New Controller from Pianoteq / Ivory / Kawai (Page 2 of 3)

I have 5 of them and they all have a velocity curve.

For the sake of argument (or lack of one), could we only focus on the ones that do.

There are many that do.

The questions surrounding this apply to the ones that do supply a velocity curve in their software.

I was at NAMM...only got a quick peek at it .... felt really nice... kinda loud there so hard to really appreciate playing with so much distraction. Hope I can find one in a quieter setting soon

If you are addressing my question, let me clarify ...

I am only addressing the velocity curve being supplied by Kawai inside their VPC controller.

That velocity curve has been approved by Pianoteq as a curve the matches their product.

It has nothing to do with an individual's personal taste and theoretically should never be modified by the user when using Pianoteq software.

It is just there to make the connection with Pianoteq more coordinated.

Here is my question:

What is so unique about the Pianoteq software as a product that makes it necessary to make modifcations to a keyboard sound ?   

I would think that Pianoteq expects midi signals 0-127.   Period.   End of story.

I would also think Kawai knows that when they build a controller.

And, not only that ...

Why is necessary to have a different velocity curve for each different piano software product ?

Don't they all expect the same thing ?     That being ... midi signals 0-127.    That is the standard.

What is it, and there must something, that makes this hocus pocus necessary ?

I have been trying for some time now to get to the bottom of this issue but to no avail yet.

All I get is discussion about different users and how hard they press keys.

Let me reiterate .... The velocity curve in the Kawai VPC controller has nothing to do with users.

It is purely a means of coordinating the piano software with the controller.

Let's keep the user out of it.

Phillipe - my experience with velocity curves is more or less along the lines of what you describe. Sometimes I've found best results by making a curve with Max that modifies the incoming MIDI info from controller (in my case a Privia 350). And then I've found that mapping the incoming MIDI from Max to another curve in Pianoteq gives best results. Actually, the Privia itself has 3 velocity settings called, I think, hard, medium, and soft. So starting with one of those, running it through Max, and then into a Pianoteq curve sometimes produces the best results. There's definitely an element of chance involved in the sense of trying various things, intuitive or non-intuitive as they might be. But the end result is the ability to have a keyboard action that more or less responds in some idealized way. For me, that "idealized" way can change from day to day and sometimes more often than that. Which is to say having the ability to tweak is very welcome.

About the high-def MIDI option in Pianoteq ... would it not make sense to transform incoming velocity information to  high-def format as the last step in the chain? With Max, I transform what controller data from the Privia to a high-def curve (using standard velocity values and CC#88 values) that I specify. My taste is to exponential shapes that emphasize softer dynamic levels.

In other words, wherever the incoming MIDI velocity values come from, if they were mapped in Pianoteq to a combination of standard velocity values and  CC#88 values, wouldn't that let us have finer-grained curves with a little more resolution? My experience with this using Max is the extra resolution that comes from this is not dramatic but it does sometimes make the keyboard and Pianoteq feel and sound a little more responsive.

On the other hand, I'm truly not sure that the extra CC#88 values that come from the Privia produce anything that maps directly to Pianoteq. In other words, I think the Privia sound engine uses that extra data for something that only it (and Casio engineers) know about. Generating random CC#8 values (in Max) to go along with the standard velocity values in Pianoteq can produce as result that's very similar to what comes from the Privia keyboard.

Is there anything you recommend to optimize/maximize usage of that extra CC#88 data? Optimize/maximize in the sense that I'm generating it in Max in response to incoming velocity data from the Privia ... so, minimally, the CC#88 data let's Pianoteq interpolate between standard MIDI velocity values.

If any of this has some sense to it, your comments would be much appreciated! Conversely, if any of it has no sense or practical use in your opinion your comments would be much appreciated!

I don't know, maybe many would indeed aspire to creating a velocity/loudness standard, I think it would be great. But it seems to be a difficult task, both from a technical point of view (what measures?) and procedure (how to bring people together to work on this?).

EDIT: this standardization problem is BTW very old and already present in acoustic pianos: there are pianos with a light touch (for example Yamaha) and some other with a heavy touch (for example Bösendorfer).

