UAD-"What Latency?" + other PT-LE latency results

[BradG]

So I've been demoing a couple of UAD plugs on an old Mackie card that somebody gave me to see if I should buy some. I do really love them and as I was using an 1176 on a final-mix vocal track, I thought I'd figure the latency and adjust for it (even though the track sounded fine to me). What I got when bouncing the track to another one for reference, was barely any latency at all - well under 200 samples, it looked like.

So I decided to do a little more thorough testing along with some of my other plug-ins and some Digi plug-ins that I was also demoing. Here are the results:

http://soundwag.com/plugin_delay.jpg

To do this test, I first recorded a click track on an audio track. I set the output of the CLICK track to bus 10 and all the other audio tracks inputs to bus 10. Then I engaged each of these plug-ins, one at a time: Digidesign 1-Band EQ, Bomb Factory PEQ-1A, Waves LinEQ, Waves RenEQ6, Digidesign Smack, UAD 1176LN, UAD Fairchild (mono), Amplitube, Antares MicMod, Waves RenAx, Digidesign 002R analog output to analog input, Track2 analog output to analog input (via ADAT), Digidesign 002R S/PDIF out/in thru T.C. Electronics M2000 set to De-Ess & Compressor.

I certainly did not expect these results. As you can see, most of the regular Waves & Digi plugs have no latency (maybe 1 or 2 samples at most), although Waves LinEQ was the big winner here with a whopping 11295 samples of latency. Both UAD plugs only returned 115 samples of latency. Strangest of all, going out of the 002R's analog OUT and back in thru one of it's INPUTS gave me nearly NO latency, although even stranger was a NEGATIVE LATENCY of 48 samples for the S/PDIF out/in through my M2000. This last one I actually repeated (and no, that's not a whole click's worth of delay - the very first returning click's waveform is actually cut off!)

Hardware specs: G5 2.5Ghz with 2.5GB RAM running OS 10.4.2, ProTools LE software version 6.9.2, Digidesign 002R clocked to Apogee Track2; ProTools' Hardware buffer = 512/ CPU = 99%

NOTE: while I have decided to buy some of the UAD compressor plug-ins, I was very disturbed by the amount of processor power that they seem to hog. Obviously these are wrapped-VST plugs, but they appeared to be stealing as much CPU as my freakin' Waves LinEQ (real hog). What is up with that? I wish they could just make these puppies native.

[norman_nomad]

Originally Posted by BradG

although even stranger was a NEGATIVE LATENCY of 48 samples for the S/PDIF out/in through my M2000. This last one I actually repeated (and no, that's not a whole click's worth of delay - the very first returning click's waveform is actually cut off!)

Negative Latency? So the sound is recorded before it is even sent? Wow... I would sell that 002r to the space program.

[THE GEEK]

Originally Posted by BradG

NOTE: while I have decided to buy some of the UAD compressor plug-ins, I was very disturbed by the amount of processor power that they seem to hog. Obviously these are wrapped-VST plugs, but they appeared to be stealing as much CPU as my freakin' Waves LinEQ (real hog). What is up with that? I wish they could just make these puppies native.

I dont have such a problem.
Id rather blame it on your LE Setup than on UAD-1



....and yeah, your negative latency is well interesting! If you could advance that I could tell my musicians: "You dont need to come, its already been recorded!"

[David Herbert]

Perhaps it was a Freudian slip, just his latent negativity surfacing. Just kidding. Thanks for your hard work. Doc

[max cooper]

Originally Posted by norman_nomad

Negative Latency? So the sound is recorded before it is even sent? Wow... I would sell that 002r to the space program.

What version of PTLE is that? I know there was an incorrect value for some internal delay compensation; sounds like something like that. I know it's fixed by now, but I don't remember when it was fixed. I believe it was an issue as recently as 6.4, and I should say that I'm not sure we're even talking about the same thing, although the only way I can see having neg. latency is from some kind of delay comp.

But I'm a little confused. I run a UAD card with PTLE (6.9.2) and the UAD latency is about a day and a half.

[THE GEEK]

Originally Posted by max cooper

I run a UAD card with PTLE (6.9.2) and the UAD latency is about a day and a half.

That is true though! I usually hit the Play Button, then go cook some dinner, have the dinner and then PT starts running. Problem is: I'm getting FAT!

[norman_nomad]

Originally Posted by max cooper

But I'm a little confused. I run a UAD card with PTLE (6.9.2) and the UAD latency is about a day and a half.

If you leave your twin and travel at the speed of UAD for 2 years, when you come back he will be a 2" tape machine.

[aussie_techie]

Originally Posted by BradG

Strangest of all, going out of the 002R's analog OUT and back in thru one of it's INPUTS gave me nearly NO latency, although even stranger was a NEGATIVE LATENCY of 48 samples for the S/PDIF out/in through my M2000.

simple, in 6.9 PT compensates for the converter latency, so going in and out is compensated for resulting in 0 delay on the recorded track. the converter in the M2000 must be faster than that in the 002, PT is compensating for the value of the 002 converter and moving the track back to far for the M2000 converter

i believe it was only 6.7 that had the wrong converter value

[Kestral]

I read on the Digidesign forums that Pro Tools LE actually does have delay compensation. The only catch is, it has compensation UP TO the buffer size that you have set.

