Please help me optimize my project and settings for realtime

[Dark Goob]

See bold section for problem statement below.

Current hardware/software:

- MacBook Pro 16" 2021, M1 Max, 64 GB RAM, 8TB SSD
- Soundcraft Delta 200 16-ch. mixer
- MOTU 896 HD and 828 Mk II with Black Lion clock
- TASCAM U-44
- DP 11.32
- 3rd Party plugins available: Metric Halo Production Suite, Waves (all)

Project:
- seven mono analog inputs (drum mics) -> bus 1/2 -> Drums Submix channel -> Main Mix
- four stereo synth analog inputs -> Main Mix
- one mono analog input (vocal mic) -> Main Mix
- one stereo analog input (vocal effects from Eventide H90) -> Main Mix
- total of 18 simultaneous analog inputs and one aux stereo bus
- one mono analog input (bass amp mic) -> Main Mix
- one mono analog input (bass pre-effects DI) -> Main Mix

How we use it:
- all our mics, synths, and effects (other than the bass amp DI and mic) go into the 16 channels of the Soundcraft mixer, then the direct outs go into the MOTU units as balanced +4 DBU
- bass amp DI and mic go into the two front-panel inputs on the 828 Mk II
- Main Mix output goes over SPDIF to the TASCAM U-44, a portable USB audio interface that lets us capture a session mix digitally to the iPhone for easy sharing to bandmates (supports 44.1 thru 96khz), I just mix down from AUM to Files app and it auto-syncs to a shared iCloud folder when I plug in my phone
- Main Mix output also goes out the TASCAM U-44's headphone port to a 4-way headphone amp to feed in-ear monitors (admittedly super ghetto, and not a proper monitoring rig, but it's what we've got for now)

Settings that have been working fine:
- Live Performance Mode enabled
- at 44.1 khz, 24-bit, and a 128-sample buffer, we can run 2 to 4 plugins per channel on the following mixing board without peaking the realtime meter (using plugs like Waves L1+ Ultra Maximizer, L2, L1 Limiter, MOTU Masterwork EQ, Masterwork Limiter, Dynamics, SubBass, Dynamic EQ, and on the final mix: Waves Abbey Road TG Mastering Live, with typically 2-3 plugins per channel):

Screenshot 2024-07-11 at 1.26.21 PM.png (527.94 KiB) Viewed 2511 times

Problem settings:
- at 96 khz, 24-bit, and a 256-sample buffer, even after switching out the CPU-expensive MOTU Masterwork EQ plugins and replacing them with Waves H-EQ Light, I still had to remove all the Masterwork Limiters, SubBass, Abbey Road TG Mastering Live, and several others, in order not to peak the CPU and cause audio pops in the live Main Mix output (and even then, it's still very borderline with the occasional pop):

Screenshot 2024-07-11 at 1.01.15 PM.png (376.6 KiB) Viewed 2511 times

Obviously, 96 khz is more than 2x the data of 44.1 khz, but in Activity Monitor, my CPU is barely taxed at all. In total it's less than 10% used. So I'm very confused why there is not enough CPU in this machine to handle the original project considering I doubled the sample buffer to 256. It feels like DP11 is simply not using all the available CPU cores.

Does anyone know if there is any settings I could disable or enable, or particular plugins that I should be using or not using, to be able to do what we are trying to do here?

I'm also asking, because I've been wanting to replace my audio interfaces with something more modern, considering the drivers aren't officially supported anymore. I've been looking at a bunch of different options, and just trying to gauge how important I should be considering DSP plugins. I was hoping the M-series chips were getting powerful enough to avoid having to worry about DSP crap anymore. I've also been looking into those new plugins that run on the GPU but haven't tried them yet.

Thanks for any advice.

[HCMarkus]

I really hate to ask this to start, but why 96k?

Realtime what? Is this for a live show or are you recording in a studio? Sounds like you are recording all the instruments simultaneously...

Pre-Gen will lighten the load on your Mac for playback. Tried it?

