New Yamaha retrofit silent unit and grade up unit.

RIP Are you considering that now? If it really performs so well, it might be an alternative to the various Cybrid options for people who would like a less challenging but still DIY alternative.

Yes, I am considering that option.

I already have the adSilent v1 system installed on my upright. So I knew exactly what to test when I visited the piano store. And I know exactly what to expect from a change to SC3.

To be more exact, I don’t need the RSC3 because I already have the mute rail installed.

The VSH3 provides only the sensors and control box without the mute rail, thus brings more value to me at a lower cost.

Regarding people considering to build their own controller : Clearly, a key sensing only would already be very good and very easy to do. For the most courageous ones (and in search of perfection), finding a way to use these sensors on the hammers would be terrific. For that, I would try to find pictures of the Yamaha grands using this new generation. Or even visit a store and take discretely some pictures :-)

Paulo164 I already have the adSilent v1 system installed on my upright. So I knew exactly what to test when I visited the piano store. And I know exactly what to expect from a change to SC3.

Thank you for all this additional information, your opinion and speculation of Yamaha motives. It definitely makes a lot of sense.

So has the adSilent v1 have the "loud note problem" like the NU1? Besides that do you find the SC3 a big improvement? It sounds you do, but you haven't commented explicitly in that regard!

Is your adSilent using the "flag" system you describe? IIRC, @CyberGene please correct me if I'm wrong, the N1 uses a light intensity LED instead.

Paulo164 Optical sensors are using lasers measuring the distance to a target (a reflective tape).

Actually LED and they are measuring intensity of reflected light (not time-of-flight of the light which is too fast). But everything you say remains the same.

RIP please correct me if I'm wrong, the N1 uses a light intensity LED instead.

There are gradient shutters (what Paulo calls flags) - small strips with gradual translucency, attached to the bottom of each key and they go inside a photo coupler (I like the toaster analogy), so it’s light intensity that is changed and measured.

But on the N1X there are also hammer sensors. They only interrupt light through metal notches and produce binary on/off signal.

RIP So has the adSilent v1 have the "loud note problem" like the NU1?

Yes, it has. Mine at home as well as the one installed on the Petrof in the store had the same issue.

I contacted Excellent piano install months ago, the German company designing the system, and they confirmed the version 2 had an improved algorithm supposed to detect false hammer return and avoid/reduce this issue.

Initially, I was expecting to test the v2 in the store but the vendor didn’t propose it spontaneously and I didn’t want to be too pushy.

My number 1 goal was to try the Yamaha 3rd gen sensor and compare it to my adSilent v1.

In this regard, the test was quite conclusive.

RIP Besides that do you find the SC3 a big improvement? It sounds you do, but you haven't commented explicitly in that regard!

Indeed, I didn’t used too much words to describe my experience with the adSilent.

This is biased by the fact I already own the system since 2018 and it was not of my interest to evaluate the system, which I already knew perfectly.

So, to explain a bit better how I would compare the adSilent v1 versus the new Yamaha system, here are some of my personal criteria making a silent system “ideal” regarding the acoustic counterpart and the winner for each :

1/ repeatability of same key: when you press a key several times at the same speed, you expect to have the same midi velocity => winner Yamaha

2/ repeatability across keyboard : when you press different keys at the same speed, you expect to have the same midi velocity => winner Yamaha

3/ gradation : when you press a key from pp to ff with equally spaced steps, how gradually does the velocity increase ? Is there a plateau ? Are the output velocity values equally spaced ? => winner Yamaha

4/ inner timing precision : when pressing 2 keys exactly at the same speed and at the same moment (ex: 2 adjacent keys with the same finger), are the 2 sounds generated exactly at the same time ? => Winner Yamaha

5/ outer timing precision : when pressing a key, is the sound of the virtual piano generated at the same time as its acoustic counterpart ? => ~draw

NB : Here it’s a draw if the adSilent is setup to trigger early the note-on event. But by doing so, you force the calculation of the velocity on a shorter course of the key, which reduces precision and increases the probability of “hot notes”.

