Gain logging

[petur]

I'm implementing gain logging and would like your opinion.

Current implementation:
- write to a logfile on every gain change
- logs time offset (in ms), left gain and right gain
- tab separated
- logfile has same name as recording but ends with .log
- file is written to disk when recording is flushed to disk or gain log buffer full
- keeps 50 gain changes in memory to reduce disk spinup

let me know what's missing/overlooked. The code is not ready so nothing to try yet.

[lenny]

- logs time offset (in ms), left gain and right gain

Might samples be better (more accurate)?

Edit: Oh and this
http://forums.rockbox.org/index.php?topic=2513.msg44075#msg44075

[Davide-NYC]

Logging should not increase disk spin-up frequency.
Whenever the HD needs to spin up to write audio the current "in memory" logs should be written to disk as well.

Was that at all clear?  

[petur]

Logging should not increase disk spin-up frequency.
Whenever the HD needs to spin up to write audio the current "in memory" logs should be written to disk as well.

That was my 5th point, probably also not very clear

[petur]

Might samples be better (more accurate)?

That is next to impossible because the data is transferred in chunks using DMA. It will be already hard to have some accurate time in ms...

[pabouk]

Thank you Petur. I think that this is a very important function for recording.

Might samples be better (more accurate)?

That is next to impossible because the data is transferred in chunks using DMA. It will be already hard to have some accurate time in ms...

Cannot the DMA controller of the codec chip return the current status (and the sample number) on request?

What I would like to have:
What about logging also the type of the gain change manual / AGC and the AGC type / preset?

[petur]

Cannot the DMA controller of the codec chip return the current status (and the sample number) on request?

What I would like to have:
What about logging also the type of the gain change manual / AGC and the AGC type / preset?

1) no: at best it could be checked every transferred chunk but this would make it quite complex because you can't do much in the dma handler. so it would set a flag, wait for the dma handler to report the number of samples and then continue writing the log file....

2) no: the type of change would have to be passed through all layers, and I wouldn't know why you'd want to know it anyway.

I'm open to discussion, however

[Davide-NYC]

Petur, buzz me whenever you have a bit of code to test. I'd be happy to help out in that regard.

[petur]

sure...

have been non-active again so there will be some delay :/

too much to do, life ought to have 48 hours in a day (at least)