Idea for Future Pedal Modeling?

Pianoteq's pedal sound is already quite good, perhaps the best out there in software pianos, but I think there's a lot more potential for the realism it could offer.

This is obviously extreme, but crank a piano's volume to the max and then also crank "pedal noise" to the max (you might even need to turn your main volume up a bit). Now depress your (continuous MIDI) pedal very slowly. You'll hear the pedal sound rapidly engage and reengage over and over again; same when you very slowly let off of the pedal. A real piano's pedal of course won't do this, and it will be louder or quieter depending on the speed the pedal is depressed.

Of course the pedal usually won't be nearly this loud, so it's not something most people will ever even notice, but when pedaling slowly in a very quiet passage, or when pedaling with a lot of force, the realism isn't quite there. And then of course there's the repedal effect that can be VERY dramatic, which isn't modeled yet.

I've also tried this with a good MIDI expression pedal (assigning it to the sustain pedal)--in case my Kawai pedal wasn't sending enough information to give a fair impression--but it doesn't actually make much difference, so doesn't seem to be an issue of MIDI resolution. I also realize this might be exceedingly difficult to improve, given the nature of MIDI...but perhaps there's some way to solve the puzzle?

---------------------

After thinking a bit, perhaps this could work:

On most real pianos, when depressing the sustain pedal there's a good bit of range at both the top and bottom where the dampers are not partially touching the strings. In other words, the "half-pedal" zone is a fairly small range; with the top and bottom ranges having no effect. Pianoteq seems to mimic this pretty nicely within the 0-127 MIDI range, with nothing from 0 to somewhere in the 20's (upper pedal range), and again nothing from around 90-127 (lower pedal range).

When half-pedaling, you're typically moving the pedal extremely carefully, because the effective range can be very small and tricky to hit, and so you're also usually moving the pedal slower than normal. I could be overlooking something important, but perhaps if Pianoteq saw a really quick pedal depression begin (in the 0-25 range perhaps), it could assume the dampers would fully lift, and use an appropriately loud damper lifting sound (depending on exactly how rapid the 0-25 range was). If on the other hand Pianoteq detected a fairly slow pedal depression in the 0-25 range...or especially if the pedal was kept above the 15-ish range and moving fairly slowly...then it could use a half pedal sound. That sound might be almost inaudible, as would a very slow/delicate full pedal, so the difference might be almost imperceptible there. So a sort of predictive pedal system. It could work the same way with the pedal release. The pedal release can be particularly loud, especially when students are still learning to control the pedal.

I think doing something along these lines could give a lot more control over the damper sound/dynamics, and go far further than any other digital piano (modeled or sampled). The next step would be to model a strong/rapid repedal, where the dampers are raised, dropped onto the strings and immediately raised again. This produces a strong effect where the dampers are made to "bounce" off the strings.

I personnally use a Roland DP10 which is working really well for half pedalling

According Philippe the noises in modern grand pianos are already  modelled.

I presume what you find as problems (not perfection) are MIDI limitations or limitations of your MIDI pedal or just matter of adjustments.

This is obviously extreme, but crank a piano's volume to the max and then also crank "pedal noise" to the max (you might even need to turn your main volume up a bit). Now depress your (continuous MIDI) pedal VERY slowly. You'll hear the pedal sound rapidly engage and reengage over and over again; same when you very slowly let off of the pedal. A real piano's pedal of course won't do this, and it will be louder or quieter depending on the speed the pedal is depressed.

NathanShirley, I can see what you mean on my Korg PU-2 Pedal. It is discrete and has three trigger points. Midi controller 64, three discrete values 38, 74, 127 and of course 0 when not engaged.

Depressing it unnaturally slow I hear three pedal noises with increasing loudness (loudness depending on Pianoteq's pedal velocity curve). I can see this three midi events in the midi event monitor. The same when releasing the pedal unnaturally slow.

Depressing the pedal fast or very fast to the bottom sometimes less controller values are scanned and triggered, for example
0, 38, 127
0, 74, 127
0, 127 (very fast)

So there are two pedal noises or just one.

joannchr wrote:

I personnally use a Roland DP10 which is working really well for half pedalling

So I'm not actually talking about the function of pedaling or half pedaling within Pianoteq, but of the sound the dampers produce. The basic function works just fine.

