Delay compensation nightmare

[TurboD]

I'm using PT9 on a Mac Mini i7, 16GB RAM, SSD with an Apogee Ensemble as the interface. I have some outboard gear I like and I have them rigged up as hardware inserts for use in Pro Tools.

I started a mix yesterday using my Compex on a drum parallel group, with signal flow like this:

All drum channels output 'bus 1-2' and come up on an auxiliary track 'KitBuss'
All drum channels have post-fade FMP sends to 'bus 3-4'
A second auxiliary track called 'KitParallel' has input of 'bus 3-4' and has the hardware insert across it.

When I was mixing away yesterday, everything was delay-compensated perfectly and rocking away perfectly in phase, no comb filtering. I open up the exact same session today, everything still plugged in the same way, and massive comb filtering across the drums. Very embarrassing when a client comes to listen.

Does anybody have a fix for this? I know there's the H/W Insert Delay pane in the I/O settings, but being in ms, it's not fine enough and if the problem is inconsistent it won't help anyway.

Please help!

[albee1952]

You need to manually enter the time offset for hardware inserts(in the IO setup menu) Also, sometimes the delay comp gets confused and closing the session can fix it.

A few more random thoughts, are any tracks showing red in their delay display? Are any tracks in record-ready? Is Delay Comp. turned on(checked) under the Options menu?

[TurboD]

Hi Albee, thanks for your reply.

As I mentioned, I know about the manual adjustment in the H/W Insert Delay panel, but it's no good for two reasons:

1. It doesn't explain why it was perfectly fine to begin with, and then not the next time

2. The adjustment is in ms, which is way too coarse an adjustment... why not samples?

Delay compensation is on and it's long (4096 samples) and nothing's in the red.

[TurboD]

Aha - problem fixed. I'll leave this here in case anybody has the same problem:

It seems that the Apogee doesn't report round-trip latency accurately when set to full 18x18 I/O bandwidth. Going back down to 10x10 I/O (i.e. only analogue, no ADAT) set it straight.