Daniel Lanois says...

[mikymike]

"What I like about analog is I can take a sound from a track, treat it with eq and an echo and print that back onto the multitrack without a conversion delay. This is a technique that I use a lot. Digital machines can't do this. Your reprint will be out of phase with the original source."
I do this all the time in Reaper without phase problems. Reaper has automatic hardware insert PDC, its called reinsert. Am I missing something here?

[warhead]

Nuendo has excellent delay compensation, almost every mix I end up sending my snare top / bottom through a single fader to the SSL Mixpander mixer, EQ'ing them together, and reprinting to Nuendo. There isn't one bit of latency, but a few years ago delay compensation wasn't available on many DAW's.

Now it is.

My hardware inserts I simply click to calculate the latency once on that channel, done. Save the setting, set for life.

War

[mikymike]

Originally Posted by warhead

Nuendo has excellent delay compensation, almost every mix I end up sending my snare top / bottom through a single fader to the SSL Mixpander mixer, EQ'ing them together, and reprinting to Nuendo. There isn't one bit of latency, but a few years ago delay compensation wasn't available on many DAW's.

Now it is.

My hardware inserts I simply click to calculate the latency once on that channel, done. Save the setting, set for life.

War

Same w/ Reaper. Maybe DL should be brought up to speed / daw's!!!

[Kiwiburger]

I wouldn't trust any auto delay compensation. The basic principle is that the audio interface reports the latency back to the DAW, which then offsets the audio track. But how certain can we be that the audio interface is reporting the actual delay? It's probably a guess - but software drivers change and may have hidden bugs.

But my real issue is that I use an assortment of various A/D and D/A, attached to my interface. Each external A/D and D/A has a different latency of it's own. I somehow doubt the each unit can report it's unique latency back through the chain, and I doubt the DAW can really figure out what the total latency is. Especially with a diverse mix of units from various makers.

So my opinion is forget auto latency systems. Just do your own experiments at different sample rates, and work out the amount of offset you really need to get perfect nulling of an OTB processed track. Then just manually apply that offset each time.

[Dirty Halo]

Yeah... and who's this Danny guy anywhooo?!!

(tongue firmly in cheeck)

-andrews

[mikymike]

Originally Posted by Kiwiburger

I wouldn't trust any auto delay compensation. The basic principle is that the audio interface reports the latency back to the DAW, which then offsets the audio track. But how certain can we be that the audio interface is reporting the actual delay? It's probably a guess - but software drivers change and may have hidden bugs.

But my real issue is that I use an assortment of various A/D and D/A, attached to my interface. Each external A/D and D/A has a different latency of it's own. I somehow doubt the each unit can report it's unique latency back through the chain, and I doubt the DAW can really figure out what the total latency is. Especially with a diverse mix of units from various makers.

So my opinion is forget auto latency systems. Just do your own experiments at different sample rates, and work out the amount of offset you really need to get perfect nulling of an OTB processed track. Then just manually apply that offset each time.

There is NO guessing, 100% accurate. The software sends a signal through the bus chain, d/a - sig. proc. - a/d, and simply measures the roundtrip time and compensates accordingly, really simple actually. There ya go, your up to speed!!!

[clip6]

I didn't get what he said about that either. He's been heavy on the Rader for years. I've never used the Rader but I'm sure you can print back on itself.

The post are being filtered so I guess just answers to questions and not disscusion.

[warhead]

I think we should make sure this thread isn't directed at Daniel Lanois, I didn't really think this thread was about him but rather the question of latency in a DAW.

On auto latency, all I know is I send stuff out (from original source files in Nuendo4) and I record back in and when I zoom in on waveforms they line up bit for bit. I always check and so far, so good.

War

[thethrillfactor]

Originally Posted by Kiwiburger

and work out the amount of offset you really need to get perfect nulling of an OTB processed track.

It will never happen.

[u b k]

Originally Posted by thethrillfactor

It will never happen.

not ever. the best you can do is take the original track and run it thru the same chain with the outboard bypassed, and use this dry bounce to replace your original.

so then you get the nice time alignment, with the conversion artifacts.

choose your poison .


gregoire
del
ubk
.

[warhead]

Even if you time align them and attempt to null, one is processed so you're going to hear the difference.

War

[mikymike]

Originally Posted by warhead

I think we should make sure this thread isn't directed at Daniel Lanois, I didn't really think this thread was about him but rather the question of latency in a DAW.

On auto latency, all I know is I send stuff out (from original source files in Nuendo4) and I record back in and when I zoom in on waveforms they line up bit for bit. I always check and so far, so good.

War

Agreed

[Drumsound]

RADAR (which DL uses) is not a DAW, but a digital recorder and editor. If you send something out of it and back into it there is latency.