Does this remain true across the range pp to ff ? Winner Yamaha

6/ Outliers : are there occurrences where the output velocity does absolutely not match the player intent ? How little/great is the probability of such flaws to occur ? Clear winner Yamaha

In case of fast repetition, the adSilent can produce unexpected high velocity values (already discussed). Also, on some keys, when pressing very fast, sometime the note is not triggered or only pianissimo, as if you were pressing the key too fast for the sensor to detect it.

7/ are all of the above qualities able to remain reliably in stressing conditions (ie. fast playing of lots of notes). Winner Yamaha

Here, some systems can struggle processing all the events in time (I am referring to data events before their conversion into midi events). In that regards, it happened my adSilent throttled processing big chords repeated fast, delaying some notes.

I was not able to test the Yamaha system as long as I did with my adSilent but I would say it can handle more data reliably.

To conclude, I can also let you know how long I played each system in the store, which reflects my interest and the pleasure I experienced for each one :

• Yamaha : ~25 minutes
• Kawai : ~10 minutes
• Schimmel : ~ 5 minutes
• AdSilent : ~5 minutes

RIP

RIP Actually LED and they are measuring intensity of reflected light (not time-of-flight of the light which is too fast). But everything you say remains the same.

Yes you are right, sorry !

I was too over-simplifying, which is never a wise choice on this forum 😉

As you suggested, measuring light intensity is an indirect way of measuring the distance, which does not invalidate my previous “theory”.

RIP Is your adSilent using the "flag" system you describe?

No, light intensity.

To my knowledge, the “flags” are used exclusively by Yamaha.

CyberGene But on the N1X there are also hammer sensors. They only interrupt light through metal notches and produce binary on/off signal.

Okay, so the precision of speed measurement itself is the same as with key-only sensors but the hammer on/off sensors add reliability regarding whether the note should be triggered or not and when it should be triggered.

Paulo164 I believe that’s correct. If I’m not mistaken, most of the time it’s the key sensors that determine velocity, however in certain (edge?) cases it’s the hammer sensors that are used instead (or in combination). I believe the reason is, the two hammer sensor points are 10mm apart and that’s too much to be used as the only measurement source because:

1. The only reliable interval for hammer velocity is after the escapement when the hammer is freely moving but that’s 1-2mm and it requires really fast scanning logic.
2. In those 10mm the hammer can re-strike the string and that can’t be detected.
3. In those 10mm there can be varying force, hence distance divided by duration may produce different result than terminal velocity at strike point.

P.S. I’m still not sure how one can eliminate the infamous loud note issue through key only sensors, unless some advanced prediction logic is implemented that models the key-to-hammer mechanical system and can predict that a repeated key is not moving the hammer. Which is why on the NU1XA they introduced hammer sensors too (if I remember correctly).

CyberGene Yes , you are probably right. By today’s standards, scanning rate is already at the lower limit we need to track very precisely the key when moving it very fast.

So tracking the hammer, which moves 10 times faster and on an only 2 mm course, is out of reach with this kind of optical technology.

This is where the new electromagnetic sensors, reacting to fluctuations of the electromagnetic field (so a direct measure of speed/energy) at a very localized point, could enter in the game.

Worth noting that the SC3 is supporting the “grand expression modeling” feature.

From Yamaha website :

The simulation technology that reproduces this variety of tonal changes is "Grand Expression Modeling." As a result, we have achieved expressive power that faithfully reproduces the detailed nuances envisioned by the performer. In addition to changing the strength of your finger's touch from the time you press the key until you release it, you can also change the tone depending on the speed of your touch and the depth of the key you press, so you can express the tone to suit a variety of playing styles, making it an authentic instrument. You can enjoy a unique performance.

As I understand it, this means that for a same output velocity value, you could trigger the note either with a small/shallow key press (which corresponds probably to staccato touch) or a deep press up to the keybed (legato touch ?).

Technically, there may be a second stream of data conveyed along the regular note-on velocity data stream that is processed by the sound engine to alter the tone character.

