Topic: What is the right acceptable Latency

Hello,

When I run at 128 samples (2,9 msec), I have some dropouts (not so much, but annoying). I hear and see these dropouts in Pianoteq (spikes), but these seem not to be correlated with CPU overcharge (checked with DPC Latency Checker). So, it is probably a soundcard or driver problem. I use a Echo Indigo IOX express card.

Now, at 256 samples (5,8 msec), it seems that there are definitely no dropouts.

So, here is my question: is 5,8 msec acceptable to have a natural "connection" with the piano? What do you think? Maybe the opinion of Pianoteq's team could be helpfull, because after all, this question is important to have the best "playability" of Pianoteq: what's the recommended latency to have the most real sensations?

Thanks in advance

SK

Last edited by stamkorg (27-04-2014 12:50)

Re: What is the right acceptable Latency

If you can't detect it and others listening do not notice it, then it is probably fine.

If you notice it, then it is unacceptable.

I do not think giving you a number will overcome those realities.

Re: What is the right acceptable Latency

ddascher wrote:

I do not think giving you a number will overcome those realities.

Yes but... no The most important is not how I feel the things but how things should be to be right. Pianoteq is a modelisation. My opinion is that the latency is in fact a real part of this modelisation. I don't know if there is a latency on an acoustic piano, but if it is the case I would like have the information.
It is clear for me that 512 samples (11,6 msec) is too much. That is ok, but I don't have enough experience with acoustic piano to judge if 256 samples (5,8 msec) gives sensations compatible with an acoustic feel... That point is my ask

Re: What is the right acceptable Latency

Every mechanical instrument has some degree of latency. Anything that's below 10 ms is not noticeable for us humans.

BTW, your ASIO latency (512, 256, etc. samples) is just one part of the equation. That is not the whole round-trip latency of your system.

http://www.presonus.com/community/Learn...io-Latency

Hard work and guts!

Re: What is the right acceptable Latency

I speculate that a TOTAL latency that is significantly different to that of an acoustic piano would feel incredibly unnatural.

Key strike, action mechanism movement, hammer flight time, string vibrations, soundboard vibrations...
There are probably known times, or some range of times for all of these and their totals.

Time from triggering an electrical sensor in an electronic keyboard to loudspeaker cone vibrations should probably be about the same.

A little research would probably find some high speed camera video timings for acoustic pianos.
Grands & Uprights may be different... perhaps very different.

Re: What is the right acceptable Latency

The latency on an acoustic is not directly comparable to latency on a DP or VST.  An acoustic gives you tactile feedback to your fingers so you're not triggering off only the sound.  While build-in DP speakers can transmit a bit of buzzing feeling to your fingers, it is done slightly after you hear the sound so it doesn't change the latency model.

Hence, the advice given above is appropriate.  Comparisons to other configurations (acoustic or not) often are not equalized in all factors ... so whatever feels good to you will work.

Perhaps the next stage of DP advancement after realistic wooden key action -- vibration applied to individual keys triggered off a different MIDI channel.  If such a DP existed, Pianoteq could include this in the model and send vibration MIDI back to the DP.

Last edited by Mossy (03-05-2014 03:48)

Re: What is the right acceptable Latency

Talking about kinds of latency...   I almost didn't believe when I watched a science TV program that said our brain have nearly 1/10 of second of latency to process visual information.

How few micro seconds can be so important for gamers when they choose a LCD/LED TV, if we all have so huge latency for processing visual information in our visual córtex????

Last edited by Beto-Music (03-05-2014 15:22)

Re: What is the right acceptable Latency

Yup - that's ironic

Hard work and guts!

Re: What is the right acceptable Latency

Beto-Music wrote:

Talking about kinds of latency...   I almost didn't believe when I watched a science TV program that said our brain have nearly 1/10 of second of latency to process visual information.

How few micro seconds can be so important for gamers when they choose a LCD/LED TV, if we all have so huge latency for processing visual information in our visual córtex????

The devil is always in the detail - 'to process information' could mean very specific things.  And often these kinds of figures are seriously out-of-date.  Life gets faster, new generations get faster.

Re: What is the right acceptable Latency

The numbers regarding brain's processing latency haven't changed by much at all.

Hard work and guts!

Re: What is the right acceptable Latency

Ok, is there a way (a protocol) to measure the real total latency, from the hit of a key to the audio out?

