[Jack-Devel] Jack and thunderbolt

Chris Caudle chris at chriscaudle.org
Fri Dec 7 03:56:25 CET 2018

On Wed, December 5, 2018 4:47 pm, liebrecht at grossmann-venter.com wrote:
> I really battle to find the optimum solution for ultra low latency
> monitoring using Linux.

For monitoring while recording, hardware monitoring is often optimum

> Since I am using Jack I have a few compatibility questions.
> 1) Is jack known to be able to work with thunderbolt 3 on Linux ?

The jackd server uses ALSA for interfacing to most audio devices, although
it does have a way to directly use some Firewire devices.  Really the
question should be is there a driver for a particular Thunderbolt device
you want to use.
My understanding is that Thunderbolt is basically PCIe with some slight
modifications to allow hot plugging cables, determining whether the
peripheral needs PCIe or DisplayPort protocol, a few other things, but the
important point being it does not define audio transport as for example
USB audio class does, so even if one Thunderbolt interface is supported by
ALSA, that does not mean that other Thunderbolt interfaces would be
supported, you would still need a driver that takes care of register level
access to the specific device you want to use.

> 2) Thunderbolt included, what is the lowest latency sound transport
> format that jack can work within your opinion on Linux
> e.g. thunderbolt / AES50 / Ultranet / Madi / etc

That question is not well formed.  Thunderbolt is a data transport used in
computers, not specifically an audio transport. AES50 and MADI are
specifically an audio transports, and no computer has a direct AES50 or
MADI connection.  So again, the question is what particular  audio device
would you like to use, and then see if there is an ALSA driver available.

The lowest latency would be PCIe connection from your processor to a
device with analog converters on board.  Many people have reported good
results from RME devices over the years.
If you happen to already have an analog converter you like with AES3 or
MADI interface, RME has cards with those transports as well.

But that is specifically just addressing the lowest latency way to get
data from your processor to an analog audio connection.  You did not
explain what you are trying to achieve, so spending money on an interface
that can run with low latency may not be the best way to achieve what you
really want.  Not to mention that there are many more system
considerations than just interface transport in determining latency, you
can spend quite a bit of time optimizing your system for low latency, but
there are still quite a few things out of your control on a modern Intel
based system (such as system management mode firmware that runs outside of
the control of the kernel scheduler).

So do you need low latency for monitoring while recording?  Probably
hardware monitoring is optimal (either with an interface that supports
hardware monitoring, or by using an external analog mixer, preamp, etc.
which supports monitoring features).
For software synthesizers?  Then system latency will probably dominate
over audio interface, you will have plenty to work on before nearly any
interface is the limiting factor.

Chris Caudle

More information about the Jackaudio mailing list