need for larger velocity scales, 0-128 to small

[Kayemef]

I've put up a few posts on this earlier in the thread.

The Kurzweil MidiBoard had something like that. It used impact sensors instead of velocity. But we don't really need impact sensors (they sound expensive and maybe a little unreliable over time), as I've mentionned earlier we only need more than the current 2 sensors on the keys.

Thinking "physics-wise", weight is a constant, so if we have the precise movement and speed of the keys, we can then calculate the pressure. Optical sensors would be great... Way better in my opinion that impact sensitve sensors who might need some recalibrating.

BTW, I don't know anything about electronics, I'm just assuming from general and theoretical knowledge.

I still think that if we get enough people interested in this, Korg might eventually consider working in this direction, especially now - they make their own keybeds, they can do whatever they want ! Korg can't change the midi protocol, or solve bottleneck issues with limited MIDI bandwidth, but they can make the most of it at the very source - the control surface itseft.

[mrk]

On some machines Yamaha uses keybeds that can retrigger. Meaning: if you play&hold a note on a piano, the dampers will stay lifted. You can release a little without the damper touching the string, but if you re-press the hammer will hit again (kinda like a sustain per note).

I would really like to see this feature becoming a standard feature on modern synths, as it's not hard to implement.

[jemkeys25]

my motif es seems to have the most responsive touch out of all my boards.

[jemkeys25]

exibity

[Kevin Nolan]

Apologies if this is already covered - I've tried to catch all the posts of this thread but may have overlooked one or two.

While Ozy is broadly correct in indicating about 9-8 well known dynamic levels on the piano and that, on first consideration seems to suggest that 128 velocity levels is adequate, actually, most competent piano players can express significantly more levels of gradation, in say legato passages and trills, for example. But far more importantly is the fact that even within, say a dynamic level of pp, no two notes played by a player sound at the same dynamic level. So to relate the traditional 8 or so dynamic levels digitally to 2E3 - 8 - sampled velocity levels would be incorrect and would sound quite unsatisfactory and obviously non-acoustic. In reality a piano player needs at his or her disposal a significantly larger number of dynamic levels to allow for that human inconsistency, which may even relate to ‘feel’.

So the emphasis should be as much on – what do we need in a piano to hear it correctly (as in historically accepted as correct under best practice) – rather than asking how many velocity levels a finger can reasonably achieve. On real pianos, it’s probably true that the number of ‘velocity levels’ realised in, say a performance of a Beethoven Sonata, is hundreds if not thousands of levels, with for example dozens of minutely different levels for pp alone (for example) , whether intentionally achieved ‘in the moment’ by the player or unintentionally achieved because they cannot retain total consistency in dynamics.

So I would suggest that there is a desire for thousands and thousands of discrete velocity levels, where by a player’s inconsistency and feel sound right on a human level. I’m confident 128 levels is nowhere near enough.

I definitely think that if thousands of velocity levels were sampled and 'sound designed' along with top players, we could then begin to consider 'tone' sampled and realisable through performance on such sampled pianos. Perhaps the V-Piano offers this already? I would expect the dynamic playing experience to be far superior to even the likes of the CP1 for this reason, even if it doesn't sound as authentic as an actual piano (though I've heard it's exquisite all round). But surely for dynamics it must leave any current sampled piano - even the VSL 128 velocity-level sampled ones - standing?

Kevin.

[EvilDragon]

Pianoteq 3 and V-Piano are the only ones that give the smoothest velocity response (without any kind of crossfades) out of any synthesized piano sound. That being said, Pianoteq 3 supports additional velocity information that's being sent by VAX-77 via CC#88, but V-Piano does not. Draw your own conclusions.

[jemkeys25]

hey kevin,
well said, the more pionts of resolution the synth can land on ,the more human sounding the results will be.

like i've said earlier in the thread" we don't know exactly why it sounds better, it just does"

[jemkeys25]

human hearing has far greater resolution than 128 specific points.

[EvilDragon]

You cannot directly link those two terms, they behave differently, because you don't necessarily have just one note with 128 velocity levels. As you play more notes, each with different velocity, the total dynamics is increased. So you get more than 128 velocity levels, globally.

[ozy]

While Ozy is broadly correct in indicating about 9-8 well known dynamic levels on the piano and that, on first consideration seems to suggest that 128 velocity levels is adequate, actually, most competent piano players can express significantly more levels of gradation, in say legato passages and trills, for example. But far more importantly is the fact that even within, say a dynamic level of pp, no two notes played by a player sound at the same dynamic level. So to relate the traditional 8 or so dynamic levels digitally to 2E3 -

of course!

when you play a trill pp/p or mf/f you never hit the keys twice at the same velocity, and reflecting that requires of course more than just 8 levels of definition!

Let's say... 10 grades for each of the 8 classical velocity ranges?

so, when trilling pp/p, let's say, C and D, you'd have 2x10x2x10=400 different combinations of velocities?

We are still at 80 considering ALL ranges. Midi allows for 128!

Can anybody here detect 400 different combinations of p/pp C and pp/p D played in a trill?

It's the same consideration I made before, now translated including the velocity effects of dragged fingers, lazy fingers, ghost notes, personal peculiarities in using the thumb harder than the middlefinger and viceversa.

Still, a scheme based on 128 points of resolution seems to me to work fine.

NEXT:

all of this is about the KEYBOARD.

Now let's go to the sound generator:

Of course 4 or 8 layers are too little.

But getting 128 sample layers for 128 velocity grades (0 = silent, piano left in a attic in a abandoned house. Sample the sound of the rats) doesn't make sense either.

The issue is moving to modeling. what happens when I hit C at 44 and D at 47? The note lenght, key release, etc etc must be taken into consideration.

