05-16-2016, 02:38 PM
(04-26-2016, 08:31 PM)allank Wrote: Hi X42,
Can you please explain the difference in MIDI timing between the 2.
My understanding was that Jack ensured MIDI and Audio were kept in good sync and that ALSA did not. Is this not the cse? .
tl;dr: the timing is the same, but [re-]alignment (latency compensation) is only possible with Mixbus/ALSA.
As long as the signal does not leave jack, all is sample-sync. The problem is at the boundary to hardware. JACK uses ALSA under the hood to communicate with the hardware.
The vast majority of soundcards treat audio and midi as separate devices and data-streams. Audio and Midi i/o is not synchronous.
Furthermore you can add multiple midi devices, each of which has its own clock.
In general this is not a big deal. Midi timing does not have to be sample-accurate. Since MIDi is ~32kbaud, a single note message spans about 1ms anyway.
However alignment can be problematic, particularly when using Midi Time Code and multiple devices.
The total latency of a system is the sum of the process-period + systemic-latency. The latter depends on a lot of things (hardware, bus-interface, etc) and is usually different for each device and system. While jack does allow to measure, configure and compensate for systemic-audio-latencies, jack offers no means to configure it for MIDI devices. Mixbus's ALSA backend does allow to measure and compensate for it per MIDI device.
(the reason jack does not offer to configure systemic MIDI latencies is because there can be multiple midi devices and they're dynamically managed.)
(04-26-2016, 08:31 PM)allank Wrote: It's a moot point for me at the moment, because the ALSA driver for my A&H Zed R16 can't run lower latencies than 256 buffers, whereas FFADO runs comfortably at 128 buffers (this seems to make a significant difference for me).. (more important at the recording stage) ..
Firewire is the only protocol that has iso-synchronous streams. Audio + Midi data-transfers are kept in sync. This allows soundcards to have identical systemic latencies for audio and midi (but requires further attention by the manufacturer: hardware buffers, transceivers, etc). Some PCI soundcards can also provide similar synchronous i/o, but only very few high-end ones do.
PS. Did you contact Takashi about the buffersize limitation of the A&H Zed R16 ? It might be a something nobody has yet flagged up. See the section "Bug report" at https://github.com/takaswie/snd-firewire-improve#readme