MIDI latency vs. USB MIDI latency

[kvotchin]

So I suspect this would come down to drivers more than anything, but...

I have two options when using the controller keyboard (Arturia Keylab MkII) - connect directly to the PC via USB, or connect a MIDI cable to my interface (RME Babyface Pro), which itself is connected to the PC via USB.

Thus, either way, USB is involved. In other words, if there is a difference in latency here, it would probably (mostly) be down to drivers, right?

Any idea which would be better? All cables are about as short as they can possibly be, btw: from 1’ to 2’, depending which sort is in question.

[Septimon]

There is no difference. MIDI itself is not causing latency. It‘s the process of producing sound when MIDI data is processed in virtual instruments

[nix808]

MIDI does take time,
inside a computer it can run at at least 15,000 hz/messages sec
- but the hardware bandwidth is far lower

MIDI in hardware muxes a UART,
it does take time

[kvotchin]

Okay, so there is no difference between MIDI cables and USB cables, when it comes to transferring MIDI data? Cool then.

I still suspect drivers might play a part though..? Don’t they always?

This was not posted because I’ve noticed any latency - simply wondering if there is any, to speak of. And how much cables and/or drivers matter, if at all.

[JCJR]

So long as midi cable is not so long and of poor quality as to cause data dropouts, midi cable length does not matter. A good quality 20 or 30 foot midi cable will have as good timing as a 1 foot cable.

You could try testing the two midi drivers to see if one seems tighter. It is not impossible that one driver could have less jitter than the other.

But as was mentioned, the daw handling of midi i/o and softsynth and audio buffer handling could possibly have bigger effect on the playthru feel.

[nix808]

I think your hardware will have some latency,
it takes a little time to compute it afaik

but -say your MIDI messages take 5ms,
a virtuoso player fluxes in time about 10 ms when trying to be radically tight

the MIDI timing is usually usable when a keyboard is approximately nicely built

test equipment would be hard to rig, you'd need a piezo to be hit by ur key

The drivers of your soundcard are the drivers in question moreso,
the MIDI latency is an electronics question afaik

I would tend to use USB, as it can cope with a rad higher bandwidth and is more modern,
that would be my guess, but I think the simplicity of the ancient MIDI data means it is practically not an issue, it's speed. Like I say I have sent a midi pulse(off message) at 15,000 hz before. For music it should be fine, like still it should be a rather mechanistic drum machine, if sending sequences

[lfm]

Of course there are delays - usb is a serial interface too.
And depending if keyboard is optimized for usb or not could make a difference.
The chips inside could be all for midi speed, then over to usb.
There are plenty ways to test this - one example below.

I have tested edrums - both
- miking audio sound as I hit a pad, which is sample accurate compensated for in daw.
- and recording midi from it how that differ

It will tell how fast unit is responding+latency through interface.

Similar can be done recording audio from when you hit a key on keybord and record midi from it.
- do midi over din midi
- do midi over usb
Record on two midi ports - one usb one din midi at the same time.

Should give a hint. If hand is too low noise, use a stick or spoon or something, same force both cases should give the same speed down for a key. Do many of course to give a possible spread. You can check and compare on those with similar velocity as recorded if you don't record two midi ports at the same time

If either way is having drivers at all - not class compliant - there could be a difference. Especially if doing audio over usb as well, thinking how priority may be done between audio and midi part.

Testing various midi interfaces I have don physical loopbacks to compare a re-recording of a midi clip over to same daw. If they differ in place on grid.

If using external midi gear response time also differ a lot how to compensate for each gear by setting sending midi so and so midi ticks or milliseconds earlier for audio to align with midi on grid.

[AdvancedFollower]

It might be a factor, but your ASIO buffer size and any internal latency of plugins etc. are going to have a much larger impact.

[Jim Roseberry]

MIDI is a serial (not parallel) protocol.
IOW, Things appear to happen simultaneously... but they actually happen in series.
This makes it more prone to timing issues.
If you have dense controller data, it can absolutely affect timing of MIDI notes.
This is especially true with older hardware.

When using the onboard sequencer, Korg O1W/FD had major issues with pitch-bend/timing.
Timing response got super sluggish.
There was a violin patch that (on its own) sounded good.
Using the onboard sequencer, resolution of pitch-bend and note timing made the patch unusable.
Called Korg about it.
Their response: "Our lab model shows the same symptoms. You need to use a computer to sequence."
The onboard processor was struggling to keep up with the load.
When pushing the polyphony limits of an older MIDI device, many suffer from poor MIDI timing.

I would tend to favor USB over 5-pin DIN MIDI connections.
More bandwidth...

[kvrcoucou]

I remember back in the days that I had real analog synth connected with midi cables the latency was greater than current usb keyboards with built in midi interfaces!

[DJMaytag]

The only real difference you’re going to notice is latency with MIDI OUT to some hardware, and getting that aligned with what you’re playing in your DAW. I have over $1000 worth of an Expert Setup for sample accurate MIDi out to my hardware’ MIDI inputs, but I plug all my MIDI outputs of my hardware into a mio4 connected via USB since there is really no way of getting sample accurate MIDI in to your computer.

[tapper mike]

How fast midi runs is depended on several factors. The absolute fastest standard midi can run is ...1msec. After that it's a matter of how well you controller and slave handle the signal exchange.

I had an extremely fast pairing of a ztar (2ms out) and a rompler. Wearing headphones the sound from the headphones would arrive to my ears faster than the physical sound from the point of contact. It was earie to say the least.

I'm getting about 3msec using midi out from the linnstrument and another 3msec from my korg krome (another rompler)

There is indeed a faster way to send midi into a computer and that's with a dedicated pci card or firewire / thunderbolt but I doubt if you could get a midi on thunderbolt.
https://forum.professionalcomposers.com ... faces/2477