Velocity issue

I have been trying to customise the velocity response of AP. Unfortunately, AP offers limited options there. Only a few preset curves. That's not enough to use AP standalone. In fact, it presents the same problem that all digital pianos have. Accidental high velocity notes during playing.
I use a StudioLogic 88 (grand touch) keyboard but nevertheless it has the same problem as any other keyboard controller. It uses velocity to function, not force. Whereas a real piano keyboard responds to force (to play louder), not velocity. It happens that velocity sensitivity gives a good approximation of a piano action but it is just that, an approximation.
This means that most of the time measuring velocity is fine for judging intended force but some notes may get momentary high velocity without being played strongly. Piano players surely know what I am talking about. These notes are heard as accidental, harsh notes during a passage and ruin the result. These same unintentionally high velocity notes are responsible (to an extent) for making a digital piano sound unnatural.
The only way to deal with this issue would be to be able to customise the velocity curve so as to limit the maximum velocity to taste. Or a way to prevent sample layers, the high velocity ones, from sounding.
Galaxy pianos have an excellent velocity editor to fully customise the curve. I do not have Ivory but I believe they have something similar too. Anyway, this is a must have for a virtual piano of this quality.
AP is an otherwise super package with beautiful samples, nice programming, great on CPU, streamlined. But this important omission leaves it behind making it very difficult to use for serious solo playing.
I hope the AP team will do something for this soon.

 

I think that the G#3 I played at 1:00 in this video is a good example of what you're talking about.

[ame="http://uk.youtube.com/watch?v=WVucj5E1Rx0"]YouTube - Chopin Etude - Op.10 No.3[/ame]

I'm sure having more flexible velocity editing facilities would help.

However I'm not sure that the problem arises because electronic keyboards respond to velocity rather than force.

According to my rudimentary knowledge of phyics: force and acceleration are directly proportional (force =mass * acceleration). It seems to me that even in a real piano the force with which a key is played determines the velocity with which the hammer moves towards the string and that it is purely this velocity (and the mass of the hammer) which finally determines the force with which the string is hit.
My difficulty is in visualising a scenario where a key is played with low force but high velocity. Is this really possible?

I certainly don't fully understand the physics involved and would be interested to know what other people think.

Apparently most electronic keyboards (including Grand Touch) lack the escapement mechanism and I was wondering if this has anything to do with the problem.

Or is it just that the the velocity response needs to be less linear (ie more curvy!)?

 

Yes, your video is quite an example. I have the same keyboard. Nice touch, I came to buy it after trying many others. But still, while playing, some notes here and there go unintentionally wild.

You are right about the physics. Velocity is an indication of energy transferred from the hand to the keyboard. My objection is about the transfer function i.e. how much of this energy is spent to hammer velocity and how much goes to hammer torque. An analogy would be a car's gearbox. You have a very responsive (to acceleration) car in 1st gear but your speed is limited. In order to increase speed you shift gears but you lose responsiveness. The engine power though would be the same in both cases. I don't know much about the details of a piano's action (though a pianist) but my guess is that electronic keyboards are over sensitive to velocity compared to a real piano.
If you use a curvier velocity setting in AP, things don't get better really. It gets less aggressive and you feel that you have to push harder. But these accidental high velocity notes escape again because although more difficult to reach, the upper velocity limit is still 127.

I come to think that staccato playing exposes the problem quite well. Try staccato (one note) on a real piano. Then try staccato on AP. Chances are that on AP you will MIDI 127 very often which of course was not your intention. Staccato is be something that should be implemented in AP (i.e. in software).
Otherwise, the velocity curve should give the option of an upper limit. With Galaxy II I have managed to get very nice results this way. However, even in Galaxy II this trick compromises overall dynamic range.

 

Well, dynamics on MIDI-keyboards are an old story. I think the problem is, MIDI keyboards measure the time between upper und lower position of the key, a real piano "measures" the impulse that is transmitted onto the hammer, that strikes the strings. Unfortunately these are two different things (but it's probably much cheaper to build it this way). For instance, it's possible on a real piano to keep a key pressed a bit for a while, before you actually play it, and this can still give you a mf-note. On a MIDI-keyboard then all you get is a ppp-note (vel=0).
The described problem with those accidental high velocity notes is a bit different, the problem here is that the complex mechanical system of a real piano isn't rebuilt properly. Often springs are used to simulate the weight of hammer and key, which might be okay for the inital force you need, but then the inertia of the system is still totally different.
Then, if you watch the necessary force over time while you play a note for a given dynamic, on a real piano you'll have a larger inital force than at the end (when the key is almost down) - the impulse transmission takes place quite early, if you just use a bit of masses and springs like in MIDI keyboards you'll definitely can't get that effect (please forgive me my bad physics english here).

Generally I found it useful (for my Doepfer PK88, which is actually pretty good) to 1. reduce the overall velocity responsiveness and 2. reduce the responsiveness of the black keys, because often they have the same "feel" on MIDI-keyboards as the white keys, which is again totally different on the real thing.

There was a thread quite a while ago about that black key thing:
http://www.nativeinstruments.de:80/forum_de/showthread.php?t=37994

Cheers,
Juergen

 

I will prepare some short audio examples to properly indicate the problem.
Stay tuned.

P.S.
I still think that this velocity thing should and could be solved in software. It's a pity that great sounding software like AP falls behind in playability because of it.

 

The idea about energy spent on torque rather than hammer velocity sounds like something which could easily be modelled by software.

Certainly springs are not very helpful in simulating the feel and response of a real piano. That's why we have "weighted keyboards". The Grand Touch keyboards (Fatar/Studiologic) even include a hammer mechanism and grade the attached weights across the keyboard in order to more accurately simulate the feel and response of a real piano.

