latency

Question borne out of curiosity and possibly boredom: Do folks here that use both hardware and software notice latency with their softsynths? It's been ages since I've actually gone hands on w/ hardware, but I always feel like there is something more immediate/direct than when using VST's - even as the time between keystroke and sound has shortened to what is arguably imperceptible limits. I don't know, it could be apophenia, but I still feel like there is . . . I don't know, some very small but real delay. 

Anyone have any actual numbers on this/science? Is there some threshold beyond which a human can't identify any delay? 

Welcoming real but also inane/snarky answers to these questions  . . .. 

I remember reading stuff about that a long time ago and my conclusion is basically that you start noticing at around 20ms. For example, if you are playing a keyboard and the sound comes out of the speakers more than 20ms later, then you are in the "this is affecting me" territory.

On the other hand, I also read that people can generally detect phase issues and tempo mismatch (i.e. if you got hihat-bass drumm out of sync) at around 4ms.

Personally I have played around with the latency setting in Ableton Live to see what happens if I go really high. The numbers there are in samples so you have to calculate how many millisecond is your 64 samples at 48 kHz.

To put into some context sound travels about 7m in 20ms, so for any instrument there is latency, even an acoustic guitar would have a latency of about 2ms.  If you amp a guitar up from the other side of the room there's 20ms.  I run at a latency of about 6ms and I don't notice, though my studio is 'modular' so depending on what my routing is this can increase to noticeable levels.  There's probably more latency between my fingers coming into physical contact with the keyboard and the pressure sensitivity triggering a note.  What (I think) I can notice more is sloppy timing.  

According to epistemological research, our minds can't perceive chronological order on a scale below 600ms, so any two things happening anywhere within 600ms of each other would appear to us to be happening at pretty much the same time.  We could probably discern two separate events from each other, but not accurately enough to tell which happened first.

I often have to compensate latency of rendered audio that I continue working with out of sheer lack of CPU power. Logically, all lookahead stuff introduces latency, but CPU intensive stuff recorded live does, too. It's a pest but I learned how to mostly work around it. I often use SpecOps, which at high FFT size introduces a shitload of latency to be compensated. At some point I made a track in which a lot of the rhythmic variation comes from switching between the track and a slightly delayed version of the track, which was created making creative use of this mostly unpleasant phenomenon of latency.

  On 10/15/2019 at 10:46 PM, drillkicker said:

According to epistemological research, our minds can't perceive chronological order on a scale below 600ms, so any two things happening anywhere within 600ms of each other would appear to us to be happening at pretty much the same time.  We could probably discern two separate events from each other, but not accurately enough to tell which happened first.

Maybe for visual perception but I defy you to find anyone who can't tell the order of two sounds that occur more than half a second apart from each other.  In acoustics, 20ms is the amount of delay between a direct sound and its reflection before the brain perceives them as two separate sounds - shorter than that and we hear comb filtering.

Latency starts to get bad enough to make it hard to actually play somewhere in the 4-6ms zone for me, but when I've tried playing with other people while monitoring through an interface, even latencies that are too short to perceive still affect the tightness of the performance (compared to acoustic or directly amplified instruments).

O, I was wrong with that comment.  600ms is actually the amount of time that it takes the brain to process sensory information.  So I guess that's the latency time of consciousness.  We're pretty slow.

  On 10/16/2019 at 5:22 AM, drillkicker said:

O, I was wrong with that comment.  600ms is actually the amount of time that it takes the brain to process sensory information.  So I guess that's the latency time of consciousness.  We're pretty slow.

That makes sense, I've read a bit about that stuff.  I think it depends on the sense - IIRC sight and especially touch are quite a bit slower than hearing. We process sound with microsecond accuracy, even though we don't consciously notice it.

The book I'm reading sez it's the same figure for both sound and sight

I find 20ms definitely makes a difference on drums. I'd love to use software with my hardware but until the technology is there can't be effed to sync it all up. Would love it if the cirklon could figure out latency per channel using a dongle like I've seen Colin working on for pitch tracking. Would get back into max then (thinking aloud maybe someone has done something in max that would do it the other way around tho) 

One more thing I remembered is that I absolutely believe that your brain can compensate for the latency really well, up to a point of course. It takes a bit of practice but you could play the same thing on a keyboard, when the sound comes 20ms or 60ms later. It's always nice to have a lower latency though..

This of course does not work absolutely in a live band situation where you need to react to people doing stuff.

  On 10/16/2019 at 5:22 AM, drillkicker said:

O, I was wrong with that comment.  600ms is actually the amount of time that it takes the brain to process sensory information.  So I guess that's the latency time of consciousness.  We're pretty slow.

Slightly OT, but I once read and article that suggested that the experience of Deja Vu is actually the brain lagging behind your perceptive array for a period of time. I.e., you experience something through your senses and the brain registers it as happening, but then you become "conscious" of it happening a few ms later. So it seems like you are remembering something that already occurred because, in a sense, you actually have. 

Thnx for all the replies here btw, fascinating to get diff takes/some science behind this.