[mikymike]

Originally Posted by Drumsound

RADAR (which DL uses) is not a DAW, but a digital recorder and editor. If you send something out of it and back into it there is latency.

That makes sense

[thethrillfactor]

Originally Posted by warhead

Even if you time align them and attempt to null, one is processed so you're going to hear the difference.

War

If it were that simple it would be easy.

Even if you just patched the D/A out to the A/D input with nothing in between the problem is that the latency is not a round number.

Also between each track there is a different latency. Out of 1 does not have the same latency as out of 2 ...etc.

But on an analog tape machine this happens as well.

The difference is its a lower number.

[TOFT]

Cubase -> fireface output 8 to a analog eq / output of the eq to input 8 (standard setup here). If I bypass the eq and record it, flip the phase on one channel, all is gone... except a tiny something if you turn things up really loud.

So Cubase is I dare say: perfect for this type of thing.

[thethrillfactor]

Originally Posted by TOFT

Cubase -> fireface output 8 to a analog eq / output of the eq to input 8 (standard setup here). If I bypass the eq and record it, flip the phase on one channel, all is gone... except a tiny something if you turn things up really loud.

So Cubase is I dare say: perfect for this type of thing.

That tiny something at 44.1khz is a 6db boost in the lows and -3db drop at 20khz.

This is more than a tiny something don't you think?

[TOFT]

Originally Posted by thethrillfactor

That tiny something at 44.1khz is a 6db boost in the lows and -3db drop at 20khz.

This is more than a tiny something don't you think?

How is this measured? 6db boost what? Is this a RME thing, a Cubase thing?

Still the EQ'ed track that comes back sounds great on my productions. Usually better then the origional. :) No problems here with delay or 6 bd boosts of anything. (I think :) )

[anguswoodhead]

Originally Posted by TOFT

So Cubase is I dare say: perfect for this type of thing.

Long live Cubase - who needs Tools anyway - hee hee

[thethrillfactor]

Originally Posted by TOFT

How is this measured? 6db boost what? Is this a RME thing, a Cubase thing?

Still the EQ'ed track that comes back sounds great on my productions. Usually better then the origional. :) No problems here with delay or 6 bd boosts of anything. (I think :) )

In the best case scenario its a one sample delay ala PTools latency correction with their 192 converters.

The problem is 2 fold:

1) If you mix 2 of the same tracks together 1 sample apart you will get a peaks and nulls at different freqs. at 44.1Khz its a 6db boost in the lows, a null at 10khz and -3.75db cut at 20khz. Not exactly what one would want to happen to their audio right?

2) The problems are added on when you do a round trip on a converter. Remember most converters are designed to take the picture once and keep it there until the very end when you want to play it back. When you roundtrip your audio in and out their is cumultive effect on the freq's themselves. All converters are designed differently. Most middle of the road converters take a -3db hit in the hi's on the first trip which in itself depending on the source may or may not be a problem. If done a couple of times though you will hear it because the hi's will become less and less until the audio starts to sound dull. That's when most people pull out the EQ's and start to add the hi's that are gone to begin with which starts to make the tracks edgy and noisy and then you get the drift.

[Kiwiburger]

There is no excuse for allowing your track to be out by 1 sample. Simply nudge it back or forward by 1 sample. Done.

Surely the worst case scenario with a DAW would to be out by half a sample. That would be uncorrectable. Unless you doubled the sampling rate. So this is probably a good argument for using high sample rates.

I don't get that it's a real problem. Who cares if it's out by half a sample. Only if you are trying to do parallel compression or saturation would it be an issue. And IMO the phase shifts through the analog gear is a bigger issue, and harder to correct.

The example giving of wanting to bounce a track OTB for eq and delay doesn't seem insurmountable. Unless it's a drum track, or something with lots of leakage, where maintaining phase integrity will be a problem. But the phase shift in the analog box is still there.

Latency is easy to understand and easy to correct.

[mikymike]

Originally Posted by Kiwiburger

There is no excuse for allowing your track to be out by 1 sample. Simply nudge it back or forward by 1 sample. Done.

Surely the worst case scenario with a DAW would to be out by half a sample. That would be uncorrectable. Unless you doubled the sampling rate. So this is probably a good argument for using high sample rates.

I don't get that it's a real problem. Who cares if it's out by half a sample. Only if you are trying to do parallel compression or saturation would it be an issue. And IMO the phase shifts through the analog gear is a bigger issue, and harder to correct.

The example giving of wanting to bounce a track OTB for eq and delay doesn't seem insurmountable. Unless it's a drum track, or something with lots of leakage, where maintaining phase integrity will be a problem. But the phase shift in the analog box is still there.

