Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Kawai Novus NV10 - Hands On

That's why you usually get offered a steep discount on the phone and then end up with no piano after sending the money.

That's nothing new on CL.

Does anyone buy unseen to ship on CL? Except for the "I'll pay you extra for shipping to me at a military APO" scams, it's a local marketplace so you'd expect to go and see/try/pick up the product yourself. Harder with a 400lb piano but a pickup would do the trick.

Oh, I love those CL scams. I've been trying to sell my YDP-181 on there. I had one lady (back in December) send me a check (obviously fake) for $1300, asking me to cash it and give $800 to shipper who would ship the YDP-181 to the east coast (from California). Upon delivery, she would send me the rest of the money for the YPD-181. After sending the check, she messaged me asking if I got it. I said yes, I cashed it and bought a bunch of cocaine. And thanked her for making my new year merry ;0

Also I had one guy (military) from the New England area say something similar. I returned his initial message saying don't bother with the scam. He said, no scam, he was very serious. I laughed and said, really, you're going to buy a YDP-181 on the other side of the continent and have it shipped ...

And the low ballers, not really scammers. Their first question - what's your best price? I say, the list price it's fair. Oh, come on dude, what's your best price?? So I ask, what's the most you are willing to spend? They return with a low number, so I return with a number about $200 higher than the original list price. Haha.

Bsharp, you can be PW's resident 419eater

To add to all of this extensive analysis, would one of the lucky NV10 owners mind checking the down weight of the action? I was interested in the N2, but felt the action was rather heavy.

Middle C: 50.0g (a stack of exactly 20 US pennies)

Mass on the front of the key, damper depressed, key drops down to escapement and no further.

The key doesn't drop on its own with 50g, I have to tap the underside of the keybed lightly 2-3x to get it started, but once it gets over the small initial resistance, it drops rather quickly--it would probably do so with one less penny (47.5g) on the stack but I didn't try that.

Wow that was FAST! Thanks so much Gombessa!

Mine is about the same, although slightly heavier at around 53g (I didn't dare tapping the underside more than gently, as people are sleeping). I guess that's inside of the typical variation and/or measuring error.

Yeah, my scale could be off as well. I'd say with different pianos, and different people, halfway around the world, a 3g difference is pretty consistent!

Today I've had the fallboard off of my NV10 again (for unrelated reasons), so I thought I might try to answer this question that was still open:

We had already discussed the answer to the first question, i.e. that no, when doing this experiment, I cannot get a note played. So now I wanted to answer the second question: It is a bit difficult to tell, but from all I could see, I think the answer is yes, the hammer is indeed thrown.

So the question is, why doesn't the note play then?

My theory is, that it is a combination of the nature of the experiment itself, and the way the sensor works.

The sensor, even though it measures hammer movement optically (and not key movement by pressure, as most "normal" DPs), still measures the velocity of the hammer, not the force with which the hammer hits the stop rail.
If I understand the technology correctly, the sensor consists of a fixed optical unit (with a laser) and a shutter that is attached to the base of the hammer shank. To determine the velocity of the hammer, the sensor probably uses the shutter/laser combination to detect two positions of the hammer shank, and the velocity is then a function of the time elapsed between these two positions.

• Very short time: High velocity -> high MIDI value. This is what happens when you play ff.
• Very long time: Low velocity -> low MIDI value. This is what happens when you play pp.
• Even longer time: Velocity too low -> MIDI value 1 (=no note played). This is what happens if you botch the pp and press the key too softly.
• Much, much longer time: Velocity too low -> MIDI value 1 (=no note played). I think this is what happens in the above experiment.

As you can see, for the sound engine, there is no difference between a botched pp where the key was pressed just a tad too softly, so that no note was played, and the last case, which I think is the case that happens in the above experiment. In both cases, the system generates the MIDI value 1, which means, "key was pressed, but no note played".