How are you monitoring while recording? Through your mixer or DP? Unless you are monitoring through DP, you can crank the buffer up as high as you want; DP will still record fine.

[Dark Goob]

I really hate to ask this to start, but why 96k?

It's fine, there are no stupid questions here

We have been recording at 44.1 khz for many years but our hardware supports up to 96 khz. We want to try this resolution out, as there are some potential benefits to doing so in terms of latency, phase accuracy, stereo image, plugin effects processing quality (particularly pitch shifting or time stretching), transient response accuracy, reduced side-effects of anti-aliasing filters on the high end of the audible spectrum, etc.

It's just generally been my experience that (in any realm of digital capture, whether it be sound, image, or video), capturing well above the distribution resolution/quality tends to consistently produce a better final result, especially if that involves baking in manipulation, filters, effects, layering, mixing, or any other form of digital signal processing.

At worst, we'd try it out and perceive no difference. I'm not expecting it to make much of a difference to be honest, and I won't be heartbroken if we give up on the idea for now.

But I'm also interested in experimenting with shifting copies of the 96 khz files to 48 khz at double duration, then filtering out everything that would have been audible at 96khz in order to use the previously inaudible but now audible portion of the frequency spectrum in the mix.

Realtime what? Is this for a live show or are you recording in a studio?

Our project studio.

Sounds like you are recording all the instruments simultaneously...

Correct... Steve Albini style, no tracking. I'm not against tracking, but I have just not found it to produce the kind of result we're looking for, for our style. I can always mix and match pieces of takes to other takes, but this sometimes involves subtle time-stretching to make things fit. In that case, 96 khz could provide a better result from the stretching algorithm. It's a bit of an experiment still, just looking to try some things.

How are you monitoring while recording?

Traditionally, we have been using the group outs from the analog mixer for zero latency monitoring through amps. This works fine for the other guys, but for me as the vocalist, it has led to feedback issues if I have the vocals loud enough to hear through my ear plugs, due to limitations of our space and setup. So one main motivation to get the upgraded Mac was to be able to transition for me to use in-ear monitors out of the DAW.

Monitoring what the DAW is recording live while we're playing also helps to avoid the situation we've had happen several times where one of the finicky cables or direct outs decided to cut out and needs to be wiggled, or where for some reason some channel or another got punched out on the DAW. It sucks to get through a session and then realize somethings was ••••ed up and now we don't have a usable recording.

Lastly, hearing what the iPhone is hearing helps me ensure we get a relatively listenable, redundant mix to the iPhone to share to the guys after practice. On the rare occasion I forget to press record in DP, that has been priceless to have.

Pre-Gen will lighten the load on your Mac for playback. Tried it?

You can't pre-generate effects for audio that hasn't been recorded yet.

Through your mixer or DP? Unless you are monitoring through DP, you can crank the buffer up as high as you want; DP will still record fine.

Yes, to reiterate, this is for the purpose of monitoring through DP, hence why I said "Live Performance Mode" is enabled. I'm well aware that cranking up the buffer will avoid the issue, but the whole point here is to avoid raising it above 256, since at 512 or higher, there won't be a monitoring latency benefit to recording at 96 khz as opposed to 44.1 at 128.

I would expect the CPU is actually maxed out, then sure we could get clipping. But the fact that the CPU is barely above 6% utilization and it is clipping out of real-time, makes me think there might be some kind of misconfiguration or a software bug of some kind. It's possible it also has to do with using these older non-supported interfaces, but I don't want to go to the trouble of ordering new hardware to fix this problem if that wouldn't help (of course most of the interfaces I'd be considering all have some basic DSP effects for use in this exact situation, but I generally hate how DSP effects are dealt with and so I have always been loathe to avoid them—a major benefit of DP to me has always been that it's not ProTools).

PS—Ugh, they really need to make this forum not log you out and then clear your draft when you log back in :/

[James Steele]

PS—Ugh, they really need to make this forum not log you out and then clear your draft when you log back in :/

Sounds good. Write to these guys: https://www.phpbb.com/

I just "work" here.

