Topic: Cool Edit Pro 2.1 time stretching

[cyfan]

I generally tape two sources at most shows:
 - SBD > JB3
 - mics > UA-5 > JB3
Really like the richness I get when I multi-track merge them in Cool Edit Pro 2.1.
Of course one is always a little longer than the other, so I have to stretch the shorter file to synch the recording before messing with levels to find the perfect fit.
But I've been finding that I seem to pick up weird digital .... um, warping when I stetch a file (and we're talking .2 seconds in a 90 minute recording). It's only noticeable in the chatter between the tracks and not during the music, but I still don't like it.

Lately, I've taken to removing tiny bits of the SBD source between songs so each track starts perfectly synched (and the delay isn't audible through the end of the track). But this is a major pain in the butt compared to stretching a file.

Anybody got any ideas were the warping sound is coming from?? (almost sounds like a DAT with dying batteries, but its only in the non music bits).

Tim

[Brian Skalinder]

so I have to stretch the shorter file to synch the recording before messing with levels to find the perfect fit.
But I've been finding that I seem to pick up weird digital .... um, warping when I stetch a file (and we're talking .2 seconds in a 90 minute recording). It's only noticeable in the chatter between the tracks and not during the music, but I still don't like it.

I bet it's simply an artifact from the stretching process.  I also suspect it's present throughout the recording, but it's only noticeable between tracks.  Any chance of an MP3 sample?

Lately, I've taken to removing tiny bits of the SBD source between songs so each track starts perfectly synched (and the delay isn't audible through the end of the track). But this is a major pain in the butt compared to stretching a file.

The only method I've used successfully is to break up the SBD into tracks and align each track individually with the AUD source.  Major PITA, but you get used to it and faster after a while.

[cyfan]

Yep. I'll try to remember to toss an mp3 link up tonight.
One thought was that stretching somehow picked up a bigger delay between the audience noise from the mics in the crowd and the light audience noise coming through the stage vocal/drum/guitar mics. Still not sure how that works in my head though. Actually, some folks who've heard those matrixes never picked up on the weird sound between songs until I pointed it out. So it's not awful. Still disturbs me.

[dklein]

Tim - I never had any luck with CEP's stretch feature.  It always seemed to introduce weird sounding artifacts as you've described.  I did find the stretch feature in Wavelab to sound good though.  I don't do this anymore because I use a multitrack device now but I've been thinking that stretching may not be the right way to go about it...

The debate (in my head) has to do with timing and the original sequence of events.  We need to stretch because clocks are different.  I used to 'preserve pitch' thinking that I didn't want to change the pitch.  But in a PCM system, sample rate, clock rate and pitch are all interrelated.  e.g. recording clock runs slow->playback sounds fast when played on a faster clock, pitch goes up.

There was only one original event - only one 'pitch', and consequently only one duration.  i.e. if the sound lasted 55min32sec545ms, then that's how long it was.  If your clock changes it, then the pitch also changes (think about playing a 44.1 recording at 48k - not only are things shorter in time, the pitch goes up).  The deviations from two recorders should be brought back into alignment.  By preserving pitch during the stretch, we may actually be messing up the true pitch.  If you follow that the clock differences cause a pitch error on playback, then what we really want to do is allow the pitch to change.

So once we establish one source as a reference (doesn't matter which one) I think the proper course of action is to resample to a corrective rate, then slap the 44.1k header on it.  This thinking came about during some video synch work.

So basically what I'm saying is forget stretching...resample to whatever rate is required.  e.g. I know the gap between my video camera and my audio tools.  If I take my 48k audio recording and resample to 47,997 and then simply change the rate back to 48k, everything lines up.  Likewise with your two audio sources - leave the one that will dominate the mix and find your magic number for the other one.  Resample at the highest quality setting, change the header back and you're good to go.  What I discovered when I was mixing UA-5/JB3 sources is that my equipment always behaved the same.  Once I had that magic number it was good for every recording.

Don't know if that made sense - ask questions.  I also made a spreadsheet for calculating the resample rate easily.  You just need to synch up one point at the beginning, one at the end and then enter the 'drift' and it will give you the number.  I can post it if this makes sense to you.  I've been thinking about this for some time and was gonna put together a proper how-to guide but your post got me going 

[cyfan]

Thanks David.
Actually that makes a lot of sense now that you've explained it.
I have been stretching with the preserve pitch selected.
I almost always tape separate mic and board recordings when bands allow it, so I'll have a good chance to test the resampling method you mentioned really soon. Perhaps hit you up for that spreadsheat you developed too.

Big thanks.

tim

[neognosis]

I also made a spreadsheet for calculating the resample rate easily.  You just need to synch up one point at the beginning, one at the end and then enter the 'drift' and it will give you the number.  I can post it if this makes sense to you.

Mind sharing that formula you worked out? 

I tried using the time stretch in SoundForge last night and couldn't get it to work for me at all, but I've never used it before so I probably told it to do the wrong thing.  I've done a few aud/aud mixes before, but this is my first attempt at an aud/sbd mix.

I did like you suggest here, found two points at the beginning and end and figured out the difference.

POINT 1:
AUD source = 2:47.977
SBD source = 2:49.786
difference = 1.809

POINT 2:
AUD source = 36:30.193
SBD source = 36:32.071
difference = 1.878

1.878 - 1.809 = 0.069 (I assume this is the "drift" you mentioned?)

I really hate this.  I am so bad at math, and also new to using SoundForge. 

Thanks for any help!