Hey,
I'm pretty sure that I discovered another bug in File Meta Source, but maybe
this is known behavior or user error.
Short Summary:
When using File Meta Source with Repeat = True, there is a very noticeable lag
(ca 500 ms) when the file "rewinds" to the beginning, at least when reading at
high sample rates.
Setup:
* GNURadio 3.10.9.2
* Python 3.11.8
* Installed through radioconda
* X86-64 running Ubuntu 20.04 LTS
* Ettus Research USRP B200
* UHD 4.6.0.0
Steps to Reproduce:
1. Create a IQ file using File Meta Sink with enough samples for approx. 1s
at the chosen sample rate (40-56 MS/s, see note below)
2. Replay the file using a File Meta Source into a UHD USRP Sink.
3. Observe the TX light on the USRP.
Observed Behavior:
The TX light blinks. Underflow messages on the console are created
Expected behavior:
The TX light stays on without interruption. No underflow messages created.
Workaround:
Introduce a block in-between the File Meta Source and the USRP sink and set the
minoutbuf value in the advanced settings to a large value, e.g. enough samples
for 1 s. Tested with Throttle block (and maximum rate set to something higher
than the sample rate), as well as a IShortToComplex block, but presumably any
block works. Essentially, this block then becomes a buffer for whenever the
File Meta Source is too slow. Setting minoutbuf on the File Meta Source does
NOT work.
I have this on video and I can also supply the GRC files for reproduction, but
I apparently can't attach this to this mail, I'm unsure about the rules
regarding attachments in mailman. Send me a mail and I'll gladly send you the
zip file.
Notes:
* Due to issue #7191, this must be done with a file with inline headers,
otherwise the repetition will not work
* This problem may not occur or may be less obvious when using lower sample
rates, I didn't check.
* This does not occur when using the File Source
* Using a ramdisk does not fix the issue
Please let me know if you have any other ideas, else I'll create another issue
on Github!
Thank you!
Adrian Winter
