How to adjust record latency with ProFire 610?

[Avi Bortnick]

Hello.
I've been testing out a ProFire 610 unit, trying to set it up for optimal record latency. I'm routing an audio file of a click out through the headphone jack, and back into input 1 on the Profire.

I can see that
a) there is latency with the recorded file (the recorded waveform doesn't quite match up with the play waveform) and
b) this latency changes depending on where I set my Playback Engine H/W Buffer Size (smaller values make the recorded track later in relation to the playback track. So far the best is 128 samples, but the recorded audio is still a bit early), and
c) When I set the h/w buffer back to a given number of samples, the latency changes from before (128 samples doesn't look the same after I set it to 256 samples, and then go back to 128 samples)

How do I set everything so that input and output line up perfectly, or at least within a few samples? (Actually, if I open the session with buffer set to 128 samples, and don't change anything, the newly recorded waveform is 10 samples too early...less than a millisecond, but still...)

My other Playback Engine settings are:
Host Processors: 2 Processors
CPU Usage Limit: 85%
Host Engine: unchecked Ignore Errors

DAE Playback Buffer: Size: Level 2, Cache: Normal
Plugin Playback Streaming Buffer: Level 2

My computer is an iMac running Snow Leopard, latest versions of everything.

Thanks Much!

[Avi Bortnick]

In snooping around, this shifting and random latency is a long-described problem with the M-Audio ProFire series. You can read all the steaming posts at the M-Audio forum:
http://forums.m-audio.com/showthread...g-Not-accurate

A real bummer for those who care about accurate input and output, without timing and phase issues. It's remarkable that M-Audio has not addressed this issue with a (supposedly) mature product.

[necjamc]

oh damn, I have never noticed any problems. But now I have tests to do tonight. UGH

[necjamc]

In testing several times, I recorded a click track out of the 2626 and back in. I set my buffer to 128 samples. It came back 410 samples late. But in a thread I started asking why it was explained here.

Originally Posted by cdavis6406

You answered yourself.
D/A conversion takes 128 one the first output
A/D conversion takes 128 on the first input back in the box to prtls.
and so forth.

Multiply your hardware delay setting by how many times your the audio leaves/returns to the box.

Add in any plugins or external hardware delay there may be.
You get the idea.

But I did this several times and got no variation.

[Avi Bortnick]

Originally Posted by necjamc

In testing several times, I recorded a click track out of the 2626 and back in. I set my buffer to 128 samples. It came back 410 samples late. But in a thread I started asking why it was explained here.

Did you try closing PT and reopening the session and trying again? Or changing the H/W buffer setting and testing again?

In any case, 410 samples should be unacceptable, and shouldn't be normal, unless you've got delay-inducing plug ins. This is almost 10 ms late (presuming 44.1k sample rate), and would sound flammy, and would make rhythmically tight overdubs difficult.

My latency is not so bad - only about 1 ms - and not really a deal breaker for recording (though it makes impulse response-based acoustical analysis impossible, but that's another story...)

All software that I know of (Live, Reaper) either automatically gets latency values from the interface and automatically adjusts, so that recorded audio lines up with played audio, and/or has a manual setting to adjust the value. I would think Pro Tools does this automatically, because so far I'm not aware of away to manually compensate, and because having timing lags this long is bad.

[necjamc]

That would be ADC Not in PT. But I did try several settings, obviously at 256 samples the latency returned was much higher. But that's what that means anyway. I was also recording at 24/96, I will try at 44.1 and let you know how the 2626 does.

[Avi Bortnick]

Originally Posted by necjamc

That would be ADC Not in PT. But I did try several settings, obviously at 256 samples the latency returned was much higher. But that's what that means anyway. I was also recording at 24/96, I will try at 44.1 and let you know how the 2626 does.

Well not really, there's two kinds of latency: the one you hear when routing your guitar/mic/keyboard through your interface, through the software, and back to your speakers/headphones. This IS dependent on buffer setting, with lower values giving lower latency.

The other kind of "latency" - which maybe should be called something else - is the timing of newly recorded audio in relation to existing audio. Properly, waveforms should line up no matter the buffer setting, input/output latency. Other software works like this, and so did Pro Tools when I use my Fast Track Pro.

The problem with the ProFire is not just related to ProTools. Similar timing errors occur with Reaper and Live - you just can't find an offset (in the preferences settings) that allows the newly recorded audio to line up with the pre-recorded audio. Again, this problem doesn't exist with my Fast Track Pro. And this issue is well-documented on the M-audio forum, and acknowledged by M-audio techies.

[necjamc]

Oh, ok well when testing I am routin out output 4 (physical output) hence the buffer laltency , then back in. When I bus internal it's perfect. I am going to run some tests right now. But if you are sending the signal out your headphone jack there will be latency coming back in. But it should never change. Unless you add something or change something. Right? I mean you are recording what you are hearing and if you have the buffer at 128 there is latency on the headphone output. And then you will record that latency.

[necjamc]

ok retest. I used audio this time, not the click track. At 64, 128 samples respectively, No latency and no changes. This is 3 or 4 tests. My latency was amplified by the click latency. So this time no prob. I will try more tests to check at different settings.

Edit: It was 10 samples of Latency. I wasn't zoomed in where I thought. I will try this test again after closing and restarting PT and changing sample rates

[necjamc]

Ok I screwed up my first test. I forgot to mute the click track and it recorded both. Retested. The recorded data from outside was 3 samples early. Several times including closing the session. but def. early (I didn't expect that.)