Can You Really Hear 1 ms of Delay? 2 ms? 5 ms? Download this File, and Find Out!

[Anderton]

I got inspired to start this thread from comments about digital latency and timing in a different thread. A/D + D/A conversion creates a delay of around 1.2 ms at 44.1 kHz, but we don't know how much delay happens because of digital processing between the A/D and D/A. So, I decided to create a test setup that involved no DSP or A/D/A conversion.

 

I loaded a one-sample pulse WAV file into a track. I then copied that to another track, but delayed it by 1 ms, and mixed the two tracks together. It's pretty instructive. I couldn't hear any timing delay with 1 ms difference between them. However, I could hear a minuscule "flam" at 2 ms. I'm not sure I would have perceived it as a timing difference if I hadn't known it was there. But as the delay increased, it seemed the delay became exponentially more obvious. Once it hits 20 ms, I can't imagine anyone who plays a percussive instrument being happy to deal with that kind of latency. Even 5-10 ms is extremely noticeable.

 

It's fascinating (well, at least it is to me) to hear the amount of delay caused by these tiny changes. I've attached a 1-sample spike you can load, so you can conduct the experiment for yourself. Just remember this is a teeny-tiny, 1 sample file, so you're going to have to zoom waaaaaaay in to even know it's there. Try it out, and circle back with what you find.

 

(This also makes a good impulse for generating cab IRs from EQs, but that's a whole other subject ) 

[KuruPrionz]

The file you posted goes from "-0.00 to -0.00 instantly. I fired up the SSL 2+ and headphones and hear only silence. Maybe you can track something musical and dupe the track on the other channel with a small delay. That would be closer to the circumstances I was describing (using a digital guitar amp). 

 

 If it's the thread I'm thinking of, I mentioned that I can't hear the latency but I "feel" it. That said, it's obvious that there is hearing involved. At the very least you'd have a high frequency phase shift (probably beyond the range of human hearing but sub-harmonics can be interesting sometimes).

[Anderton]

12 hours ago, KuruPrionz said:

The file you posted goes from "-0.00 to -0.00 instantly. I fired up the SSL 2+ and headphones and hear only silence. Maybe you can track something musical and dupe the track on the other channel with a small delay.

 

I just downloaded the file to check, it's a 4-sample WAV file with a 1-sample pulse at the beginning. When I play it back, I hear a short click in my headphones. If you don't hear the click, I'm not sure what the problem might be. Maybe try moving the pulse a few seconds later on your timeline, and then see if you hear a click when playback goes past the pulse.

 

I could pad the WAV file with a bunch of silence after the pulse so that the waveform is easier to see and move around, but you're still going to need to zoom way in to see the pulse.

 

If I track something musical. copy it, and separate the tracks by 1 ms, I'll hear a flanging effect. The reason for using a single-sample pulse is it's the shortest audio waveform I could think of that would allow me to hear clearly whether there was a delay or not. This test doesn't address the concept of feel. I just wanted to hear the sound of super-short delays, and at what point two sounds go from sounding like a single event, to sounding like they have a definite gap between them. I found it pretty instructive

[KuruPrionz]

16 minutes ago, Anderton said:

 

I just downloaded the file to check, it's a 4-sample WAV file with a 1-sample pulse at the beginning. When I play it back, I hear a short click in my headphones. If you don't hear the click, I'm not sure what the problem might be. Maybe try moving the pulse a few seconds later on your timeline, and then see if you hear a click when playback goes past the pulse.

 

I could pad the WAV file with a bunch of silence after the pulse so that the waveform is easier to see and move around, but you're still going to need to zoom way in to see the pulse.

 

If I track something musical. copy it, and separate the tracks by 1 ms, I'll hear a flanging effect. The reason for using a single-sample pulse is it's the shortest audio waveform I could think of that would allow me to hear clearly whether there was a delay or not. This test doesn't address the concept of feel. I just wanted to hear the sound of super-short delays, and at what point two sounds go from sounding like a single event, to sounding like they have a definite gap between them. I found it pretty instructive

I only hear silence, I've tried it multiple times, direct from my Mac mini, through the SSL 2+. 

 

Also, for my purposes - (attempting to substantiate "feel"), I would need to be triggering the signal and comparing a non-latent sound with a latency of short duration. 

I guess it could be done on the computer but I doubt mine is fast enough to provide 1ms latency playing a distorted guitar tone through a plugin (I also question if the plugin can round trip that quickly!!!). The SSL 2+ provides a "Monitor Mix" knob that blends input signal with the return from USB 2 so that problem is more or less solved when I am tracking. It's the feel when I play live that I am mentioning. 

