Wow, that are small dimensions!
Taken from your excellent Photos:

For greater completeness, it may be useful to understand the graph of the time intervals between the first two sensors (1 -> 2) and the last two (2 -> 3) while pressing the keys with different weights applied: 150g, 200g, 300g .
The measurement accuracy is about 0.1ms (about 10 complete scans are done in one millisecond).

The octave A0-A1 sounds much softer and is evident from the graphs. Also this fact being persistent with Kawai electronics and with another custom electronics I've used means that probably the rubber strips are the weak point but with Kawai electronics there is no calibration compensation, just a simple midi offset which is another thing.

Furthermore, the graphs are similar to each other to the point that they can be considered the basis for calibration. After the calibration there will then be the various key response curves to have an effectively calibrated keyboard.

Btw, cauldron, when the time interval with a test weight is sensored between 8 and 10 mm key dip in a VPC1 - at which position in mm is the let-off simulation bump?

I would expect the let-off simulation being active somewhere between 5 an 8 mm key dip. This implicates: The lower a test weight, the higher is the influence of the individual friction at each let-off nose, when the time interval (average velocity) is sensored later.

The let-off point is exactly between sensor 2 and sensor 3, after sensor 3 there are a few millimeters of key end. The value was calculated by measuring the electrical resistance (2 MΩ) of each of the three contacts starting from the diode and with the HD74HC138P integrated circuits removed as visible below:

Sensor 1 closed

Let-off start, sensor 2 is here, just after

Sensor 3 closed

Key end

I would expect the let-off simulation being active somewhere between 5 an 8 mm key dip. This implicates: The lower a test weight, the higher is the influence of the individual friction at each let-off nose, when the time interval (average velocity) is sensored later.

YvesTh wrote:

Hi,
Velocity curve is not weight curve...
It is the velocity of the finger hitting the key that creates the velocity of the key and therefore the outgoing midi velocity, not the weight of the finger (your finger always has the same weight). Of course the resistance of the key slows down more or less this velocity during the key stroke and it is important that this resistance is homogeneous on the whole keyboard. However I think that it is by pressing all the keys at the same speed (with a straight edge for example) that we can highlight a bad calibration of the sensors independently of the pressing force.

Translated with www.DeepL.com/Translator (free version)

F = ma
or in long: Force = Mass x Acceleration.
(Mass = 140g = Constant ) + (Acceleration = G 9.81m/s² = constant) = constant Force.

To whom where you talking by the way?

Hello,
I was commenting on the article as a whole and addressing everyone and no one at the same time.
You are absolutely right about your comment. Your formula is correct without resistance but the key has a resistance and the acceleration is not going to be 9.81m/s2 during the stroke. If the key had no resistance it would sink at the same speed with any weight (light or heavy). The reality is that with a very light weight it will not sink, but with a heavier weight it will sink very fast.
However, it is the speed of the key (and its acceleration because it starts from zero speed) that creates the velocity of a midi keyboard or the velocity of the hammer in a real piano.
Another remark about keyboard actions, it seems to me that in Roland action like "Pha-4 standard" the sensors are activated by the key itself, but on VPC1 I think that the sensors are activated by the hammer (closer to real piano). With Pha-4 you have to do almost the whole stroke to have a signal, with vpc1 the hammer is thrown at the beginning of the stroke and a half stroke can product a sound I mean. What do you think about it ?

Translated with www.DeepL.com/Translator (free version)

Hi,
Velocity curve is not weight curve...
It is the velocity of the finger hitting the key that creates the velocity of the key and therefore the outgoing midi velocity, not the weight of the finger (your finger always has the same weight). Of course the resistance of the key slows down more or less this velocity during the key stroke and it is important that this resistance is homogeneous on the whole keyboard. However I think that it is by pressing all the keys at the same speed (with a straight edge for example) that we can highlight a bad calibration of the sensors independently of the pressing force.

Translated with www.DeepL.com/Translator (free version)

F = ma
or in long: Force = Mass x Acceleration.
(Mass = 140g = Constant ) + (Acceleration = G 9.81m/s² = constant) = constant Force.

To whom where you talking by the way?

Exactly, the MIDI encoder only measures the time interval between sensor 2 and sensor 3 which correspond approximately to the last 3mm at the end of the key stroke.

Really? Where are the three trigger points of a VPC-1 exactly?

For example in an old Kawai ES3 the time interval is taken between key depth 5 mm and 9 mm (midi off and on):

The Kawai ES3 has a double sensor. Although the Kawai VPC-1 has a triple sensor I would expect similar dimensions with just one more trigger between 5 mm and 9 mm.

EDIT:
Oh, sorry, I saw your older posting at https://forum.modartt.com/viewtopic.php...47#p983647
It looks like the time interval in a VPC-1 is taken between 8 mm and 10 mm in your photos.

Hi,
Velocity curve is not weight curve...
It is the velocity of the finger hitting the key that creates the velocity of the key and therefore the outgoing midi velocity, not the weight of the finger (your finger always has the same weight). Of course the resistance of the key slows down more or less this velocity during the key stroke and it is important that this resistance is homogeneous on the whole keyboard. However I think that it is by pressing all the keys at the same speed (with a straight edge for example) that we can highlight a bad calibration of the sensors independently of the pressing force.

Translated with www.DeepL.com/Translator (free version)

Exactly, the MIDI encoder only measures the time interval between sensor 2 and sensor 3 which correspond approximately to the last 3mm at the end of the key stroke.
With a stiff bar a constant velocity is applied to all keys in the cluster and the divergence ratios between keys (MIDI velocity) are more significant.

In the worst cases I think that the divergences are due both to rubber strip construction tolerance problems but also to a lack of sensor de-bounce management as highlighted at the beginning of the thread.

