[BC] dueling algorithms and audio quality

[Dana Puopolo]

Marty:

Like it or not, cheesy converters ARE used in broadcast facilities all the
time! Why? Because managers are holding pennies with Vice Grips (tm) these
days!

Yes, linear is the way to go, but as you said, linear takes bandwidth. How
much? It's an easy calculation:

44,100 x 2 (channels) x 16 (bits) = 1411200 kbps = 1.411 Mbps

This will barely fit into a T1.

If you use 48 kHz sampling, it gets too tight:
48000 x 2 x 16 = 1536000 kbps = 1.536 Mbps

As many know, a T1 is 1.544 Mbps.

One way that some conmpanies are dealing with this is by using FLAC
compression. FLAC stands for Free Lossless Audio Codec. It is freeware; no
license fees are required. FLAC typically does to audio files what ZIP does to
programs; it compresses them without losing a single bit. Typically, the
compression ratio is 2;1, meaning that you now have HALF the bandwidth
requirements of linear. The tradeoff is latency-it takes extra time to
physically do the compression and decompression.

-D

From: "Acuff, Marty" <macuff at orban.com>

On Thursday, October 01, 2009 13:28 Dana Puopolo wrote:

> Actually you are only partially right, Chip, because
> eventually digital HAS to become part of the analog 
> domain-and that's frequently where mucho quality is lost.
> ..
> 
> Just keep it linear and digital is OK.
> 
> --chip

This would only be true if either the A-to-D or D-to-A converters are cheesy,
OR if the digital transport path ISN'T linear.  
What chip says is true, perhaps a bit oversimplified.

It's important to differentiate between linear and non-linear digital audio. 
Linear digital is the PCM WAV that you get from a direct rip of a CD.  Linear
is not "compressed" (i.e. bit-rate reduced). In a properly designed digital
system there is a bit-for-bit data transfer. It is a transparent system.  But
there's a price tag: the bit rate is very high.  
Too high for some STLs.  
For STL systems that cannot pass the high bit rate of linear digital audio,
they use "compression" (i.e. bit-rate reduction). Many of the bit-rate
reduction algorithms are lossy, permanently removing audio data.  Others are
"lossless" algorithms.  
My point is this: loss of quality DOES NOT occur because of a linear digital
audio transport system; the loss of quality usually occurs because of bit-rate
reduction, or improper configuration of the transport system, or (God forbid)
a crappy A/D or D/A conversion. 
Cheesy converters have no business being used in a broadcast facility.  
Sometimes out of necessity, bit-rate reduction is unavoidable.

But please, don't shoot the messenger ;-)

	-- Marty

--