Subject: Re: Re[2]: [linux-audio-dev] peakfiles and EDL's
From: Richard C. Burnett (burnett_AT_tality.com)
Date: Wed Feb 28 2001 - 22:12:03 EET
Yeah, Samplitude has a utility to automatically remove DC offset, its not
even a plugin, built right in. Probably for what you say, and depending
on the age of the sample or the quality of the A/D converter. When I was
working with one, the input had to be from .5 volts to 5 volts, so I had
to build a circuit to bias the signal (and diode to clip when
overdriven) the input. The problem here is where is zero? the A/D has no
idea, because it was up to me to pick the input circuitry and up to me to
play with the output. I think this is why you sometimes see offsets, and
the fact that poorly designed filters/effects could be doing something.
As also stated, most of the good converters have a digital filter built in
to remove that, but, they are probably just as many crappy converters
being used. The one I used was from Harris and cost $100.00 a piece in
quantities of 1-100, and it had not internal filter, but this too was a
general purpose version, not one built specifically for audio data. But
again, if it wasn't a problem, then people wouldn't be seeing files with
DC offsets and the tools to remove them wouldn't exist.
FWIW, Rick
On Wed, 28 Feb 2001, Frank Neumann wrote:
>
> Hi again,
> Robert Schrem wrote:
>
> > <ExtensiveExplainMode DetailLevel="max">
>
> [...]
>
> > As far as i know this isn't a issue for some years now -
> > the today converters are more sofisticated and don't produce
> > DC offsets. They have digital high pass filters built
> > inside (with very low filter frequency). This avoids that
> > any constant offset will be output by the converter.
> > (This also means, that you can't measure DC voltages directly
> > with an AD converter build for audio purposes).
> > </ExtensiveExplainMode>
>
> Thanks for that nice historical explanation. Still, I did stumble across some
> cases of files with DC offset here and there - you know, I'm sometimes out
> hunting for drum .wavs on the 'net, and at times I wonder why files that I
> play stop with a "click" though their end clearly does a fade to 0.
> I wrote a tiny tool for myself to check, and found that some of them were a
> good deal off from the 0 point. Good to have a DC offset corrector handy
> then!
>
> > Anyway: Even if you would care about DC offset you would
> > still recognize ANY clipping if we only compute the
> > absolut maxima of an set of samples. For positive signed
> > samples as well as for negative signed ones.
>
> Hmm..I might have misunderstood your definition of "maxima" here. I simply
> think of "maxima" as "largest positive value in range" and "minima" as
> "smallest negative value in range". Only looking at the largest positive
> values then would not show me when a negative value hits the bottom of my
> signal range.
>
> Now I understand your idea is to do "abs(x)" for all samples in the range,
> at the same time storing the max value of these, and then only plotting the
> "positive half" of the samples. Ok, got it :-). Maybe people are just more
> used to finding a peak representation of a waveform that looks like the
> original waveform - with values both above and below zero.
>
> (Someone tell me when I'm making a fool out of myself here :-),
> Frank
>
+------------------------+-----------------------+
| T a l i t y | +------+ |
+------------------------+ +----+-+ | |
| Richard Burnett | +-+ | |
| Senior Design Engineer +---+ +----+ |
| burnett_AT_tality.com | | |
| | | |
| Phone: 919.380.3014 | |
| Fax: 919.380.3903 | | |
+------------------------------------------------+
This archive was generated by hypermail 2b28 : Wed Feb 28 2001 - 22:31:22 EET