2

I have a USB MIDI keyboard attached to a Windows PC (it is recognized as a standard MIDI device).

I use a virtual software synth VST-plugin (standalone or in a DAW software)

I also have a (rather new) soundcard with ASIO drivers.

Suppose that I play a note on the keyboard, what is the latency delay of the MIDI Note-On message before it reaches the VST-plugin ? Clearly it depends on the hardware, but can it be an issue (i.e. introduce a latency > 10ms) ?

Every information I found on Internet refers to latency that is introduced later: when the synth renders the sound on the audio interface (and in this case the ASIO drivers can drastically reduce the "final rendering" latency).

Improve this question
4
  • 1
    I sort of have this same question and I also sort of think that it answer is something like "yes, sometimes" or "it depends". I just started playing a software synth with a USB MIDI controller through a thunderbolt interface. My audio latency is less than 1 ms (based on the number of samples in the buffer and clock rate). But it seems like if I play a large enough number of notes at the same time, or very closely spaced, there is a tine amount of latency added on some of the notes. Like maybe the USB MIDI buffer is filling up.
    – Todd Wilcox
    Commented Jan 27, 2017 at 17:02
  • 1
    Interesting. Check out this SOS article in which it says, "a MIDI Note On message will itself last nearly 1ms, and a six-note chord could therefore emerge spread over 6ms." which seems to be exactly what I'm experiencing. So that should be a problem no matter what kind of MIDI interface one uses. The question I'm asking myself now is, how have I never noticed that before?
    – Todd Wilcox
    Commented Jan 27, 2017 at 17:05
  • Another article: soundonsound.com/techniques/solving-midi-timing-problems. And it looks like a Dutch graduate student measured the latency of USB 1.0 of around 10 to 15 ms. doc.utwente.nl/56344/1/Korver03adequacy.pdf Even though that's USB 1.0, the details of the experiment and theory behind it are interesting in that the driver used for the device(s) seems to be an important ingredient, and USB in any version doesn't necessarily have tight time constraints. So it may be that a MIDI DIN connection to a thunderbolt MIDI interface performs better than USB MIDI.
    – Todd Wilcox
    Commented Jan 27, 2017 at 17:29
  • see also - music.stackexchange.com/questions/6619/…
    – Stephen Hazel
    Commented Jan 27, 2017 at 23:28

1 Answer 1

Reset to default
3

Suppose that I play a note on the keyboard, what is the latency delay of the MIDI Note-On message before it reaches the VST-plugin? Clearly it depends on the hardware, but can it be an issue (i.e. introduce a latency > 10ms)?

Yes

Depending on the MIDI implementation, the drivers of the MIDI interface, and the bandwidth of USB or other interface bus, it is possible for MIDI message latency to be more than 10 ms.

Factors:

  • MIDI messages themselves last about 1 ms. 10 MIDI messages sent at once would mean one of the ten messages will arrive about 10 ms after the first message does, even if no other latency is introduced.
  • Some data busses, notably USB 1.0, can be quite slow an introduce additional latency.
  • The bulk of the real-world latency in a modern computer MIDI system seems to be driver-dependent, so poorly written drivers can introduce significant latency.

It's not clear if there is a typical advantage of one type of MIDI interface over another (e.g., direct USB MIDI, MIDI DIN to USB interface, or MIDI DIN to Thunderbolt or other type of interface).

Sources/references:

http://www.soundonsound.com/techniques/solving-midi-timing-problems

http://doc.utwente.nl/56344/1/Korver03adequacy.pdf

http://www.soundonsound.com/sound-advice/q-usb-too-slow-midi-interfacing

Improve this answer
1
  • 2
    There is the additional issue that there may also be an internal delay between hitting the note & the MIDI signal even being generated. I'm too long out of the business to know the modern values, but back in the 90s when I used to do this kind of thing, 7-10ms was "good enough for drums, ch10", anything else would be de-prioritised behind that & might be as late as 30ms if the [serial protocol] stream was busy. A considerable amount of my time & effort was spent making this as unnoticeable as possible ;)
    – Tetsujin
    Commented Jan 29, 2017 at 13:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.