FS#10199 - Dynamic Range Compression

[saratoga]

That patch doesn't actually work.  I'm not sure why he posted it before actually implementing compression. 

[Chronon]

Oops.  I missed this: "This is a cleaned up version of the interface test. Functionality is still forthcoming."


====

It looks like some functionality has now been added.  Also, I updated the topic title to refer to the current patch rather than the obsolete feature request.

(clarify my edit)

[Blue Dude]

I just posted a first attempt at FS#10199.  It does work, but it's not ready for prime time yet.  I've loaded it on my Sansa E280 and it behaves itself so far.

[CCCMikey]

Excellent Would you like any of us to trial it and provide feedback yet?

Just converted to RockBox last week and loving it

[Blue Dude]

I'm working on bugfixes right now.  I'll post something tomorrow that should work better.

I've got it running on my Sansa E280 at the moment, but I've found several issues that I'm addressing.  I'll post here when a better version is out.

Update: an improved version is posted.  This one has smoother output and most bugs worked out.  Give it a try and leave feedback here please.

[CCCMikey]

Just saw this post - since I read it before it was edited, I didn't get the reply notification - will go see if I can figure out how to install it now

Edit: n00b alert: This exceeds my knowledge. I assume the file is at http://www.rockbox.org/tracker/task/10199?getfile=20037 - which yields a text file starting with

Code: [Select]

Index: apps/pcmbuf.c
===================================================================
--- apps/pcmbuf.c	(revision 21841)
+++ apps/pcmbuf.c	(working copy)

I'm guessing this file doesn't go into RockBox utility, but is either copied to the device or is 'built' somehow? Feel free to ignore this post if I'm too far below a reasonable knowledge to proceed.

If you'd like an example of a loud 'n' quiet file that would test this, you could try http://mpegmedia.abc.net.au/local/sydney/200907/r397504_1862526.mp3

[soap]

CCCMikey:
Flyspray contains (for the most part) Patches

[nerdrunner]

As I said in another post, a really good way to implement this would be to utilize the _microphone_ to sample the background sound & adjust the volume & dynamic range compression to match this noise floor.  This is an analogy to a program that you used to be able to get for the Palm Treo that utilized the Treo's builtin camera to sample the amount of ambient light & use this info to set the brightness & contrast on the Treo's display.  Similar ideas are used in some TV's to adjust brightness & contrast.

[Llorean]

Adjust volume based on background noise is just a recipe for hearing damage. Compression might make sense though.

[nerdrunner]

Here's a high level description of how a microphone closed loop dynamic range compression system could work.

1.  Implement a simple system to sample the microphone & compute a crude dB number for the ambient background noise.

2.  Implement a system to compute the current dB number from the playing track; this should already be available from the VU meter.

3.  Provide an exponential decay with a time constant sigma1 to filter the background noise.

4.  Provide an exponential decay with a time constant sigma2 to filter the playing track.

5.  Instead of using the volume control to set the absolute dB, utilize the volume control to set the _difference_ between the filtered playback dB and the filtered background noise dB.

6.  Yes, different players, different earphones & different microphones will all affect these numbers, but you can utilize the "volume" control to provide enough headroom above the background noise to hear the podcast and/or music.

7.  If your ears are getting blasted by the earphones with such an algorithm, then you should probably leave the environment, because the earphones are only a few dB ahead of the background dB.  When I'm riding on the subway and/or riding my bike in traffic, the earphones may well have to kick up pretty loud to still be understandable.

8.  The settings of sigma1 & sigma2 should be user controllable.

[soap]

7.  If your ears are getting blasted by the earphones with such an algorithm, then you should probably leave the environment, because the earphones are only a few dB ahead of the background dB.  When I'm riding on the subway and/or riding my bike in traffic, the earphones may well have to kick up pretty loud to still be understandable.

The appropriate response from a hearing protection point of view is to always be listening to the same, safe, volume level and to control the background noise be it though passive or active noise reduction.  A few (3) dB over 80ish traffic noise is twice as damaging!

[CCCMikey]

Dynamic Range Compression and Ambient noise-based volume adjusting are two rather different things.

Compression is (would be) most useful for those podcasts which are recorded raw and thus have the laughter much louder than the sick joke that caused it While this would be particularly useful for cars, etc, where there's a high amount of background noise so you end up getting deafened when someone laughs!

Ambient noise-based volume would also be useful in cars, etc but for a different purpose. Also, I wonder how many devices actually have micrphones out of the box.

Testing patches is currently out of my league at this stage I think, being a Windows guy.

[Chronon]

CCCMikey, you can build just fine on Windows.  You just need to set up a build environment as described in the docs index.

nerdrunner: I also don't like the idea of changing volume level according to sound pressure at a microphone.  I sometimes use IEMs and the volume difference between sound out there and sound "in here" can be quite dramatic (20 to 35 dB).  How is the algorithm supposed to know how loud to drive the phones when it doesn't know the actual sound pressure in my ear canal? 

[Blue Dude]

A stable patch is posted at FS#10199.  I'm interested in feedback from users of all kinds of targets.  Sansa E200 works great but I'd like to hear from everyone.

[fredwbauer]

I just updated my Sansa Fuze to version 22895 and I saw the compressor option. It's GREAT. Thank you! Thank you! (Now I can hear the other half of my music over the road noise.) This really makes my Fuze more useful.