Hello dewster,

Yamaha and Casio pianos send

NoteOn
NoteOn
NoteOff
NoteOff

Reaper had problems with that.
Ableton has Problems with that.
Galaxy had problems with that. (These problems are fixed in the current update and it will probably take full advantage from these sequences in the next major update)
Synthogy has no problems with it.
Pianoteq has no problems.
Kawai CP136 has no problems with it. It even has such sequences in the inbuilt MIDI music and that was how I discovered the problems. It can, however not generate such sequences from the keyboard.

Thats all I know. I dont know what percussion machines do, sorry. If I play percussion on the keyboard, then it sends Noteoff.

The main problem is this:
Say, a pianist plays
NoteOn C Velocity 100.
Then the keyboard bounces a little bit and sends
NoteOn C Velocity 20.
This was not intended by the pianist, it just happens by bouncing.
Then the previous note is muted.

With percussion sounds this probably would not be a major problem.
It is also not a problem with 2-sensor keyboards because these dont send MIDI events when they bounce. (Provided that it is a well constructed keyboard)

With Reaper this happens:

Reaper converts the MIDI sequence during recording.
Internally it stores the notes in a Note-Velocity-Duration format. As soon as it sees a NoteOn it terminates the previous tone.
So when it receives NoteOn,NoteOn,NoteOff,NoteOff then it will record and playback NoteOn,NoteOff,NoteOn,NoteOff.
Again, with percussion instruments this will not matter. It is not totally correct, but nobody will hear a difference. If the percussion instrument ignores all NoteOffs then there is no audible difference at all.
With a piano sound this leads to muted notes even then when the VST instrument can handle NoteOn,NoteOn sequences. The VST will never see these sequences and so it cannot play them. The consequence are muted notes.

Peter

Yes, these keyboards need a "compatibility" mode it seems. In this mode, the third sensor would still be active, but a Note-Off would automatically be inserted JUST before every partial repetition Note-Off. This would retain the benefit of being able to repeat slightly faster (and for smaller key returns) however it would have a very small and subtle disadvantage: tone generators that produce different timbres for successive, repeatedly struck strings that are not damped inbetween each strike would no longer do this. (this is typically done by simply overlapping multiple voices)

Greg.

Btw, has anyone inspected the MIDI from a Roland PHAIII action yet?

Greg.

Intrigued I took a look at both my Casio and Kawai. I don't have calipers just a metal tape measure and short-sighted eyes.

On the Casio it's pretty consistent to say that the gap between the black C# and D# is about 1mm on average more than the gap between the groups of 3 black keys. It's not everywhere but more times than not, so hard to say if it's by design or a manufacturing assembly tolerance thingy. The Kawai is pretty consistent overall I would say using eyes and the tape measure.

Same gap on both my Casio and Kawai. I don't believe that strip of felt is functional but just for appearance, at least on these keyboards.

Maybe this bouncing effect: "Noteon with very low velocity shortly after a high velocity keystroke" could be recognized by software and surpressed. I think a real piano action cannot do this and if it could it wouldnt be audible. So it could simply been surpressed by software or firmware. This however requires a lot of fine-tuning andd I fear this is not done for digitals, only expensive acoustics get this treatment.

I wouldn't be so sure about real pianos not doing this. Real grand pianos can repeat for quite small key returns. I have been told in the piano tech forum here that it is actually less than 50%. That's the whole point of the third sensor - to allow a digital piano to behave closer to a real grand piano.

Certainly, in my brief testing of some real grand pianos, I could repeat for quite small key returns. I didn't actually measure the returns though.

Do you see any problems with my suggestion for a compatibility mode?

Of course, it would be best if everyone just started designing for tri-sensor actions. We need to go forwards. Tri-sensor actions are just the BEGINNING - we want much more than this!

Greg.

No. It could solve some problems with badly misbehaving software.
It would however not solve the problem that a very loud keystroke can be muted by a following very soft keybounce.
It would even prevent a software solution because it removes the information about half-releases.

I'm not very concerned about that at all. When I play a note, or a chord, very loudly, I don't notice any key bounce whatsoever. The keys are held down very firmly indeed. I don't think key bounce would occur very often at all. Even if you're right, if a very soft note is played on a real piano after a forte note, I think it could still dampen the previous note to some extent - sometimes even to a great extent, simply because the hammer itself acts as a damper when it comes into contact with the string! ;^) I suspect it's a non-issue.



