Piano Sample Libraries - pedal, sympathetic resonance etc. in 2012...

[Douglas Whates]

Wow! It's been a while since I posted here! Everything looks exactly the same - I almost feel like I'm in a time warp...!

Anyway, I'm looking for some expert advice regarding piano sample libraries. Specifically, what are the options for recreating sympathetic resonance and pedal behavior?

With the pedal, I'm thinking of things like re-pedaling, being able to catch the tail end of a staccato note with the sustain pedal, full continuous control with the sustain pedal (i.e. from slightly sustained sound to very sustained sound).

Sympathetic resonance, I'm thinking about the way already held notes resonate in sympathy when notes within the same harmonic series are played.

Are Xavier, Olivier and Kornel still working with these ideas...?

Looking forward to hearing all your thoughts and wisdom!

Douglas.

[EvilDragon]

For the most detailed piano behavior emulation, physical modeling is the way to go, not samples.

[Douglas Whates]

For the most detailed piano behavior emulation, physical modeling is the way to go, not samples.

For me, the best sampled pianos *sound* more convincing (i.e. true to a real piano) than physically modeled pianos. I think physically modeled pianos can *feel* more convincing (or more like you're playing the real thing) when triggering from a good keyboard, but in general terms they sound more synthetic to my ear.

[EvilDragon]

That's a heated and very subjective discussion. I'm on physical modeling side, both sound-wise and behavior-wise.

[Douglas Whates]

That's a heated and very subjective discussion. I'm on physical modeling side, both sound-wise and behavior-wise.

From a raw sound point of view, I am yet to be convinced by a physically modeled piano. I could compare the sound to the visuals of very high end CGI where you can't quite put your finger on what it is that looks unrealstic, but something just isn't right. There's just something in the inherent sound that is jarring and, for me, never pleasant to listen to.

The best sampled pianos, on the other hand, can be very pleasing to listen to and occasionally convince me for a minute of so that I'm listening to a real piano.

(I should point out that I'm talking about solo piano here, or piano that is very exposed in a mix. If it's less prominent in a mix, then I think any number of synthesized or sampled pianos are convincing enough.)

[EvilDragon]

Even the best sampled piano doesn't behave like a real piano, so that loses at least 50% of the expressivity of the whole thing IMHO... And to me, expressivity is paramount.

[Douglas Whates]

Can you point me to what you consider to be the best audio clip of a physically modeled piano?

[Gregjazz]

Sorry to interrupt here, but just wanted to say how good it is to see you here on the forums, Douglas!! Been a while.

I've experimented with sympathetic resonance in pianos--the tricky part is adding thresholds so that the polyphony doesn't get too crazy. That's because whichever notes are held are available for resonation, as well as their own harmonic nodes themselves. You have to draw the line at some point, otherwise those easily multiply. And then you have the top octave and a half or so of strings without dampers, so those are continuously open for sympathetic resonance--though that can be modeled through convolution (though that would mean muting them during sampling as not to double the resonation).

Limiting the harmonics to maybe 3-4 nodes helps a lot, but I guess the key is to have an intelligent polyphony handler for variable detail in the resonance.

[EvilDragon]

Can you point me to what you consider to be the best audio clip of a physically modeled piano?

Well, I use Pianoteq 3 Pro exclusively, and demos are widely available. For classical and inspirational work, and even sound design, it is unmatched.

[Douglas Whates]

Can you point me to what you consider to be the best audio clip of a physically modeled piano?

Well, I use Pianoteq 3 Pro exclusively, and demos are widely available. For classical and inspirational work, and even sound design, it is unmatched.

I'll have a listen to the demos again, but I think I have to respectfully disagree. I think the theory behind physical modelling is great (and it certainly saves the hard-drive!). Modeled and sampled both have their respective compromises, and from what I have heard and tried, I'm happier to accept the compromises in sampled pianos over the compromises in modeled pianos.

