Yamaha NU1X "loud note" fix seems to cause another problem

eddiepiano

Hallo dear PianoClack members,

I discovered your forum as an alternative to pianoworld and thought I would share the problem I have with my new Yamaha NU1X, as some of you may be interested in this.

I copied and translated the description of the problem out of the Mail to my dealer:

While playing, I noticed that some tones that I play forte / fortissimo do not sound as expected, but quieter and more muffled, as if I had only played mezzo forte or even mezzopiano. This does not refer to specific individual keys, but can occur with all keys. At first I didn't think anything of it, but it became more and more annoying to me.

Today I tested it very systematically: I played several keys fortissimo one after the other. For example, the first two notes played sound as one would expect - but the ln the third key is much more muffled and quieter, so as I would have only played it mf. If I then play this key one or two more times, even with very high intensity, it only stays with mf instead of f / ff. If I now take a short break and start again, then the key that has just caused problems and did not want to be louder reacts normally as f / ff. If I then repeat the whole thing again with other keys, the next few keys sound again quite normally ff and suddenly dull / mf again.

I wanted to take a closer look at the MIDI outputs of the keys and connected the NU1X to my laptop with Pianoteq. The program allows me to see the MIDI output (between 0 and 127; the higher the louder) of the keys being played. And then I could observe the following: When I play a key f / ff the MIDI output is around 110-120. If, however, suddenly there is one key that is quieter, the MIDI output is much lower, at around 70, as if it were limited by the software. Since the sensors work with every key, and this problem can occur with every key, I assume a software-related problem.

Before the firmware update 2.10, the NU1X was affected with the so-called "loud note issue", where sometimes very loud notes suddenly appeared while playing. Examples of this can be found on youtube or forum.pianoworld. The cause of the problem was probably due to the fact that the NU1X only uses one key movement sensor, which does not work as precisely as three-sensor systems. The problem does not occur with the N1X / 2 / 3X or with the Kawai NV-5/10 with three (Kawai two) sensors incl. hammer sensor. After the problem was first denied by Yamaha, the "loud note issue" was "fixed" with the firmware update - apparently a kind of algorithm was added to the software to detect very loud MIDI velocities and reduce them to avoid the sudden occurrence of loud notes... I guess this mechanism contributes to the fact that the software in the Yamaha sometimes recognizes very high MIDI sensor velocities as "loud notes" and reduces them, which explains the sudden duller tones. Again: The problem does not relate to certain individual keys, which could then be assumed to be a fundamental fault in the sensor or the like, but affects all equally (or not).

Until I discovered this problem it was a fantastic instrument. But my impressions really make sense now in the retrospective, in some pieces where I had do play f/ff I sometimes had the feeling my strike didn't translate very well in a nice, loud, growling/piercing tone. I thought it was my playing and I have to adjust to the new instrument…

Today I visited the next big piano store in Berlin to test if this happens with their NU1X - it does! It seems to be a systematic error due to the "loud note issue" firmware-update "fix". The software/algorithm seems to detect intentional hard strikes sometimes (and way too often) as an unintentional "loud note" and lessen the MIDI-value, resulting in a too soft sound... But it's a false alarm. I can't believe it, how can Yamaha still produce and sell these pianos with this systematic error due to their non-sufficient one sensor system? And I can't believe how nobody has noticed this until now? It's less noticeable than a sudden out of place loud note of course, so for Yamaha this may seem to be a better "solution" than the loud note issue.

I will send it back - that the keyboard responds accurately to my input is the minimum requirement for any piano, and even my Roland FP-30 can do this.

Seems that there is no alternative to a NV-5S/10S or N1X, but they are much more expensive…

burkey

eddiepiano faulty firmware brought down Volkswagen, brought down Boeing, and now Yamaha!

That's one more reason not to own an NU1X.

They better not have used the single sensor design in their upcoming models.

CyberGene

eddiepiano
I’m on the road, so can’t elaborate now but I think what you describe makes perfect sense and is a sad consequence of the general design of that piano 😢

David B

eddiepiano I will send it back - that the keyboard responds accurately to my input is the minimum requirement for any piano, and even my Roland FP-30 can do this.