Thanks Phillipe, let me try to explain again. My comments come from using the hi-res option and seeing incoming velocity values  with three additional decimal points in the PT MIDI window. Although I hear and feel a difference between standard and hi-def resolution, I pretty much get the same result (in sound and feel) whether I use CC#88 values "as is" from the Privia or alternately by generating random CC#88 values in Max and and adding them into the standard MIDI stream. So this isn't a critique of Pianoteq. It's a more that I'm wondering what exactly the Privia sensors are measuring to produce CC#88 values.

As a counter example, if each incoming velocity value (from 0 to 127 in standard resolution) is squared and divided by 128, the CC#88 value could come from the remainder of that (modulo 128) - and then the shape of the velocity map is a curve rather than a line. The map still only has 128 values but those additional CC#88 data resolve the map to a finer grain than possible with standard velocity values only.

So that's the context of my question - suggestions you might have for velocity maps that transform standard MIDI resolution to a finer grain (by using CC#88 to represent modulo values).

Any suggestions you might have much appreciated! Thanks in advance!

I would be interested to hear from anyone here who has played both RM3 AND TP40WOOD. I have never played the RM3 so am not able to compare.

It is noteworthy that the VPC1 is compatible with Fatar triple pedal. (Not sure of spec though).

However, it seems that the FLK LACHNIT boards only accommodate 2 pedals? Hopefully I'm mistaken.

The VPC1 sure looks tempting at €1,299!

Lastly, has Modartt played both the VPC1 and the LACHNIT boards?
If so, what is their/your opinion of both please?

Kindest Regards,

Chris

Does it do high resolution-midi?   I do not find mention of this on the website.

I applaud Kawai for NOT wasting time with high resolution MIDI. It is the second most UNimportant thing in the world to me, with the most unimportant thing being global warming. ;^)

Greg.

Yes, it would more typically be about 0.5dB, which is very small indeed. (this doesn't include the change of timbre, of course)  However, the other aspect is that it is VERY VERY difficult to strike any individual velocity! 128 steps is IMHO more than enough to capture the physical resolution of a human player. Try it yourself using the MIDI monitor in Pianoteq - see if you can hit any individual velocity with any consistency whatsoever, and observe how much each strike deviates from your target. ;^)

On the off chance you want access to tonal shades that lie inbetween the existing 128 steps, those can be triggered in a random fashion by the instrument, by adding +/- half a step of noise.

Greg.

And that attitude is exactly why, in my opinion, the earth is doomed.   The majority of the earths inhabitants live for today and are unwilling to heed the warnings. 

On a similar note ... Did you know that if you place a frog in a frying pan with only an inch of water in it, you can very slowly heat the water to the point of boiling and the frog will not jump out ?    Same principle ... the frog does not notice the water temperature rising very slowly ... and when he eventually does ... the water temperature has drained him of his strength so he is unable to jump.   

We have the advantage of being able to check the water temperature periodically and get a heads up ... but alas .... we ignore it.

Curious...
I wonder if there is a discount for folk (e.g. US) who already have pianoteq licenses with unused registrations.

Also, re: previous comments on the need for a huge hard disk to store Ivory.
Depends on one's definition of "huge", but I think Ivory could fit quite well on a modest SSD and the price of those has come down a lot in recent years.

I am not a good enough player for PTQ anyway.
At my level I need the faster response of olde fashioned analogue synthesis, my hands/mind/brain can't cope with PTQ's latency.
Yes, I have an i7 laptop for it and Windoze 7, I even have a hybrid hard disk with 8 Gig of SSD cache on the front end - it is STILL too slow (for me).
but that is a whole different subject.

Well, I'm sorry that I was flippant. But I can understand how high resolution midi can be valuable in some situations--with pedal response levels, etc. And for modelling in general--being able to get more steps around the level at which a partial reaches an audible amplitude, for example. And, in general just having more amplitude stages, which would change the exact amplitudes of all of the partials in relation to each other on any note and on chords. Wouldn't the difference be like a change in sampling rates, or at least bit depth?