[Douglas Whates]

Sorry to interrupt here, but just wanted to say how good it is to see you here on the forums, Douglas!! Been a while.

And nice to see you're still here too - so funny to see familiar names cropping up all over the place! Would you believe... it's been almost four and a half years since I posted here. That's how much ns_kit7 took out of me; I needed a 4 and a half year break from sampling!

I've experimented with sympathetic resonance in pianos--the tricky part is adding thresholds so that the polyphony doesn't get too crazy. That's because whichever notes are held are available for resonation, as well as their own harmonic nodes themselves. You have to draw the line at some point, otherwise those easily multiply. And then you have the top octave and a half or so of strings without dampers, so those are continuously open for sympathetic resonance--though that can be modeled through convolution (though that would mean muting them during sampling as not to double the resonation).

Limiting the harmonics to maybe 3-4 nodes helps a lot, but I guess the key is to have an intelligent polyphony handler for variable detail in the resonance.

Yes, it's a tricky one. And also tricky is how best to create the actual source sound of those resonances so they are convincing, not sin-wavey sounding?

Convolution is not something I have any experience of doing. I have used convolution reverbs, but have no idea how they are made. How does it work with regards to dampers-up on a piano?

[Gregjazz]

And nice to see you're still here too - so funny to see familiar names cropping up all over the place! Would you believe... it's been almost four and a half years since I posted here. That's how much ns_kit7 took out of me; I needed a 4 and a half year break from sampling!

I believe it--that's a monster library! Speaking of which, I need to install it again.

Yes, it's a tricky one. And also tricky is how best to create the actual source sound of those resonances so they are convincing, not sin-wavey sounding?

Capturing the attack is the most difficult part. Obviously, you can just play the note an octave below staccato, while depressing the key, but even that way you'll have to chop off the beginning part of the samples. Another way would be to use something besides the piano itself to excite the string. For example, an ebow, though I'm guessing that would return a pretty sine-wave-ish result.

Convolution is not something I have any experience of doing. I have used convolution reverbs, but have no idea how they are made. How does it work with regards to dampers-up on a piano?

Convolutions are a way of recreating any acoustic space, which includes EQ and resonant nodes, etc. It's fairly accurate, too, up to a point (as soon as you get into spaces that have volume-dependent anomalies, although those tend to be fairly subtle nuances).

Technically you could recreate all sympathetic resonance using convolutions, but it would just get pretty CPU intensive--particularly due to the length of the convolutions. And you would have to have a system of swapping out impulses based on what's needed, because simultaneously running an impulse for each of the keys is impractical.

[Douglas Whates]

RE: Pianoteq. I had another listen to the Pianoteq demos. While I can hear some benefits to using it and I admire the theory and programming, I think it's hard to ignore the elephant in the room; i.e. the terribly synthetic and plasticky tone.

RE: convolution. How would one go about making an impulse response of the sustain? That's where I'm a bit confused. I've researched it a bit and I just can't work out how you would capture an impulse response resonance of the piano with dampers up without also capturing the room reverb.

[egbert]

RE: Pianoteq. I had another listen to the Pianoteq demos. While I can hear some benefits to using it and I admire the theory and programming, I think it's hard to ignore the elephant in the room; i.e. the terribly synthetic and plasticky tone.

RE: convolution. How would one go about making an impulse response of the sustain? That's where I'm a bit confused. I've researched it a bit and I just can't work out how you would capture an impulse response resonance of the piano with dampers up without also capturing the room reverb.

That's a job for an anechoic chamber. Some sample sets for hardware pianos are recorded in them - Kawai used to have photos of the recording for the MP9x00 series where they claimed to have done this. And welcome back Douglas - still a fan of your library.

[EvilDragon]

I think it's hard to ignore the elephant in the room; i.e. the terribly synthetic and plasticky tone.

Well, I don't hear that "plasticky" tone you mention. My ears are perfectly fine I assure you.