Re: What is the right acceptable Latency

Hi all,

In my own experience latency (or how it feels) is dependent on many things.

In live stage those milliseconds matter more than perhaps in home studio: PA system is farther away from your playing position and this increases latency. So if your piano is having little bit too much latency, in stage you'll probably feel it worse.

Most of all it is still about getting used to some kind of latency. If you play a lot with instrument with more or less latency, it will probably feel right to you. Of course latency must be inside acceptable level. I think it's about adjusting our brains... But it may become problematic if you have to chance from one latency level to another. For example, playing a lot with real grand piano and then suddenly with software piano with latency. Our drummer has said to me that it's very difficult to play mixed set of real drums and e-drums (with latency). On the other hand, to him it's no problem to play just these e-drums.

IMO latency is a big part of the picture, how piano emulation succeed to feel like a real thing. Still I think quite many of us have different idea what is acceptable and what is not. Personally I am quite intolerant here. I like to have as low as possible latency with PTQ.

Cheers,
P

Re: What is the right acceptable Latency

Beto-Music wrote:

Talking about kinds of latency...   I almost didn't believe when I watched a science TV program that said our brain have nearly 1/10 of second of latency to process visual information.

How few micro seconds can be so important for gamers when they choose a LCD/LED TV, if we all have so huge latency for processing visual information in our visual córtex????

I think that is why athletes in response games learn to "read ahead" and the very best of them do it a lot better than the ones they beat.

Tennis players for example watch their opponent more than they watch the ball, the ball moves too fast so it will land before they can watch it and track it.
They LEARN what the result of their opponent's movements will be and they get to where the ball will most likely go and what sort of spin will be on it - hopefully, though they lose the point when they anticipate wrongly. 

Watching TV... the delay hardly matters, everything in the pipeline is delayed the same amount, the story gets to the brain in the right sequence, however late that may be.

Last edited by tractor_music (05-05-2014 19:19)

Re: What is the right acceptable Latency

EvilDragon wrote:

The numbers regarding brain's processing latency haven't changed by much at all.

If that's true then as I said, the devil is very much in the details.  Any gamer will tell you that 1/10s (ie. 100ms) is an eternity.  So it really depends what kind of information processing is being talked about, and just quoting a number out of context means very little.

Simple example, it will take us much longer to identify a moving shape, than to recognize a flashing light in the dark.

Re: What is the right acceptable Latency

Hello All,

I wholeheartedly concur with Mr. DDascher that if you can feel the latency, then it's too much; otherwise it's fine.

From a qualitative standpoint, if you are able to hear the sound by time the depressed key hits the keybed (or bottoms out), then latency is not a problem.  I have read somewhere in the distant past that two distinctly produced sounds do not sound as two sources, until the delay is on the order of 25 milliseconds (1/40 of a second). This figure was mentioned as an amount of delay that reverb units had to beat, in order for the reverb to sound as though it was associated with the original sound source (instead of a separate slap or signal).

Cheers,

Joe

Re: What is the right acceptable Latency

From what I read on Piano Forum, about 10ms total latency is maximum acceptable value for most of the people. And I agree with this number. Of course there are some people who don't mind 15ms and there are some who claim 5ms is too much (probably the same who can hear difference between 96kHz and 192 kHz audio :-) Numbers like 100ms are completely wrong of course (when talking about latency).

So general answer is 10ms.

Re: What is the right acceptable Latency

jarosujo wrote:

and there are some who claim 5ms is too much (probably the same who can hear difference between 96kHz and 192 kHz audio :-)

In other words - they're full of -.

Hard work and guts!

Re: What is the right acceptable Latency

stamkorg wrote:

Hello,

When I run at 128 samples (2,9 msec), I have some dropouts (not so much, but annoying). I hear and see these dropouts in Pianoteq (spikes), but these seem not to be correlated with CPU overcharge (checked with DPC Latency Checker). So, it is probably a soundcard or driver problem. I use a Echo Indigo IOX express card.

Now, at 256 samples (5,8 msec), it seems that there are definitely no dropouts.

So, here is my question: is 5,8 msec acceptable to have a natural "connection" with the piano? What do you think? Maybe the opinion of Pianoteq's team could be helpfull, because after all, this question is important to have the best "playability" of Pianoteq: what's the recommended latency to have the most real sensations?