Would that be replicated by a 128/layers sampler. No friggin way.

My final conclusion:

128 is fair for HARDWARE KEY resolution, it's not been exploited yet deeply.

Curves have far worse resolution than 128

128-layer samples are, on the contrary, not adequante to rendering the sound of an acoustic instrument.

(indeed, remember that when you have a 4-layers sampler you don't connect velocity just to 4 layers. you connct it to 4 layers plus VCA values plus VCF values plus effect [EQ, eg] modulation). That makes for far more nuances that 4.

But to get a "mapping" of a paino behaviour, modeling is necessary.

Do we want better samplers? Nooooo!
We want mo-de-ling! mo-de-ling! mo-de-ling!
We want it all, and we want it NOW!

Either that, or analogue (whose continuous signals CAN answer to the theoretical 128*128 combinations of a two notes trill...)

[peter_schwartz]

(indeed, remember that when you have a 4-layers sampler you don't connect velocity just to 4 layers. you connct it to 4 layers plus VCA values plus VCF values plus effect [EQ, eg] modulation). That makes for far more nuances that 4.

It's about time someone mentioned this.

But to get a "mapping" of a paino behaviour, modeling is necessary.

Ideally, and in theory, yes. Still, that doesn't mean that you can't get perfectly good results from a whooooole lot of gear that currently exists.

Do we want better samplers? Nooooo!
We want mo-de-ling! mo-de-ling! mo-de-ling!
We want it all, and we want it NOW!

Modeling and/or resynthesis baby!

[jemkeys25]

right, when you play a note you not only have the sound , but every parameter that affects that sound, filters,lfo,funs,aftertouch, tuning, on and on, into the deepest pages of editing, and all those many parameters are also rounded to the nearest whole number between 0-127, not able to show their full potential, we are being robbed of the full sonik soundscape that these instruments are capable of, because they still use 30 year old technology. and to settle and say 128 steps is good enough, well i don't agree, and you shouldn't either, would you want your cancer treatment from 1983 or 2011? I know this isn't a matter of life and death, but the principles are the same.

[peter_schwartz]

right, when you play a note you not only have the sound , but every parameter that affects that sound, filters,lfo,funs,aftertouch, tuning, on and on, into the deepest pages of editing, and all those many parameters are also rounded to the nearest whole number between 0-127, not able to show their full potential...

Sorry jem, but that's not correct. If that were the case it would be impossible to generate any electronic sound that had any degree of smoothness in its attack, filter sweep, LFO shape, and so on. Everything would sound like PacMan graphics look.

There's an important distinction to make here. The values available in MIDI messages are much coarser as compared with the extremely fine increments of change that take place in an EXi or a computer plugin during sound generation. Say you have this synth sound that evolves very slowly over time. When you trigger a note via MIDI, the "coarse" values of note number and velocity have their impact on the initial pitch and volume (and maybe filter). From there, the EXi creates this evolving sound with very fine increments of change.

So MIDI resolution and the nature of sound generation in an EXi just can't be compared because the MIDI part of it barely has anything to do with the sound generation itself.

[ozy]

would you want your cancer treatment from 1983 or 2011? I know this isn't a matter of life and death, but the principles are the same.

you are missing something really basic.

I DO use OLD treatments, and CURES from the middle ages.

It's called "analogue", and its most important value is:

using continuous, not discrete, values.

If you are pissed off by 1/127 resolution, you'd be pissed of as well by 1/256 resolution.

If you want to go into serious resolution (1/4000, 1/8000).... that's the minimoog's software resolution.

On a pure modular analogue, there NO "resolution" at all.

The answer is simple. go analogue. welcome to the club.

I sit in front of an analogue driven by a CV keyboard, I coudln't care less of "resolution" and 127th.

This is where I as you: I DO have the daily chance of using non-stepped keys. I DO know the difference. Believe me, I know very little players who could really use it.

Also about programming: you talk of being "robbed" [I don't like that kind of language btw] because "deep" in menus there are 1/127 scaled parameters.

Do you REALLY program new patches fron scratch digging into those parameters? Is that what you do? Be honest, please.

Or is it about just grabbing a filter knob and hoping it goes "gnewwww" faster while you riff?

Because in this case, stop worrying about the limits of your radias a buy a moog.

Another point: you scorn 1/127 resolution in digital menus.

Did you do the math, and calculate how many variation you can get by applying a 1/127 velocity paramenter to a 1/127 envelope paramenter which drives a 1/127 filter cutoff parameter, with a 1/127 resonance paramenter?

Do you now what the math formula is, right?

it's 127*127*127*127 = 260.144.641

And this is ONE voice. I didn't even mention waveforms or vca or LFO or a second env

Call me when you have exausted all possibile variations.

my point is:

since scarce resources are involved in the calculation (adding resolution takes power),

if I was asked to choose between

a) revamping midi, rewriting software, rebulding all the infrastructure, in order to go from 1/127 to 1/256 theoretical resolution

b) spending the same amount of time and money in producing new modeling algorhytms which USE deeply the existing 1/127 resolution capabilities,

I'd go for the latter withour exhitation

[jemkeys25]

yes I know how many combinations there would be, not as many as say 0-1280, 0-12800, and so on. and guys i'm not pissed about anything, the midi manufacturers association spoke about achieving more exspressive instruments by updating the midi code, and it sounds like your trying to talk them out of it, I've been buying keyboards over thirty years, using every incantation of the latest thing, and yes i do program into the deepest pages of sounds, thats what it's there for. I've used non programmable keyboards. , and your math is wrong, you only have 128 possible combinations, the sound and every parameter is rounded to the nearest whole number from 0-127, once you reach 128 velocity ,which is easy enough to do, playing harder won't give you anything, does that happen with say a piano.