What latency are you happy working with?

[BONES]

We can get up on stage and my bandmate can play drums samples from my laptop just fine with 256 samples of latency - a roundtrip time of 23ms or so. A long time ago, Scott Solida did a big test of his basement full of hardware and compiled the results. He found that most hardware synths have latency in the 15-20ms range and absolutely nobody ever complains about latency in hardware. Why? Because they don't know it's there so they don't notice it. OTOH, everyone knows about latency in computer set-ups, so they think they can feel it. It's nonsense. If you're up on a big stage playing your guitar through a Marshall stack, you'll experience more latency than the average computer musician.

[zvenx]

This came up in a discussion on gearspace recently. It varies from person to person. Some people especially drummers and singers hear the comb filtering effect and it really throws them off above 4ms.

I myself use 64 which is different values on my two computers (both under 10ms RTL). At 64 I can play without quantizing on virtual instruments or play in the pocket on live guitars and basses.

I can feel the difference at 128 and I definitely can't record at 256. That would throw me off and require lots of quantizing.

https://ask.audio/articles/monitoring-l ... can-you-go

A tool to measure rtl on your system: https://oblique-audio.com/rtl-utility.php

And a paper on the issue:
https://www.semanticscholar.org/paper/E ... 881b55e16d
Rsp

[BONES]

What does "play in the pocket" mean?

[zvenx]

https://my.artistworks.com/blog/nathan- ... ing-pocket

Rsp

[BONES]

Good Lord!

[mystran]

Some DAWs insist on playing live MIDI notes "as soon as possible" which effectively means accurate timestamps are ignored live and everything rounded to the beginning of the next available block. This lowers the average latency slightly, but means there is one full block of jitter. Personally I find such jitter highly irritating, because it's essentially random so you can't really compensate (not to mention that what you hear on later playbacks don't match what you heard when you were playing). In a DAW that insists on doing this, even 256 samples at 44.1kHz feels kinda terrible for me, where as if we add pure latency (eg. with a delay plugin) with no jitter whatsoever I can easily tolerate at least another 20ms just fine.

[PAK]

Yes. I'm surprised Mystran is the first to mention MIDI time stamps and jitter. If you're using sub 10ms latency, and something still "feels off", it's nearly always something related to jitter messing with your note timing - and there's all sorts of variables which can influence it - Whether software, drivers, OS, or hardware. If you have a PCIe card, which has MIDI I/O, you may want to try that as an alternative input to USB, and see if you get any improvements.

When you set an audio cards latency you can frequently add a couple of milliseconds on top for the "real" play input latency, since something like a keyboard still has to scan the keys, convert to MIDI, and send etc. Some older digital hardware might not always be the quickest at doing this either. Then there's the time it takes audio to travel from the speakers. So the "real" latency is always higher than it says.

I like to stay below 10ms on the card personally. I can certainly "feel" a setting of 20ms. However, below 10ms, it's rapidly diminishing returns, at least for virtual instruments. Going below 256@44/48 is probably just increasing your CPU load needlessly. That said, if you're outputing a mix to someone (EG they hear their own voice with FX real-time) then you certainly want that I/O delay to be as low as possible. Otherwise I think 5 or 6ms is fine for VST's..

In terms of what people can "really" notice, regarding timing differences, the brain is a complex thing and there is no simple answer. EG Our ears might not pick up on a few milliseconds of timing difference. But change what's fed to the left and right ear, by those same few milliseconds, and suddenly timing differences may be very apparent.

Another factor is the input device itself - Many keyboards don't have particularly good velocity response, and there can be other things, such as inconsistent key travel amounts required, which will influence not just the "playing feel", but your perception of the timing.

[Scotty]

Some plugins , particularly look ahead compressors or limiters, verbs add latency. Any kind of dsp like Universal Audio will also add latency. I've had drummers complain that they can hear latency at 256 and when I toggle back to 128 they say it is fine. Some of the plugins on the cue mixes were adding to the latency so I track at 64 samples live off the floor and not have problems. When I am mixing I increase from there. The whole signal chain comes into play. Drummers / percussionist appear to be the most sensitive . Midi latency is detectable. If you play a Roland drum kit that has both USB and Midi outputs you can easily test the response. USB out of the Roland engines has less latency than the 5 pin midi dins. You can track both simultaneously on some modules (TD-30) and you'll see and hear the difference. Most drummers would compensate instinctively but if they are playing electronic kits they can hear that the audio from their strikes in real time as heard in the room is different than what they are hearing in their headphones. Enough latency can throw some of the better players off in my experience.

[EnGee]

I have 2.3 ms input and 5.9 ms output latency with the 64 buffer 48kHz. I'm fine also with the double size of buffer/latency.

[Kapsy]

Edit: Made a new post here, as my reply was off topic:
viewtopic.php?t=594182

[umd]

In the soundcard's driver, it is pointless to go below 4ms, in my opinion.

But of course there's a lot more latency to consider in the signal chain. As long as you don't feel a lag when you're playing... I would not think too much about it.

Sometimes if I could just hit the right note, even with 100ms of ( real world ) latency, I would be happy.

[Bombadil]

Logic @ 64 samples, 44.1khz, RME FF802, M1 mini, 4.9 ms RTL. I can go up to 128 comfortably.

[jackoo]

I have 5.2 ms latency by default, and about 12ms for really big projects.
But my brain adjusts just fine up to 15-20ms, I can play without issues.

[lfm]

It depend on daw I used.
- I was ok with the setting 128 samples buffer, 48k
- but then I tested Reaper and wondered what is this delay
- turns out Reaper add a buffer(or buffers) depending on pdc in project you run
- so this became 256 in reality monitoring through Reaper.
- so I started with 64 samples buffer to remedy that
So roundtrip been around 4-5ms or so. When my drumming gets as tight as that I would just be happy to go to 32 samples.

Then StudioOne is similar, it always add an extra processing buffer.

I don't think Cakewalk(Sonar) or Cubase(without AsioGuard) do that.

Samplitude was a whole Masters exam to figure out how it worked in this regard.

Doing recent calibration with RME Digiface USB, which is only optical Sonar adds some expected delays for outgoing DA and ingoing AD converters. So 64 samples buffer become roundtrip 184 samples anticipated.

Using that setting and doing metronome loopback analog style which was too much and clips placed early and reduced that so clips are placed 58(entered in sonar) samples later and clips were placed on grid. So this included DAC on out and Audient preamps over ADAT. The higher roundtrip daw think you have the earlier i places clips as compensation. No delay and clips can be placed as samples arrive in recording. If what you listen too is later it places earlier.

So not being disturbed by delay is one thing, but good if as sample accurate where clips are finally placed too. I am probably a few samples off, hard to say which metronome sample to place on grid, middle or as soon as metronome audio start.

[EnGee]

I'm currently between 128 and 256 48k or 44.1k in Windows 11 with Cubase/Live. Reason can do 128 or even 64. On M1 Ventura, things gets a little bit better. 64 or 128 with Logic/Live. Roundtrip with Core Audio in M1/Ventura is much less than ASIO in Windows 11 for Tascam US 4x4 HR (my main audio interface) about 5 to 6ms less with buffer 128.