Good for you and thank you for sharing this fault for others to be aware of. Hopefully, Yamaha will be motivated to address this issue as more noise is made (pardon the pun) by owners and potential buyers.

God Bless,
David

Frédéric L

eddiepiano (quoted) The cause of the problem was probably due to the fact that the NU1X only uses one key movement sensor, which does not work as precisely as three-sensor systems.

It is not pertinent to compare hybrid numbers of sensors with usual digital piano action. Alpha pianos has only 2 sensors (hammer pressure, damper), the N1X only 2 sensors (key position, hammer strike).

The issue is that there is nothing on a NU1X to measure if the escapement action has been rearmed and if we can play a sound.

The fact that 3 sensors on usual digital piano is enough deals with the absence of escapement.

vagfilm

Welcome to the forum. You are one of the firsts to come here through a google search (was it a google search?).

I don't have a nu1x so I will fire up some ideas just for troubeshooting: your ff notes are done with fast or slow repetition? Do you think that the muffled notes occur when playing a key that has not yet completely stopped and may still be bouncing? Try to see if you find a pattern like "it occurs everytime I press a key at ff and the key is still halfway up", or "... when the key is bouncing after reaching the top". If you can pinpoint the exact conditions, maybe the engineers can fix it... But since you cannot access the sensor values (only the final midi output) your hands are tied. Alternatively, if you really pinpoint the cause, there may be an action regulation tht solves it (less/more rebound, faster/slower key reset, etc).

eddiepiano

vagfilm Thank you! No, I didn't find with with google. I discovered the forum a few weeks ago when that ban thing at pianoworld happened. I think someone in the modartt forum mentioned it. I just didn't sign up until now.

I'm very sure it's a software/sensor issue, not a regulation issue (the action is regulated very well and even). The muffled notes also occur without repetition, for example when hitting a new, not already played note with ff intensity.

vagfilm

I was not saying that it was a regulation problem. I thought maybe it was a firmware problem related with the starting position of the key: if the key travel was less than expected, the firmware would translate your midi 110 into midi 70 to avoid a loud note. But if the problem can also occur at first strike, then it is as you suspect: either a sensor problem (not likely) or a consequence of the loud note safety firmware (very likely). Maybe CG has another clever idea...

eddiepiano

David B Yes, that was my intention.

vagfilm I am note sure. The NU1X was delivered with the newest 2.1 Firmware (that supposedly fixed the loud note issue). Now that I agreed with the dealer to return the NU1X I will not start to mess around with the firmware. But it really would be interesting to see if the "muffled ff note issue" (thats how I call it) disappears with the return of the "loud note issue" in an older firmware, as it would prove that this problem was introduced with the "loud note detection" algorithm in the 2.1 Firmware update.

QuasiUnaFantasia Unfortunately I think so too.

I am really frustrated. I decided for the NU1X after many months of testing all DPs from Yamaha, Kawai, Roland and Casio. Just to discover this.

vagfilm

If you do a factory reset, will that take you to the original firmware or the revised? Probably the revised... And can you flash the original firmware?

QuasiUnaFantasia

My guess is that Yamaha is using some AI to figure out how loud to make the sound, and that this AI is not infallible. But given that there has not been much criticism of the NU1X since the (presumed) AI was included in a software upgrade, I have a feeling that Yamaha is going to ignore this issue with a vengeance.

CyberGene

I'm finally at my computer and can elaborate more.

