Plug-in Latency Question

[muxlow]

Preface: I did search for this first; lots of stuff on latency, but I didn't see anything specific to this.

Just did a little experiment and confirmed that audio going through an aux track with Custom 59 + Live Room B is coming out ever so slightly delayed.

I understand plug-ins can cause latency, but isn't that what Setup > Configure Audio System > Configure Studio Settings > Automatic Plug-in Latency Compensation is for?

Thanks for any light you can shine on this

[muxlow]

Ah, this bit from the manual is interesting:

"For example, if an effect is instantiated on an aux track or instrument track, it always runs in real time."

[bayswater]

Ah, this bit from the manual is interesting:

"For example, if an effect is instantiated on an aux track or instrument track, it always runs in real time."

That means it doesn't use the pre-generation mode by which DP calculates the effect in advance to spread out the CPU load. Delay compensation should still be working.

[daniel.sneed]

[...]Just did a little experiment and confirmed that audio going through an aux track with Custom 59 + Live Room B is coming out ever so slightly delayed.[...]

Hum, hum, hum... just in case: AFAIU, DP Delay Compensation is intended to compensate plugin calculation delay in previously RECORDED tracks, while playing back these tracks.
That is to say, there's no such thing when it comes to hardware input audio. In hardware realtime audio, any plug will induce delay.
BTW, of course, same applies to realtime playing a VI with a controller. In the VI itself and in any plugin on the VI track.

[muxlow]

So running a plug-in real-time vs. pre-gen has no impact on plug-in delay compensation. Interesting.

I just did some experimenting with a MIDI instrument in a V-Rack:

EXPERIMENT 1:
Chunk MIDI track -> V-Rack Aux track with EZ Keys -> Chunk audio track (recording) –

Result: No latency (well, 4 samples of latency)

EXPERIMENT 2 (trying to make latency):
Chunk MIDI track -> V-Rack Aux track with EZ Keys, Custom 59 +Live Room G -> Chunk audio track (recording)

Result: Big latency on first pass, no latency on subsequent passes

So aside from the first pass weirdness, it seems okay.

I also noticed on my DI experiments (DI -> Vrack amp model) that an early recorded pass had significant latency, but subsequent recorded passes did not.

[muxlow]

[...]AFAIU, DP Delay Compensation is intended to compensate plugin calculation delay in previously RECORDED tracks, while playing back these tracks.
That is to say, there's no such thing when it comes to hardware input audio.

That's how I understand it as well.

I'm running tracks of recorded DI to various amp models that I keep in a V-Rack (to maintain consistency across songs), then routing those out to a Bounce track back in the song chunk (stereo audio with the Input Enabled always on).

Works great, except every once in a while I feel that it's not "right". Hence me digging into latency, which probably isn't the problem.

But now I just want to understand how it works

[Don T]

Hello,
I have my students test latency for all plug-ins, the results are mixed.
We test latency by bouncing one prerecorded track to another through the plug-in. It should show zero latency when compensation is turned on BUT that is not always true. Latency cycle count is provided by the plug-in programmer and it is not always correct. We've had some plug-ins be late with compensation off and early when compensation is turned on OR still late just not as late. Pitch correction software like autotune do not correct latency because it is variable depending on how far off the pitch is plus it cross fades between corrected and bypassed sections with notable cross-fade time artifacts.
IN real time tracking mode there is always latency because of buffering. A/D & D/A conversion added to hard drive (or not) plus any data communication protocol conversion buffering (FW, USB, SPDIF, AES).

Only way to know is to test them. Sometimes a plug-in will have latency that disappears when it's re-scanned so errors in the preferences are possible.

A lot of the bundled plug-ins (MAS) have zero latency even when compensation is turned off. Same with buses, the latency is known and therefore always compensated for.

Also, is the LiveRoom plug-in adding accurate latency for mic position?

Don

P.S. we wont be running tests on DP9 until the school pays for the upgrade.

[muxlow]

Hello,
I have my students test latency for all plug-ins, the results are mixed.
We test latency by bouncing one prerecorded track to another through the plug-in. It should show zero latency when compensation is turned on BUT that is not always true. Latency cycle count is provided by the plug-in programmer and it is not always correct. We've had some plug-ins be late with compensation off and early when compensation is turned on OR still late just not as late. Pitch correction software like autotune do not correct latency because it is variable depending on how far off the pitch is plus it cross fades between corrected and bypassed sections with notable cross-fade time artifacts.
IN real time tracking mode there is always latency because of buffering. A/D & D/A conversion added to hard drive (or not) plus any data communication protocol conversion buffering (FW, USB, SPDIF, AES).

Only way to know is to test them. Sometimes a plug-in will have latency that disappears when it's re-scanned so errors in the preferences are possible.

A lot of the bundled plug-ins (MAS) have zero latency even when compensation is turned off. Same with buses, the latency is known and therefore always compensated for.

Very interesting. How do you handle latency once you know it exists? Nudge everything recorded through the plug-in back X samples? That seems both obvious and horribly tedious.

Also, is the LiveRoom plug-in adding accurate latency for mic position?

Now THAT is a distinct possibility. I'll try testing that; it has close & far "mics"...maybe there'll be a slight difference.

[Don T]

Hello,
I just keep mindful of how I'm using the plugin. If phase accuracy is paramount I'll either not use the plug or I'll instantiate it on the related track and adjust it not to do anything. And yes if I am sure of my plug choice since the latency is known, I can nudge by that amount. A few samples doesn't make too much difference in a vox track. Drum overheads are another matter entirely.