So if you have buffer set to 64 samples, then it will compensate as much as 64 samples.

If you move it up to 2048 samples (that's the maxmium at least on my Pro Tools M-Powered w/Firewire Audiophile interface), then it should compensate for as much as 2048 samples of latency.

I haven't tested it out yet as my studio has been torn apart so I can rack everything in nice neat racks instead of having it strewn out on a table but if you could try it do this:

1. On the menu bar, select "Setup -> Playback Engine..."

2. Set the H/W buffer size to 2048 (or whatever is the highest on your LE). Then run the test. Looks like that Waves LinEQ plug won't be compensated for, even with 2048 samples, but the other ones should be.

[norman_nomad]

As far as I know, PTLE does NOT compensate for any kind of latency. But I'm still working on 6.4... so maybe it's changed.

An easy way to test.

1. Import a wav file
2. Duplicate that file
3. Put a plug with latency on ONE of the tracks
4. Check the sample delay on that track by ctrl clicking the volume value
5. Nudge the non-plug-in track forward by the sample value delay of the plug-in track
6. Invert one of the tracks and check for null

[jeronimo]

I never heard/read anything about PT LE 6.9 having delay compensation....

[max cooper]

PTLE has a compensation, not for plugins, but for it's own converters, I believe. There was a thread here about it, but I can't find it at the moment.

There was an issue in one version of PTLE where the internal delay comp was incorrect, but it's been fixed as of (I think) 6.7.

[usamike]

Here's a reprint of a post I made in 'music computers' a month or so ago.

So I did some tests, recording a click to an audio track, adding different plugins that I use alot, then recording the processed signal to another track, as suggested above. All tests done at 24 bit, 44.1 K. Here are the delays in samples.

Plug-in Latency:

Renaissance Channel: 65 samples @44.1K
Ren comp: 64 samples @ 44.1K
C4: 69 samples @ 44.1
Renn De-ess: 64
Amplitube: 64


BF 1176 and Renaissance EQ4 gave no latency, doesn't seem possible but it explains why when I bus drums out to an aux channel w/compression, Renaissance Compressor gives me horrible phasing but BF76 sounds great. So next mix, I will try nudging back audio as per these tests and see what happens. Comments?

[BradG]

Yeah usamike, those figures are what I'm seeing displayed using option-click (Mac). However, I don't trust them, although I haven't yet done a full test for null using them. But I feel safe in saying that it wouldn't work as I can see with my eyes that the LinEQ is being delayed 11295 samples, yet PT is displaying a delay of 2679.

PT also displays 0 delay for both of the UAD plugs. There is a very tiny delay of only 129 samples on these plugs (I re-measured the offset since posting that JPEG), it is certainly not 0 (zero). Yes, this will cause phasing on a send, but there is no way that you are going to notice timing problems on an insert with this latency.

So it's pretty obvious that there is some kind of compensation happening here, at least as far as the converters go. The most obvious example of this is on the S/PDIF where ProTools seems to be compensating for latency that never occurs, thus the negative offset of that returning track.

One useful product of this test is that I will continue to use the 002R's out/ins for my external hardware effects. I had been planning on switching over to the Apogee (I've got the 8D/A card in my Trak2), but I think I'll stick with the 0 latency on the 002R.

As for my small UAD latency, what can I say? I guess I lucked out, although, as I said, it looks like I am paying for it with CPU power. it would be cool if other UAD users out there could try a similiar test, at the very least, let us know what your software is telling you their delay is.

[cap217]

what HW buffer did you wrap the VST's to? If you set it really low then you will have no delay (very little) but it will take up a SHIT load of power!

[BradG]

Originally Posted by cap217

what HW buffer did you wrap the VST's to? If you set it really low then you will have no delay (very little) but it will take up a SHIT load of power!

I have no idea how I first wrapped those, using VST-RTAS v1.0 but you're obviously correct as they used to give me a huge CPU hit.

Since I first made this post, I've upgraded to PT 7 and VST-RTAS v2 and I am now getting the crazy latency that is making my wonderful UAD chorus plug-ins COMPLETELY USELESS.

Where/at what point in the wrapping process do you choose the "HW Buffer" size? Can I re-wrap just the chorus plugs with a lower HW buffer?

[kats]

What I do t make it ez is just set up 4 aux tracks and use the plugs on the busses. htis way I don't have to piss around too much with the delay comp plugs. Although for the most part I only usually use the plate 140.

And while on the subject, since I mostly use outboard eq and comp I just measure the delay once (in my case it's 2550 samples) and enter that into my nudge field. So if I want to eq my kick for eg, I just have to highlite it - hit the KP- and insert the outboard and viola. It's always the same delay as long as you stick to one buffer size. It's an extra step, but that's digi for ya.