Re: [ARSCLIST] The Hope of Audacity Was--Re: [ARSCLIST] Seeking recommendations for oral history digitization equipment (fwd)

[Eric Jacobs <EricJ@xxxxxxxxxxxxxxxxxxx>]

On Sunday, August 17, 2008 6:14 AM, Mike Hurst wrote:

> However, reading Goran Finnberg's comments re the work dome by George 
> Blood, I am surprised to find that different combinations of software 
> and hardware can produce different results. Am I to understand that 
> Adobe Audition, for example, will capture audio differently when used 
> with an M-Audio card, for example, than it would when combined with an 
> E-MU or SoundBlaster card? does it then follow that the same card 
> combined with different software would produce different results? Is 
> this something I should worry about? What is the extent of this 
> discrepancy and how best can I avoid it?

Worth considering:

   - the ADC has an analog section, which can influence the sound
     ahead of the converters by introducing analog noise, 
     distortion, cross-talk, etc.).
   - not all converters are equal in terms of their digital precision
   - clock synchronization between the ADC and the computer can be 
     an issue (whether an internal or an external ADC)

So in the first case of Adobe Audition used with a variety of
capture (ADC) cards, there's a good chance that you will indeed 
hear subtle (or perhaps not so subtle) differences and get 
different results, and the differences will be due to the cards 
themselves as well as how they interact with the software (ie. 
synchronization).

On the flip side, using the same card with a variety of software
is an interesting question.  Each software package should be 
receiving the same bitstream, particularly if they are going
through the same hardware interface with the same drivers.

I've not tried this, but doing the test of multiple software
packages with a single ADC would seem to be tricky.  To do 
this test correctly, you would need all the software packages 
to capture the same digital audio stream simultaneously (ie. in
parallel) on the same clock.  To do otherwise will introduce
either slight input signal differences or quantization 
differences on the part of the ADC - a natural part of 
digitization.  This test would require either a powerful 
workstation so that none of the software packages are starved 
for resources (ie. buffers, CPU) as each software is running 
at the same time, or you need a well-tested digital distribution 
system to multiple computers.  Some software packages may take 
exclusive control over the audio stream in order to guarantee 
sufficient computer resources and to prevent other processes
from introducing audio or noise to the capture stream, making 
such a parallel software test impossible on one computer if 
this is the case for a particular software package.

I've not seen the George Blood presentation, but would be 
interested in hearing about the detailed test and
measurement methodology.

To your questions: (1) is this something you should worry
about, and (2) how best can you avoid these differences?

Regarding (1): Yes, differences in audio capture are something
you should worry about.  "Worry" is perhaps the wrong word
here - "Understand" might be better, so that you can make
decisions on how you can best capture your audio to meet
your needs (access, preservation, budget constraints, etc.).

Regarding (2): I would encourage doing your own tests.  Even
if the technical testing and measurements are beyond your skills 
or means, your ears (along with others' ears) are still a good 
tool (although perhaps constrained by your monitoring equipment
or the acoustics of your environment).  Do lots of A-B testing 
between the source and a variety of capture combinations.  Even 
if you can't identify the root cause of the differences, you'll 
be able to identify elements in the capture chain which sound 
more right or true.  And you'll learn a lot and fine tune your 
listening at the same time (always good things if you're 
involved in audio).


Eric Jacobs

The Audio Archive, Inc.
Tel: 408.221.2128
Fax: 408.549.9867
mailto:EricJ@xxxxxxxxxxxxxxxxxxx