Iq tool buffered

[vladisslav2011]

Implement buffered file_source/file_sink to improve IQ recording/playback success rate in case of slow media/heavy background activity.
Includes #1013.
This is the last PR, created from splitting #1008 into smaller parts.
May close #1098
May close #1017

[vladisslav2011]

I think, there is better way to implement buffered file_source: reuse GNU Radio vmcircbuf for buffering, add throttle block functionality and, maybe, do type conversion inside of buffered file_source. All of these tasks are relatively lightweight and do not worth wasting dedicated thread and set of output buffers.

[MightyPork]

I just tested this. It works better than the old system, looping also works great, but the slider is somewhat erratic.

• I think it's limited to 1 second of resolution, so fine movement in shorter recordings is not possible
• when you click on the slide to jump to that positionm, instead it starts playing from the beginning, or jumps to the end.

Peek.2022-02-18.21-39.gqrx.mp4

[vladisslav2011]

• instead it starts playing from the beginning, or jumps to the end

That's because slider left click granularity is 1 minute, but your recording is just 7 seconds long. Try clicking the slider with a middle mouse button or test it with a bit longer recording (2-3 minutes at least).

[MightyPork]

Ah, middle click works as left click normally would on a slider. This could use some explanation in the dialog, I would never find out. I rarely have recordings longer than a minute due to the file size.

How about << and >> buttons to do this 1 minute skip, perhaps at the ends of the slider?

[vladisslav2011]

This could use some explanation in the dialog, I would never find out.

That's expected behavior to me. Left click does a page skip (like on scrollbars), middle click moves to exact position (like on scrollbars too).
Adding buttons is possible, but changing page skip would break keyboard navigation.

[stefanino-ch]

Synced qgrx master branch as of 2023-07-20.
Integrated this pull request.

During compilation I get:
[ 26%] Building CXX object src/CMakeFiles/gqrx.dir/applications/gqrx/mainwindow.cpp.o /home/stefan/sw-dev/gqrx-2023-07/gqrx/src/applications/gqrx/mainwindow.cpp: In constructor ‘MainWindow::MainWindow(const QString&, bool, QWidget*)’: /home/stefan/sw-dev/gqrx-2023-07/gqrx/src/applications/gqrx/mainwindow.cpp:124:34: error: ‘DEFAULT_FFT_SIZE’ is not a member of ‘receiver’ 124 | d_iqFftData.resize(receiver::DEFAULT_FFT_SIZE); | ^~~~~~~~~~~~~~~~ /home/stefan/sw-dev/gqrx-2023-07/gqrx/src/applications/gqrx/mainwindow.cpp:132:37: error: ‘DEFAULT_FFT_SIZE’ is not a member of ‘receiver’ 132 | d_audioFftData.resize(receiver::DEFAULT_FFT_SIZE); | ^~~~~~~~~~~~~~~~ gmake[2]: *** [src/CMakeFiles/gqrx.dir/build.make:221: src/CMakeFiles/gqrx.dir/applications/gqrx/mainwindow.cpp.o] Error 1 gmake[1]: *** [CMakeFiles/Makefile2:233: src/CMakeFiles/gqrx.dir/all] Error 2 gmake: *** [Makefile:136: all] Error 2 13:57:16: The process "/usr/bin/cmake" exited with code 2. Error while building/deploying project gqrx (kit: Desktop) When executing step "Build"

DEFAULT_FFT_SIZE is definitely missing in all the receiver files I have.
For me it looks like there's a dependency to a previous pr?

