Re: [ARSCLIST] Cassette obsolescence - digitizing standards

[Jerome Hartke <jhartke@xxxxxxxxxxxx>]

> -----Original Message-----
> From: Association for Recorded Sound Discussion List
> [mailto:ARSCLIST@xxxxxxx] On Behalf Of steven c
> Sent: Wednesday, February 22, 2006 9:25 PM
> To: ARSCLIST@xxxxxxxxxxxxxxxx
> Subject: Re: [ARSCLIST] Cassette obsolescence - digitizing standards
> 
> ----- Original Message -----
> From: "Dave Bradley" <db65@xxxxxxxxxxx>
> > At 09:17 PM 2/21/2006, you wrote:
> > >Now, just so I understand...
> > >Do 1412 and 1426 represent a value which expresses the value of the
> > >height of that particular "step" when an analog sound file is
> > >converted to digital sound? So that there are 44,100 of those values
> > >in a second of a digital sound waveform...which more or less
> > >approximate the analog version of the waveform?
> >
> > There are 44,100 of those values, of which any one of them could be
> > anywhere from 0 to 65,535. And while I get the dig that you're
> > implying about the inaccuracy of the digitization process, to say
> > that because it's not entirely accurate justifies leaving in other
> > large errors is an awfully big jump that I certainly wouldn't be
> > willing to make.
> >
> No dig intended...I just get lost trying to see in my own mind what is
> happening when analog files are digitized, what with all the different
> terms that are thrown around! My understanding is that, while a series
> of steps can never be an exact duplicate of a curve (shades of calculus!),
> there is a certain point where we poor analog(?) humans can't hear
> any difference.
> 
> However, now I see that the 44.1 (or whatever) is the number of samples
> taken, and the 16 bits are the number that evaluate to the "height" of
> each step...so each one has something to say about how accurate the
> stepwise picture is in comparison to the analog curve...
> 
> Steven C. Barr

Both analog-to-digital (A/D), and its inverse digital-to-analog (D/A),
conversion require filtering. When digital audio is "played", the output of
the D/A converter is filtered to remove the "steps", resulting in a
continuously varying "smoothed" analog output that can closely replicate the
analog input. Fidelity to the original depends on the initial sampling
frequency, i.e. how close together the "steps" are, and also bits per sample
that determines how accurately the recorded amplitude represents the
original amplitude and also determines the noise floor.

16-bit samples represent 65,536 possible values. If noise is the
quantization error of one-half bit, then the noise floor is 102 dB below the
maximum possible (peak) output. Similarly, amplitude uncertainty of one-half
bit represents amplitude distortion of 0.0008%.

Although A/D and D/A conversion may superficially seem artificial, its
fidelity usually surpasses that of purely analog means such as records and
analog tape. Ears accustomed to records and tape have "learned" to expect
certain colorations, and digital audio may seem sterile when these are not
present. I started playing with mono "hi-fi" in the 1940's, and initially
thought that stereo was unrealistic. How things change!

Digital capture and storage usually provides error detection and correction,
advantages not present for analog means where scratches, debris, hum, and
the like impair playback.

Jerry
Media Sciences, Inc.