It occurred to me that it would not be too difficult to write a small stand-alone application which captures midi events and transforms them before passing them onto AP.

I wouldn't mind having a go at this myself sometime but, meanwhile, a quick google came up with this:-

http://www.musiclab.com/products/vmmanual/vmman.htm

I've not tried it and wouldn't be surprised if there were other similar utilites out there.

I suppose an improvement to this could involve adding a facility to single out black notes for special treatment as nana suggests.

 

I don't think this is just a software issue.
The idea with the "weighted keys" is much better than playing against springs, of course. But on a real piano you've got the inertia of the key AND the inertia of the hammer - and these two are not constantly connected. The piano mechanism is more like a catapult. (I remember the FATAR -keyboards - whatever the brochures say they try to simulate - they are far too "wackley", IMHO, like riding a horse on ice ;-)
The only data the MIDI-keyboards usually measure are the times between key-up and key-down. I don't think this is enough, I think for a real realistic keyboard simulation you should measure the position of the key over time. Then the rest is software, agreed.
For instance think of the effect, you play a chord and then slowly (or very slowly) release the keys and you can exactly determine the point the damper contacts the strings (so no chance for "release velocity"). This is totally impossible to realize with the usual time-between-up-and-down-method.

We dealing with crutches, sigh...

Cheers, Juergen

 

I take the point about the piano mechanism being more like a catapult. Though judging by the Fatar schematics I'm not entirely sure that the key and hammer are constantly connected. I'll have to open mine up to have a look.

According to a thread on the "piano world" forums it seems that the Doepfer PK88 actually has a Fatar controller at its heart! Of course - they may be wrong.

It's also worth noting that Fatar keyboards have made progress over the last few years. My original SL880 has a lot lighter and less realistic touch than the VMK188+ "Grand Touch".

I suppose, in the end manufacturers (and buyers) have to decide how much of the mechanism it is worth reproducing before the law of diminishing returns kicks in.

 

For the time being I get away with manipulating MIDI velocity through Logic's environment. Apart from setting a velocity response curve in the plug-in, Logic can also reduce maximum velocity reaching the sofware instrument. But I need Logic to play AP. In standalone, which is the most direct way to practise, this can't be done.

 

I have prepared some mp3s which clearly show the effect caused by unwanted high velocity notes. You can find them at:

http://idisk.mac.com/wiser_guy-Public?
(Mac users can always browse my iDisk directly from the Finder)

The mp3s are small excerpts of solo piano pieces, the kind of which would be normally played by an AP user. I tried to make it as diverse as possible. They are arranged in pairs. [title].mp3 is the dry result of the AP plugin whereas [title_mod].mp3 is the same but with a customised velocity curve before the plugin. This velocity curve is made in Logic's Environment and as we have already discussed, AP lacks the ability to draw such a custom curve severely impairing playability when used as standalone.
Every example bears no audio manipulation at all. No limiter, no equaliser, no levels. Just the AP 'Studio' reverb. All pieces have been done on the Bechstein. Firstly, the high velocity effect is more prominent there and secondly I am tired of hearing that the Bechstein is good only for pop or blues. The Bechstein is perfect for any style, it holds a unique character and in AP sounds and records absolutely amazing.

You can see from the mp3s that some pieces (notable the rags) suffer more than others (the FBlues may actually be better without the custom curve). Someone, after hearing these examples, may say that I have overrated the effect of a high velocity note here and there. Believe me, it's ten times worse to hear these when playing. They directly affect the taste and feeling of the virtual instrument at a time when the pianist struggles to forget that he plays a virtual instrument and tries to concentrate on the music.

I wouldn't mind if it was about a mediocre digital piano. But AP sounds fantastic and can be used by a pianist. That's why I wish someone from the NI team listens to these examples and plan a fix in a future update of the software.

 

Those examples are quite revealing - as well as being played very nicely. The velocity modded ones certainly sound less dynamic but in a good sense. They are more pleasing to the ear, less wild, more musical.

Apart from stray notes like the one in my example, I hadn't appreciated that this velocity issue has a subtle effect on the music as a whole.

 

David, thanks for taking the time to listen.
I am not an expert in samplers but I know the piano sound. And I am sure that a good sampled piano should be way more than just great sound. This may be enough for mixing as a side instrument but it needs a whole lot more to play solo. And it's paying attention to details like these (stray high velocities) which set it apart from just samples. Which make it convincing as an instrument.

Galaxy II have solved this issue but they have others, staccato being the most important for them. AP has great attacks for staccato or non legato, especially after hearing a recorded performance with it.

And yes, you are right. Sound affects the way you play. It should. You cannot become a good pianist by just playing on a nice instrument. But if you are a good pianist, you need the best instrument you can get your hands on.

 

Stray notes on my Roland 300SX is something I struggle with more than anything else, and I usually manually go in and fix them all afterthefact, but the idea of a custom velocity curve is new to me and would be perfect. Is there any plan to create a 1.1.4 patch for Akoustik Piano?

My makeshift solution while I'm playing is to decrease the dynamic range of the piano a great deal and play by feel instead of by sound. That way I'm not jarred by what I hear and I usually like the result.

 

Hi all. I came to this topic as a new user of AP looking for a solution for the issue you all have dissected very nicely. Although not a trained pianist, I believe I have good "ears" and very quikly noticed the "spikes" as I will call them, that appear while playing. (SL990). My first thought turned to mechanical keyboard error, even though I had not experienced this with other programs, after ruling that out with test procedures, I then tried to better my "feel" for the program, thinking it might be due to my playing style. Alas I am here, fearing that the previous descussions do indicate a flaw in an otherwise VERY nice program. Being more of a sound engineer than musician, this problem, as others have mentioned, sadly will relegate this program to a "background" instrument in usefullness. NI- I hope your listening.