But, yes, on the other hand, I do worry that high-resolution midi will drive someone to create a sampled piano with several thousand layers.

About MIDI:

I have never tried hi-res MIDI (has anyone?) but I have felt that with this 127 level system it's not capturing nuances like you have with real things. Of course it's whole packet which matters: you must have a decent controller, software/instrument to catch these subtle nuances, good sound system. And yes, good enough fingers to perform these nuances. Anyway I strongly believe that this ancient 127 level system should definetily updated to this millenium. And as we well know, it's not only about piano velocity: all synth parameters would benefit from this smoother curve.

This been said, bad latency is something that will easily destroy everything if it exists. This too is very subjective and depends much on personal experiences. At the moment I am 99% of time playing real acoustic piano or hardware piano. This being my "normal" latency, everything which is bigger feels very unnatural to my fingers. On the other hand, there was a time when I used mostly softwares like PTQ and Ivory and I kind of adapted my own system (brain and fingers) to little more latency. This is how it works. Whole thing is to fool our brain that we are sitting in front of the real acoustic grand and our fingers are firing real hammers and so on. Process (to fool) is easier if we don't play real things very often.

Before you even bother trying high-def MIDI, do the simple test I suggested - try to hit any velocity repeatedly, whilst monitoring the velocity with a MIDI monitor. (Pianoteq can do this, of course).  You will probably notice that you simply CANNOT hit any given velocity repeatedly, in which case I assert that standard definition MIDI is already sufficient to capture the physical resolution of a human.

There's another test you can do, which is to see if you can HEAR the difference between standard MIDI steps, for a Pianoteq instrument. Feel free to give yourself every chance of success - increase Pianoteq's dynamic range to max, disable the limiter, set the soft hammer strike strength to zero, and the forte hammer strike to 100%.  IMHO the difference is absolutely NEGLIGIBLE.

No matter how you look at it, standard MIDI is sufficient IMHO.

For instruments that behave in drastic ways, where even a single standard MIDI step can cause a very noticable change of timbre, you may well argue that since a single step has a large audible change, that you are missing out on the timbres that lie inbetween each of these steps, and I agree - that is certainly a possibility. However, if I'm right that 127 steps is enough for a human's physical dexterity, that simply means that the best you can do to trigger the "intra-steps" is to trigger them purely randomly, and you may as well just have the instrument generate those intra-steps in a random way, by adding +/- half a step of random noise to the standard velocity, to create the quasi high-res velocity signal for this wild instrument.

Greg.

The general argument for hi-def midi is that standard midi is only 8 bits of dynamic range whereas human hearing is sensitive to dynamic ranges of perhaps 18 bits which quite a mismatch.   You can find some high def midi files on the Ecompetion site:  http://www.piano-e-competition.com/midi_2011.asp   It might be interesting to listen to these with and without the high-definition enabled (in Options->Midi you can select the dialect of Midi, hmm I see there the Ecomp midi files are in XP format which is actually not quite the same as hi-def midi which uses controller 88).  At any rate the Ecomp XP files probably can give you some idea of what hi-def midi offers (or not).

I totally agree with you. In fact, the 63_64 test might convince me we don't even need 128 velocity steps. Related to that (or maybe not) is no one plays a piano such that certain requisite velocity values have to be hit dead on - dynamics are relative. A pianissimo today might be a piano tomorrow (no pun intended!)

Another test might be to see how much difference in velocity values are needed before a difference in dynamics can be heard. In other words, what's the threshold where we hear the difference? But things (other than MIDI velocity values) could influence that result too ... Re: establishing a threshold - another experiment could be to add CC#88 values to passages see if they can be heard (and if so when and where).

My experience is high-def with rapid single line passages doesn't add extra dynamics, at least not that I can hear. But it sometimes make Pianoteq sound a little less digital. A little. In testing the Privia 350 through headphones in a store, the sales person and I noticed that high-def made rapid single note lines seem more lifelike and "warm" than they would sound w/out high-def. It was subtle - and the description is subjective. Some might describe the difference as "negligible." Another sales person in the store (the owner actually) didn't hear any difference at all! For him, the difference was LESS THAN NEGLIGIBLE! Perhaps a good reverb or other processing external to PT could get to the same result (a warmer sound).

