Latency of amp modelers

[Rex]

Could someone explain in layman's terms so my small brain can understand how this is possible? I believe Cliff is telling the truth, I just don't understand how it works that a 20 ms IR doesn't add any latency.

You can have 20,000 ms of reverb tail, but that doesn't mean that you have 20,000 ms of latency. IRs are the same way.

 

[stdout]

I'd be very keen on knowing more about this, because to my knowledge it contradicts most of the scientific literature.

As you say - a paper published in June of ‘21 reports that subjects can detect half a millisecond delay in sound. And for binaural hearing which our brains use for things like sound location and selectively picking out sounds in an auditory scene, our auditory brain functions at the level of 10’s of microseconds.

‘Of Sound Mind’, a very approachable book by Nina Kraus subtitled ‘How Our Brain Constructs a Meaningful Sonic World’ and was published in 2021, has citations for a host of the published literature. Kraus heads the Brainvolts lab at Northwestern University. Among the criteria she set out for her research were

• a biological approach that could reveal sound processing so subtle we aren’t even conscious of it.
• be able to capture how the sound mind processes sound ingredients like pitch, timing, timbre, …
• to get this information without requiring active participation from the listener.

 

[mr_fender]

Some IR's, like some of the FullRes IRs, have a bit of silence at the beginning of the impulse. Those do add a bit of latency. Any alignment adjustments made there in the cab block can also add latency as well.

 

[sprint]

Some IR's, like some of the FullRes IRs, have a bit of silence at the beginning of the impulse. Those do add a bit of latency. Any alignment adjustments made there in the cab block can also add latency as well.

which, if already using hard panned L/R IRs, is a cheap way to get the haas effect - just crank the "distance" of the IR on one side.

 

[FractalAudio]

As you say - a paper published in June of ‘21 reports that subjects can detect half a millisecond delay in sound. And for binaural hearing which our brains use for things like sound location and selectively picking out sounds in an auditory scene, our auditory brain functions at the level of 10’s of microseconds.

‘Of Sound Mind’, a very approachable book by Nina Kraus subtitled ‘How Our Brain Constructs a Meaningful Sonic World’ and was published in 2021, has citations for a host of the published literature. Kraus heads the Brainvolts lab at Northwestern University. Among the criteria she set out for her research were

• a biological approach that could reveal sound processing so subtle we aren’t even conscious of it.
• be able to capture how the sound mind processes sound ingredients like pitch, timing, timbre, …
• to get this information without requiring active participation from the listener.

That paper is talking about group-delay. Group-delay is NOT latency. It is phase distortion which is something entirely different.

Latency has zero phase distortion. Research has repeatedly shown that humans cannot detect latency of less than 5 ms.

 

[stdout]

That paper is talking about group-delay. Group-delay is NOT latency. It is phase distortion which is something entirely different.

Latency has zero phase distortion. Research has repeatedly shown that humans cannot detect latency of less than 5 ms.

This is the abstract for the 2007 paper ‘The Effects of Latency on Live Sound Monitoring’ presented to the AES.

A subjective listening test was conducted to determine how objectionable various amounts of latency are for performers in live monitoring scenarios. Several popular instruments were used and the results of tests with wedge monitors are compared to those with in-ear monitors. It is shown that the audibility of latency is dependent on both the type of instrument and monitoring environment. This experiment shows that the acceptable amount of latency can range from 42ms to possibly less than 1.4ms under certain conditions. The differences in latency perception for each instrument are discussed. It is also shown that more latency is generally acceptable for wedge monitoring setups than for in-ear monitors.

 

[yek]

This is the abstract for the 2007 paper ‘The Effects of Latency on Live Sound Monitoring’ presented to the AES.

A subjective listening test was conducted to determine how objectionable various amounts of latency are for performers in live monitoring scenarios. Several popular instruments were used and the results of tests with wedge monitors are compared to those with in-ear monitors. It is shown that the audibility of latency is dependent on both the type of instrument and monitoring environment. This experiment shows that the acceptable amount of latency can range from 42ms to possibly less than 1.4ms under certain conditions. The differences in latency perception for each instrument are discussed. It is also shown that more latency is generally acceptable for wedge monitoring setups than for in-ear monitors.

As I read it, that study is about the effect of latency when combined with the original signal. Yes, latency of just a few ms (delayed signal) can cause a comb filtering effect when combined with the original signal. But that's not the discussion here.