I wonder if this additional data stream is exposed outside of the Yamaha box, in the standard midi message (maybe on a specific CC like it is for hi-res velocity). And if yes, if program like PianoTeq could take advantage of this. I don’t think so as it is non standard and proprietary format of Yamaha (which does not collaborate with Modartt, as we see no Yamaha nor Bösendorfer approved models).

Anyway, if proven to be effective, this could give Yamaha the edge over even high quality VSTs that cannot access some intricate details hidden in hardware.

I finish just by saying that I was aware of this feature before my visit to the piano store. So I specifically tried different kinds of touch (ie. staccato vs. legato) to see if I could detect a change in the sound. I confess I was not able to clearly notice this like in a A/B test. But maybe the outstanding feeling I had when playing the SC3 was simply the result of this grand expression modeling taking part in the recipe, contributing to the magic without necessarily being able to intellectualize it.

Thanks again for this additional wealth of information, we really appreciate your time and effort to explain your experience with these instruments.

I still need to internalize everything you say which is quite a lot and very interesting, and will come to this thread once and again to do so. It may even inform/change my plans for my own DIY project.

Small correction / addenda:

Paulo164 I wonder if this additional data stream is exposed outside of the Yamaha box, in the standard midi message (maybe on a specific CC like it is for hi-res velocity). And if yes, if program like PianoTeq could take advantage of this. I don’t think so as it is non standard and proprietary format of Yamaha (which does not collaborate with Modartt, as we see no Yamaha nor Bösendorfer approved models).

Yamaha does not directly collaborate with Modartt, but does have a so-called "dialect" of MIDI with proprietary extensions. I am not sure about what is the license for using those extensions, but PianoTeq does use them (I believe they call them something like "MIDI Disklavier XT version"). So what you say is possible, and anybody with the instrument can easily verify what is happening by inspecting the MIDI packages (if anybody wants I'm willing to help).

That leads to another question (sorry) which you might have an answer for: the Japanese specs for the RSC3 say "no" on the MIDI line. I assume that means "no" for the DIN-MIDI plug, and I was expecting to see a line saying MIDI-over-USB with a "yes" but there isn't such a line (even though there is USB-to-host and USB-to-device and the former might mean what I mean by MIDI-over-USB). Do you know that for sure?

CyberGene Which is why on the NU1XA they introduced hammer sensors too (if I remember correctly).

You do. From https://usa.yamaha.com/products/musical_instruments/pianos/avantgrand/nu1xa/specs.html#product-tabs the specs for the NU1XA say

Non-contact continuous detection electromagnetic inductive type

for both Hammer and key sensors (the NU1X is also included in the table and the only one without hammer sensors). The N1X and N2X both have $Non-contact optical fiber$ for both hammer and key sensors. Interestingly the N2 has no sensors either for keys or for hammers, so it must be working by telepaty.

RIP That leads to another question (sorry) which you might have an answer for: the Japanese specs for the RSC3 say "no" on the MIDI line. I assume that means "no" for the DIN-MIDI plug, and I was expecting to see a line saying MIDI-over-USB with a "yes" but there isn't such a line (even though there is USB-to-host and USB-to-device and the former might mean what I mean by MIDI-over-USB). Do you know that for sure?

I had exactly the exact same interrogation :-)

So I asked very specifically to Queue99 if it was possible to connect the RSC3 to an iPad through the USB-to-host plug (located on the rear panel I think) to send MIDI data via wired connection and not bluetooth MIDI and his answer was "yes" without any ambiguity. The VSH3 has this same ability but, in addition, includes also the legacy 5-pin DIN IN/OUT connectors (which I would not use anyway in my specific case).

RIP Yamaha does not directly collaborate with Modartt, but does have a so-called "dialect" of MIDI with proprietary extensions. I am not sure about what is the license for using those extensions, but PianoTeq does use them (I believe they call them something like "MIDI Disklavier XT version"). So what you say is possible, and anybody with the instrument can easily verify what is happening by inspecting the MIDI packages (if anybody wants I'm willing to help).

Okay, thank you for this precious info.

