Latency test of 23 amp modelers

[Will Chen]

A device has 5ms latency.
I put something into it and it comes out 5ms later.

I set my digital delay for 5ms and 100% wet.
I put something into it and it comes out 5ms later.

Someone says they don't think they could tell 5ms latency.
Someone else suggests using a delay pedal to simulate it.

A device has 5ms latency, I put something in and it comes out 5ms later. No mention of how it is monitored, so I guess it is monitored directly from the device so an absolute 5ms.

I monitor my tube amp from 5 feet away we'll forget about atmospheric impact and just call it roughly 4.5ms of latency plus the 5ms 100% wet delay, so 9.5ms latency in this scenario.

So...what are we comparing again?

EDIT: The only scenario where anyone can beat a low latency modeler via headphones is a miced up amp with the mic kissing the grill, feeding an analog mixer, with wired headphones. In this scenario, the signal in your headphones will hit your ears BEFORE the sound from the cab 5 feet away can travel to hit your ears.

 

[PLysander]

A device has 5ms latency.
I put something into it and it comes out 5ms later.

I set my digital delay for 5ms and 100% wet.
I put something into it and it comes out 5ms later.

Someone says they don't think they could tell 5ms latency.
Someone suggests using a delay pedal to simulate it.

No, the original suggestion talked about relative differences. Even at 5ms, those are relatively easy to pick up. See f.ex.:



Now, if you want to stick a 5ms hard analog delay in your signal path, ask someone to hit the footswitch without you looking. You might be surprised. Or, if you want to simulate the experience, just take 6-10 steps away from your speaker while playing.

 

[12TameMen]

See f.ex.:

I should clarify, whenever I bring up latency it's related to the feel in my hands relative to what I'm hearing that's the issue. That's why just an audio example doesn't work.

 

[PLysander]

I should clarify, whenever I bring up latency it's related to the feel in my hands relative to what I'm hearing that's the issue. That's why just an audio example doesn't work.

Noted. The clip was intended to be a an example of relative delays, which most people can pick up easily - even when that delay is very very small.

A good test for latency sensitivity would be a clip of a metronome at, say, 100 BPM (~590ms between notes), randomly spliced with another of a metronome with 600ms between notes. I think there's good reason you won't find a lot of these on YouTube.

 

[Digital Igloo]

The worst unit here (Ampero II) is clocking at 10ms latency. For reference, that's is roughly what you get standing less than 4 meters away from a speaker.

The human brain is weird. If your eyes see that your playback system is 20 meters away and everything sounds like it's 20 meters away, your brain will compensate. But if your eyes see your playback system is, say, 3 meters away, but what you hear makes it appear as if it's 20 meters away, that can totally mess with your head. And your timing.

Same thing for bands who play huge festival stages, where it takes a while for the drums to hit your ears. Good guitarists and bassists learn to compensate and when necessary, follow what they see the drummer play.

I never tested this but I expect that on helix you get less latency using one DSP path as well

Yep, one path on Helix should have the same latency as HX Stomp.

Is there a scientific study about the minimal amount of latency that is perceivable?

I suspect it has very much to do with the type of instrument played. As someone mentioned earlier, musicians playing instruments with slow attacks (like brass instruments) don't suffer throughput latency nearly as much as, say, drummers. But singers seem to be the most sensitive. I was a vocal producer/editor for my first 5 years in LA and have worked with session singers and rappers who can absolutely, positively hear 5ms latency—it can totally affect their performance, and they're never shy about telling me as much. Once we set up a special monitor mix for them via analog (0ms) or control panel (sub-2.5ms), the problems instantly disappear.

If Steve Vai says he says he can hear 5ms, I totally believe him.

 

[benadrian]

when I saw Leo Gibson’s latency test I immediately knew what will happen here

The Line6 boys will grab the crown out of whoever had it and declaring themselves the winners by any arbitrary chosen reason. The Fractal boys will grumpily argue that latency isn’t a problem for anyone and never was.

Obviously, everyone making digital gear wants to have the lowest possible latency. I will say that al the top end modelers, in my personal opinion, do a fine job with latency while maintaining their top quality sound.

This is always a funny subject. As people have said here, latency really comes into focus when switching a device in and out of the signal path which can add a non-trivial amount of latency. Non-trivial means, to me, something that is noticeable to a player that throws off their hand to ear coordination.

Also, as people have said here, latency is perceived much differently when it's within a closed system. Modelers with an "always on" few milliseconds, standing 15 feet from an amp, etc; all that can get easily lost in funny workings of the brain.