I don't understand this. The whole reason for the compatibility mode is so that software that cannot interpret half-release re-strikes can be allowed to behave more normally! Once the software is fixed, we then flick the switch on our controller back to normal tri-sensor MIDI behaviour. Are you saying that if the controller provided a compatibility mode, that software designers would be lazy, and not bother to implement tri-sensor behaviour?

Greg.

Maybe the Casio bounces less. This would also solve the problem that I mean.
I came to this conclusion because I found this sequence Hard keystroke,Soft keybounce and no Noteoff inbetween very often in my MIDIfiles.
I played MIDIfiles quite often with Galaxy for testing purposes and for pleasure and had prematurely muted notes and found this as reason.

Peter,
Where did those MIDI files come from? AND, how do you know those soft notes were not played intentionally? (I'm not saying they were, but I'm interested to know how you can be so certain, that's all)


Greg.

Most came from the inbuilt demos of my Kawai.
Others from Yama forums played with clavinovas.

I im certain because it sounded right with the Kawai inbuilt sound and with Ivory and sounded wrong with Galaxy and because I examined the behaviour of Galaxy and the MIDI sequences carefully.
BTW. I participated in the beta tests and reported the problems and it was improved. Next version will hopefully be fully compatible.

Edit:
Maybe the Kawai demos where originally purchased somewhere and played on a Kawai Silent piano or Yamaha Grandtouch. Dont know....

Peter,
Thanks. What I'm also asking, though, is how you can be so certain that the player did not intentionally play those soft notes straight after the the forte notes. How do you know that those notes were not intentional.

I don't notice any bounce on my MP9000 either. The keys stay down VERY firmly if I play them forte. They do not budge.

Just out of curiousity, were the "offending" forte notes played staccato, or just normally?

Greg.

Because these bounces where so soft that normally they vanish under the preceding loud keystroke. They would not have been audible. And because I know for Example Gounods Ave Maria and these bounces had no musical sense.

That's what I was after - thanks. ;^)

It's very intriguing indeed!

I'm very curious now how much damping occurs on a real grand, for these partial-release forte/pianissimo repeats. I'm sort of expecting there to be more damping for higher notes than lower notes.

Greg.

Btw, I'm not yet agreeing with you that a softly played note that is played just after a forte note should not make the note appreciably softer. I think it often should. (I may try to test this on a real piano). If I am right, that means that your MIDI files are not being interpreted correctly. ;^) (and it then follows that the original performance data, in whatever form it was in, may not have been translated properly to SMF.

Greg.

I did a quick test of a real grand. I played a forte note (MF to F - I never tried FFF), and then tried my best to play a very soft note shortly afterwards, releasing the key partially so the strings were not damped. It's tricky doing subtle testing in a store, but it seems that the soft note EITHER plays at the intended, soft level, OR it plays somewhere between the intended level and the level of the decaying first note at the instant the second, soft note is played. I.e - it is possible that the volume of the first note does contribute to the volume of the soft note. Sometimes, the timbre takes on a bit of a "twang" sound for the second note, due to the fact that the hammer is striking strings that are already vibrating, and the hammer does not completely deaden the strings first.

I was never able to cause the note to be deadened to a great extent, so I was wrong about that. If there is deadening, it's pretty subtle I think.

I guess properly modelling repeatedly struck, un-damped strings might be a tricky thing to do.

Peter - re: your MIDI files, I'm wondering whether there is something wrong, somewhere. If some of them were played on a Disklavier or similar, perhaps the hammers didn't actually make contact with the strings for the soft notes? Maybe the key/hammer movement was faithfully recorded, but the fact that the hammers did not strike the strings may have been lost. (or, perhaps these MIDI files should be interpreted with a velocity curve that filters out all velocities below a certain value? Maybe the Kawai DP is doing that, and that's why it sounds normal! )

If any DP I come across has "bounce", I'm really going to look into that carefully. I don't think it's right.

Greg.

On a real grand, yes, it is indeed possible to strike a key, and lift it just enough to be able to strike it again without having the damper stop the vibrations between the two hits. However, I have to say, I have never run into any actual musical need to employ this subtle maneuver. In typical key re-strikes, the damper does fall and silence the strings in between the strikes (unless the pedal is down, of course).

