qpwgraph works just like qjackctl

See https://docs.pipewire.org/page_man_pw-cli_1.html See https://gitlab.freedesktop.org/pipewire/pipewire/-/wikis/Migrate-JACK

Manage IO connections. Takes the place of jack_connect, jack_disconnect and jack_lsp

https://docs.pipewire.org/page_man_pw-link_1.html

https://docs.pipewire.org/page_man_pw-mididump_1.html

See also https://gitlab.freedesktop.org/pipewire/pipewire/-/wikis/Config-PipeWire#setting-buffer-size-quantum

This can be configured in ~/.config/pipewire/. Copy files from /usr/share/pipewire

Applications can be set to use a certain buffer size / sample rate with:

# buffer size over sample rate
PIPEWIRE_QUANTUM=192/48000 Ardour

This will determine the latency:

44.1kHz:

48kHz:

https://www.reddit.com/r/rhythmgames/comments/1gamrnc/low_latency_guide_for_linux_using_pipewire/

By default pipewire handles multiple sound cards timing by resampling. In pavucontrol you can set the Pro Audio profile on the card which, amonng other things will assume all devices share the same clock.

https://gitlab.freedesktop.org/pipewire/pipewire/-/wikis/FAQ#what-is-the-pro-audio-profile

Instead of letting pipewire handle timing by resampling, which I found to be less than perfect, we can use a module-loopback to introduce a delay on one or more devices.

See also pulseaudio for how to set that up

https://support.system76.com/articles/audio/

See also https://github.com/apmiller108/audio …and

• https://www.reddit.com/r/linux_gaming/comments/1gao420/low_latency_guide_for_linux_using_pipewire/
• https://docs.pipewire.org/page_man_pipewire_conf_5.html
• https://codeberg.org/rtcqs/rtcqs
• https://manual.ardour.org/setting-up-your-system/the-right-computer-system-for-digital-audio/

I used these scripts to configure my audio interfaces. See also https://github.com/apmiller108/audio/tree/main/config/wireplumber/main.lua.d

Different hardware clocks can create problems. Pipewire will correct for different clock rates with adaptive resampling. I found this could create audio problems (eg, pops and clicks) when applying audio processing on the output streams.

From the docs:

Source, sinks, capture and playback streams contain a high quality adaptive resampler. It uses sinc based resampling with linear interpolation of filter banks to perform arbitrary resample factors. The resampler is activated in the following cases:

The hardware of a device node does not support the graph samplerate. Resampling will occur from the graph samplerate to the hardware samplerate. The hardware clock of a device does not run at the same speed as the graph clock and adaptive resampling is required to match the clocks. A stream does not have the same samplerate as the graph and needs to be resampled. An application wants to activate adaptive resampling in a stream to make it match some other clock. PipeWire performs most of the sample conversions and resampling in the client (Or in the case of the PulseAudio server, in the pipewire-pulse server that creates the streams). This ensures all the conversions are offloaded to the clients and the server can deal with one single format for performance reasons.

Setting the clock.name property on the output nodes of both interfaces to the same value seemed to fix it for me. By sharing the clock.name pipewire will not apply adaptive resampling.

The name of the clock. This name is auto generated from the card index and stream direction. Devices with the same clock name will not use a resampler to align the clocks. This can be used to link devices together with a shared word clock.

In Pro Audio mode, nodes from the same device are assumed to have the same clock and no resampling will happen when linked together. So, linking a capture port to a playback port will not use any adaptive resampling in Pro Audio mode.

In Non Pro Audio profile, no such assumption is made and adaptive resampling is done in all cases by default. This can also be disabled by setting the same clock.name on the nodes.