Re: advice on multi-track recording
Re: advice on multi-track recording
- Subject: Re: advice on multi-track recording
- From: Brian Willoughby <email@hidden>
- Date: Wed, 27 Apr 2011 16:19:17 -0700
MPEG was not designed for recording raw, unmixed, unmastered,
individual tracks. Instead, MPEG was designed as a final delivery
format with the assumption that the audio content has already been
mastered to standard levels and thus which uses the available 0 dBFS
efficiently. I'm not saying that MPEG was designed to rely upon the
current trend that has resulted in the CD loudness wars (that's too
much level), but that the music content should certainly be within
about 6 dB of full scale.
MPEG was never intended as a recording format for audio that is
intended to be processed later. It is purely for final delivery and
playback without further processing. As such, MPEG is entirely
inappropriate for multitrack recording.
If you want to produce 8 single track MPEG audio files, then you will
need to apply some kind of manual (mastering engineer) or automatic
gain control to the tracks. If you don't, then the noise performance
of these MPEG tracks will be much worse than the typical MP3 that you
might buy from an online music vendor.
To be specific, a 8-track voice recorder is not typically going to
have optimized recording levels for all channels. Even if it is a 24-
bit converter, you will not get full 24-bit performance unless a
highly skilled operator is manning the controls. Thus, when you
convert to MPEG, all of the psychoacoustic models for masking and
frequency-dependent noise analysis go wacky. I have tested this with
8-track voice recordings, and the noise floor is much worse that a
standard 'music' MP3. You might get better results if your recording
levels are set much higher than the test facility I worked with, but
then you risk clipping the individual tracks.
What I have done is to focus on reducing file size in two ways, but
avoiding MPEG or AAC.
First, voice requires only 8 kHz to 10 kHz at the high end. If you
were recording at 44.1 kHz, then I would suggest 22.05 kHz. Since
you're recording at 48 kHz, then I recommend that you down-sample
this to 24 kHz using an AudioConverter. This leaves you with 12 kHz
for your voice recordings, which is plenty. This will cut your PCM
CAF file size in half.
Second, Apple Lossless will reduce your bit rates another 50% or so.
Because ALAC is lossless, you will not suffer from increased noise
just because your raw tracks are not very loud.
If you absolutely must have MP3 format, then I suggest that you need
intelligent mixing and mastering algorithms in your signal chain.
As for the slow speed of your PCB CAF to 8-track MPEG conversion, I
think you may need to focus on optimized file access API. As a
sanity check, have you run your 4-hour 8-channel 48 kHz CAF files
through Logic in offline bounce mode to MP3? That might be a
reasonable gauge of what is considered efficient coding for file
processing speed. If your code is as fast as Logic, then you might
not be able to improve it by much. If you're slower than a Logic
bounce on the same material, then perhaps you can optimize your file
access.
Brian Willoughby
Sound Consulting
On Apr 26, 2011, at 23:40, Iain McCowan wrote:
I have a software application that records 8-track audio (48 kHz)
direct to disk using the ExtAudioFile API. The application is
voice recording e.g. for meetings.
My goal is to end up with 8 single track mpeg audio files.
My solution to date has been to record to an 8-channel uncompressed
PCM CAF file, and then at the end of the recording, split into 8
encoded single channel files.
This works reliably, however it can take a long time for multi-hour
recordings (e.g. 4 hours x 8 chan x 48 kHz), which users don't like.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden