Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Kawai CA67/CA97/CS8/CS11 Serious Bug?

First the the keyboard action velocity output it is most likely due to the nature of how the hardware works.

While on PC with gigahertz processors and gigabytes of memory something might be "very easy" to modify it might not be as simple for a digital piano.

For example one might assume the velocity-layer blending DSP in use by Kawai is able to distribute the velocity layers freely over the input range: So for example you have a layer above 120, so move it down to 100, problem solved.

But there is good chance, that this is simply be not possible with the existing hardware. And together with the limited sensor scanning you end up with an inaccessible velocity layer and nothing you can do about it. At least not without redesigning the entire sample-based synthesizer from a ground up. Do you really want to pay for the latter?

The most simple fix for you is to set Local Control to off, then process Note-On events through MIDI OUT and IN via PC. Of course, you will add latency this way.

Its was the most advanced available in a hardware instrument then, as the new Modelled engine is now.

I always have an instrument in my home. Because I play everyday, and enjoy it greatly. I am a consumer of our instruments. I am also as critical as anyone can be when I wish to see improvements.



Again though, you cannot compare a VST and a hardware piano. Some VSt's have up to 16 multi sampled layers covering a greater range than the 4 (maybe 5) level multi-sampled instrument in your Kawai.



But it is not a bug. Call it 'bad design' or whatever you will. But it is operating as intended. That is definitively not a bug. The distinction in terminology is important. Engineers design things a certain way for a reason. Longevity could be one. Limitations of the built in sound engine could be another. Trying to keep the pianos at a certain price point in many cases, comes with limitations of what can be done. This doesn't make those limitations 'bugs'.

This knowledge comes from years of experience of working for a manufacturer of digital pianos.



And Kawai would likely be grateful for the feedback, to possibly implement it in the next sound engine, or a new firmware.

But you're making a feature request. Not reporting a bug.

Jay

There are also very poor quality vsti with less layers, it's all the same principle and you keep making up facts out of thin air like 'Kawai intended this' and 'this instrument has 4 or 5 layers' (even if they are true you made them up out of thin air)

Pocomoto.

I'm not attacking you personally. I am providing a very well educated perspective on your comments though.

I've been involved in digital Pianos for more than 2 decades; the last 13 working with a manufacturer, including product management for our Pianos. This isn't thin air. This is borne of a lot of experience.

I don't know specifically how many layers your piano has. But based on what I know of the Kawai sound engine, my educated guess is likely a good one.

Jay

JR, you're wasting your breath (er: typing). Some people just want to argue ...

(a) Even when arguing over nothing important. None of this velocity nonsense has any bearing on the quality of this (or any) piano. It's pointless.

and

(b) Even when the "facts" are not facts ... as in "cannot reach that last high-velocity sample".

I will state again -- make the MIDI file and use an audio analyser to determine if there REALLY is an inaccessible layer. But that's hard and takes effort!!! Well there's a 100000000000+ post thread (DPBSD) on this forum where people have been doing exactly that for 150+ digital pianos and VSTs. I half-remember running across an analysis on the CA63 which identified 6 layers. If the the 17/67/97 has a similar 6 layer engine, then it's highly unlikely Kawai reserved 17% of the sample space just for velocity 127.

But until somebody does this analysis, this entire thread is just one giant guess.

@Jay Roland: I wasn't taking your response as a personal attack, I just wanted to point out you make assumptions and then state them as facts.

@MacMacMac: Are you playing on this piano? NO, do you suffer from poor dynamic range and expression on your piano? NO. So far all the people who own any of these models and contributed to this thread have noticed the same weakness in forte range while some people who don't own these pianos apparantly like to devalue this observation based on nothing.

@MossySF: I have no previous experience with spectrum analysis, but I believe to hear a timbre change significant enough between 110 - 127, I don't think it is just at 127. The first 1-20 steps are mostly skipped by the escapement (something I noticed recording with pianoteq) so it is somewhere in ~16% of the range left. Honestly if I believe to hear it exactly at 113, I don't know what difference it makes, some people will just keep assuming I didn't hear it right while for people who hear a difference the exact position doesn't matter (for them it just matters the weakness in sound exists).

I have the same experience with my CA17. When I set the touch curve to light and play fff I can hear a sample layer I can't reach when I set it to "normal", even if I hit the keys really hard. It must be an other layer because it has a very different tone, less mellow, and more power.
I don't know if this is "normal" if when put on the "normal" setting, you can't reach the highest sample layer recorded in your instrument.

Tim

You may hear a timbre change but how do you know it's a separate sample versus audio manipulation? This is what spectral analysis is for. If you hear it's exactly at 113, pull up the graphs for 110/111/112 versus 113/114/115 and compare visually. If it's a separate sample layer, you can usually detect the visual differences because the same sample will keep the peaks & valleys in the same order. If don't know what to look for, post the images here and let the wisdom of the PW crowd chime in on whether it's a separate layer or not.