Everyone seems to have a story. Mine is working with very outspoken, analog-loving, retro guitarists. I think people should use, love, and play whatever gear makes them happy. However, I've been in studio situations where guys have gone on and on about latency and feel and how terrible and unnatural it is for them. In al cases they were talking about the time between plucking a string and hearing the sound come back at them. Then, they'd sit in the back of the control room doing guitar overdubs 15-20 feet from the studio monitors (with the amps mic'd up in iso booths on the other side of the studio). I'd never say anything because it's not my job to be their science teacher (unless asked), it was my job to make them a great sounding record, which includes making them feel comfortable and accepted in the studio.

So, from a sheer data and science perspective this is interesting and can be used to push and direct future development. From a musical perspective, it's still a very individualized experience and luckily we all as players can mess with our gear to give ourselves the best personal experience.

Cheers!

 

[Karl Houseknecht]

yet they can all play together cohesively

But they aren't, right? That's kind of the whole point of having an orchestra with multiples of each instrument. The overall sound is filled out by virtue of the sounds of individual instruments arriving at the listeners' ears at different times, and the minute variations in playing by each of the musicians.

 

[Will Chen]

I was a vocal producer/editor for my first 5 years in LA and have worked with session singers and rappers who can absolutely, positively hear 5ms latency—it can totally affect their performance,

I suspect that is because they can hear their voice in their head and the delayed sound in the cans and this is more akin to hearing the relative latency between signals, but who knows.

 

[12TameMen]

A good test for latency sensitivity would be a clip of a metronome at, say, 100 BPM (~590ms between notes), randomly spliced with another of a metronome with 600ms between notes. I think there's good reason you won't find a lot of these on YouTube.

That would convince anyone!

Someone talked earlier about playing behind the beat. This video has ZERO to do with latency.
But for anyone who's not entirely sure what behind the beat means, there's no better example on YT than this.
And as a bonus Beato is a hot topic right now!

 

[PLysander]

The human brain is weird.

Oh, i absolutely agree. Psychoacustics is a fascinating field, from what little i dabbled in. I keep coming back about the difference between absolute and relative sensitivity because i recall being struck by that when i had to study on the subject. For example, the human ear can discern relative pitch with accuracy approaching some strobe tuners, but can't tell apart absolute frequencies differing by ~4Hz.

If your eyes see that your playback system is 20 meters away and everything sounds like it's 20 meters away, your brain will compensate. But if your eyes see your playback system is, say, 3 meters away, but what you hear makes it appear as if it's 20 meters away, that can totally mess with your head. And your timing.

Also agreed. But note that 20 meters implies a delay of ~60ms which is roughly half the average human reaction time to sound. This is 10x to 30x the figures presented in Leo Gibson's video.

If Steve Vai says he says he can hear 5ms, I totally believe him.

I'm just a puny mortal and i have no trouble believing pretty much all claims about Steve Vai, who operates on a completely different musical level than me. I'm just finding hard to believe he'd be bothered by the same amount of latency you'd experience by walking a couple meters away from the stage monitor.

The thing with psychoacoustics is that if you expect to hear a difference, you will hear it, and no amount of evidence short of a true blind A/B test will convince you otherwise. We're all weird machines.

 

[Jay Mitchell]

The human brain is weird. If your eyes see that your playback system is 20 meters away and everything sounds like it's 20 meters away, your brain will compensate. But if your eyes see your playback system is, say, 3 meters away, but what you hear makes it appear as if it's 20 meters away, that can totally mess with your head. And your timing.

The extreme scenario you propose would require that the signal processing add 17 meters' worth of delay. That's 49 milliseconds +/- depending on atmospheric conditions. That much latency is always easily detectable, regardless of any visual cues a person might have.

Same thing for bands who play huge festival stages, where it takes a while for the drums to hit your ears. Good guitarists and bassists learn to compensate and when necessary, follow what they see the drummer play.

Now that is pretty farfetched. It would require that band members constantly watch the drummer, and it would put them substantially out of sync with what's coming out of their floor wedges. Learning to compensate, yes. Constantly watching the drummer, I'd say close to never.

I was a vocal producer/editor for my first 5 years in LA and have worked with session singers and rappers who can absolutely, positively hear 5ms latency—

"Session singers and rappers" are, with very rare exceptions, wearing headphones. As I've pointed out several times in the past, latency in a vocalist's headphone signal causes multisource interference (aka "comb filtering") with the sound of their voice via bone conduction. As you've observed, this can be debilitating; but it's an entirely different scenario than latency that affects someone playing an instrument.

If Steve Vai says he says he can hear 5ms, I totally believe him.

Here's something that, as far as I can tell, has been absent from this discussion: There's a yuuge difference between detecting 5ms total latency - which occurs when your amp is a few feet away - and detecting 5ms - or less - added latency when you're already close to a neurogical threshold. E.g., the difference between ~12ms and 15ms total latency can be easily detected by many people.