 

[Fenderbendr]

I recently tried a Helix again, and to me, it is noticeably more latent than the FM3 under my fingers. Even with just an amp and an IR, it felt very stiff and not nearly as immediate as the FM3, no matter what amp I used. The feel of the Fractal amps are much bouncier and better too. I can’t believe people think the Helix sounds just as good. Maybe if you add three compressors and two EQ’s with “secret settings” to the Helix you can get it close, but I’m not interested in that.

I can pull up an amp and a cab on the FM3 and it will sound and feel great.

I know the test results say the latency is similar, but to me, it doesn’t translate to a similar feel and response as the FM3.

 

[sprint]

p

I recently tried a Helix again, and to me, it is noticeably more latent than the FM3 under my fingers. Even with just an amp and an IR, it felt very stiff and not nearly as immediate as the FM3, no matter what amp I used. The feel of the Fractal amps are much bouncier and better too. I can’t believe people think the Helix sounds just as good. Maybe if you add three compressors and two EQ’s with “secret settings” to the Helix you can get it close, but I’m not interested in that.

I can pull up an amp and a cab on the FM3 and it will sound and feel great.

I know the test results say the latency is similar, but to me, it doesn’t translate to a similar feel and response as the FM3.

I'm still perplexed wrt how much a "lively feel" has to do with latency. At one point I was convinced that higher latencies of 12ms or so always result in a dead feel, however, I'm unsure again having discovered some configs (ie running some plugin presets in 4cm with Axefx) that have a good feel despite not-so-great latency (12-13ms). And in the opposite direction, I've played directly thru some real amps that felt stiff and lifeless despite 0 latency. From my experience so far, I'd say latency is a factor in a lively feel, but there seems to be other variables, to the extent that some good tones can be had with mediocre latency (maybe some fx combinations that cover it up? dunno).. Almost universally though, once latency hits 15+ or so, whatever I'm playing through starts to feel dead. For me, being able to hear a latency gap is irrelevant - the feeling of it is what's material wrt guitar imo, so I don't see the point, wrt guitar, of studies that focus on peoples' ability to hear a gap. Though, I guess perception of "feel" is related back to hearing, if I was to study the effects of latency on guitarists I'd be inclined to set up a series of presets at differring latencies and fx combinations and start quantifying multiple players reactions to those presets via terms like: liveliness, bouciness, dead, stiff, connectedness, expressiveness..., rather than focusing on ability to hear a gap. Maybe some of these studies do this - I haven't read them extensively.

 

[FractalAudio]

This is the abstract for the 2007 paper ‘The Effects of Latency on Live Sound Monitoring’ presented to the AES.

A subjective listening test was conducted to determine how objectionable various amounts of latency are for performers in live monitoring scenarios. Several popular instruments were used and the results of tests with wedge monitors are compared to those with in-ear monitors. It is shown that the audibility of latency is dependent on both the type of instrument and monitoring environment. This experiment shows that the acceptable amount of latency can range from 42ms to possibly less than 1.4ms under certain conditions. The differences in latency perception for each instrument are discussed. It is also shown that more latency is generally acceptable for wedge monitoring setups than for in-ear monitors.

This paper is talking about combining direct and delayed sound. It says:
"Another imposed constant is the amplitude ratio of monitored and direct sound. In order to account for the most artifacts arising from small amounts of latency."

So the crux of the study is how much latency is necessary for it to be detectable when combining direct and delayed sound. IOW, how much comb filtering is perceptible in a live scenario when there are two sources of sound: one direct and one delayed.

This is the problem with the internet and videos like in the OP. Now latency has somehow become the most important feature of an amp modeler even though it's been long established that most people cannot detect anything below 10ms and even the most attuned performer can't detect below 5ms. And people are determined to "prove" that sub-ms latency matters and go to great lengths to find evidence and use papers as evidence of this even though the papers have nothing to do with I/O latency.

We now have two papers cited in an attempt to prove that latency below 5ms is detectable and neither of these papers is at all relevant.

 

[FractalAudio]

Oh, and never mind that the speed of sound in air yields about 1ms per foot of distance from your speakers. So if you're 10 feet from your speakers you've already got 10ms of latency. Do you really think another couple ms is going to matter? If so, move two feet closer.