You could talk to the famous "Doris in Accounting." She handles all board refunds.

[bayswater]

If you dare.

Doris in Accounting.jpg (67.84 KiB) Viewed 2428 times

[James Steele]

If you dare.


Doris in Accounting.jpg

That's her! I told her maybe this year, I'd see if we had the budget to get her one of those fancy adding machines with the paper tape!

[Phil O]

You have spent a lot of money already on gear. A decent pair of closed back headphones would not break the bank and would solve some of your issues. You don't need expensive boutique cans. I use Sony 7506s for my vocalists all the time and I've never had an issue with feedback. They cost around a hundred bucks. No one has ever complained.

I have an old analog Soundcraft board that I use just for zero latency monitoring. You already have the gear. As HCMarkus already mentioned this gives you the freedom to set your buffer in DP to whatever you need without having to worry about latency. There's no need to monitor through DP unless there's a particular effect you need for monitoring (can't imagine what that would be for vocals).

As far as noisy cables is concerned. Really? Dude, fix them. Just do it.

Your friendly neighborhood Phil

[HCMarkus]

A decent pair of closed back headphones would not break the bank and would solve some of your issues.

When I do remote keyboard sessions where I'll be in the room with a drummer, I always bring along my in-ears... allows me to hear very clearly yet keep volume under control as I'm not fighting headphone leakage.

[Dark Goob]

Dupicliate removed

[Dark Goob]

You have spent a lot of money already on gear. A decent pair of closed back headphones would not break the bank and would solve some of your issues. You don't need expensive boutique cans. I use Sony 7506s for my vocalists all the time and I've never had an issue with feedback. They cost around a hundred bucks. No one has ever complained.

I have an old analog Soundcraft board that I use just for zero latency monitoring. You already have the gear. As HCMarkus already mentioned this gives you the freedom to set your buffer in DP to whatever you need without having to worry about latency. There's no need to monitor through DP unless there's a particular effect you need for monitoring (can't imagine what that would be for vocals).

I respect your opinion, and agree that we could go back to using the analog mixer for all our live monitoring, as we did for many years prior to acquiring the M1 Mac. We have plenty of headphones, etc.

But for two years we have been using DP11 for live monitoring at 44.1 khz with zero issues on the M1 Max Mac, and we enjoy the benefits of this kind of workflow.

The problem I'm seeking help with in this thread is not, "if can we make live monitoring via the DAW work," but rather, if we can make it work at 96 khz.

Obviously if we try 96 khz and it's enough better to be worth using, we could always go back to analog monitoring if that's the trade-off. But I don't want to give up trying to make it work if there could be a technical solution that I'm just missing here.

In that respect, lets refocus this discussion on the original question of whether the Audio Performance meter clipping with less than 10% CPU utilization in Live Performance Mode is an expected behavior, in case perhaps there might be a certain option in settings, or a certain plugin, that might be causing a bottleneck.

Objectively, I feel like doubling the buffer ought to be enough to compensate for doubling the sample rate unless the CPU is close to maxed out already at 44.1/128 samples, which it isn't. So could there be something like, an Intel plugin running via Rosetta 2, or using Audio Units and not just MAS plugins, that could be bottlenecking the CPU? Or some setting that we have turned on that could be preventing DP11 from utilizing all the CPU cores?

One weird anomaly I've noticed is that bypassing plugins does not remove them from the Effect Performance list and does not negate their impact on the Audio Performance meter. Disabling input on the channel mostly reduces the impact, but it's like they'er calculating for bypassed effects unless the entire track is disabled. Maybe I just need to contact MOTU support directly.

As far as noisy cables is concerned. Really? Dude, fix them. Just do it.

We don't have any noisy cables.

[James Steele]

This topic is a bit involved and I didn't want to dive too deeply into it, but I saw this:

You can't pre-generate effects for audio that hasn't been recorded yet.

Are you trying to use plug-ins in DP in realtime *while* recording? Like you would if you were using an Apollo interface, or Avid Carbon... or some other DSP assisted interface?