 

[PLysander]

Here's something that, as far as I can tell, has been absent from this discussion: There's a yuuge difference between detecting 5ms total latency - which occurs when your amp is a few feet away - and detecting 5ms - or less - added latency when you're already close to a neurogical threshold. E.g., the difference between ~12ms and 15ms total latency can be easily detected by many people.

This is a good point. Now i want to see that Vai quote in context, because i know he routes his guitar through many digital (and analog) devices, both live and in the studio. I can see latency figures adding up to something discernible in such a scenario.

Not a problem for a regular modeler user though.

 

[John Quinn]

I think it's awesome that Leo did this but pretty much all of them have imperceptible latency when used alone.

For me 2 ms is the starting point I hear - then when you get up to 6ms - 8 ms I can hear that plainly - but I doubt these boxes are hitting that numbers - people would complain about the feel being off - and other things.

 

[Jay Mitchell]

Not a problem for a regular modeler user though.

Usually not. However, consider guitar => digital wireless => modeler => powered monitor with DSP. That adds up.

I bought a digital wireless rig ca. 11 years ago so I could do in-the-audience area sound checks while playing. The first time I connected it in my rehearsal space, it drove me nuts. I used it exactly once after that. No way would I want to try to play with the total latency that system had. The modeler's latency was under 2ms, the monitor's about .5ms. I never bothered to quantify the wireless rig's latency, but, whatever it was, it put me well beyond my tolerance threshold.

 

[12TameMen]

No way would I want to try to play with the total latency that system had.

Not directed at you.

Raise your hand if at some point in the past you thought maybe a Bluetooth speaker would be convenient for guitar so you tried one......

(hand raised)

 

[Digital Igloo]

The extreme scenario you propose would require that the signal processing add 17 meters' worth of delay. That's 49 milliseconds +/- depending on atmospheric conditions. That much latency is always easily detectable, regardless of any visual cues a person might have.

The specific values aren't the point. The point is that musicians in large venues are able to easily —perhaps subconsiously—compensate for playback systems that are far away. Most touring bands use IEMs these days, but not all of 'em do.

Now that is pretty farfetched. It would require that band members constantly watch the drummer, and it would put them substantially out of sync with what's coming out of their floor wedges. Learning to compensate, yes. Constantly watching the drummer, I'd say close to never.

I didn't say anyone's watching the drummer all the time. But massive festival stages come with their own delay issues, and again, there are methods by which professional touring acts compensate. Bandmates glancing at each other, especially on big hits, is one of them.

"Session singers and rappers" are, with very rare exceptions, wearing headphones. As I've pointed out several times in the past, latency in a vocalist's headphone signal causes multisource interference (aka "comb filterring") with the sound of their voice via bone conduction. As you've observed, this can be debilitating; but it's an entirely different scenario than latency that affects someone playing an instrument.

You're assuming every other situation doesn't also have multisource interference. Whether you're playing a tuba, a guitar, a drum kit, or your vocal cords, your monitoring can mess with your head—and if egregious enough, your timing—if throughput latency is too high. I suppose if studio monitors, FRFR speakers, or a power amp+cab is loud enough to never hear or feel your strings in the room, maybe. But the notion that physical distance in space (where you can SEE and easily compensate for that distance) and latency through a digital system (which is wholly unnatural and sometimes unpredictable) applies the same sort of impact to one's playing is disingenuous. Musicians have been compensating for distance through the air for thousands of years; DAW latency (or latency through digital hardware) is a relatively new problem, and can be very jarring if you're not used to it.

 

[PLysander]

Usually not. However, consider guitar => digital wireless => modeler => powered monitor with DSP. That adds up.

Yeah, i can see that.

Then again, that's likely the live setup for many professional bands on the road, right now.

 

[12TameMen]

The point is that musicians in large venues are able to easily —perhaps subconsciously—compensate for playback systems that are far away.

Absolutely. You start out in a bedroom and then slowly keep graduating up in the sizes of acoustic space you need to fill, all while entertaining. First time I struck a loud chord in a rather large auditorium I nearly soiled myself.

The good ones learn to do it like you said - subconsciously.

Zeppelin in 77 at the old EnormoDome.
seattle kingdome

 

[timbuck2]

do modeling amps(all in one) have less latency?? seeing as everytjing in in one from preamp to effects to poweramp to speakers?

 

[12TameMen]

do modeling amps(all in one) have less latency?? seeing as everytjing in in one from preamp to effects to poweramp to speakers?

A lot of it is the A/D and D/A converters. Each box has to have at least one of each so normally an all in one system is going to be better than multiple individual boxes.