High Resolution MIDI Velocity (CC#88) (Page 2 of 2)

Well, it suffice to set to round all extra bits to 0 to approximate the velocity.

Between the expanded velocity and the touch acceleration, I felt, heard rather quickly a difference with other pianos and picked this one for that reason. Also, I craft my piano playing and sound according to those details, and I would not go back to anything else for any price. I believe that the acceleration plays a significant role too, even if the number of states seems meager (about 60, according to measures).

Well I can also hear and tell the state of a piano, and it's voicing and the pianist's technique and emotional and intellectual state. I think the more one is professional and experienced and pushing the exigeance, the more, those details appear clearly.

I understand that others may not hear that or nor believe that, that's OK.

What feature of the Pianoteq model - or any other virtual piano - is programmed to respond to acceleration?

You can put me firmly in the skeptical camp regarding the value of high-resolution velocity. Did you know that repeated isolated notes rendered by Pianoteq with constant velocity from a DAW can vary in loudness by 1.5dB or more? I suspect even the most perfectly built and regulated escapement in a real piano driven by an ideally controlled electronic actuator would not be significantly more repeatable than that. It's part of what makes a real piano more engaging than a virtual one, and Pianoteq more engaging that a lot of sampled pianos.

Sorry Pianistically that is an immediately obviously silly comparison. Just four layers Vs 127.
Flawed reasoning in that argument.

I hear you, but the point I was trying to make was just to illustrate that any continuous data set is superior to a sample of discrete values  in a context where the data set is used to produce a sound, more even so with Modelling as opposed to Sampling.  It is particularly audible with a sustain pedal hence the analogy  Feel free to call it a straw man argument.

It is obviously more subtle and less audible with extended midi, but still valid. The 'chaotic' nature of the various interactions, the variability inherent by the imprecision of the fingers don't change the fact that continuous > discrete. Moreover, as  mentioned , an important number of keyboards don't transmit the full range of midi 1.0 data. My own keyboard (MP11) generates values between 25 and 106 which is not that impressive. It is ok when you play with the internal sounds as it probably corresponds to the limit of the velocity samples they recorded, but with Pianoteq, it is definitely inferior to a portable Roland piano I use, that can generate the full range and very noticeable ( one would expect that as you lose 40% of the whole range of  values)  .Continuous intermediaries values would allow me to produce a much better corrective curve.

Hi Pianistically - my MP11 is getting to a point where it's no longer as fresh as it was when new, but even so I'm experiencing quite good precision, from 8 up to 121 in velocity (I do have to truly try for >fff to get there)..

But that said, I find the Pianoteq velocity curve fine to stretch the range well into the physical key action's limits. The engine does interpret things really well, even with old MIDI - but some future MIDI (matching dpiano + vst) could be interesting for sure in future years.

For new users or passers-by, if not getting a full range from your dpiano, just create a velocity curve to fit the entire 0 to 127 in-between for example 25 and 106 or 8 to 121 (or something quite different, to make it 'feel' like you're playing with light or heavy action.. limit the top, or bottom ranges to get clear/crisp low velo tones, or softer fortissimo etc.)

I'm not sure if anyone posting here needs or wants to read that kind of statement (whether agreeing or not) but, and I'm really not sure how much my opinions have changed in some years gone by..

Generally, I'm not against any future improvements to MIDI. There will always be reasons for it, beyond just velocity. I'm happy with what I have though.

Re any kind of future work with MIDI.. I do like the idea that more data can mean 'better', perhaps not just about interpolating nuances better, but in just having more direct anchor-points for creative tools to set 'this is correct velocity' for any given action which is therefore perhaps more accurately computed to output something specific - not a 'tween' or 'guess' at an 'in-between range' - or just by supplying a lot more stochastic data points between smaller 'in-between' states of play.

That may seem like too much - but IDK, there's more to future MIDI spec than just velocity - and the other aspects, like new storage points for new data could be used to include a slew of other things/readings about acceleration or whatever - I don't know if 'better with future MIDI' really has much to do with just covering making velocity more accurate per se..

Although I personally don't crave or feel a need for much more detail inre velo levels, for sure others have much more pianistic prowess than myself.. and there are plenty of inputs in my world along the lines of "I'll take up dpianos when they get to some next level - currently they still feel plastic and unresponsive"..

