26.500, "temperature_amb_C" : 32.400, "is_probe" : 1, "probe_batery" : 3, "mic" : "CRC"}

This causes an issue with samples & tests, since run_test.py runs rtl_433 without -f argument and only a signal from the probe (see the first sample) being read. A second signal from the booster is skipped and not being covered by the test.

[zuckschwerdt]

That's about the discriminator then (the stage after the demod that decides on 1 or 0 bit state).
Default is -Y classic and above 800 MHz (FSK_PULSE_DETECTOR_LIMIT) the -Y minmax is used.

[n6ham]

Could you elaborate please, what does this mean practically for the sampling/testing purpose?
Should I record samples somehow else (with extra arguments) to ensure readout completeness in tests, or just accept the incomplete readouts & coverage and be with it?

[zuckschwerdt]

The samples are most likely fine. We don't have any mechanism to switch the discriminator for test samples. Maybe we need to extend the protocols meta file or better yet run both options and use the outputs as a "quality" indication.
This isn't well explored yet, sorry.

[n6ham]

Okay, then I'll change merbanan/rtl_433_tests#497 status from "draft" back to "ready for review"

[n6ham]

Would it make sense to update the test script to parse the frequency out of the sample file name, then add -Y minmax to the command if the frequency is above the threshold (800M)?

Alternatively, we may need a way to override parameters for specific tests (i.e. if a file named in a certain way exists in a folder with test samples - read the values from. I'm not comfortable with blindly appending the content of the file to the command, but we may parse specific attributes from it and validate them before appending.

If the latter idea sounds reasonable - I'll draft a PR

[zuckschwerdt]

Would it make sense to update the test script to parse the frequency out of the sample file name, then add -Y minmax to the command if the frequency is above the threshold (800M)?

I think we are already doing something like that. rtl_433's main uses file_info_parse_filename() to parse the center_frequency. The sdr_callback() decides the discriminator based on the frequency[] though. That might be an oversight, I'll investigate.

[zuckschwerdt]

Yes, center_frequency should be used in sdr_callback(). This fix will add 13 fails to the tests. I'll PR the change but hold off until we figure out how to add some selector to the tests.

[n6ham]

This fix will add 13 fails to the tests. I'll PR the change but hold off until we figure out how to add some selector to the tests.

Um, do you mean this change needs to be enabled at an individual test level so we won't break existing tests?

[zuckschwerdt]

The change #3554 is a fix to get parity between live and file inputs. The then breaking tests are really things that are not working, at least for live input with default settings. It needs to be fixed at some level, not sure where.

[merbanan]

Issue addressed.