It was me who first proposed a workaround to the loud note problem on the PW forum:

Since the issue happens when the key has not been released fully but just below the top position, the sensing logic assumes it has been released fully and will be ready to detect a repeated strike. However mechanically the jack has not been reengaged under the butt and the key is then very easy to be pushed with a high velocity (no weight to throw under your finger) and so the sensing logic is fooled into thinking this is a regular strike with a high velocity. So, my suggestion here is that this can happen in different situations. Of course one of them is to do it deliberately as I have demonstrated:
https://www.youtube.com/watch?v=VpxJ5WrduE8

However it can also naturally happen in quick repetition, such as e.g. single key repetitions or trills when the key hasn't returned fully and you're pressing it again. And so for this particular case my proposal is for the sensing logic to detect the time between two or more activations of the same key (even if there is another note in between) and if one of the repeated strikes has a velocity higher than say 125 while the duration is shorter than a threshold (hence the repetition is quick), then assume this is the "loud note issue" happening, so then take the velocity from the first strike (or average it for multiple such strikes) and apply it to the "offending" one. This can work in many situations but unfortunately may affect genuinely loud strikes...

Anyway, a member on PW implemented my proposal (I forgot his name, wasn't that actually @xooorx ?) and published it on GitHub by just using a script interpreting the MIDI data, similar to @vagfilm MIDI filter. He said it was working OK for him and soon he was contacted by Yamaha and they provided him with a beta firmware that implemented the same logic for him to try. Later on they made it official and also created one for NU1 (but it's not published, people should ask Yamaha for it).

As has been discussed so many times, there's no easy solution to this problem if using only key sensors. I believe the only simple and good solution for Yamaha would be to add mechanical repetition springs as in https://pianoclack.com/forum/d/171-feurich-high-speed-kamm-upright-action

This way the key will repeat as a grand piano and the mechanical jamming won't occur in the first place.

vagfilm

CyberGene the sensing logic assumes it has been released fully and will be ready to detect a repeated strike

But eddiepiano stated that he could create a muffled note by a single ff stroke, without repetition. So, if this is reproducible, it is note a jack issue. My bet is that at some fffff strikes, the Yamaha logic calculates very high velocities, and as a safety measure assumes that to be a "loud note issue" and puts the note at m... (you can hear Yamaha engineer say: "no, no, no... no pianist can play this hard... if it is this high speed it should be a misreading due to repetition, and in that case make it a midi 70 and he should be happy").

vagfilm

Frédéric L The cause of the problem was probably due to the fact that the NU1X only uses one key movement sensor, which does not work as precisely as three-sensor systems.

This is not correct. Even if it is only one sensor, it is an optical sensor that is able to continuoulsy measure distances, and is used to calculate velocities between multiple distance points (how many, I am sure that Yamaha did not disclose). So, it should be, at least, as precise as a 3 sensor (if it calculates 3 distances) and has the potential to be much more precise if it calculates from more than 3 distances. And, if I remember correctly, Yamaha may use angular readings of the key rather than distances... but it boils down to the same principle. So, it is definitely not a "one sensor" system.

CyberGene

vagfilm But eddiepiano stated that he could create a muffled note by a single ff stroke, without repetition. So, if this is reproducible, it is note a jack issue.

Hmm, if Eddie really meant that a single fff note would be muffled without a recent prior strike for the same note, then something is very wrong with the Yamaha algorithms.

I had other problem with my NU1X which was ultimately the reason to return it: two of the keys in the middle consistently produced slightly higher velocities than the surrounding keys, which could be confirmed by pressing the keys together with an object and observing the MIDI velocities. I reported that to Yamaha and they sent me instructions on how to activate the key calibration procedure but it didn't help. Unfortunately the NU1X has its quirks. Luckily my N1X is perfect in that regard.

Frédéric L

But my point is that the number of sensors is not the key factor. NU1X has 1 analog sensor which count for 3 (1 for the damper, 2 for the velocity), and the Alpha has a damper sensor (1?) and a hammer pressure which could count for 2 since it is analog and measure the velocity.

No the key point is the measure of hammer movement which should verify the hammer has been moved. A configuration : 1 key sensor (damper) and 2 hammer sensors (hammer, velocity) could be more pertinent than 3 sensors at the keys like the shaded gray flag of the NU1X can emulate.

vagfilm

Sorry Frédéric: I quoted eddie from your post and somehow the quote was atributed to you.

eddiepiano

Hey, thank you all for the answers and the clarification regarding the sensor systems!

CyberGene Hmm, if Eddie really meant that a single fff note would be muffled without a recent prior strike for the same note, then something is very wrong with the Yamaha algorithms.