Translated with www.DeepL.com/Translator (free version)

EDIT: Hahaha you were talking about the let-off simulation by yourself on the first page?

Yes, I know what the let-off simulation is. It can't be activated or deactivated by a Kawai setting. But now I undestand you just meant you didn't cut those rubber noses. Inverse logic, you kept "Let Off Activated". Thank you :-))

thank you cauldron for your personal insight of the actions, the details worth a lot to me - makes me again think the VPC1 is the best choice for my desired goal.
Lets see what Kawai will reply!

The Action got some "hook" with a flexible counterpart that causes a resistance when releasing the key, to simulate an acoustic pianos behavior at release.
I've read somewhere (I think its in this thread, check the first site) that some people deactivate this.

Because the VPCEditor wont run in my OS (works with WINE but is not recognizing the MIDI connection) I dont know if people refer for a setting that can be deactivated in the software (which I kind of doubt) or if they refer for physically removing it.

EDIT: Hahaha you were talking about the let-off simulation by yourself on the first page?

BTW:
The idea to use points instead of lines makes a lot of sense - looks way cleaner! Never considered it as I blindly followed the mass

Greetings.

In similiar threads a test weight of 500 g per key had been very reasonable (which I don't want to discuss in this thread). Probably your curve would be flatter with that higher weight. But thank you, that you have chosen 140 g to keep it comparable.

You wrote "Let Off Activated". What does that mean?

There are some things that are subjective. I can speak from my direct experience without filters. I can't tolerate MIDI velocity divergences higher than 14 (therefore +7 or -7 with respect to neighboring keys).

I have tried all the Kawai keyboards:
RM3 Grand (MP10)
RM3 Grand II (VPC1)
GF1 (MP11/MP11SE)
GF2 (CA67)
GF3 (CA79)

GF1 is too light for classical use, better RM3 Grand and even better RM3 Grand II. GF2 is even better with a very smooth feel. I prefer GF2 to GF3, GF3 is a little grainier as if you can hear the mechanical elements moving more, someone might like it more. (You can also apply modifications to the mechanics, you have to try them). You don't have many options as a stage piano, no GF2 and no GF3 unless you take the raw action and repackage it as a stage piano.

Yamaha NW-GH3 (Natural Wood Graded Hammer) is also very good for me but it has one problem: there is a strange feeling of the key returning as if pushed by a spring.

The first thing to do is to use the keyboard as an end user. If it performs well for your purpose then there's no need to worry. The type of virtual piano used and the musical style might mask this problem but if you are a classical pianist and if you use a sophisticated virtual instrument with 100 real sample levels (=Vienna Symphonic Library) there are many small nuances that bring out any problems of calibration.
I had calibration problems on two central octaves and I changed the rubber strips and electronics under warranty and things improved apart from the first octave A0-A1 which sounds much softer.
If it becomes difficult or not guaranteed to calibrate with the hardware components (rubber strips, ...) then it becomes relatively easier to compensate for the divergences through a new custom MIDI encoder that implements this function through the software since the Kawai firmware does not does. So official support can help you if you are lucky but it is not guaranteed to work fully.
So I repeat that if the keyboard responds well there is no reason to worry.

The keyboard is weighted differently from A0 to C8 therefore for calibration purposes it is more meaningful to play a cluster of white notes or black notes with a stiff bar to keep the pressure constant on all the keys in the cluster. The third graph has acceptable values between 95 and 105, it does not seem comparable with the other two much more unstable.

Thanks for your reply!
I am a bit too lazy to try again and press registers with different weights, maybe another day - as I think my single key measurements represent quite good that there are some registers that have a tendency within them.
First of all: The (old) third graph was the RELEASE! Value of the Keys - not the Value on Pressing. Im not sure if you saw that?

Second: Ive read somewhere, that a good pianist can play aprox. 10 different velocitys by hand. To me, this translates roughly that 128 MIDI Velocity values offer a precision within a human of 12.8 (or +/-6.4) that all are kind of "one playable velocity" next to another playable velocity. So, if my velocitys would be within this +/- 6 range, I would be more than happy and call it a day.

To better display the min/max values, I split the Black and White keys for my updated graphs: For the white keys the lowest value is 43, my highest value is 70. This means to me that I theoretically could be more precise than the Keyboard itself between the 2 worst keys: Fitting roughly 2 - 3 different playable velocity's within this range. Realistically it is not that worse: The most keys hover around 45 to 60, which would be quite precise enough so I wouldn't be able to play 2 different speeds within this. But at least the few keys that are really off would require some fix. I measured these keys more than 3 times so it was not necessarily a user mistake that I pushed or pulled the test weight more than with all the other keys.

Same for the black keys:
Lowest is 30, Highest is 61 on Press - with the first Octave being a little bit lower and the 5th octave being a little bit higher. In there I see more constant values than with the white keys.

I'm in doubt if the GF3 Action of the MP11SE would have been a better purchase - but I guess I will contact Kawai now and ask for their opinion because the Keyboard is brand new and no key have been pressed more than 100 times yet.
I dont want to return it, because for the price it costs in Europe (~1300€) I think its a great value, as the MP11 would already be in the 2200€ range - and some Lachnit is maybe great in value but I simply can not ever afford this within the next decade

Dont be confused: I updated the charts in my previous post. They all feature now the suggested 0 to 127 scale by User groovy, which makes the graph appear more flat. Previously the third picture was the "Off Chart for All Keys", which I now skipped and always display On + Off graphs in each picture.

Cheers

The keyboard is weighted differently from A0 to C8 therefore for calibration purposes it is more meaningful to play a cluster of white notes or black notes with a stiff bar to keep the pressure constant on all the keys in the cluster. The third graph has acceptable values between 95 and 105, it does not seem comparable with the other two much more unstable.