 

I'm not sure I need further data since I've found a solution for the problem (don't use digital guitar amps). 

 

I may download and mess about with your file at some point. You are correct that running the non-latent sound with the 1ms latency sound will cause some form of flanging. Obviously, I don't hear that flanging when playing any guitar amp with a solid body electric guitar. I don't currently own any digital guitar amps to try it with an acoustic guitar and there is probably a proximity effect caused by differing distances. So it's remarkably complex. 🙂

[Anderton]

44 minutes ago, KuruPrionz said:

I only hear silence, I've tried it multiple times, direct from my Mac mini, through the SSL 2+. 

I've padded the file with silence to make it longer, see if this works. You should hear a click when playback goes past the file's beginning.

 

Craig

[KuruPrionz]

32 minutes ago, Anderton said:

I've padded the file with silence to make it longer, see if this works. You should hear a click when playback goes past the file's beginning.

 

Craig

Got it, that played on my Mac mini speakers. It's a quick click but it's there. 

[Anderton]

2 hours ago, KuruPrionz said:

Got it, that played on my Mac mini speakers. It's a quick click but it's there. 

 

Have fun finding out just how bad 15 ms of delay is!

[KuruPrionz]

28 minutes ago, Anderton said:

 

Have fun finding out just how bad 15 ms of delay is!

I already know that!!! I used somewhere around 11 ms years ago. Faster computer, faster hardware and software and I'm running about 4.5 ms. Yes, you can hear that too. 

But when I blend in the original input signal I can deal with it. 

The file you've uploaded does not have a simple basis for comparison. I suspected it was longer than 1 ms but there is nothing to measure against it without doing a deep dive. 

[RABid]

I remember testing a few things about 15 years ago when I started getting more VSTi's. I could comfortably compensate up to 7 ms, but would notice recording issues when recording MIDI parts using my compensation, then converting those tracks to audio. At that point the latency would be gone and the part would be rushed. 

[Anderton]

5 hours ago, KuruPrionz said:

The file you've uploaded does not have a simple basis for comparison. I suspected it was longer than 1 ms but there is nothing to measure against it without doing a deep dive. 

 

Right, you have to create the basis for comparison. As mentioned at the top, "I loaded a one-sample pulse WAV file into a track. I then copied that to another track, but delayed it by 1 ms, and mixed the two tracks together. It's pretty instructive."  The experment is that you can load the same pulse WAV to a second track, then move it around to create a variable delay compared to the original track. That's when I realized that two sounds separated by 1 ms still sounded like one sound. Perhaps the ear has a characteristic that's like the eye's persistence of vision.

[KuruPrionz]

11 minutes ago, Anderton said:

 

Right, you have to create the basis for comparison. As mentioned at the top, "I loaded a one-sample pulse WAV file into a track. I then copied that to another track, but delayed it by 1 ms, and mixed the two tracks together. It's pretty instructive."  The experment is that you can load the same pulse WAV to a second track, then move it around to create a variable delay compared to the original track. That's when I realized that two sounds separated by 1 ms still sounded like one sound. Perhaps the ear has a characteristic that's like the eye's persistence of vision.

When I mentioned "feeling" the difference of a digital guitar amplifier, I was playing my guitar for a good while. I can only speak for myself, when I am immersed in playing music I do not "think" in a verbal sense, there are truly no words. This is a huge part of the joy of playing music for me, escaping from my own blabbering brain to another place. It's still my brain but it doesn't verbalize, I don't think of things in terms of the names of the notes, intervals, or any sort of verbalization of expression. 

It just "flows" through me, I can't really explain it in words since they are not there when it is happening. 

 

Point being, perhaps placing a waveform in a single "blip" and adding a second one slightly delayed from the first is a different thing entirely in terms of the way I perceive it when compared to the act of immersing into music. The "blip" is more of a scientific approach and less of an inexplicable wonder (if that makes any sense?)

 

[harmonizer]

For a point of comparison: The speed of sound is pretty close to 1 foot (30cm) per ms. When I do the audio mix of our covers band, I delay the kick track by .003 or .004 (3ms or 4ms) because the kick sound also appears in the overheads. Kick mic is in front of the drum, about 1 foot from where the beater strikes. The overheads are 4-5 feet above where the beater strikes, and get an opposite phase kick sound coming out the back (I reverse the phase of the kick mic track). After I delay the kick track by 3ms or 4ms, the kick sound becomes much clearer in the mix.

