"Pre-Echo" Offline bounce problem

[frenchman]

Hi,
I have juste reported in "Idea scale" a problem when bouncing offline, in certain configurations the reverb precedes the source and this is a total disaster.
I though everyone should be warned : I have produced a couple of masters that way and am very angry at discovering this, since in certain mixes you won't hear this right away in the middle of everything
Below you will find a image that shows my test session so that you can eventually reproduce this

Pro Tools 2018-3, mac ox 10,12,6, Six core 3,5GHz Mac Pro

I have devised a test sessions with :
-One mono source track with a short pulse, with bus 2 send on it
-A reverb return on aux track with a D-Verb plugin
- A stereo dummy audio track with a plugin that produces a fair amount of processing delay (for convenience I used Avid's RT Flanger but it will do the same with others)
Offline bounce import track clearly shows that something is wrong, whilst real time bounce is ok

I hope this is going to get fixed FAST
Thanks for listening
Mark Haliday
Paris, France

https://drive.google.com/open?id=1D_...z1-_MDLuFecIRO

[Drew Mazurek]

can you post the session so one of us can repro?

[frenchman]

Originally Posted by Drew Mazurek

can you post the session so one of us can repro?

Yes sorry for that !
I didn't know how to do it, that is why I posted a link to a screen capture of the session

Here is the link to a zip file containing the session
https://drive.google.com/open?id=1ZB...hyaXTDROEAk-Px

[frenchman]

Originally Posted by Drew Mazurek

can you post the session so one of us can repro?

Thanks for your answer,
It's below

FYI, I purposely changed the plugins to use only Avid plugins, so this is not a plugin software problem, but apparently when doing offline bounces something goes wrong in the delay compensation in the buses

https://drive.google.com/file/d/1ZBu...hyaXTDROEAk-Px

[midnightrambler]

I've just downloaded and tried this and it's working absolutely correct here. I don't have the flanger plugin though so I substituted it for Pitch II which has a high delay compensation (4096).

http://duc.avid.com/attachment.php?a...1&d=1523444645

[home_studio]

Works OK with me.
Pro Tools 2018-4, mac ox 10.11.6, Six core 3.5GHz Mac Pro

[Drew Mazurek]

All good here as well, and I have all the same plugins.

[off the wall]

Originally Posted by frenchman

Thanks for your answer,
It's below

FYI, I purposely changed the plugins to use only Avid plugins, so this is not a plugin software problem, but apparently when doing offline bounces something goes wrong in the delay compensation in the buses

https://drive.google.com/file/d/1ZBu...hyaXTDROEAk-Px

I get the exact same (WRONG) result as you. Reverb is pre the source on the offline, OK on the realtime.

Weird.

https://drive.google.com/file/d/1Can...ew?usp=sharing

[BScout]

I think this might be an HDX thing. For those without issue, were you non-HDX?

I'm seeing the same bug but I think(?) I know why.

First off, the Dverb plugin can be anything (including an EQ or bypassed Dverb) and it'll still do it but then you can clearly see the distance between "clicks".

Second, the Flanger channel is sending a stereo signal into a mono (sub path) bus (Bus 2). If you change that bus to a stereo bus (Bus 1-2), the problem goes away. I think what we are seeing is delay comp only happening on one leg of the bus (bus 2) while bus 1 is staying undelayed. But since we are going out center pan, it ends up looking like the undelayed signal and the delayed signal as two clicks with distance between them (the lack of delay on A1 path preceding the delay comp A2 path)

And the reason I think this is an HDX thing is HDX online uses the DSP mixer but offline uses the system processor. So "native" users are using the same mix engine whether offline or online and it is passing the delay comp info for both paths but HDX users are switching engines where the info isn't getting passed.

Tested the theory and that's what's happening. If you change the source to a stereo (duplicate mono audio file into a stereo track and keep them hard-panned, you can see the right channel delay comp correctly and the left not)

[BScout]

I don't think anyone has seen this bug before because it is really odd (with tons of uncertainty) to route a stereo channel to a mono bus on a send.