Don't forget that inaccessible sample layers might still be needed for the DEMO function to make the instrument sound good in the store to people who can't play the piano.

Alright here is a spectrum shot of a recording of three relevant notes at velocity 103, 125 & 112. I typed the numbers on top of the image in paint so you see which is which:



The entire difference of what you see between 103 & 125 is not available by normally playing the piano when the touch is set to normal, it is only fully available when touch is set to light+ and as I already pointed out this makes ppp-mp so sensitive it is impossible to maintain a proper melodic line in this area.

Also note on normal there is already an extreme amount of extra force needed to go from 103 to 112 and the difference doesn't look as significant as between 112 and 125.

EDIT: This was recording of default SK-EX sample.

There's a lot of technical debate going on here, and i'm not sure i get it all right. What i think i've got is there is a last layer of fortissimo samples, which are uploaded into a number of Kawai pianos, and which are not usable by player.

If this is correct, then why are they in the piano, given that they surely consumed resources in being recorded and then occupying memory in the piano?

What is the purpose for which they are phisically provided but then cut out from normal use? Thanks

This is the big question. Although the sounds are available, but only on 'light' keyboard touch. That is how the users know it exists. But it is all but impossible to activate in the more useful 'medium' or 'normal' touch setting. That's the problem as I understood it.

There is a pleasant brighter, metallic sound which occurs when playing loudly with the Light+ touch curve / sensitivity setting but not with the other settings - I can confirm the CA97 has this issue/feature also.
It doesn't bother me that much but it would definitely be a good enhancement if Kawai were to fix/amend this in a FW update, as it does seem like this very nice sounding sound layer should really be there for the Normal touch setting.
I hope Kawai James can come of out of woodwork and comment on this.

Good data. Could you add a liitle more info? There is a significant increase in upper partials as the velocity changes from 103 to 112 to 125, which makes sense if the saturation point (the level beyond which it is not possible to make an acoustic piano louder) is set close to 127.

What note were you playing?

Is it possible to send MIDI NOTE-ON messages for a few more samples in the range 100 to 127 - say 100,105,110,115,120,125 and record them in audacity as you have done here? Question: How do you identify a layer change? I don't know what to look for.

I find this interesting and useful info. Thanks.

i should add while fiddling with pianoteq today i was able to hit the keyboard with somewhere above 120 on normal touch and hear the sound is available there, but in order to do that I kept my arm completely straight (so it becomes a twice as long lever) and focused all force on a single key. Of course this has nothing to to with piano playing.

Typically velocity curves on digital/software pianos are just that - curves. The max value should always be reachable and at the same force you apply on the keys. It's only the values in between 0 and 127 that are affected by the curve. For example in a light touch curve you'd be able to easily reach higher values with even a light touch, and then when you increase the force which you strike the keys with, the actual MIDI values won't increase that much, there will be saturation. And with a heavy touch setting it's the opposite: you will reach higher velocities harder, you'd need to apply a lot of force.

What Kawai has implemented (I take what OP states for granted) is instead a change of the max MIDI value produced, without changing the actual curve, keeping it linear, i.e. they apply something like a compression/expansion of the MIDI values range. I'm not saying that's not a valid approach, however it also limits the timbre variation and that's wrong. In fact, I have created a thread some time ago, where I discussed the difference between timbre variation, sound volume dynamics and MIDI touch curve mappings and that it's a common problem I've found in all Kawais I've had: normal touch curve is way too sensitive and jumpy, so I need to switch to heavy, however the actual timbre range is reduced and the sound gets dull. I've said that literally hundreds of times on this forum and there hasn't been much of a reaction from James or any other Kawai users besides few sporadic confirmations and complaints from a very few users.

What is needed here is simply a touch curve editor like the one included in Kawai's VPC1, then everyone can produce her or his favorite curve. Make these curves downloadable and one can even share them.

By the way I still think the existence of an "inaccessible sample layer" is not proven. What the spectrum graphs above show could be the result of a velocity-dependent filter. I am not saying that it is, just that it could be. More analysis in DPBSD style is needed.

Note as well that on an acoustic there are also a number of rather inaccessible "layers" of sound - most of us don't want to break strings so we stay away from the sound they make just before hammers are breaking them.

maurus,

No graphs are needed to 'prove' the existence of this layer, just sit in front of a CA67/97/CS11 etc, put it on 'Light+' play forte and hear it for yourself. It's very obvious.

I'm not an expert on spectrum analysis, but I notice more similarities between the 103 and 112 and some clear differences in 125.

But probably more important I'm not necessarily trying to prove the difference is caused by the existance of an additional velocity layer, anyone can see from this picture the sound of 125 is waaay more fat than from 103 and 112 and if it is caused by an effect or something else I still want to have it.

PS The note was F2