Re: Users and the development process (was: Re: See ya!)

On Wed, Nov 15, 2006 at 08:54:10PM +0100, Florian Berger wrote:
> No one is talking about squeezing. I want to go for a certain RMS-Peak
> ratio to actually avoid overcompression, a thing recommended by
> professionals[1] and hard to do without an appropriate meter.

Well, it's recommended by Bob Katz in the referred-to article, but that
is still only his personal opinion.

Actually RMS-peak ratio is very dependent on the shape of the waveform,
eg. a square wave has a RMS-peak ratio of 1, but a sine wave has a ratio
of around 0.707. The `ideal' ratio (if such a thing exists) depends on the 
particular program material. I find it fairly easy to avoid over-compression
just by listening and looking at the output meter - if it stays near the top
and doesn't move much, it's probably overcompressed. Of course it also
depends on the dynamics of the input signal, but you've got the
meters on the compressors to allow you to keep an eye on that.

Bob Katz's article also referred more to film sound (where there are
defined standards for playback levels) than stereo audio, where playback
level is (quite rightly) undefined. As JAMin is aimed at stereo audio
and not multi-channel film sound, isn't trying to pin down defined
amounts of compression just overkill? Music itself isn't rigidly defined,
so it seems a little pointless to try being too rigid in your approach
to processing it.

> ... 
> But the projects I've dealt with so far do it another way. They want
> to build a tool that is useful not only to them, but to end users. So
> the end user's opinion is a critical measurement of how they succeeded.
> Of course, again, there is no obligation to work this way. But in my
> mind it produces the better tools.

What end user are you aiming at? One user's fully featured tool can be a 
bloat monster for someone else who just wants effective basic features.

> ...
> Oh. As for "any of the work": Yes, I won't go, learn C and study DSP
> algorithms to fix it myself. You see, there _are_ developers in the
> project who know about all that stuff intimately.

You're assuming they have the time. Developers have lives too!
Last year I wanted to be able to use AMD64 SSE optimisations in Ardour,
but they were only available for x86. Rather than pester the Ardour
devs, I learned enough assembly language programming to port the
existing 32 bit assembly code to the AMD64 platform. Result: AMD64 users
can now record and mix 33% more channels in Ardour than before for the same
DSP load. It took a couple of weeks (I'm a sound engineer not a professional 
programmer), but I learned a lot of useful stuff. Maybe you should
try it - it doesn't hurt (much)!

> ... 
> For the -30dBFS:
> D > Probably because if you are compressing audio that quiet, then
> D > adding makeup gain to bring the peaks to 0dBFS, you will be raising
> D > the noise floor a lot too. You'd be better off re-recording.
> 
> I do not want to add makeup gain. A 24 Bit recording with -15 dBFS peak,
> -30 dBFS mean level and a noise floor somewhere around -100 dBFS is not
> at all bad in a technical way. If I want to do some leveling with a 1.5
> ratio and a threshold of -45 dBFS, the software should not get in the
> way.

A mean signal-noise ratio of 70dB on a 24 bit system could be described
as not particularly good (yes, I realise that practical microphones
can't do much better than that, but I would argue that the recording
system should have a much better S-N ratio than the signal source).
Expecting a mastering compressor to handle such a low threshold is
a little unreasonable IMHO - by the time a mix is ready to master I'd
expect levels to be high enough to avoid the need for such extreme
settings.

John

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV