Overdubbing, Latency and Headphones

[fl]

I'd like to learn how other Native users, presumably with some sort of ASIO audio interface, deal with sessions where a performer will be doing punch-ins on a track, while getting a headphone mix of the backing tracks combined with their own live, latency free performance, enhanced with added reverberation.

How do you set this up? What's the routing look like in Pyramix? Any extra equipment required?

[SoundKlang]

I can not completely answer your question, but maybe offer some usable ideas for multitrack arrangements.

I have been fighting latency for a long time. My personal opinion is: latency always wins. The latency problem does not occur with the playback. The musician interacts with what he/she hears in the headphones, so it dos not matter if the playback was delayed. The problem occurs only with the recorded signal, that is the latency of the AD, the ASIO input buffer, Pyramix' internal processing and then back out through the ASIO output buffer and the latency of the DA. So if you have a buffer size of 4096 samples, that is perfectly okay for the playback. The whole task is to find a latency free solution for monitoring the recorded signal.

If I imagine a perfect solution:
At no stage of the project may latency become an issue. When using plugins in mixing it can become necessary to increase buffer sizes, so the perfect solution must allow for final overdubs or correcting recordings with increased buffer sizes. In my opinion, this would not be possible without additional external hardware. An external analog mixing desk (a digital mixing desk would reintroduce latency issues) with lots of aux sends for different headphone mixes (if needing to record different musicians at the same time) and effect processors would be needed. Also a certain number of DA and AD channels would be needed to turn Pyramix into a digital "tape" machine. All the details would be depending on the specific needs. I am not sure if there is a "one size fits all" solution for this task.

- - -

In real life I have to use a much simpler solution. There are two basic assumptions in the following:

- main focus is the optimum result in the mix, not in the recording situation
- recording one musician at a time

I am coming from a musician's perspective here. All the other band members are always too loud. I only need playback for orientation, but my own voice or instrument raised high above the others. So for example for recording acoustic guitar I have an in ear microphone in my left ear (the ear which is not facing towards the microphone). The second in ear microphone is muted (put some gaffa tape around it, so I always know which is the one for the left ear and I do not have to care about un-/muting the right channel) and fixated so it can not "clonk" against the instrument or microphone. There is practically no spill from the headphone playback (which is very important when doing timing corrections or replacing that one "bad" note with another good one).

This gives me a somewhat natural recording situation: I prefer listening by ear(s). Monitoring the recorded signal through effects does not help the expressiveness of a performance. But I am aware of this approach being highly depending on the style of music or the performer.

I can live with the imperfection of the recording situation because I am satisified with the results I can achieve afterwards. In my experience, recordings for popular music multitrack arrangements almost always need editing. Manual levelling, slight timing corrections, fret noise and so on. I also do manual editing of the timing of a recording in regard to the meter, so that a somewhat edgy bass groove can turn into a laidback piece.

This will not work for recording excellent musicians. I may be wrong in this point, but from the results I am getting while editing, I assume at least some of the excellency in some recordings does not necessarily need to happen while recording. In the end it is only popular music ...

[klaukholm]

A simple solution we use is the Direct analogue outputs on our Horus or Crookwood in combination with a summing mixer.
With the SSL Sigma or the AMS neve 8816, you even get faders.
True 0 latency cuemixes

[fl]

A simple solution we use is the Direct analogue outputs on our Horus or Crookwood in combination with a summing mixer.
With the SSL Sigma or the AMS neve 8816, you even get faders.
True 0 latency cuemixes

I figured that the solution to this question would involve splitting off the new performance's signal before it enters Pyramix, so the performer hears it with no delay, but then there is still the question of providing the backing tracks.

With your method, do you find that you have to manually shift overdubbed recordings in time to compensate for their late arrival into Pyramix? Or does Automatic Delay Compensation magically take care of all that?

The way I thought of configuring Pyramix is to have two Mix busses, one with the Repro button activated, both sent to the headphones. Send only the track to be overdubbed to the "Repro Buttoned" Bus, all other tracks and effects returns to the other Bus. The performer hears their own direct sound from a split sourced "before" Pyramix, from the input stages of whatever interface is being used, or from a high quality mixer/splitter as you've mentioned. The performer hears their own recorded track (from earlier) up until the punch-in point, whereupon the Bus is muted when Pyramix goes into Record, leaving them to hear only their live, latency free split combined with the backing tracks and effects. Added reverb is still sourced from the recorded track, so the performer hears that - with a little extra, um, "pre-delay", which is quite likely not a big problem.

This still seems a little cumbersome in Pyramix with the two Busses - have I missed something? Is there a simpler way? I have seen those "Direct Monitoring Input" mixer strips are available, but I gather those only give latency-free playback when you're using Merging hardware, Mykerinos or Horus. I'm using Native with an RME Fireface.

[fl]

My personal opinion is: latency always wins.

