Need digital audio expert. Nulling files.

[steenamaroo]

Hi everyone. I'm in a situation and it's got me baffled.

I'm working on a session and during tidyup I accidentally delete the vocal track in it's entirety.
(Tried data recovery - Garbled wavs)

They're not in the audio folder - I did it properly!


Now, it's not the end of the world; I can have the chap re-sing it, but I'm thinking smart and realise that I have a recent stereo bounce with vocals and a session save point from exact that moment which is now missing vocals.
The bounce region is still selected in the session and everything!
I'm obsessive compulsive about that. I ALWAYS select bounce region, save, don't touch anything, bounce, quit.

So I rebounce the stereo wav but this time without vocals.
Line up the two wavs in a daw using the grid, flip polarity of one and BAM..I've got my vocals back, right?

Well...All acoustically recorded instruments cancelled out as expected and left nothing but vocals, but virtual instruments that come in later don't cancel!?
I can't get my head around this at all.

The virtual instruments appear to be out of time by about 9ms in the vocal-less bounce and even if I manually compensate, they still don't null. It's like they're just slightly different somehow.
I can completely rule out changes to the session. Even the file properties show date created and date modified as being identical.

I'm using ProTools 10.

I know everything is lined up perfectly and the volumes are perfect etc.
Bounce settings are the same; I never change them.

I can't think of anything that would effect VIs only.
One of them (addictive drums) introduces 128ms delay, but I have/had ADC on.
The other's (xpand) don't, so that rules out ADC anyway.

There's very little else I could have changed. I can't see how buffer size or anything would come into play.

Any thoughts?


EDIT: Just found out, no matter what way I set the buffer sample size, the acoustic instruments always cancel.
The VIs, however, cancel to varying degrees depending where the buffer is set.
I'm currently bouncing at all settings for a bit of comparison.

[John_Toolbox]

Did you
1. Print the Virtual instruments to an audio track, and then do the bounces?

or

2. Do the two bounces with the virtual instrument loaded, but not pre-recorded as audio?


I have a feeling you did #2, and what's more than likely going on is that the virtual instrument is designed with a bit of randomization in it, so the round robin(random sample selection) aspect of the instrument has created two slightly different bounces. The only 100% sure way to get a VI to null is to print it to audio first, and then use that audio to create your full bounce and instrumental bounce.

[steenamaroo]

Hi John,
Thanks so much for the reply. That makes sense I guess.
I've been doing option 2 across the board.


I've discovered that there's up to a 10ms time difference in VI's comparing a bounce with 256 buffer to 1024 buffer, whereas the acoustic recordings are the same regardless.

I still can't achieve complete VI null though, so I'll try what you're saying now.

I'm wondering though, the fact that the original bounce with vocals used 'live' VI's (ie, not printed) means that I'll just never get a match again?

[John_Toolbox]

Originally Posted by steenamaroo

I'm wondering though, the fact that the original bounce with vocals used 'live' VI's (ie, not printed) means that I'll just never get a match again?

Unfortunately, I think that will probably be the case. What I would do first, is solo one of these VI's, and do a bounce of it(playing back as midi), then do another bounce, but name it differently. Now check and see if these two files null. I have a feeling they won't.

[steenamaroo]

Man, that's some logical thinking.
I'm embarrassed I didn't think of that.

Ok, so I bounced a bar of drums twice and they didn't quite null so that kills the whole idea.
At least I know though. Thanks for taking the time to help out.


Separate issue but I'm shocked that buffer settings introduce a delay between VIs and real instruments!

Seems like 10ms difference between kick and bass could really make a difference, no?

[John_Toolbox]

This has me somewhat baffled, and since I don't host my VI's directly in Pro Tools I'm not actually sure what to expect (In my workflow I use logic as a sequencer, slave PCs for my VI's, and only use PT for audio editing and mixing). So I asked some of my fellow duc moderators, and here are a few things to consider with this problem:

1. There used to be a bug in PT 7.4.2 that caused erratic ADC behavior. It could be worked around by toggling ADC off and then back on again. I'm not sure what other versions (if any) of pro tools this problem existed in, I haven't experienced it firsthand as I went from 6.4.1 TDM (no ADC) directly to PT9HD.

2. ADC has a finite number of samples available, which varies from system to system(I think it's 4096 for older systems, and about 4x that for HDX and HDN systems). By increasing your buffer, you might be exceeding this limitation, which would cause the VI's to go out of sync.

3. Stig(duc moderator "PT Lover") mentioned to me that when you print a VI individually, it can be slightly off, and need to be re-aligned. I haven't used any VI's in Pro Tools since the samplecell days(long before ADC ever existed), so I haven't actually tested this myself.


All in all, I think that the only way to make sure that your workflow produces consistent results that null, you're going to have to do some more experiments with ADC printing your VI's to figure out what needs to be done.

[steenamaroo]

Thanks again John.

If you are interested in pursuing this, I can rule out no.2 and possibly no.1 for you.

No.1
I can't think of anything that would suggest erratic ADC behaviour.
The only track making use of ADC is addictive drums at 128 and the discrepancy measured was 9 or 10 samples (I said ms by mistake earlier).
Thinking about it, 10 samples is nothing.....I only noticed because I was zooming in to sample level to line things up.



No.2
My max buffer size was 1024 and ADC setting was 4096.
The max delay introduced by any plug was 128, so 1152 would still be well within the ADC limit.



I'm happy that I'm never going to hear the difference so I'll go about my business. :)
Really though, thanks again for taking the time.

[John_Toolbox]

10 samples is not a lot, I agree... you said 10ms before, which while usually not noticeable by itself, can contribute to a noticeable delay if other factors are present.

The real question is whether or not it's a consistent offset. With VI's, it's hard to know what is actually causing the latency you measure, because there can easily be 10ms of dead space at the beginning of one sample, but the next one will be edited right at the beginning of the waveform without the 10ms of flatline.

If you can measure a consistent offset with every bounce, I believe you can adjust the "global MIDI offset" somewhere in the PT preferences, which would essentially shift all MIDI earlier, so that your MIDI remains quantized to your grid without playing back the VI's with a delay. I'm not sure if this only affects external MIDI tracks, or if it covers instrument tracks as well.... But this feature is really more designed for a studio with external midi boxes, where there is a known latency of a hardware MIDI interface that is consistent for all MIDI devices.

I think the simplest way to deal with this is just bounce the VI's when you've committed the MIDI edits for them, and then nudge/edit the rendered audio if necessary on a case by case basis.

[steenamaroo]

You're a wealth of Knowledge, John.

Thank you.

[Mark Waters]

Just so you realize when you delete a file on a computer is not really deleted.. the pointer to it is simply removed, and the file will remain intact until its randomly overwritten.

There are many "undelete" programs out there... $30-$70 for one is usually a lot cheaper/faster than wasting hours trying to finagle a track back.