I should have tried, during my visit to the store, to connect my iPhone via bluetooth to the SC3 to monitor incoming MIDI events. Too bad…

So for now, what would be interesting to know is :

1. what MIDI messages are sent by the SC3 unit ?
2. does Modartt has any plan to exploit these additional non standard infos to improve their engine just for the happy few owners of Yamaha "Grand Expression Modeling" compatible hardware ? I have a strange feeling that my question is… self-answering :-/

Paulo164 Worth noting that the SC3 is supporting the “grand expression modeling” feature.

Paulo164 I should have tried, during my visit to the store, to connect my iPhone via bluetooth to the SC3 to monitor incoming MIDI events. Too bad…

Paulo164 what MIDI messages are sent by the SC3 unit ?

According to the manual, available at https://jp.yamaha.com/files/download/other_assets/4/1618034/ta3_tc3_sh3_sc3_en_om_c0.pdf at page 53

On an actual acoustic piano, subtle changes in the sound can be produced by changing your
playing touch from when you press a key to when you release it. For example, when you press
a key all the way to its bottom, the key hits the keybed underneath and the noise reaches the
strings, changing the sound slightly. Furthermore, the tone when the damper is lowered onto
strings to mute the sound is changed by how quickly you release the key. Grand Expression
Modeling technology recreates these subtle changes in the sound that respond to your touch.
This allows you to add accents by playing strongly or add resonance by playing softly to
produce superbly expressive sound. A crisp tone can be heard when playing staccato as well
as a lingering sound produced when releasing the keys slowly.

and then page 64

When transmitting MIDI data from this instrument to an external MIDI device, unexpected sounds may occur because of data related to the Grand Expression Modeling function

I am convinced that it does send these "internal" information as MIDI. Now for how VST companies could use those, your guess is as good as mine.

EDIT: the MIDI pdf file at https://jp.yamaha.com/files/download/other_assets/5/1784935/ta3_tc3_sh3_sc3_en_mr_b0.pdf has lots of information (probably too much 🤣 ) and it's not clear to me if this Grand Expression is included (there is a CC called "Expression" which could be it, and a quick Google search hits mostly Steinberg stuff talking about "note expression")

RIP

Ah, nice find !

To my modest knowledge, Yamaha is the only keyboard manufacturer that would enrich standard midi data with this kind of subtleties.

So it would be worth for Modartt (or any other VST company, although I think only a modeling approach could benefit from this) to implement this subtle changes in sound in their own way.

Queue99 I am interested in obtaining a Yamaha silent system. Thank you James

RIP

I have found some very interesting information regarding Grand Expression Modeling, which may be also very good news for us, enthusiasts :

• SC3 transmit “key acceleration” in CC#19 coded on 7 bits
• SC3 also transmit “expand velocity” (hi-res velocity) on CC#88 coded on 7 bits (so a total of 14 bits for velocity)

This is mentioned in the MIDI reference spec sheet of the Silent SC3/SH3 & Transcoustic TA3/TC3 models :

https://fr.yamaha.com/files/download/other_assets/5/1784935/ta3_tc3_sh3_sc3_en_mr_b0.pdf

So there are strong evidences that SC3/SH3 are sending midi events containing the 2 components of Grand Expression Modeling, namely “key acceleration” and “expand velocity”.

Some precisions of a CLP785 user on Modartt’s forum (CLP785 sharing the same specs as SC3) :

https://forum.modartt.com//viewtopic.php?id=5897&p=2

Yamaha CLP785 records them (extended velocity with controller 88 and touch acceleration with controller 19) and transmits them through midi (I receive them on Cubase, and I can read them too in the midi files I recorded with only the CLP).

Only 7 bits per controller is used (128 values possible), so we have 14bits velocity and 7 bits acceleration, which is 21 bits of resolution per key touch (2097152 values, or more than 2 millions).

With hours of recording, I checked, all 14 bits of velocity are used, and all 7 bits of acceleration are used.

The acceletation seems to have more frequently the lowest, middle and highest values, which I believe are the most negative value, 0 and the most positive value.

I can tell the difference, and the nuances are really really great to hear.