Words of doom! Don't crush my youthful heart!

The whole task is to find a latency free solution for monitoring the recorded signal.

That, or some way of "automatically" shifting the newly recorded track in time once recording is completed, by some amount that is the sum of all latencies. (And by "automatically" I'm resigned to the fact that it means me doing it manually.) At the same time the performer must be provided with a headphone feed this is actually what they are doing at the same time as they are doing it.

So for example for recording acoustic guitar I have an in ear microphone in my left ear (the ear which is not facing towards the microphone).

Surely this was a typo and you meant to say "in-ear headphone".

The second in ear microphone is muted (put some gaffa tape around it, so I always know which is the one for the left ear and I do not have to care about un-/muting the right channel) and fixated so it can not "clonk" against the instrument or microphone.

Okay, here you've lost me. The right 'phone is muted, so it's essentially an ear-plug with, um, gaff tape on it. Wouldn't it be better to leave that ear open, so that the natural acoustic performance can be heard clearly? Sorta like the practice of sliding one headphone "half-off" the ear when doing vocal recording?

There is practically no spill from the headphone playback (which is very important when doing timing corrections or replacing that one "bad" note with another good one).

Ack! Leakage (ah, it takes me back...). Yes, of course. Get me to tell you the story about the opera singer who was so loud in her headphones that she squeaked herself by turning her head sideways for an instant. There were tears. And guilt.

This gives me a somewhat natural recording situation: I prefer listening by ear(s). Monitoring the recorded signal through effects does not help the expressiveness of a performance. But I am aware of this approach being highly depending on the style of music or the performer.

Show me a singer who doesn't need a little 'verb...

I can live with the imperfection of the recording situation because I am satisified with the results I can achieve afterwards. In my experience, recordings for popular music multitrack arrangements almost always need editing. Manual levelling, slight timing corrections, fret noise and so on. I also do manual editing of the timing of a recording in regard to the meter, so that a somewhat edgy bass groove can turn into a laidback piece.

Sure, yeah, no question. But. Surely the best results are achieved by providing the best possible monitoring environment, so that the performer can key into just those subtle details that we're trying to perfect. Better performance = better recording, right?

This will not work for recording excellent musicians.

Of course I - kaff, kaff - never work with anything but.

I may be wrong in this point, but from the results I am getting while editing, I assume at least some of the excellency in some recordings does not necessarily need to happen while recording.

Well, there's a can of worms probably best left undisturbed. However, I will say that some of the main attractions to recording classical music as opposed to multi-tracked, multi-layered pop music, are that the musicians are competent at the very least, and quite often have long experience playing the piece being recorded. They know their parts, we all know how the music is going to go, since we all have scores. My job is to get all the technology in place so that it intrudes as little as possible, and then get the hell out of the way while the magic happens.

My issue is the "disconnect" created between the performer(s) and the process. They're out there in the hall or church or whatever, doing their thing and I'm collecting it like some creepy stalker lurking in the shadows. For the most part, they have no idea what the recording sounds like since they're not hearing the results - until later - maybe - when they're out of the environment and not actually producing the sounds. Because they're out of the loop, as it were, I've found that a lot of classical musicians approach recording with an attitude more on the side of "let's get this over with" and less "we're here to create art".

Plus, how are they going to appreciate my genius if they don't hear it?

From back in my pop recording days, I remember that the better the performer can hear what they are doing, while they are doing it, the more subtly nuanced the performance can be. Can be. (There's always the danger of performers disappearing up their own, um, yin-yang.) As a beneficial by-product, this raises the quality of communication between the artist and the recording team - we're all on the same page, hearing the same thing. In short, we're all equally involved.

In the end it is only popular music ...

More words of doom! Oh, where are the smelling salts, I'm feeling quite faint...

Yeah. Sure. But it's a game of inches, as the saying goes, and every tiny improvement contributes one way or another. Doesn't it? I mean if it didn't, I'd be using Pro Tools through a Behringer...

Yes, we have marvelous ways and means to manipulate performance, but if you do what you can to squeeze better performances out of your musicians, doesn't that make the whole job easier, and the results better anyway?

[SoundKlang]

It is sincerely not my intention to speak any words of doom.

You were right, I meant headphone, not microphone. The right in-ear headphone is fixated outside of the right ear. "... so that the natural acoustic performance can be heard clearly". I use in-ear headphones because they produce less spill, maybe not the best solution for vocal recording.

I do agree with "get out of the way", "we're here to create art", "subtly nuanced performance", "hearing the same thing", "every tiny improvement contributes". That's the ideal, but after building that perfect re-/recording chain (for my purposes), adding the perfect monitoring solution is not possible and not needed at this moment. What I am doing here must be "no budget" compared to what you are doing. The only musician recorded here is currently myself (no excellent musicians involved).