Beto-Music wrote:

According Philippe the noises in modern grand pianos are already  modelled.

I presume what you find as problems (not perfection) are MIDI limitations or limitations of your MIDI pedal or just matter of adjustments.

Yes, the pedal sound has been modeled for some time now, I'm just thinking of ways the model could be made to sound better and respond more realistically to the speed of the pedal.

I figured this would be a bit difficult to explain clearly... If you follow the directions below, you should be able to hear what I'm getting at:

NathanShirley wrote:

This is obviously extreme, but crank a piano's volume to the max and then also crank "pedal noise" to the max (you might even need to turn your main volume up a bit). Now depress your (continuous MIDI) pedal VERY slowly. You'll hear the pedal sound rapidly engage and reengage over and over again; same when you very slowly let off of the pedal. A real piano's pedal of course won't do this, and it will be louder or quieter depending on the speed the pedal is depressed.

Let me see if I can summarize and clarify what I wrote about before:

With a real piano, if you depress the pedal very slowly you won't hear any sound as the dampers lift from the strings without disturbing them. If, on the other hand, you depress the pedal very rapidly you'll hear a fairly significant sound. That sound will be significant even if you don't bottom out the pedal.*

*NOTE: If you bottom out the pedal by really stomping on it, the resulting thud will resonate throughout the piano giving a really loud sound. I believe Philippe said he didn't think the bottoming out of the pedal should be included in Pianoteq since part of that sound is created from the pedal slamming down into the felt/wood underneath, and the MIDI pedal used when playing Pianoteq would also create a sound when stomping on it. I understand that perspective, but since the sound then resonates through the piano, I think it would make sense for Pianoteq to model it. (End side note.)

So with Pianoteq you don't really get any variation in the volume of the pedal sound. If you press the pedal very rapidly you hear more or less the same volume as if you press it very slowly.

Now, when you release the pedal on a real piano the range of volume is much greater still. If you release it very gently and slowly you won't hear much of anything other than the strings as their vibration is gently stifled. If, on the other hand, you release the pedal very rapidly, there will be a very strong sound as the dampers fall onto the strings, resonating through the instrument. This is true even if you don't let the pedal come fully up (but still let the dampers hit the strings), so in other words you don't let the pedal's mechanics slam into anything. If you do let the pedal come up fully and very rapidly (say if you depress the pedal and then slide your foot off the pedal so it snaps up quickly), then the sound will be even louder still. But that might get back to the question of how far the model should be taken which I brought up in the note above.

Again, in Pianoteq the sound is basically the same volume whether you release the pedal slowly or rapidly.

And finally, if you repedal extremely rapidly on a real piano ("repedal accent"), the resulting thud as the dampers hit the strings and then "bounce" off can be quite dramatic. You start with the pedal depressed, then very rapidly release the pedal, but the instant the dampers hit the strings you rapidly depress the pedal again (so the thud of the dampers resonates through the undampened strings). I suppose some would argue this might fall into the "extended technique" category, but since it doesn't require reaching into the piano or drumming on the case, and it's a natural effect of pedaling in a specific way, I would argue it would make sense to eventually model.

The sustain pedal is such a fundamental part of a piano, I think it deserves significant attention. That added realism would absolutely blow other piano software away.

Hi Nathan, yes, this question is indeed a difficult question. We found for now no way to make it work as you would like without having at our disposal a continuous MIDI message, which is *very* far from what most pedals provide (we need something like 100 values per second to start having something nice and smooth). The reason is simple: we cannot read in the future, so we don't know if at the next time step the pedal will go up or down, and unfortunately there is no workaround, except ignoring the received steps and providing a uniform noise (possibly depending on a calculated velocity), or introducing a large delay which is of course not desired. Another way to go would be to try to guess from the music itself, but I'm not sure that would lead somewhere. Your ideas look interesting, we did have thoughts along that line, but then there are counterexamples where it won't work. Nevertheless, we will have to think about it to see how we could improve further the sustain pedal behavior.