Agreed. It may be that only the most demanding works require it. However, it's a definite feature (double-escapement) that was added to the grand piano action, and there must be some reason for it. ;^) I have read that Erard invented this.

Greg.

Note I suspect it's usually used without the player being concious of it. The player will simply be repeating extremely rapidly, and if you examine the key motion (and dampers) very closely, you will see that the key is only returning partially, and the dampers are not coming into contact with the strings. I.e - it's simply there to help the relatively slow, sluggish action of a grand piano.

Greg.

I don't think so. I think that when a player repeats rapidly, even though they key may only be returning partially, it is generally still coming up enough to damp the strings.

Try this test on a grand piano: Do a very rapid multiple repeat on a note. Now do it again, but with your other hand (or a friend's hand), hold the damper up to stop it from coming back down on the strings between strikes. I bet you'll hear that it sounds very different. When you use your hand to stop the damper from descending, the fast repeats sound less distinct as one hit bleeds into the next. When you do it "normally," the damper falls between each hit, and allows each strike to sound clean, percussive, and sharp. This is the normal and desirable behavior, as far as I know.

No - I completely disagree. Yes, usually it will damp, but repeat fast enough, and it will not. This is one thing that seperates the grand piano from an upright piano. An upright piano does not have double escapement. (I think some very advanced ones might - e.g Seiler).

I DO agree that it's a subtle feature.

Greg.

(I'll quote my piano tuning/technician book later, if it will help. I am not a piano tuner, but I have a book

Greg.

I just went through the process of trying to understand the double-escapement action as I could hear the Casio PX-330 modelling this when I demoed it, although at the time I didn't know what it was called. I believe Greg is right, on a grand piano action, the double escapement action allows a note to be re-struck without the damper re-touching the strings. While it may well have been developed to allow a higher repetition rate, I believe speed has nothing to do with that attribute, and it is purely a function of the travel of the key. There are some good on-line virtual models of the action if you do a search.

Pete

Pete,
It was done for speed. It allows the key to be pressed down again after it has returned only a certain amount. It saves time. Without the double escapement, the key has to return further up, which takes extra time, ultimately reducing the maximum repetition rate.

Greg.

Btw, this probably goes without saying, but I am not trying to say that a Casio tri-sensor action will necessarily repeat faster than every, or any brand/model of two sensor action. That would be silly. However, for any given action, a tri-sensor version of the action would generally be expected to repeat faster than it would if it had only two sensors. The reason for this is that when there are only two sensors, the top sensor has to be placed in a position that results in a good compromise between repetition speed, legato, and staccato performance. With tri-sensor action, the two upper sensors can be spread out, to bring the overall behaviour of that same action a bit closer to that of a real grand piano.

Greg.

Ah ha!!! I forgot about TRILLS!!! Yes, the shallower return would definitely benefit trilling!!

Extract from: http://www.learnpianofree.com/different-types-of-pianos

"One noticeable advantage that the grand piano action has over the vertical action is that all grand pianos have a special repetition lever in the playing action that is absent in all verticals. This repetition lever, a separate one for every key, catches the hammer close to the strings as long as the keys are played repeatedly and fairly quickly. In this position, with the hammer resting on the lever, a pianist can play repeated notes, staccato, and trills with much more speed and control than is possible on a vertical piano. Because of this, piano manufacturers claim that a piano player trill notes faster on a grand than on an upright."

Greg.

Greg, I understand it was developed for speed, indeed I said so in my post, however a by-product of that action is that it allows the key to be repeated without the damper touching the string. As I said, while it may have been developed FOR speed, speed is not a REQUIREMENT for the action to repeat as explained. So long as the key is not released up past a certain point the double escapement action on a grand piano will allow the key to be re-struck ad infinitum as slow or fast as desired without the damper touching the string.

Pete

Pete,
Thanks for the clarification - yes - I fully agree. (I thought we had established that a long time ago, though. )

Greg.

I think the point is this:

When the key is 50% released and stroken again, that means double escapement is in effect, then the behaviour must be the same as if sostenuto would have been active for this key.

It is clear that the second stroke will dampen the previous stroke a little bit, but not much. A very small bounce would barely touch the string and not dampen it.

The point is: the repetition behaviour with 50% releases must be the same with and without sostenuto applied because in both cases the damper does not touch the string.