But we happen to use the same technology, so I want to get back with a few more thoughts:

I wonder if a true zero latency with Ravenna/Horus is possible. For example in Pyramix native (1 stereo strip, 10 mono strips, 5 VCA strips and 1 stereo mix bus, no plugins, no re-routings) I get the following:

ASIO buffer size:
128 = 0.7 ms @ 176k4 -> 1.4 ms ASIO input to output latency

Pyramix settings declare:
Input 168 samples = 0.95 ms
Output 373 samples = 2.11 ms

The status bar of the project window declares 3.1 ms latency, too. Pyramix native adds 1.7 ms of input to output latency.

If I add at least one stereo subgroup bus (for example for the reverb), latency increases to 3.8 ms.

If I would add the VS 3 Dynamics plugin to a subgroup bus (or an aux send bus) the input to output latency would vary depending on the lookahead delay parameter of that plugin.

The manual of my AD/DA declares 0.7 ms latency for AD and DA @ 176k4 -> another 1.4 ms input to output latency.

Between the ASIO interface and the AD/DA I use reclocking devices, I have no information about their latency (if any).


Also I read various times, that using VST plugins does change the latency behaviour of the Ravenna/Horus MassCore. The question would be whether you use VST plugins in your projects.
Assuming the arrangement is almost ready, we got the sound we wanted and for getting that VST plugins were used (for example for reverb). But now one final overdub or punch in is necessary. Only, by using VST plugins the latency of the project became way too high.

If the composition, recording and pre-mixing process is not clearly separable, there is always a configuration possible, which could raise the latency too much. In my opinion a "perfect" solution would need to rule out the latency problem right from the beginning. That is, one needs to be able to do whatever is needed without caring about the resulting playback latency. That leads to a solution which does the monitoring of the recording outside of the box and with no noticeable latency - so probably in the analog domain.

- - -

I do need to manually correct the timing of recorded clips.

I do not use a "repro" mix bus, but the track's (or a dedicated recording track's) monitor icon "Auto" together with the record icon "Autopunch". That way the track plays back before the MarkIn / after the MarkOut locator and inbetween these locators, the recorded signal is played back. The playback affects all send effects of the corresponding mixer strip, no matter whether outside (the track's signal) or inbetween (recorded signal) the MarkIn and MarkOut. (The playback of the recorded signal has the full input to output latency - the reason for me to monitor "by ear" as described above.)

[fl]

Well, my brilliant idea of using two busses, one having the Repro button enabled, was completely undermined by the Auto-punch function. I had hoped that the muting would only take place for the duration that Pyramix was actually, you know, recording, but instead, it mutes the track from the time the Record button is pressed, which is prior to the cursor passing the Punch-in point.

Since the whole point of the exercise was to have the pre-recorded material audible as a lead-in and out for cueing, then hearing only the zero-latency feed from my audio interface's input while punching, this is not a solution. Pyramix only allows me to mute the whole track, not just the punch duration. If I leave it unmuted, I get the latency delayed input signal that it passes in addition to my zero-latency signal from the interface - double attacks on everything (or, if I reduce the interface's latency to the point where it's clicking and popping, phasiness). It's all or nothing.

It was so much easier with a multi-track tape machine! The Soundcraft 760 we had at the studio where I started out even had a nice feature they called "Dab", where you could be rolling, and with one's middle finger pressing down the Record button, use the index finger to depress "Play" for the duration of the punch, and then release it and be back out of Record. You could literally "dab" syllables in words with enough practice. A very slick feature on an otherwise undistinguished machine.

So, I'm back to the drawing board with my original question. How are Native users performing punch-ins? Is there any alternative to just sucking it up?

[SoundKlang]

"It was so much easier with a multi-track tape machine!" Yes, those buttons and footswitches were made for the minds of musicians, not computer specialists. I miss them sometimes too...

If I understand correctly, you want to route the audio input in RME's TotalMix directly to the audio output of your RME hardware that feeds your headphones? And you want to be able to play along with the output of the Pyramix mix bus in sync? I think a workaround could help. You could add another track dedicated for the recording. It should be possible to assign that track to the same mixer strip, if not then to an an additional mixer strip. If then the part that you want to replace of the original clip was muted (cut and mute) and the recording track was muted in it's track header (or set to playback only, the icon with the triangle only), there should be no disturbing output from Pyramix. But your input signal send by TotalMix to your headphones should still be audible.

[fl]

Yes, that would work - you get the lead-up to the punch point, then nothing, because nothing's there.

I'm just not sure how Pyramix will behave when you have more than one track sent to the same mixer strip, and then you place one of them in Record Ready (and I'm not in front of Pyramix now to try it out). It does seem to need a fair amount of fussing, and I was rather hoping for a fuss-free solution, but, whatever works - I'll test it out tomorrow.

In the meantime, I've emailed Merging support to see if they could change the interaction of the Auto Record and the Repro button functions - it could be a huge issue, or it might be as simple as changing a line of code. I don't know, we'll see.