Logic 9 / Apogee Intermittent Latency / recording delay

[30,000 Days]

Having followed several threads on this Forum and others including this one:

 

http://www.logicprohelp.com/viewtopic.php?t=72319&highlight=logic++recording+delay

 

I am still totally confused about what seems to be an intermittent recording delay in logic 9. I'm using an Apogee Rosetta 800 over firewire, L Pro 9.1.3 , Snow Leopard, Mac Pro Intel 2x quad 2.93 with 11gb RAM, UAD- 2 Quad PCIe. I've been using digital audio for many years, and on this system for the first time I have an intermittent recording delay, which happens on some projects and then seems to be fine on others.

 

The weird thing is, that when it happens none of the advice anyone gives me, or any of my own rationale, will solve the problem. Low latency mode, turning off SDC, lowering buffer, software monitoring off, the whole lot, and it's still there. The most infuriating thing is that on my system, altering the recording delay seems to change nothing. I have carried out rigorous tests to determine the recording delay, which runs at about 73 samples, and so set it at -73 in the hope that everything will land in the right place- but nothing changes.

 

I have the delay on totally fresh projects with no tracks, plug-ins or instruments running, and at other times I can have a fully loaded track running UAD plug ins and have no problem at all.

 

I have considered upgrading to a PCIe card for the apogee, but intuition and spme outside advice have suggested that this might not solve the problem, even if it slightly improves it.

 

The Rosetta driver is V 2.5.2fc1 firm ware v 3.8.

 

Having run far cheaper MOTU systems for years, I am surprised to have this happening with an apogee card, but maybe the apogee is not the source of the problem.

 

Does anyone have any suggestions to where to head next?

[ski]

From what I've read, Apogee systems report their latency to Logic, and thus I imagine (keyword: imagine) that setting your recording delay to any value at all would be overridden by the communication going on between the Apogee driver and Logic, behind the scenes, invisible to the user. I'm speculating along these lines because on those systems which report latency to Logic, a recording delay setting of zero results in perfect placement of audio recordings in Logic. If you compare this to systems which don't report latency to Logic, a recording delay setting of zero usually results in audio being placed late, warranting a negative value be set for delay compensation. My conclusion, therefore, is that a rec. delay parameter itself with respect to the Apogee system is meaningless, and that any value would be ignored.

 

Now... on a less imaginary front, on my RME system (which reports latency to Logic), there's one situation where I have to set the recording delay to a non-zero value, and that's when I'm digitally looping back audio from the RME itself. If you're not familiar with RME stuff, they have a mixer application called TotalMix. In short, all of Logic's outputs appear on this mixer. If I'm using multiple outs from Logic, I can digitally mix them down to a single pair of virtual outputs on this mixer and loop them back into Logic to record a 2-mix. When I do this the looped back audio ends up ahead unless I set the recording delay to +75 samples.

 

"So where is this all going?" he asked. Well, I'll tell ya... is there any chance that the way you're routing audio into Logic that it's taking a circuitous path (as is my loopback situation) rather than a direct path (which would be latency compensated)?

[30,000 Days]

Interesting. I just go into the Rosetta ( via Preamp abviously) which goes straight into Core Audio via the driver, I presume. I do use the Apogee low-latency monitoring software, but didn't think that it would cause any kind of hindrance to the input speed....?

 

Can't think of any other way anything would be slowing it down...

[ski]

I'm not familiar with the Rosetta, so, a question... is this connected to your system via FireWire? If not, what's the connection? (Based on your answer, I might have a hunch...)

[30,000 Days]

Yes, It's connected by firewire.

[ski]

OK, here's the next theory... Per research I did on this a while ago, I learned that FireWire audio drivers have a built-in 64 sample "safety buffer". I'm now wondering/theorizing if the Apogee driver is overcompensating for that. I'm suspecting this because the amount you calculated for latency (73) is very close to the safety buffer size. And I've read numerous reports that in some cases it's possible for audio systems to place audio ahead even though they're (supposedly) reporting the correct delay amount to Logic.

 

You said you did extensive testing, but FWIW did you do a hardware loopback test (patching the outputs to the inputs per the test outlined in this link)?

[30,000 Days]

Yes, I followed it to the letter, a few times. Sometimes I got a delay, sometimes not.

[minus1]

I've had the same issue on certain projects as well. Never found a solution, my work around was to make a bounce of the project and then track in a new session. And you are right unfortunately pci doesn't help. Luckily it's been a while since this last happened.

[playagibson]

Make sure all plugins in the output channel are bypassed, while recording.

I made that mistake once. I checked everything else before I discovered that's a no no while recording. It stopped the latency immediately.

[redlogic]

Make sure all plugins in the output channel are bypassed, while recording.

I made that mistake once. I checked everything else before I discovered that's a no no while recording. It stopped the latency immediately.

Even this is not foolproof.

 

If Plug-in Delay Compensation is set to All, bypassing plug-ins on busses, auxes, and outputs will not eliminate the latency that they create. You must actually remove these plug-ins from the Insert slots to eliminate latency.

 

 

If Plug-in Delay Compensation is set to Audio and Software Instrument Tracks, bypassing plug-ins on busses, auxes, and outputs will eliminate the latency that they create.

[playagibson]

Make sure all plugins in the output channel are bypassed, while recording.

I made that mistake once. I checked everything else before I discovered that's a no no while recording. It stopped the latency immediately.

Even this is not foolproof.

 

If Plug-in Delay Compensation is set to All, bypassing plug-ins on busses, auxes, and outputs will not eliminate the latency that they create. You must actually remove these plug-ins from the Insert slots to eliminate latency.

 

 

If Plug-in Delay Compensation is set to Audio and Software Instrument Tracks, bypassing plug-ins on busses, auxes, and outputs will eliminate the latency that they create.

 

Cool !

I just learnt something new !

[ski]

Yes, I followed it to the letter, a few times. Sometimes I got a delay, sometimes not.

 

The delay shouldn't be variable as you described unless.... the drivers themselves are flaky (or should I say, it's not unknown that some FW audio drivers will exhibit different amounts of latency upon each reboot of a computer).

 

I've never been a fan of FW-based audio systems even though I'm running one myself (RME Fireface 800). I don't get variable latency because the RME drivers are rock solid (except for when they cause my audio to stutter, and one day soon that little problem is going to be solved by my dropping this unit off of my second story balcony and into the gullet of a giant carnivorous plant which I'm growing just for the occasion. Anyway...)

 

PCI-e based systems are, in my experience, the most reliable. For years I worked with a MOTU 2408 (connected with a PCI-e424 card) and the most variation I ever had was one sample! It was either -27 or -28, and I'd regularly do tests on it to be sure of where things were at. So I guess all of this blather is to say that if a PCI-e based audio system is something you can consider then it might be the very thing that saves your sanity so that you don't have to second-guess whether what you played is in time or not.

 

Other than that, the only thing I can suggest is that before you start any given session that you do a loopback test using a completely blank song with no plugins and with PDC turned off. Tabula rasa and all that. Calibrate your recording delay, and then start making music.