These two articles (in academic jargon) may be interesting:

http://www.speech.kth.se/prod/publicati...s/3095.pdf
http://www.ncbi.nlm.nih.gov/pubmed/11508980

They're about something called "melody lead." Which is why it is that a melody stands apart from an accompaniment and how understanding that might have practical results in computer-modeled piano playing.

... I think the test you set up are an excellent start ...

Chris, may I ask if you happen to recall the model name/number of the Kawai instruments you auditioned?

Just to add my thoughts on the higher than 7 bits velocity (0-127 is 7 not 8 bits), I looked at the difference between the output of a MIDI XP file from the Disklavier and the standard one of the same piece. There is a small visible difference in the wavefom, but try as I could, I can't hear it.

Also if you look at the velocity values in the MIDI output pane as a file plays, you will notice that the two least important digits are 00, 25, 50 or 75 while the first decimal has all 10 values, giving 128*10*4=5120 different values, a far cry from the theoretical 16256 steps announced in the CC#88 spec.

I think that if Yamaha outputs only these, it is because of the lack of a sensor being able to capture more than 10 times the normal 0-127, the other digits looking more to me like an arbitrary interpolation. On top of that, there is the velocity curve that changes the audible way the resolution is dispersed.

So actually, I believe that except if a much more precise sensor becomes available in the controller, we are currently limited to around 10 times the current range, which is an inaudible improvement, at least for me...

the privia 350 outputs a full range of CC#88 values (14-bit). the disk clavier is a different beast with far less resolution than 14-bits. in the pt midi monitor window we can choose high-res, standard or diskclavier xp.

i think you have to try this stuff, hands-on w/a keyboard. otherwise it's like trying to compare and discuss  keyboard actions.

IMO while it's true that I cannot control the exact velocity with my fingers - nor in 0-127 mode or in hi-res - this does not prove that bigger range does not matter. More smoother curve might still give more "life" or "depth" to performance or playing feeling. Also, there's also a chance that wider range can have effect in subconscious level. I know this can sound speculative and it is in fact, but I remember reading that our perceptual system is perhaps more sensitive lower level in perception. We see and hear more nuances than we think. To prove my thesis, you should have a test to choose more "lively" performance of two MIDI renderings of same performance (one in normal and one in hi-res). I wouldn't be suprised if we don't hear that much difference but actually feel "something" and pick up hi-res... But as said, I am just guessing here.

Is there maybe some equivalence to sampling: it's said that CD freq 44000 is some kind of perceptual limit but still quite many want to use and hear bigger sampling freq. Are they just imagining things?

I don't really follow your logic here. IMHO we should strive to allow the musician to have as much control over the sound as possible, and if high resolution MIDI were shown to provide more control, I would be all for it. If repeated notes sound too similar, then it's a simple matter for the musician to vary the key-press velocity a bit from note to note!  Just btw, if the number of steps is too few, sometimes repeated notes would have too much variation from note to note, because the player would not always be able to strike the key right in the centre of the velocity range corresponding to any particular "step" - they might be playing very near the boundary between steps, such that two adjacent steps would be triggered in a random fashion from note to note.

Greg.

I think I understand what you're saying now. There are two cases to consider:

1. The player is capable of playing successive notes consistently enough on, say, a real piano, which has infinite "resolution". In this case, we need a MIDI resolution that is high enough not to cause audible differences between steps.   For the FXP I uploaded, it seems that standard MIDI will suffice.

2. The player is not capable of playing successive notes consistently enough on a real piano, and there is more variation than the player desires.  In this case, we need to somehow process the velocities to reduce the variability.  We can try reducing the MIDI resolution, however as I said before, there is a risk that the player will sometimes play too close to the edge of the steps, and will trigger TWO steps, not just one. Over time, though, the player will learn the behaviour of the instrument, and may be able to play near the centre of the steps, when he wants every note to sound identical. Another way to achieve this would be to give the player a knob to select the desired velocity, and then throw away the velocity information from each note.

Greg.