That's what I'm not sure if I understand. You have a MOTU 896 HD. It connects either by Firewire or USB? So the only onboard DSP on those units I think is like most other MOTU interfaces: the only way to get low latency effects during monitors (and maybe recording) is by using the effects supported in CueMix and run on the interface's DSP. Those are very minimal. If you're trying to use plug-ins on an audio track in DP as you record in realtime, seems like there's going to be no way around latency.

I guess I'm not understanding here, but for as long as I've been using MOTU interfaces, I have never recorded through plugins as I just couldn't run my buffer low enough on my machine to monitor with low enough latency. I pretty much used CueMix (or now the AVB Mixer since I have a newer 828es) to patch into a outboard delay or simple reverb unit (when tracking vocals) just as a confidence booster that I'm only hearing my headphones but isn't recorded. I do all the effects AFTER the audio is recorded. I never attempt to go through software plugins in realtime while recording.

One way I've gotten around that now, is I've purchased a couple of good quality outboard *analog* compressors which don't add ANY latency and I can patch them between my external mic pre and an input on my interface and record that.

Anyway... sorry if I'm not understanding, but without an interface designed for realtime use of plugins during recording, not sure how you can do this? Even UA removed the special low latency mode from their UAD-2 plugins that ran on UAD-2 cards and Satellites. You have to have an Apollo interface to track through their plugins.

[EMRR]

The direct answer to your question - DP isn’t efficiently distributing the load to the available. CPU, and maybe never has. I’ve seen DP processor bottlenecks for years with CPU barely used, even more apparent with Apple Silicon. It’s been mentioned here that some more recent macOS have improved core sharing efficiency, but I strongly believe there are many things MOTU could do to improve task distribution, based on the evidence I see

I’ll add too that I think we are odd men out trying to process a bunch of live inputs, and that’s not been a focus of DAW development, well, ever. Modern digital consoles do it fine; different game.

Tangent - tracking latency - I’m always building the future mix from first playback, so there’s always a pile of plugs including reverbs. Higher buffer settings are essential in that case, I can’t imagine it otherwise. I never set lower than 1028, with an M1 processor, and my monitor path for musicians is fully analog via a split before the AD.

[Pentatonik]

Hello,
James Steele had explained it well. At their set 96 khz, you can only do this through a second mixing console, the CueMix. Or just as described above via an analogue mixer with external effects.There is a reason why there are still large studios with ProTools HD systems, possibly with analogue mixers.These live from recordings of complete bands and orchestras.
They certainly had also under 44.1 khz latencies, but they did not enter so strongly in restrictions.
PS.
I just listened to your songs. Why do you want to take the shots in 96 khz? For classics with high fluctuating dynamics, this makes more sense.

Greetings, Denis

[James Steele]

Hello,
James Steele had explained it well. At their set 96 khz, you can only do this through a second mixing console, the CueMix. Or just as described above via an analogue mixer with external effects.There is a reason why there are still large studios with ProTools HD systems, possibly with analogue mixers.These live from recordings of complete bands and orchestras.
They certainly had also under 44.1 khz latencies, but they did not enter so strongly in restrictions.
PS.
I just listened to your songs. Why do you want to take the shots in 96 khz? For classics with high fluctuating dynamics, this makes more sense.

Greetings, Denis

Yeah... if money were no object, I can see why some people sink a LOT of money into ProTools HD systems. I found this YouTube video interesting where this guy, Colt Capperune, sold his Apollo and invested heavily in the HD system. It's the only way around the latency issues that are part of the trade off of using a native DAW.

[Phil O]

As far as noisy cables is concerned. Really? Dude, fix them. Just do it.

We don't have any noisy cables.

I think you do.

Monitoring what the DAW is recording live while we're playing also helps to avoid the situation we've had happen several times where one of the finicky cables or direct outs decided to cut out and needs to be wiggled, or where for some reason some channel or another got punched out on the DAW. It sucks to get through a session and then realize somethings was ••••ed up and now we don't have a usable recording.