Yes, I mean it like this. A f/ff/fff strike on a note can send the right MIDi value and sound like it's supposed to, but a f/ff/ff strike on the same note can also only send a MIDI value of 70ish and sound muffled. It can happen on any key and it can happen with or without repetition. But this disbalance between input and output only happens when striking the key hard in the f+ range, thats why I think the software/algorithm that "fixed" the loud note bug detects a supposedly "loud note" (false alarm) and softens it. It does not happen the other way around (making a softer input louder as as it is) nor are they any other irregularities with die MIDI output when playing vom ppp - mf, it's very precise. And it's also very precise in the f+ range when the algorithm isn't involved.

Today I practiced on my good old Roland FP-30 (+ Pianoteq) and it was a good feeling to get reliable sounds also in the high dynamic range… But I missed the real action and especially the escapement.

Frédéric L

What puzzle me is that the N3 service manual (sorry it is the only Aventgrand Service Manual I have), has no procedure for calibration, only a « test » which verify if each threshold is at the good height. But if the test fails, we are out of luck.

Note : there are 4 thresholds…

eddiepiano

TomLC Eddiepiano,: So you saved $5grand? 😆 Or will you get something else?

No, I think I have to spend more, as I'd really like to have a real acoustic action under my fingers. The real escapement makes playing much more comfortable and authentic for me. Also, although PHA-50 and GF3 are pretty nice and have longer pivots, there are in my opinion not that much better than the PHA-4 in the FP-30 (in combination with Pianoteq), I can't justify spending 2000 - 4000 € for this felt sidestep. The real next step up is an acoustic action. Sadly, there is no alternative to the NV5S/10S or N1X now. I would love to have a VPC1 like MIDI Controller in a minimalistic cabinet with a grand action, but to a more affordable price than NV10S or N1X.

spanishbuddha eddiepiano Do you have, or can you make a small video of this behaviour? Thanks

Yes, I will try to capture it. To make it as clear as possible I will try to film my fingers hitting the keys as well capture the screen and tone of Pianoteq so that one can see the MIDI values and tone and cut everything together.

Del Vento I am certainly not a experienced virtuoso player - although my teacher stated that I could have developed to a concert pianist if I had started as a child. 😄 I play (early) intermediate pieces at best at this moment. As good and technically clean as possible, but still. But to notice this problem I don't have to be a virtuoso, as this problem does not only occur when playing fast repeated notes but also when hitting a new note that has not already been played. As I written earlier:

eddiepiano Yes, I mean it like this. A f/ff/fff strike on a note can send the right MIDi value and sound like it's supposed to, but a f/ff/ff strike on the same note can also only send a MIDI value of 70ish and sound muffled. It can happen on any key and it can happen with or without repetition. But this disbalance between input and output only happens when striking the key hard in the f+ range, thats why I think the software/algorithm that "fixed" the loud note bug detects a supposedly "loud note" (false alarm) and softens it. It does not happen the other way around (making a softer input louder as as it is) nor are they any other irregularities with die MIDI output when playing vom ppp - mf, it's very precise. And it's also very precise in the f+ range when the algorithm isn't involved.

So it can occur randomly on any note that is played f/ff.

Del Vento But, Yamaha could have made lemonade out of the lemons, by making a settings for this "fix" to be implemented as different "modes". I proposed "practice" as the name for the fix "off", so you'd notice your poor technique with a loud note, and a "performance" name for the fix "on". Not sure if @eddiepiano would have tolerated the loud note issue (like I do), but at least he could have had another choice rather than simply returning the instrument.

I think I would have tolerated the loud note issue, because I have no trouble playing acoustic uprights and let the key release fully before hitting again.

Del Vento As for cheaper alternatives, if you have time/stamina, you could engage in the Build your own hybrid piano which is the natural next step of the Cybrid.

I am fascinated by your projects and have read the Cybrid thread, but unfortunately I don't have the technical or workmanship skills to realize a similar project.

TomLC

Eddiepiano,: So you saved $5grand? 😆 Or will you get something else?