Beating Pro Tools Delay Compensation tip...

[BScout]

Plugins with look ahead (dynamic eqs, look ahead limiters/compressors, most "AI" plugins) add massive latency numbers that have to be compensated for. Avoid as much as possible, freeze/commit these, or for ones with adjustable lookahead settings, reduce lookahead buffer as much as possible.

Linear eqs add massive latency numbers (because the filters have to be calculated forward and reverse to be linear.) Use min phase or natural phase.

UAD hardware plugins add massive latency (due to communication back and forth to the hardware) when used as a plugin inside a DAW (the Spark versions don't)

Finally, using a bunch of high latency plugins on one track is bad for compensation. But spread out across multiple parallel tracks (must be parallel -- if it's on one track that busses to another track and so on with plugins on each, it's the same as stacking all those plugins on one track) can actually be good. Because if all the tracks have latency due to plugins, the lowest latency number doesn't have to be compensated (subtracted from the latency comp) for all the tracks.

(if you are dealing with AAX-DSP there is also the trick of keeping those together so processing isn't getting swapped often between DSP chips and computer processor)

[anguswoodhead]

Originally Posted by BScout

Plugins with look ahead (dynamic eqs, look ahead limiters/compressors, most "AI" plugins) add massive latency numbers that have to be compensated for. Avoid as much as possible, freeze/commit these, or for ones with adjustable lookahead settings, reduce lookahead buffer as much as possible.

Linear eqs add massive latency numbers (because the filters have to be calculated forward and reverse to be linear.) Use min phase or natural phase.

UAD hardware plugins add massive latency (due to communication back and forth to the hardware) when used as a plugin inside a DAW (the Spark versions don't)

Finally, using a bunch of high latency plugins on one track is bad for compensation. But spread out across multiple parallel tracks (must be parallel -- if it's on one track that busses to another track and so on with plugins on each, it's the same as stacking all those plugins on one track) can actually be good. Because if all the tracks have latency due to plugins, the lowest latency number doesn't have to be compensated (subtracted from the latency comp) for all the tracks.

(if you are dealing with AAX-DSP there is also the trick of keeping those together so processing isn't getting swapped often between DSP chips and computer processor)

Thankyou

[Coconut_Head]

Originally Posted by BScout

Finally, using a bunch of high latency plugins on one track is bad for compensation. But spread out across multiple parallel tracks (must be parallel -- if it's on one track that busses to another track and so on with plugins on each, it's the same as stacking all those plugins on one track) can actually be good. Because if all the tracks have latency due to plugins, the lowest latency number doesn't have to be compensated (subtracted from the latency comp) for all the tracks.

I've never heard of the lowest not needed to be compensated, but I'd be delighted to find out if this is true... So clarify this for me, please:
Example: 3 tracks
Track A = 17,000 samples
Track B = 16,000 samples
Track C = 1,000 samples

Are you saying that because they are all above 1,000 samples, Track C doesn't need to be compensated, and track A and B are compensated 16,000 and 15,000, respectively?
Would that mean you CAN have a latency above the DC limit of 16,383 samples (Track A in the example) so long as the difference between the lowest and the highest is underneath said limit?

If that's NOT what you're saying, then Track A's DC would be shut off and instead not be compensated at all, right? So Track C would then get 1,000 samples of compensation, Track B get 16,000 samples of compensation, and Track A get 0 samples of compensation?


I have more questions regarding DC, but those are all I'll ask here. Thank you in advance!

[BScout]

Originally Posted by Coconut_Head

I've never heard of the lowest not needed to be compensated, but I'd be delighted to find out if this is true... So clarify this for me, please:
Example: 3 tracks
Track A = 17,000 samples
Track B = 16,000 samples
Track C = 1,000 samples

Are you saying that because they are all above 1,000 samples, Track C doesn't need to be compensated, and track A and B are compensated 16,000 and 15,000, respectively?
Would that mean you CAN have a latency above the DC limit of 16,383 samples (Track A in the example) so long as the difference between the lowest and the highest is underneath said limit?

If that's NOT what you're saying, then Track A's DC would be shut off and instead not be compensated at all, right? So Track C would then get 1,000 samples of compensation, Track B get 16,000 samples of compensation, and Track A get 0 samples of compensation?


I have more questions regarding DC, but those are all I'll ask here. Thank you in advance!

Yes. It isn't exceeding the limit though (that's what you are misunderstanding).
Delay compensation limit is the limit on compensation not on a general concept of delay. If one track is 1000 samples and the others exceed that number, the other tracks get the lowest subtracted as their compensation number.
No "limit" is exceeded because it didn't happen.
Delay compensation is a relative number. You can watch this by simply adding plugins to tracks (if you have the delay comp entries open on the Mix window.) If you delay Track A a lot but then add plugins to Track B that increases that Track B's delays, then Track B doesn't need to compensate for the amount of delay already added by the new plugins. Only the difference.

I should make mention, by default, an input-enabled or record-enabled track in Pro Tools has zero delay compensation (you have to put it in blue-mode to compensate). So adding (or enabling) such a track can throw the math off.

[TNM]

Hi, are we absolutely certain that ADC has not been increased to 32K samples at 44/48k?

I have cloned an audio track with a drum loop and soloed the two.

I put a UAD Multiband (dsp) comp on one and everything is perfectly in time.
If I add a second UAD Multiband, it's now 33K samples and the two tracks are out of time.
So I removed one and kept one in, and kept adding high latency UAD plugins till my 2 DSP ran out and it showed 28K sample delay, and the two tracks are perfectly in time

Maybe it's a bug and NOT supposed to behave this way, as the figure IS indeed red, but it's definitely at 44K, checked both in PT project info and UA Apollo Mixer.
Edit, I seem to be getting reliable compensation up to about 18K yet I get clicks and pops in the audio in this situation. Weirdest thing I ever encountered in PT to this day.