| Author |
 Topic 
|
beetle
Location: USA
Posts: 2591
|
 Posted - Wed Jun 27, 2001 7:32 pm
|
|
|
|
My card produces about a -.001% DC offset. I know the improtance of centering your wave, and I do it, but at that degree, is it really necessary?
|
|
jonrose
Location: USA
Posts: 2901
|
|
Posted - Thu Jun 28, 2001 12:06 am
|
|
|
Perhaps not, at that low of a level...
I would think it'd only start to cause problems if the DC offset level was quite a bit higher, especially where the beginnings and endings reside, if they're not topped and tailed. Then it might actually be something to cause concern (might actually cause audible clicks at those points).
I've never had any particular problems with an offset this small.
All the best... -Jon
:-)
_________________
 |
|
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Thu Jun 28, 2001 1:37 am
|
|
|
The only time that an offset that small could make a difference is if you are recording a full-amplitude symmetrical wave. One peak would clip a fraction (0.001%!) before the other. And we do this all the time, don't we? So that's not an issue.
As for zero-crossing clicks between the offset and digital silence... well, that's an interesting one. In samples, we are talking 65,536*0.001=65.5 samples of baseline offset on a 16bit recording, likewise 1048@20bit and 16777@24bit, so I think that jonrose could well be right about this - I'd rather do away with it, especially on a card with a low noise floor, as this much offset is significantly greater than the amount of dither you might use, I believe.
Actually, that's quite staggering when you think about it - a 0.001% offset at 24bits uses a quarter of all the available samples at 16bit!
Steve
_________________
 |
|
|
|
younglove
Posts: 314
|
|
Posted - Thu Jun 28, 2001 10:11 am
|
|
|
Steve, Steve, relax. That's .001%, which is
a factor of .00001. For higher bit depths,
it's still the same percentage because the
"total available samples" also would go up.
.001% is not much, but I wouldn't tolerate any higher. You can always record with
"Adjust for DC on record" checked on. Try
recording a few seconds without DC adjustment on record, then highlight the first half of that and record over it with
adjust on. See if the new recording is quieter.
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Thu Jun 28, 2001 11:25 am
|
|
|
| Quote: |
Steve, Steve, relax. That's .001%, which is
a factor of .00001. -younglove |
And he's quite right of course. I blame the heat! But that does give rise to a slightly interesting theoretical question, because that makes the original error 0.65 of one sample, and how exactly did anybody measure that?
For anybody else that's confused, that means that we're talking 1.04 samples at 20bit and 16.7 at 24bit.
The chances of hearing the effect of this offset, especially the 16bit one, is ZERO! This does make a bit more sense (it's cooler now), and since thermal drift, electrical noise, etc in your soundcard will all be present at a greater level than this anyway, how can the software justify the 0.001 percent error claim at all?
Steve
Edited by - SteveG on 06/28/2001 11:26:59 AM
_________________
|
|
|
|
| |
Topic
|
All Forums
DC bias offset