 

[la szum]

As you say - a paper published in June of ‘21 reports that subjects can detect half a millisecond delay in sound. And for binaural hearing which our brains use for things like sound location and selectively picking out sounds in an auditory scene, our auditory brain functions at the level of 10’s of microseconds.

‘Of Sound Mind’, a very approachable book by Nina Kraus subtitled ‘How Our Brain Constructs a Meaningful Sonic World’ and was published in 2021, has citations for a host of the published literature. Kraus heads the Brainvolts lab at Northwestern University. Among the criteria she set out for her research were

• a biological approach that could reveal sound processing so subtle we aren’t even conscious of it.
• be able to capture how the sound mind processes sound ingredients like pitch, timing, timbre, …
• to get this information without requiring active participation from the listener.

Thanks for the shout. Really excited to check that book out.

 

[steadystate]

Reading comprehension is becoming a lost art.

 

[Rex]

Reading comprehension is becoming a lost art.

Understanding nuanced material has been trying to become a lost art for millennia.

 

[steadystate]

Understanding nuanced material has been trying to become a lost art for millennia.

We used to have Shakespeare, now we have twitter.

 

[stdout]

So the crux of the study is how much latency is necessary for it to be detectable when combining direct and delayed sound. IOW, how much comb filtering is perceptible in a live scenario when there are two sources of sound: one direct and one delayed.

Further down in the paragraph you quote it says - Conversely the keyboard, electric guitar, and electric bass instruments have little or no direct sound. Although they will have no physical artifacts such as comb filtering, psychophysical effects may still be present.

And again there’s no effect from comb filtering when using IEMs. But believe it or not I have no axe to grind here and also no latency issues with my FM3. What I do question is whether the 5 ms floor claimed as absolute is indeed definitive. It wouldn’t be surprising if that number originated in telephony where for comprehensibility latency is a major consideration. But musicians are a non-representative subset of the population when it comes to hearing and listening. And even within this subset there is great deal of variability. So when some musicians speak of ‘feel’ when discussing analog and digital I wonder if they may be referring to a subliminal effect of latency perception. But I’ll drop this in good faith now that the acolytes are flocking in.

 

[JoKeR III]

One man's latency is another man's sag.

 

[la szum]

Further down in the paragraph you quote it says - Conversely the keyboard, electric guitar, and electric bass instruments have little or no direct sound. Although they will have no physical artifacts such as comb filtering, psychophysical effects may still be present.

And again there’s no effect from comb filtering when using IEMs. But believe it or not I have no axe to grind here and also no latency issues with my FM3. What I do question is whether the 5 ms floor claimed as absolute is indeed definitive. It wouldn’t be surprising if that number originated in telephony where for comprehensibility latency is a major consideration. But musicians are a non-representative subset of the population when it comes to hearing and listening. And even within this subset there is great deal of variability. So when some musicians speak of ‘feel’ when discussing analog and digital I wonder if they may be referring to a subliminal effect of latency perception. But I’ll drop this in good faith now that the acolytes are flocking in.

I love Carl Sagan's admonition that "Lack of evidence is not evidence of lack."

Or as Donnie Rumsfeld put it, "there are the unknown unknowns." We don't
know what we don't know that we don't know.

We humans do seem to have a tendency towards absolutes and close-mindedness
---even when history has proven time and time again that it is VERY presumptuous on
our part to ever really do so.

I for one appreciate the study, the inquiry, and the ongoing research. Since we
are the subject to the object of all of our studies, discerning how much that is
the case, and to what degree, and the kinds of variability that exists within how
we perceive the "differences that make all the difference" seem like valid lines of
inquiry. To me.

Look forward to reading the book you cited above. Thanks for sharing.

 

[GlennO]

now that the acolytes are flocking in.

With all due respect, insulting the people who disagree with you doesn't prove them wrong.

 

[stdout]

With all due respect, insulting the people who disagree with you doesn't prove them wrong.

I have no problem with people disagreeing with me and I’m willing to be wrong. Cliff’s criticism of my use of the first paper is fair. And I appreciate he took the time to read part of the second paper though I think he stopped short of the point before responding as I pointed out in my last post. On the other hand I do take offense when someone piles on with condescension without adding anything to the conversation. And I don’t think I’d be the first to observe that some of the behavior seen here verges on the cult-like. In my opinion Fractal products are second to none. But I guess it’s my problem that like Groucho, I’d never want to belong to a club that would have me as a member.