The Latency Thread?

pipelineaudio

Shredder
Messages
1,502
Since Tonex, the issue of having extra AD and DA hits is yet again in the news, Maybe we could make a database to see just what we are getting.
I plugged in the Mooer pitchbox and made a little test

I'm getting a reading of 66 samples or 1.375 milliseconds

cGFiM07vQP.png


Here's Helix on an empty preset (Helix native sometimes shows [I think] up to 16 samples at 48khz). I get 87 samples or 1.833 milliseconds, and an interesting inversion for you nutty absolute polarity purists

Helix Floor with a pretty full featured amp, IR, and FX setup, showing 101 samples or 2.104 milliseconds.

I'm assuming the FX send and return will be similar but too lazy to test that at the moment.

The implication here, is that you will be adding 2 milliseconds plus whatever your tonex or whatever adds if you stick that in one of Helix's loops. In my case, because I don;t really have a way to switch bypass pitchproof, I would for convenience sake want that in one loop and tonex in another, I'm getting pretty close to the edge of my tolerance by adding 5 milliseconds plus whatever Tonex is.
 
Leo Gibson measured Tonex pedal as about 3ms latency.

I used to run Helix Floor even in a 7CM stereo amp setup just for shits and giggles. So that's 7 A/D/A conversions. I still did not find its latency to be an issue in any way.

For me wireless + VST plugins is the situation where latency can grow big enough to become a problem. Even then you mostly notice if you play shred stuff where you find you are not quite hitting your notes etc.
 
I used to run Helix Floor even in a 7CM stereo amp setup just for s***s and giggles. So that's 7 A/D/A conversions. I still did not find its latency to be an issue in any way.

I always found funny how the same people obsessed by latency tend to be the same who'll build a pedalboard with 5 digital stompboxes, just to plug it on a tube amp.
 
I always found funny how the same people obsessed by latency tend to be the same who'll build a pedalboard with 5 digital stompboxes, just to plug it on a tube amp.
I think my second banning from TGP came from daring to ask the latency purists for evidence of their claims. I'm documenting this with hard numbers, but on record as not being super sensitive to it until about 11 msecs
 
Just because others have a different experience than yours doesn’t make them snobs. That’s bunch of bullshit.

I'm kidding, of course. This topic has been beaten to death (and multiple times too) in here, but the TL;DR is that most people seem to care about latency only when they're actively looking for it.

5ms is a ridiculously fast response time, y'all. You can safely enjoy playing your guitar.
 
Last edited:
Anyone interested in DAW latency and testing what latency starts to actually bother you?

I am looking at the final Roundtrip number which in my case is 7.6ms as reported in my DAW with a buffer size of 256 at 96kHz.
In my opinion under 10ms is low enough to monitor my voice in headphones without noticing latency, your own voice is the most critical latency test since you naturally hear your own voice instantly all your life, in other words hand-ear latency is much more tolerable than voice-ear latency.

daw latency.png



A test you can do in your DAW is to use a simple digital delay set to 100% wet with no repetitions so you can simulate latency and check what value starts to bother you if the current roundtrip latency of your DAW is imperceptible.

Another test I like to do with DAW guitar plugins is I loop the processed output to a second input with a patch cable, quickly tap the guitar input plug with metalic object and measure the sample distance between the two inputs, this automatically removes one AD+Buffer stage when measuring and gives you real world Analog-In to Analog-Out value or what you actually experience when playing.

In my case the measured roundtrip latency with Helix Native is 11.3ms and without is 11ms, that's slightly longer than the driver reported roundtrip of 7.6ms in my DAW settings (which might not be accurate in the first place) and totally fine for live voice or guitar monitoring, in my personal experience of course.
EDIT: With buffer size of 128 samples I get 8.7ms measured analog roundtrip latency.

plug tap.png


Without Helix Native:
Without Native.png
 
Last edited:
Another good option, even though it doesn't come with as clear numbers, would be to test e.g how much you are bothered if you enable pitch shifting on a modeler. It often has the highest latency penalty out of any effect.

I find that I'm still alright playing with virtual capo on my Axe-Fx 3 enabled, but I think it starts to get in the way if I try to shred. Totally fine for some rhythm playing tho. So I can now understand those "Steve Vai can spot latency" things because he plays fast leads a lot at a way higher level than me.
 
Another good option, even though it doesn't come with as clear numbers, would be to test e.g how much you are bothered if you enable pitch shifting on a modeler. It often has the highest latency penalty out of any effect.

True, but note there're many reasons pitchshifting / virtual capos throw players other than latency. There's tracking stability, and the disconnect between the acoustic sound your instrument makes (and what you feel under your fingers) and what comes out of the speakers - which can take a while to adjust to.

So I can now understand those "Steve Vai can spot latency" things because he plays fast leads a lot at a way higher level than me.