Why do I think that this last case is what happens in our "experiment"? Because if you look at what happens with the hammer during the experiment, you can see that when you slowly press the key up to the resistance of the letoff, the hammer is already moved upwards quite a bit. (Which is of course totally logical. After all, the letoff is the moment when the hammer is decoupled from the key, so the hammer must already have moved at this point. If it hadn't, and the key was decoupled before the hammer even starts moving, then the hammer would never move at all!).
And since the hammer has already moved upwards quite a bit, the sensor has probably already passed the first measuring position, where the measurement for the elapsed time starts. Only we, while doing this experiment, hesitate so long until we finally press down the key fully, that it takes a (relatively) loooong time for the hammer to finally reach the second measuring position. Depending on how slowly you "feel" your way down to the letoff point and how quickly thereafter you then press down the key, the time between the two measuring positions can be even several seconds (but certainly at least some tenths of a second). And this is of course an elapsed time that never happens during normal play. So while the hammer eventually reaches the second position, the system calculates a veeeery low velocity, hence the MIDI value 1 = no note played.

So my theory is, that the way we are performing the experiment, in combination with how the sensor works, actually makes it impossible to measure what we want to measure, as we artificially increase the time between the two sensor positions so much, that the result must be "no note played".

So is that a problem when actually playing the instrument?

I don't think so. Although my technique is not good enough to reliably play "from the letoff", I assume that when a pianist who has mastered the technique does this, then he (or she) will not slowly and gently press the key down (=taking quite a bit of time) until he consciously feels the letoff, and then, after a reaction time delay (after noticing the letoff notch feeling), finally press down the key fully. That method, even if mastered, would only allow to play maybe one note every second or so. No, I think that a pianist who plays with this technique plays this much more fluidly. He has a "feeling" where the letoff is, without having to slowly approach it as we do in our experiment, so the total of the key press is still relatively quick, regarding the time between the two measuring points of the sensor, which is interpreted as a low, but not too low, velocity, which results in a low MIDI value >1, which in turn produces a pp or ppp note - just as desired.

OK, now poke holes in my theory!

JoBert, thanks for this explanation. Can you elaborate more on the optical sensors bit, if that’s visible? Where are they located, what do you mean by shutters, etc? As you probably know I was researching the possibility to build my own hybrid controller. I’ve contacted a piano technician and we’re currently waiting for a good opportunity for a broken grand piano with acceptable and preserved action. In the meanwhile I researched sensor technology. I’m not sure how AG and NV which is why I ask you about the latter. I also checked the Alpha piano and it uses force sensors that the hammers hit. The force sensors are offered by a German company:
https://www.hoffmann-krippner.com/polymer-strain-gage/

In the materials they say this is a cheap and convenient solution and for my purposes it looks like so, but they haven’t yet answered to my enquiry about price. Wondering if they will ever do since I am just a hobbyist and not a company.

So I need to consider if optical sensors are feasible solution. And how Kawai have implemented it exactly.

Interesting, so the hammer does fly! Thanks for that JoBert.

While I agree there is some timing element or interaction with the the key sensor, I feel if the hammer flies and hits the rail, there needs to be a sound regardless of velocity (after all, it hits the string right?).

What is REALLY interesting is that MIDI registers the hit. CFX shows the key stuck if "played off the jack" with sufficient force, though there is also no sound made. Play it "silently" and no strike is registered. So there is something very conscious being decided here for some reason.

I agree it doesn't affect regular play (at least not my play) but it's interesting to note.

I'm sure they will answer your enquiry, Gene - after all, a lot of single person hobbyists become startup companies at some point, and startups can become volume orderers at some point too...

I found this picture interesting:



Looking at that, this printed circuitboard appears to be able to take a lot of mechanical force! I wonder what the life expectancy of such a material is? That is quite a lot of force and displacement that a piano hammer could create (especially a bass hammer). If you decided to use the hammers that come with your piano action, you would have to compensate for the extra force that would be detected by the strain gage polymer depending on which key it was - that sounds pretty complicated. You mentioned you wanted to eliminate the "grading" anyway, so maybe it would be easier to attach identical hammers, or hammer replacements so that the sensor response is the same for each key (less programming work). But that would require you to adjust the key weighting by removing/drilling key leads so that each key has the same key weight, shank length, and hammer mass (more mechanical work). Plenty to consider there in terms of where you put your time. It's all manageable though.

