Ultra-low latency, jitter & CPU isolation

Flavors & Remixes Ubuntu Studio
, ,
1

Hi,
I’ve been learning about and tuning for low latencies for quite a while on multiple distros. I have the basics of pipewire and jack down along with the .conf files, bootloader options, and some rt-kernel stuff.

I’m some way into tuning cpu core isolation, affinities, rt priorities, interrupt and irq management. I’m even pulling things like NVMe interrupts off of audio-isolated cores.

I now have a fast system. but..

…We’re on Linux. How far can I tune a framework 13 (AI340) laptop? I want an industrial grade, ultra-low latency, dedicated master ‘sequencer’ that works as a very realtime extension to my large eurorack system via VCVRack and Bitwig (and soon, Ardour). My tower PC runs higher end DSP stuff so this tuning is very much about latency, not throughput (although squeezing out what I can).

I have an RME RayDAT running via a Razer core X enclosure (ThB3-PCIe) with an almost reliable round trip latency of 2.367ms with admittedly very low track counts before xruns. That latency includes system > Raydat > Expert Sleepers ES-10’s ADAT to DA/AD converters > Raydat > System.

No matter how far I go, the hole gets deeper and more complex - it’s extremely difficult to keep track of, never mind setting up again after I break something!!

So I’d like to ask; is there a guided route into this kind of ‘industrial appliance’ type of setup through Ubuntu (Studio)? As the official framework distro I think this is the best place to ask for help and/or pointers. Most guides online describe very generic ‘usb/pipewire/producer’ setups, not this. I’m finding some possibly useful tips from industrial server/ machine builds, though.

I don’t have any official education in computing, just a couple of years in Linux, and this stuff is really complicated - so some input, pointers, advice would be really appreciated and hopefully helpful to other realtime audio users since I’ve found very little on this specific topic.

Thank you for reading!!

TLDR: Industrial spec realtime audio?

2

Hello, good thinking but it seems to me you have coupled your tower indirectly(as far as i can read from your story). Today we use pipewire and just one timer for all things audio. Maybe more direct link tower with laptop is my thought on this. But i have never heard of a - nor seen a industrial grade laptop but beleive it can exsist. I just bring my bigtower where i can. The main sollution in your case (most proberbly) would be to use the timer of your laptop and let the clocks of your other equipment listen to it. It is cristal-tuning and some ubuntu guys figured it out how alsa had always explained..and then came pipewire :guitar: Not that i really know your setup but i hope this is useful.

3

Thanks for chiming in.. The tower isn’t time critical, I work there separately from this machine. The Framework laptop will work like a eurorack sequencer for sequencing, recording, cv i/o - No plugins or any real dsp. It’s a live jamming machine really. Since all the synth gear is controlled and recorded from the laptop, it needs to be as realtime as possible - It automates lots of synth modules which are sub-1ms in most cases.

Pipewire doesn’t play nice with my RayDAT and I prefer Jack2 for its slightly lower overhead, latency, and simple setup away from system audio (I give my interface to jack exclusively).

As far as ‘industrial grade’: I mean the linux set up, not the Framework hardware itself.

I hope that clears my post up a bit.
cheers