To me my MP11 really suffices because, maybe I'm not aiming at a top-tier specialist repertoire. I respect the opinions of those who are perhaps many layers better pianists than I'll ever become at this time in my life. But, personally happy with current tools - but going in circles now perhaps.. but I can't disregard someone's informed opinion about "I can feel and accurately attain exacting velocity on real grand pianos, but less so on current dpianos"... and who knows, things could get really interesting for anyone aiming high, if some confluent factors collude from manufacturers and vst creators following new specs to their conclusion - probably over the next half-decade for some common above-low hanging fruit to ripen. But yeah, just off-the-cuff-observation-and-guesswork there.. I really am not way up on a lot of what's happening with anything to do with high-res or some future specs inre MIDI 2.0 etc. But - never say never - I'm sure a lot of us Pianoteq users will be here talking about some new implementations of things which, some will say they cannot 'hear/experience', whilst others will swear by it

I couldn't readily find any freely available MIDI files using high-res velocity, only rendered audio, and no comparative recordings. So I did some testing with manually generated CC88 messages in Cakewalk. As I suspected, the presence of high-res messages was swamped by Pianoteq's inherent variability in any normal performance context. After listening to some recordings where I generated CC88 messages with values matching the velocity of the note (i.e. CC88=55 at velocity 55) and failing to hear any material difference, I created a clip of individual notes at constant velocity with CC88 increasing by one every measure from 1 to 127 and rendered it to audio. I could just make out the increasing trend in peak amplitude amid the 'noise' of Pianoteq's inherent variability for base velocities up to about 10 (I used Pianoteq 7 to eliminate the contribution of the recently implemented individual damper noise).

Here's a screenshot of two stereo tracks rendered from MIDI notes with a base velocity of 16 (with amplitude scales zoomed identically to make the transients more visible); can you tell which has the increasing CC88 values? And how often do you actually play at such low velocities?

Move along, folks ; nothing too see (or hear) here. ;^)

I have two problems with that analysis - the 90dB figure is not for a single note, and the lowest level you can get is not 90dB below that. Every note being played offers 127 different levels, so the net level of multiple notes being played together (or sequentially with overlap) has more than 127 possible gradations. The default dynamic range of the Pianoteq NY Steinway D Classical preset is 40dB, so your linear dB increment becomes only .3dB which is considerably closer to the absolute limit of what a human can detect under ideal lab conditions with extremely close listening to (I imagine) individual, tonally simple and invariable sounds in a largely anechoic environment. And at most we would need just one more bit of resolution to meet that standard.

In any case, my point is not about what's detectable when the only source of variation is the changing value of CC88, it's about whether you can discern that variation in the context of the Pianoteq's inherent variability in level and tone with unchanging velocity.

And I do think it matters whether a performer can exercise any meaningful control to that degree, but that remains undetermined so far as I know...?

My current Roland RD-700NX easily achieves the full range of MIDI velocities. My first keyboard, a Roland RD-300s, also went easily to 127, but was limited to a minimum velocity of 4. Rather than setting a curve with a different slope I would just offset the velocity in my DAW so the effective range was 0 to 123 as I cared more about being able to hit ghost notes and silent notes than being able to shatter glass. ;^)  In between those, I had an Alesis QS8 that also generated the full range of velocities.

This isn't personal for me, either. I only intended to contribute to a friendly and thoughtful debate, and try to shed some light with an actual example. One reason I think the debate is worthwhile is that "more" is not always "better" when it comes to editing. If high-res MIDI had been implemented as an increase in the bit depth of the velocity component in the existing Note On message definition, I'd be more inclined to accept it as being potentiallly beneficial with no down side. But the complexity and potential pitfalls of dealing with velocity information that's no longer tied to it's respective note event when editing MIDI in various ways makes it less inarguably beneficial.

I just did a quick test of the RD-700NX's onboard sound engine which, unlike Pianoteq, generates consistent loudness at a given velocity. My slightly customized "Dry Clear Studio" patch has a dynamic range of about 58dB. With a MIDI file playing very open 6-note chords around the middle of the keyboard with velocity increasing one step every measure, the range of loudness (Youlean-measured LUFS) from velocity 54 to velocity 74 was 5.8dB > 0.28dB/step. Pretty close to the expectation. I deliberately tested around the center of the velocity range because loudness change is typically not linear across the dynamic range - the jumps will be larger at lower velocities and smaller at high velocities.

All that said, the change in loudness with each step was audibly clear to me across the range. I'm quite sure I could reliably tell the difference (and direction) between any two velocities played back to back in a blind test. So there's a point in favor of high-res velocity. ;^)

It would depend on the octave and the frequency response of a given piano in a given octave - I very much doubt this would be even across the keyboard.
This would likely be most audible in the presence region of 4-6khz. Softer piano with a dark treble likely to be less easy across the  bass notes.
This is still nonetheless probably meaningless. Rarely do we listen to single keys alone. It's even less likely that fractions of a decibel between key presses on an individual note could make or break a performance, at this level what you can detect and what affects you emotionally I believe are completely different things (even to those who believe it's likely to be important to them) even in the slowest delicate slow passages, in the world's best hands, from the greatest interpreters of a given composers intentions.
Timing differences and much more gross dynamic differences than this are what makes a difference to a performance.

Piano dynamics are inherently percussive steps anyway. How important being just about able to tell extremely subtle sound pressure level changes at a per key level on one key is to a musical performance well I really don't believe it's anything whatsoever to do with any lack of realism in what we're hearing - that's still to come in even better modeling.

I still want this to be a non issue. Every controller being on HD midi across the board for every control.