Here's what I got from the merge in receiver.h:

`
//<<<<<<< HEAD
static const unsigned int DEFAULT_FFT_SIZE = 8192;

//=======
enum file_formats {
FILE_FORMAT_LAST=0,
FILE_FORMAT_NONE,
FILE_FORMAT_CF,
FILE_FORMAT_CS8,
FILE_FORMAT_CS16L,
FILE_FORMAT_CS32L,
FILE_FORMAT_CS8U,
FILE_FORMAT_CS16LU,
FILE_FORMAT_CS32LU,
};

struct iq_tool_stats
{
    bool recording;
    bool playing;
    bool failed;
    int buffer_usage;
    size_t file_pos;
 };

//>>>>>>> pr1016
`
Note: // in front of the git tags are just for display purpose

Don't deleting the line:
static const unsigned int DEFAULT_FFT_SIZE = 8192;

Makes the code to compile. But not shure if there is still an issue with it.

[stefanino-ch]

@vladisslav2011 something is wrong with the recordings. If I do a record of one Minute, a file length of 10 seconds is displayed in the IqRecorder window. If I play the file it shows data for 10 seconds, that's it.
Or in other words the recorded file is (60/10) 6 times shorter than it should be.

[vladisslav2011]

There may be bugs from recent changes or previous incomplete rebase, missing recent fixes to IQ formats implementation.

While I still think, that the new plotter is not ready yet, maybe it's time to rebase my fork, rearrange commits and update old PRs.
I have not decided yet if I should include the new plotter and fix it's issues (heavy performance impact from averaging, references to absolute units, markers) or just revert it and subsequent fixes...
I think, I'll update my PRs today or tomorrow.

[stefanino-ch]

There may be bugs from recent changes or previous incomplete rebase, missing recent fixes to IQ formats implementation.

While I still think, that the new plotter is not ready yet, maybe it's time to rebase my fork, rearrange commits and update old PRs. I have not decided yet if I should include the new plotter and fix it's issues (heavy performance impact from averaging, references to absolute units, markers) or just revert it and subsequent fixes... I think, I'll update my PRs today or tomorrow.

Just found that the shortened recording is already an issue in the main branch and not based on this pr.

[vladisslav2011]

@stefanino-ch I've rebased all my pull requests on top of current master.
Feel free to tag me in case you found, that something got broken during rebase.

[stefanino-ch]

@vladisslav2011
I used your pr during a long time in a recording system. The different file formats and the buffer helped a lot in therms of saving disk space and recording stability.
Unfortunately this system died, I've tried now to reintegrate your pr on the current main tree but get:

Auto-merging CMakeLists.txt
Auto-merging src/applications/gqrx/mainwindow.cpp
CONFLICT (content): Merge conflict in src/applications/gqrx/mainwindow.cpp
Auto-merging src/applications/gqrx/mainwindow.h
CONFLICT (content): Merge conflict in src/applications/gqrx/mainwindow.h
Auto-merging src/applications/gqrx/receiver.cpp
Auto-merging src/applications/gqrx/receiver.h
Auto-merging src/qtgui/freqctrl.cpp
CONFLICT (content): Merge conflict in src/qtgui/freqctrl.cpp
Auto-merging src/qtgui/freqctrl.h
Auto-merging src/qtgui/iq_tool.cpp
CONFLICT (content): Merge conflict in src/qtgui/iq_tool.cpp
Auto-merging src/qtgui/iq_tool.h
CONFLICT (content): Merge conflict in src/qtgui/iq_tool.h
Auto-merging src/qtgui/iq_tool.ui
Auto-merging src/qtgui/plotter.cpp
Auto-merging src/qtgui/plotter.h
Automatic merge failed; fix conflicts and then commit the result.

Updating all the code goes way off my programming knowledge (I was into all those files).

I wonder, are you maybe thinking about an update to get it to work on the current tree or is this illusoric?

[vladisslav2011]

@stefanino-ch
I've rebased the PR on top of current master. Tag me if something got broken during rebase.

[stefanino-ch]

@vladisslav2011
I've integrated and tested today your PR update. No issues at all works smooth and saves a lot of recording space again. Thank you very much for your effort.

@argilo
The PR helps to save a lot of disk space while recording. As it is at the moment, and as far I could test, it integrates without any issue into the main tree. I vote for an integration very soon, as it might help also many others.