Re: linear pcm: 16 or 32 bits?
Re: linear pcm: 16 or 32 bits?
- Subject: Re: linear pcm: 16 or 32 bits?
- From: Bob Lang <email@hidden>
- Date: Wed, 11 Feb 2009 23:13:45 +0000
Hi Joel
On 11 Feb 2009, at 19:01, Joel Reymont wrote:
The hardware default for LPCM seems to be 32 bits per channel and
thus 4 bytes per packet and per frame.
What happens if the audio format is set to 16 bits per channel? Less
audio fidelity?
Here is a vast simplification of the issue:
As you reduce the number of bits per sample, so the signal to noise
ratio gets worse. The SNR is approximately 6dB per sample bit, so 16
bit samples have 96 dB SNR. This increases to 144 dB for 24 bit
samples. If you reduce to just 8 bits per sample than you only get 48
dB of SNR - which is very noisy.
Now 96 dB (16 bits) is actually very good, but it does come with a
sting in the tail - it only applies to signals whose amplitude uses
the full 16 bit range (-32767 to +32767). If you record lower
amplitude signals then the effective number of bits you're using gets
reduced and the SNR gets worse - each time you halve the amplitude you
take away 6 dB from the SNR. This gets to be an issue when you're
mixing different signals since you often need to bring low amplitude
signals into the mix - these therefore have a lower SNR.
This becomes less of an issue with 24 bit samples (which includes 32
bit floats) because you have so much extra headroom to begin with.
Of course in practice there are other factors that come into play and
complicate the issue.
Bob
_______________________________________________
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