Long live Steve Vai, but he complains a lot about digital latency while he's perfectly happy running an AxeFX in his Synergy FX loop :stirthepot

Single-digit ms latency is low enough as to be, quite literally, physically imperceptible for humans. Which is when we wade into the terrain of psychoacoustics, where subjective perception plays a major role, and things are more complex than just scientific measurements.
 
True, but note there're many reasons pitchshifting / virtual capos throw players other than latency. There's tracking stability, and the disconnect between the acoustic sound your instrument makes (and what you feel under your fingers) and what comes out of the speakers - which can take a while to adjust to.
I don't disagree at all, it was just something that is easy to try out to experience a straight up higher level of latency.

Long live Steve Vai, but he complains a lot about digital latency while he's perfectly happy running an AxeFX in his Synergy FX loop :stirthepot

Single-digit ms latency is low enough as to be, quite literally, physically imperceptible for humans. Which is when we wade into the terrain of psychoacoustics, where subjective perception plays a major role, and things are more complex than just scientific measurements.
Again, I agree but to be fair, the Axe-Fx is really low latency already.
 
All I'm going to say about latency is this:
Compare side-by-side
Lower will always feel more immediate/responsive.
Play thru your favorite amp-sim plugin at 10ms RTL... then at 2ms RTL.

James mentioned using a delay plugin set 100% wet to simulate different latency values.
That's a great way to test where it really starts bothering you.

I know for triggering drum samples, if RTL gets above ~3ms, it starts to feel "off".
Above ~5-6ms, you can start feeling the lag.

IME, With higher transient signals, latency is more obvious.

Some folks have perfect pitch... and are more sensitive to pitch (intonation) than those who don't.
I have no doubt some folks are more sensitive to latency than others.
 
James mentioned using a delay plugin set 100% wet to simulate different latency values.
That's a great way to test where it really starts bothering you.

Yeah, but have someone else switch the delay in and out for you - as in, a real blind test. You'll be surprised.
 
Last edited:
Yeah, but have someone else do the switching for you. You'll be surprised.
I've been dealing with DAW related latency issues for ~30 years.
It's a subject that I'm extremely familiar with. ;)

When my career started, we had to monitor off the console (while tracking) to avoid 10-12ms latency.
At that time, some PCI audio interfaces offered hardware based monitoring... but it was spartan compared to what's available now.

~25 years ago, I was hired to tech some sessions in Nashville (Memphis Horns, Terry McMillan, Tabitha Fair, etc).
Those session players would absolutely not tolerate 10-12ms of latency.
We tried, they bitched. My job was to make the sessions run smooth... to keep the DAW/machine out of their way.
At that time, that meant avoiding software based monitoring.

I'll repeat myself.
Try playing your thru your favorite amp-sim plugin at 10ms RTL... then immediately at 2ms RTL.
Quickly switch back and forth.
There's zero doubt in my mind which feels more responsive/immediate.
I've done these tests 1000s of times.
When triggering drum samples, it's extremely obvious.
If you've got a Vdrum kit, load something like Superior Drummer 3 and play with 10ms RTL.
Now, play Superior Drummer 3 with 2ms RTL.
Guarantee you can feel the difference.
At ~10ms, you can actually feel the Vdrum hit slightly before the sound is generated. That gets distracting/annoying.
At sub 3ms, that disconnect between the Vdrum hit and sound generated/heard is much less obvious.
 
Try playing your thru your favorite amp-sim plugin at 10ms RTL... then immediately at 2ms RTL.

About a year ago i spent days obsessing over ~7ms latency when recording Helix Native in my DAW. Many rounds of tears and ASIO configuration changes, i managed to bring this down to ~2ms, feeling great about it.

Some time later i discovered most config changes were reverted by mistake and i spent months recording with ~5ms latency, while thinking i achieved half of that, and i couldn't tell a difference.
 
Last edited:
At 120 bpm a 1/256th note duration is 7.8ms.
120/60/256*1000

I am not trying to disprove anyone's experience but unless the drummer is a machine he's not playing 1/256 notes on the beat at 120bpm, neither can anyone have a sense of feel/groove for such note length be it digital or analog.

Was the reported 10ms an actual measured analog RTL ie. Drum Pad to Headphones?
Can be easily measured by micing the drum pad, looping back the processed midi trigger sound, and counting sample distance.

EDIT:
Steve Vai can't groove to 1/256 notes at 120bpm either.
 
Last edited:
I am not trying to disprove anyone's experience but unless the drummer is a machine he's not playing 1/256 notes on the beat at 120bpm, neither can anyone have a sense of feel/groove for such note length be it digital or analog.

Honestly... this. It seems most people don't realize how small of timeframes we're discussing here.

Double digits, a.k.a 10ms, is normally mentioned as the threshold where digital latency becomes noticeable for some. For reference, that's roughly the same experience as standing ~10ft (3.3m) from your guitar speaker.
 
Back
Top