Clips During Recording vs in Cleanup Audio

[bshonbeck]

I have found there is a big discrepancy between the number of clips indicated while recording and the number of Clips found when scanning for clips in the Cleanup Audio page. I set my recording level externally on a Parasound Zphono USB interface.  Invariably the sound samples I choose on the album don't represent the loudest passage on the entire album so I end up with some clips being shown while recording.  However when viewing the recording in the Cleanup Audio page and doing a Scan For Clips, it often (always?) finds much fewer, if any, clips in the recording. 

Until I discovered this, I would often re-record an album side that had a dozen or more clips counted while recording.  As an example, today I was recording Cat Steven's Peace Train and, as above, it had at least 10 clips shown while recording (I reset the meter when they occur).  I was going to re-record it but on a whim, I checked in the Cleanup Audio page and it didn't show even one.  Several peaks were very close to 0db but none apparently exceeded it.

Can someone give insight?
Thanks

[Paul Sanders (AlpineSoft)]

Thanks, I'll take a look, although it's probably not audible in practise.  We're probably using (slightly) different criteria to report them in the two places in the code.

[bshonbeck]

I agree, these likely were not audible.  However their high count causes me to think my input level is too high so I lessen it.  If the clip count is incorrectly high, then I have reduced the dynamic range of the recording unnecessarily.  For now, I will view the clips and clip count on the Cleanup Audio page to determine if a lower input level and re-recording is indicated.

[Paul Sanders (AlpineSoft)]

We've looked into this now, in fact, and the definition of a clip was slightly different in the two cases.  This will be corrected in the next release.

As for setting the recording level, the difference between the two clip detection criteria was very small, so I would be guided by the number of clips reported while recording.

[bshonbeck]

Thats good news!  So if I understand correctly, the next release will show more clips in the Cleanup Audio page to match what was tallied in the Recording window.

Can you give some guidance as to how many clips could be tolerated before the input level should be reduced?

Thanks!

[Paul Sanders (AlpineSoft)]

Actually, the exact numbers are unlikely to correspond because the way clips are detected differs in the two places.  But at least the same signal level is now being used.

The answer to your other question is 'how long is a piece of string?'  Maybe you can start recording and if you see the clip count going up reduce the recording level slightly.

[bshonbeck]

Did this change make it into the latest release?  I did not see it listed and wondered if it came under the "A few other bug fixes and improvements" category.

[Paul Sanders (AlpineSoft)]

Yes, we did make a (minor) change, but the algorithm for detecting clips is not the same in the two places in the code, so you will still see a discrepancy.

[bshonbeck]

Is it correct that the detector in the Cleanup Audio window is accurate (since it deals with the actual digitized waveform and there is no headroom above 0db) and the one in the Recording window is more conservative, thus potentially indicating more clips?

I am sure there must be technical reasons why they can not be the same but it sure would be nice if they were.

[Paul Sanders (AlpineSoft)]

I would say the detector in the Cleanup Audio window is more accurate, yes.  It uses a threshold to decide where a clip begins and ends (in order to decide how much audio to repair).