Latency is easy to understand and easy to correct.

100% agreed. All I know is I bounce otb quite often, drum bus included, and w/ auto pdc I have no discernable phase shifting. If this can't be done w/ radar, Id say dump radar for a daw, you can still use radar converters!!!

[Volodia]

I'm surprised by those statements . I mix ITB with an analog insert on the mix bus . I mix to a track in Pro Tools . When I want to check my recall I play the tracks and the mix(out of phase) and they null ( apart from the non linear stuff like reverb and modulations) .

It works perfectly . I use it on a daily basis almost .

[MickeySmid]

Originally Posted by Volodia

I'm surprised by those statements . I mix ITB with an analog insert on the mix bus . I mix to a track in Pro Tools . When I want to check my recall I play the tracks and the mix(out of phase) and they null ( apart from the non linear stuff like reverb and modulations) .

It works perfectly . I use it on a daily basis almost .

Me too... 100% in cubase.

I think with 2 tape machienes the difference will be bigger, cause it's motors running :) (but no ad da conversion, so that is way cooler)

[jarbar]

I use parallel compression with outboard comps with CuBase and an Emu 1820m. The new compressed track is usually slightly out of phase (though less than a sample it seems), but sometimes it just makes it sound thicker. If it's adding a bit of weirdness to the highs, a few nudges and it's usually fine. And the track's so compressed that I don't notice (or care about) added lows or reduced highs or whatever the converter's doing.

I also do a lot of analog work and love its many imperfections. It's nice to know that some digital imperfections work in a good way too.

Yup, gotta love imperfections.

[thethrillfactor]

Originally Posted by Kiwiburger

Surely the worst case scenario with a DAW would to be out by half a sample. That would be uncorrectable. Unless you doubled the sampling rate. So this is probably a good argument for using high sample rates.

Here i agree.

At the higher sampling rates the latency is shorter and where it does affect the audio is further up the freq range where you don't hear it as much even though its there.

The only issue is i don't know any DAW systems with the horsepower to do a 200 track production at 96khz.

Originally Posted by Kiwiburger

I don't get that it's a real problem. Who cares if it's out by half a sample. Only if you are trying to do parallel compression or saturation would it be an issue. And IMO the phase shifts through the analog gear is a bigger issue, and harder to correct.

Here i totally disagree.

Who should care?

Everyone who is working this way should.

We are talking about 25 microseconds(at 44.1khz) in digital compared to 3-7 microseconds in analog. That is a huge difference sonic wise. And the reason you should care is if you add a bunch of tracks this way it will effect the over all sound. Its part of the reason that people who use analog gear in DAW mixes feel that the "summing is broken" or that DAW mixes come out sounding flat and 2 dimensional.

Also any information shared among outputs will change. For example a drum kit that has leakage among all the tracks. This is where most parallel processing is done when mixing drums. There is a sonic impact difference between an analog insert parallel in a DAW compared to an analog mixer(even something like a Mackie or Behringer).

Or how about bussing to a comp or an EQ? Again you have information shared among 2 D/A outputs and 22-25 microsecond shift between the tracks will affect the freq response.

So what do you normally do in this situation?

Most people over process.

Add more hi's(the air freq's which are attenuated) and hi pass the bottomn octave(for the bass that is boosted).

What you have now is a real edgy, forward sounding, "digital"sounding" and "non warm" mix.

Originally Posted by Kiwiburger

The example giving of wanting to bounce a track OTB for eq and delay doesn't seem insurmountable. Unless it's a drum track, or something with lots of leakage, where maintaining phase integrity will be a problem. But the phase shift in the analog box is still there.

Latency is easy to understand and easy to correct.

Read above.

The best case scenario when mixing hybrid ITB with analog gear is either:

1) Use plug ins for parallel processing

2) If you use analog gear for inserts and parallel processing make one round trip only and try not use the analog gear on tracks that share sounds.

3) If you go out to analog processor(s) for the final mix come back on the best hi end converter you can get to either an external recorder or back to your DAW.

4) Split out your mix on different tracks of a D/A to analog console and mix on that. As much as you can try not to share info on the D/A's themselves. The more seperation the better so 24,36, 48 D/A's. Do all of your analog processing in the analog domain and come back for the final mixdown either to tape or the best converter you can find at the highest rate you can record.

[jammybastard]

...and all this math is rendered moot when it's converted to an 192k mp3.

[Jules]

DAW Latency is a pain, its no wonder Lanois doesnt dig it..

What if you wanted to route out one element (like a snare) of a multi mic set up and re print it..

Who can blame him if he doesnt want to rely on computer processing to get it 'close to where it is supposed to be' give or take a sample here or there..?

These sample differences mess things up to some peoples ears.