Subject:Can I correct a sampling snafu?

Posted by: Doug_Marshall

Date:5/19/2010 6:06:31 AM

I blew it at a recording session last week and I'm wondering if I can rescue the wave files that are messed up. In short, I used a sound a sound card with a random sampling error problem. The errors are immediately visible in the waves as a spike that is one sample wide here and there. They are approximately twice as positive or negative as they should be, with resultant hash in the audio. I tried the Click Removal tool and it doesn't seem to detect the errors (I suspect that the clicks it is designed to find are longer than one sample).

Is there a tool that might be able to detect and interpolate the correct value for these random errors? Thanks for all help.

Doug

Subject:RE: Can I correct a sampling snafu?

Reply by: rraud

Date:5/19/2010 9:42:28 AM

Post a sample on the internet so we may make a more accurate assessment.
- Have you tried the click removal tool in NR-2.
- You can try to draw the glitches out with the pencil tool.
- Copy/paste alike sections of a song or segment if possible.
- Try the 'Repair' process under "Tools"
- Other third party restoration plug-ins may work also. Download some demos and try em' out.

Depending on which version of SF you have, some of the above tools may not be available or have advanced parameters.

Subject:RE: Can I correct a sampling snafu?

Reply by: Steven Myers

Date:5/19/2010 9:52:27 AM

Sometimes iZotope RX does a great job of guessing what should be there instead of that spike you're seeing.

Subject:RE: Can I correct a sampling snafu?

Reply by: Doug_Marshall

Date:5/19/2010 9:31:32 PM

Thanks for the suggestions. I've discovered that the Tools\Repair\Interpolate command can handly fix the problem one sample at a time but doesn't seem to be able to distinguish the errors in larger chunks - and it is limited to 0.5 second selections, so is totally impractical.

I downloaded Izotope but the audio didn't play properly for some reason.

Subject:RE: Can I correct a sampling snafu?

Reply by: MarkWWWW

Date:5/20/2010 5:25:27 AM

If there are only a few of these spikes then I would try using the glich finder (Tools|Find). Then repair them either manually or using Tools|Repair.

If there are a lot of them, then I would look into whether the above approach could be automated using Scripting. (I don't know enough about it to know whether this is feasible, but it might be.)

Mark

Subject:RE: Can I correct a sampling snafu?

Reply by: Doug_Marshall

Date:5/20/2010 8:37:54 AM

The glitch finder can find some of the errors but it still seems that by eye is best, though not practical.

I uploaded a snippet of one of the bad files here: http://www.marshallandogletree.com/index.php?cID=129

Believe me, I am grateful for all help!

Doug

Subject:RE: Can I correct a sampling snafu?

Reply by: jackn2mpu

Date:5/20/2010 9:22:11 AM

Two things to try - a narrow filter or something that can take a noise print and apply that to the file to subtract the noise. Either way you may have to go through the file more than once as trying to take the offending signal all in one pass could take out too much of what you want to keep, particularly overtones. If you take out too much of that the sond will be dull.

Jack

Subject:RE: Can I correct a sampling snafu?

Reply by: MarkWWW

Date:5/20/2010 3:01:49 PM

Ah, now I've seen the file I can see that this is far worse than I was expecting. I was expecting a small number of big spikes that would be relatively easy to deal with. But the file has a vast number of small spikes, which I don't think are going to be easy to deal with by normal methods.

If you aren't able to just do the recording again with decent gear then it might be worth trying some of the more severe Smooth settings of the Process|Smooth/Enhance tool, but you may feel that as well as smoothing out the spikes it does other things to the audio that you may not be able to tolerate. Worth a try though.

Mark

Subject:RE: Can I correct a sampling snafu?

Reply by: rraud

Date:5/20/2010 4:52:59 PM

I agree with Mark, you can try the smooth process, but I think your SOL.
If you re-record the piece I think the 96k sampling rate is overkill if the final destination is the internet, even for CD in this genre. 24bit is good since there's a wide dynamic range.

Subject:RE: Can I correct a sampling snafu?

Reply by: Doug_Marshall

Date:5/20/2010 5:36:11 PM

Thanks very much. I will try the smoothing process and see if anything is salvageable. I do plan to record this again soon because I didn't have a lot of hope for these files. From the audio point of view, this organ (which my company built), is quite unusual because of the 64' pedal stops - a genuine 8 Hz is reproduced by a couple of Thigpen rotary woofers. Anyone interested can view the organ and hear the sound on our installation page at MarshallandOgletree.com. This organ is Opus 4.

Subject:RE: Can I correct a sampling snafu?

Reply by: musicvid10

Date:5/20/2010 6:06:41 PM

Toccata in D minor. I will definitely check out the organ on your site. I'm more of a B3 guy who never fully mastered the pedals.

As far as the audio sample, there doesn't seem to be a lot of program material above 5KHz. Although not a terribly effective solution, applying a high shelf (low pass) will take out some of the chirping artifacts.

Other than that, I have been know to tackle some irreplaceable recordings at the sample level, but it is very tedious, sometimes taking days.

Subject:RE: Can I correct a sampling snafu?

Reply by: Doug_Marshall

Date:5/20/2010 10:25:54 PM

At the sample level the Repair/Interpolate command works perfectly. But I think it'll actually be quicker to spend a day recording it all again rather than spending even longer staring and clicking. So, I guess I'll lay this one to rest. At least I have the organ all pre-programmed and ready to go.