I'll be interested to follow this project, Gene!

I think this is the point where actual hammer position differs from the lowest sensor on a DP.

On a DP, lowest sensor, whether on hammer or on key, alone can signify a silent press. However, I'm not convinced that the same thing can be said of an acoustic piano hammer. Assuming the lowest sensor position is the point where the hammer meets the string, there should *always* be a strike anytime the hammer reaches that point, regardless of any previous hammer position. A hammer strike is always a momentary "point source," the hammer never rests against the string. So imo there should be a minimum volume which which always sounds if a hammer ever reaches the strike point, even if there is zero other input to consider, right? Maybe I'm missing something about how a piano action works, but that's my thought.

But that's the thing, the piano doesn't know that the hammer hits the rail, because there is no sensor that register a rail hit.
Essentially, there's only a sensor that registers the hammer movement at two positions, and calculates the velocity from the elapsed time (the reality how that is measured may be more complex, I'm not an expert on these sensors). If that velocity is high enough so that the hammer would have reached the string (if there were a string), you get a MIDI value >1, otherwise you get =1. Im sure that this is calibrated quite well so that the minimum velocity required to simulated a "string hit" is essentially identical to the actual velocity in an acoustic piano with the same action.

What we have here is a discrepancy between the measured velocity (=veeery slow) and the actual velocity after the moment of letoff (=fast enough to hit the rail).

However, in my opinion that discrepancy never happens while actually playing the piano (even with advanced technique). So if this happens only in the artificial situation of an experiment, then I'm not losing sleep over it.


That's actually not very visible even with the fallboard and/or the back panel removed, but I can refer to the following picture from the Novus announcement page (I've posted that before):

The upper picture shows the action with the actual sensor bar removed. What you still see are however the sensor shutters that are attached to the base of the hammer shanks (the little black window-like fins on top of the shank bases). If the sensor bar is installed (as shown in the second picture) then these shutters will disrupt the light of the sensor laser (one sensor per hammer) which can be measured and used to calculate the velocity (note that the yellow laser beams in the second picture were added for illustration purposes, and are not visible in real life).

You wrote that while I was writing my response above, so I didn't see it before.

Now I understand your thought process.

My guess is: It would be much too complicated to calibrate the second sensor position to fall exactly together with the rail hit. The tolerances would just be too tiny: Just a tad misaligned, and the sensor does not register a note at all, even though the hammer actually hits the rail, i.e. the rail would stop the hammer before the second sensor position! Then that key would be permanently silent, which would be a catastrophe.
I think that for reasons like this, the second sensor position is reached before the hammer hits the rail. And if that is so, then this second position is only good for measuring velocity, but not good for making assumptions if the string (rail) was actually reached or not. Or rather: That assumption can only be made based on the measured velocity, not based on the fact that the second sensor position was reached.

Yeah I've seen those pics I have wondered before if there is a difference with the Kawai I HAS optical sensors and Yamaha's "continuous" optical sensors though from Yamaha's pictures they also use similar open shutters.

From the pictures there are *two* sensors per hammer. One closed shutter (smaller) and one (larger) open shutter with a window. I'm assuming that the open shutter position is the exact point of strike against the rail (why would it be anything otherwise? You'd just need to interpolate a hit). Which is why I would also think that any time that shutter is tripped, you can reliably register a tone.

Also, remember if you just barely raise your finger from the letoff (a fraction of a mm), you can trigger a tone no matter how long you wait before the strike. That suggests a few possibilities, but it makes a pure timing difference less likely (why register a tone with >>time from letoff but not >>time just above letoff, when the midi signal registers the same input?)