Thanks in advance

SK

Hi. Ive thought alot about this questions about 10 years ago when I started playing with my electrical piano connected to my computer. I have a roland fp-5 and play alot of classical music and jazz.

When it comes to playing chords for a pop song, I might settle for a high latency but when I play fast pieces of classical music I feel I need a latency as short as possible, otherwise I loose the rythm or something like that.

I could definitly play around with 5-6 msc but Ive prefer it lower then that. I upgraded soundcard to get lower, now im at 3msc running an RME Hammerfall hdsp 9632 at 64 buffersamples. Feeling greath about switching to this soundcard instead of my old M-audio delta audiophile 2496 which didnt have that much strenght.

I guess it all up too taste and the music you play. If it feels good, it is good.

Hope you get it just right and play some beautiful music doing it !

Re: What is the right acceptable Latency

stamkorg wrote:

Ok, is there a way (a protocol) to measure the real total latency, from the hit of a key to the audio out?

Two ways:

1.  With a microphone and a headphone, and a recorder:

put the microphone near the keyboard;

Connect the headphone to the output of the computer's soundcard;

put the headphone near the microphone;

Set the computer's output level so that the sound from the headphone is roughly as loud
. . . as the sound of a key hitting the bottom of the keyboard.

Play some notes (high notes -- two octaves above middle C) on the keyboard,
. . . and record both the physical keyboard noise and the sound from the headphones
. . . on the same track.


Use Audacity (or any other recording software) to measure the time between the "key-down" noise and
the tone from the headphone.

Second way:

Using a stereo recorder, record on two tracks:

. . . the output from a digital piano (high pitches) -- assumed "zero latency"

. . . the output from Pianoteq.

Use Audacity to measure the difference in "start time" of a note on the two tracks.

.           Charles

PS -- neither of these ideas is original to me.  I found the second one much easier to use.  When a key "hits bottom", it makes a _thump_, not a _click_.  The "thump" doesn't have a well-defined start time.

Re: What is the right acceptable Latency

I found this today,
http://www.google.com/url?sa=t&rct=...mp;cad=rja

I hope that works as a link

EDIT
OK, I tested it just after I posted this reply.
It works but,,, The browser then asks if you are willing to take the risk of downloading the file and if so what you want to open it with.

It is a .pdf and I think it is harmless.
From a cursory reading, hammer flight times are in the order of several tens of milliseconds to a few hundreds of milliseconds according to  key velocity.
END EDIT

Now I'm trying to remember if sample players and/or pianoteq artificially delay key strikes according to their velocity - it would seem that they should.

Last edited by tractor_music (07-05-2014 17:22)

Re: What is the right acceptable Latency

Indeed, "learn ahead" it's a automatic "kind of compensation" that our brain use to do with images. In the TV programa the science behind it showed a test to someno point in a creen the exact isntant a ball (that moves left to right quite fast) reached the right side. Most people tested pointed a bit before the ball really reaches the right side, cause the brain try to compensate the latence deficience by predicting the next move.

And our body have another latency issue, this one is from our nervous system, since the impulse Voyage in the nerves at a speed of about 350km/hour, if a remamber well. Like a indy race car. If you feel something in your finger it travled by such speed until reach your brain.

With so many latency issues, brain processing for images, sound speed, nervous signal speed and the hardware latency itself...  I ask myself if the diference between these latencies can became relevant when someone starts to play extremely fast and could generate some small confusion.

tractor_music wrote:
Beto-Music wrote:

Talking about kinds of latency...   I almost didn't believe when I watched a science TV program that said our brain have nearly 1/10 of second of latency to process visual information.

How few micro seconds can be so important for gamers when they choose a LCD/LED TV, if we all have so huge latency for processing visual information in our visual córtex????

I think that is why athletes in response games learn to "read ahead" and the very best of them do it a lot better than the ones they beat.

Tennis players for example watch their opponent more than they watch the ball, the ball moves too fast so it will land before they can watch it and track it.
They LEARN what the result of their opponent's movements will be and they get to where the ball will most likely go and what sort of spin will be on it - hopefully, though they lose the point when they anticipate wrongly. 

Watching TV... the delay hardly matters, everything in the pipeline is delayed the same amount, the story gets to the brain in the right sequence, however late that may be.