[Nathanael_I]

This is why Avid can still sell HDX systems for tracking. The new hybrid mode is brilliant. Tracking @96Khz gives .7ms of latency. Done and sorted. Once you take the tracks out of record mode, you can go back to native and typical buffers, where latency no longer matters.  
 

My previous Dante/digital mixer setup was 1.7ms from mic to IEM. 
 

concert organists play in huge churches with 9 sec reverb tails and up to 2sec of latency…. You play and keep playing and eventually others hear it. Humans are very adaptable. 
 

But 96khz and DSP mixers get you there. 

[ElmerJFudd]

I’ll play the game and post back.  But the only time I find it bothersome is when playing AU/VSTs live if the buffer is not low enough.  Especially percussive/hard attacks, as we get with piano.  It generally feels quite acceptable at single digit ms.  But somewhere around 10, 15, 17, 20ms or so it starts to feel noticeable. More so than any air travel delay one gets when sitting at an acoustic.  

[ElmerJFudd]

2 hours ago, Nathanael_I said:

This is why Avid can still sell HDX systems for tracking. The new hybrid mode is brilliant. Tracking @96Khz gives .7ms of latency. Done and sorted. Once you take the tracks out of record mode, you can go back to native and typical buffers, where latency no longer matters.  
 

My previous Dante/digital mixer setup was 1.7ms from mic to IEM. 
 

concert organists play in huge churches with 9 sec reverb tails and up to 2sec of latency…. You play and keep playing and eventually others hear it. Humans are very adaptable. 
 

But 96khz and DSP mixers get you there. 

They do, because of the pipes could be quite some distance from the organ.  It’s a known annoyance that they deal with.  Would it feel better if everything were tighter?  Yes. 

[Anderton]

6 hours ago, harmonizer said:

For a point of comparison: The speed of sound is pretty close to 1 foot (30cm) per ms. When I do the audio mix of our covers band, I delay the kick track by .003 or .004 (3ms or 4ms) because the kick sound also appears in the overheads. Kick mic is in front of the drum, about 1 foot from where the beater strikes. The overheads are 4-5 feet above where the beater strikes, and get an opposite phase kick sound coming out the back (I reverse the phase of the kick mic track). After I delay the kick track by 3ms or 4ms, the kick sound becomes much clearer in the mix.

 

Here's a fun experiment I show people at workshops: Have one track with drums, and another with bass. Line up the transients so they start at the same time.

 

Move the bass a couple milliseconds ahead of the drums. It sounds louder, and the music sounds more melodic. Slip it behind the drums a few milliseconds. The bass sounds softer, and the music more percussive. I think this underscores how much our perception depends on what happens in the first few milliseconds of an attack (if it didn't, Roland probably would have never sold any D-50s, LOL).

 

After all...was that grass rustling the wind, or a sabre tooth tiger looking for lunch?

[Nathanael_I]

Another fascinating latency thing….  Synth attack time settings and when you have to play to have it “in time”.  Compared to a piano, the feel is different. Co,posers similarly have to adjust the start time of samples to get the attacks right on string samples.

 

it’s all simple… until it’s not. 
 

 

[ElmerJFudd]

8 hours ago, Nathanael_I said:

Another fascinating latency thing….  Synth attack time settings and when you have to play to have it “in time”.  Compared to a piano, the feel is different. Co,posers similarly have to adjust the start time of samples to get the attacks right on string samples.

 

it’s all simple… until it’s not. 
 

 

Oh ya.  When playing slow strings live you have to strike way ahead of the beat, by feel.  if you sequence where it’s quantized you drag the midi ahead or if you’re using audio you drag the audio ahead.  Actually sting players just bring the strings up to volume with pressure and speed of the bow to make it sound right.  And it’s different per note depending on frequency and our perception - and musicality.  

[Munscio]

Hi. I've been obsessed with round-trip latency for a long time. As a guitar player I concluded that I'm happy with anything less than 6ms round-trip latency. Between 6 and 13ms, I could not hear at all. However, the difference in feel on the instrument is very noticeable!!! Above that, it's pretty unusable...

When it comes to computers (Core Audio/Asio) things are not what manufacturers say. Check this long thread in GearSpace

https://gearspace.com/board/music-computers/618474-audio-interface-low-latency-performance-data-base.html

 

Hope that helps 

Danny Bullo