Edit: as to the sensor being slightly before the rail, that shouldn't change things as the rail serves no tone generating purpose (it only physically stops the hammer). The point of tone generation should be the instant (or +x ms) the hammer trips the bottom sensor, just as on any DP, even if the rail is 2mm away. Imo there little point in measuring the end point of a strike and assuming it doesn't hit the string.

It could very well work the way you describe but that really seems non-ideal to me, particularly when Kawai and every other dp maker has demonstrated significantly higher sensor precision with standard DP actions.

I was thinking a lot about how I would implement optical sensing in my DIY project (although saying this it sounds like it’s all so serious whereas I’m really hesitant and rather pessimistic) and I went through almost exactly the same thoughts I read in your last posts Gombessa and JoBert For instance, if having two sensors, one should be very close to where hammer hits the rail but not too close because it might miss real hits and not too far since it would think there is a hit while the hammer won’t actuslly be hitting the rail. And then that even if that’s the case I can actually know from the calculated velocity whether the hammer is expected to hit the rail or not...

A particular problem I think of is how to equalize sensor distances in a DIY project. In a commercial piano they are both manufactured with tight tolerances and probably calibrated. But I would struggle with that. I prefer the Alpha piano approach with strain gages but I’m also thinking if it is durable and whether it’s precise enough, as well as producing instant enough reading. After all, that all sensing thing is far from trivial.

That makes sense. I guess I had still misunderstood what you meant...


I think that is actually quite easy to explain. The first sensor is right there at the letoff point. If you feel the letoff distinctly, then you have already passed the first sensor. If you back off by a fraction of a mm, where you don't feel the letoff as obviously, then you are again "before" the first sensor.
As a result, in the first situation, when you obviously feel the letoff, you are already past the first sensor. If you now hesitate to press the key, you artificially increase the time between the two sensors to such a long time, that the calculated velocity is impossibly slow, so MIDI value = 1.
In the second situation, just a tiny bit before the previous point, you are still "before" the first sensor. You can now hesitate as long as you like, because you hesitate before both sensors. And as soon as you press the key, the hammer will then pass both sensors and the two sensors together can properly measure the velocity, so MIDI value >1.
The problem is, that the letoff point (as represented by the notch that you can feel) is a region, not a dimensionless point. So the position of the first sensor in relation to the letoff "region" is always a compromise. Put it early, when you barely start to feel the letoff? Or put it late, when the letoff notch is very obvious? Or maybe very late, juuuust before the letoff occurs? I can't begin to guess the advantages/disadvantages of these, but the Kawai (Novus?) sensors seem to be placed quite early in the letoff "region" (although that could probably be changed by regulating the letoff?).
(One thing I noticed when I opened the fallboard today: It is very, very difficult to hold he hammer still at the letoff point. Because of the levers involved, even the tiniest movement of the key, just like the fraction of a mm you mention, results in a hammer movement that is obvious to the naked eye - and if I can see it, then I'm sure the sensor can see it. So it is very possibly that such a tiny fraction can result in a movement that results in passing, or not passing, the first sensor.)

And that's also why I'm not losing any sleep over this behavior of the sensors/action: The first situation, where you artificially hesitate between the two sensors, does not really happen during normal play. You cannot be "hesitating" right on the letoff notch for every note, or you could only ever play Andante or Largo . And playing "from the letoff" without that artifical hesitation works fine even with the sensors in the Novus.

The problem with this is what the midi data suggests. If you are at the letoff (past the first sensor), you can still get a midi strike (or not) depending solely on how hard you press the key from that point (though still, no note sounds even if the strike is registered). That means it's impossible (super unlikely) for it to be a 2 sensor system where the first sensor is before or at the letoff.

When you confirmed that the hammer does fly to the rail when played off the jack, it strongly suggests to me that the fact that no note sounds is purely a software decision. Kawai could have registered velocity 2 (or higher) when played off the jack, and could still register velocity 1 if played softly enough for no note to be triggered.

But yes, this is all academic and doesn't bother me in any case; I have yet to encounter a situation in real playing where the nv10 doesn't behave like a real piano