Digital Piano a LONG way off from realism

  • Thread starter Thread starter Toddskins
  • Start date Start date
Toddskins

Toddskins

Member
Just a vent post.

I have spent dearly for piano realism on my Kurzweil K2500XS, and recently upgraded to an even better Steinway Piano sample than those I previously loved. I have listened and played with other manufacturer's machines, too.

BUT...

Just not anywhere close to a real (good) piano.


For those of you who are not finatical about this topic, but wonder why the science of samples (and sampling) will still be many a year off from sounding like the real thing, here's why:


Theoretically, it is possible to mathematically map out a real piano algorithm, but what it will take is exhausting, mentally, let alone the development capability.

When you sit down at a great piano and strike a chord, that chord resonates off the other strings AND the wooden sounding board, bringing out mathematical harmonics from other strings' frequencies that are in the played chord's spectrum.

What that means, in practical talk, is that an extremely complex sound is being manufactured just from the simple playing of say, 3 notes that make a cool sounding chord - the original 3 notes, plus all the harmonic overtones from sympathetic neighboring strings (not to mention the sounding board itself). As you add more simultaneously played notes to the chord, the complexities grow exponentially!

In Sampling, each note is sampled individually, even if is has a dozen strike points, it is still a digital recording of a single note. The sample does NOT record harmonic overtones for diads or chords, obviously. Therefore, all those beautiful rich harmonic overtones that are made dynamically when the pianist plays, do not get created by the playing of a digital keyboard.

IF you understood what I just wrote, you can see that one day, mathematically, it is feasible to create an algorithm that could accomplish what I just explained, but it would be deep to say the least. Maybe not a Google, but close.
 
Best one I've heard is a Yamaha Motif........

There was also a $5,000 dollar Yamaha upright grand, looked AND PLAYED EXACTLY like the upright grands in all the practice rooms at the university...... sounded absolutely authentic and perfect......


Until the sales person walked up and turned it off.....

Not that I've heard everything there is to hear, but other than those two, everything else sounds obviously fake. Even the non-musicians off the street will tell me it aint the real thing. I think the way the notes decay always gives it away, there is something very unnatural about a decaying piano sample.
 
I think one of the problems also isnt' just the quality of the samples, but how we are hearing them. When you play an acoustic piano the sound you hear is the strings vibrating with the color and reflections of the wood. When you play a digital piano you hear sound coming out of a speaker cone that has been rehashed in a thousand ways.
 
I think you nailed it. I came to the same conclusion not long ago. When you play a combination of notes, it causes other notes to sound as well. You would not get all of the sympathetic vibrations from a sample unless you sample every possible combination of notes (at least all of the humanly playable combinations). Even then, what does the sampler do when you sustain one note, then add additional ones? How would you introduce additional tones in a natural way?

Not easy.
 
I think it also has alot to do with where the sound is coming from. Your speakers may have as good a stereo image as any, but a piano (or any acoustical instrument for that matter) is gonna emit sound differently from every little part. Just like how a guitar neck sounds different than a guitar body at the heel, or at the bridge, or at the headstock.

I think, to be honest, it won't ever get there. But that's ok, keeps us going to concerts :)
 
What also has to be taken into consideration is that the player hears the piano differently than the audience. Only the player is sitting at the keyboard and therefor, is the only one that hears the piano from that vantage point. The samples are recorded from audience perspective, something a player never hears when he is playing. The feedback that a player gets from a real piano is another thing that is missing from the digital piano.
 
2 sends worth

curious i keep finding myself agreeing with far.... as a classically trained pianist ive always foud it strange to sit out front sometimes.... much different acoustic signature.... but things are getting pretty damn good with things like giga around as developers are doing more and more with depth of sampling (more # of strikes & pedal up/down) will it put steiny/bosendorfer out of business ... probably not.
 
Theoretically, it is possible to mathematically map out a real piano algorithm, but what it will take is exhausting, mentally, let alone the development capability.
There IS such a thing as Physical Modeling. It's moved into adolescence and someday soon all of your objections and issues will be moot.

Computers are getting expotentially more powerful, and the jump from 32 bit to 64 bit is just the beginning. Huge hard drives will be able to store 48bit samples at 192kHz --- and that will be just the beginning.

I predict that within 5 years there'll be a sampled/modeled piano that will fool the pickiest most sensitive ear.
 
Within Reason

There is a Grand Piano patch for the NN-XT in Reason that is simply beautiful sounding -- and the closest to the real thing that I've heard so far.

It sounds amazing with just a little reverb and I often get asked how and where I miked that awesome sounding piano. It's definitely worth looking into.

Cheers, Rez
 
Todd, I completely understand where you're coming from. But it's the same with any sampled instrument... I will even veture uot and say it's the same with any emulation of any instrument be it acoustic or electronic, analog or digital. You may make it close, but the "essence" will be missing.

As a piamist, I too find that I dont react the same way to sampled piano as I do with the acoustic ones. Forget about capturing all the little details... just the fact that the sound comes out of spakers rather than the soundboard has major ramifications.

Stop killing yourself and irritating yourself over it, and accept the fact that you are NOT playing a piano, just a frozen recorded representation of one... once you're over this, you start to realise that the sampled piano can do things the acoustic can't... for example pitch bend, vibrato and filter sweeps :D Go further and put it through flangers, distortion and leslie. Stop making it appologise for being a sample, instead blatantly treat it as such. Specially since you've got a K2500XS. You can get a lot of milage out of the on board pianos once you mess with them with VAST blocks such as SHAPER (a little goes a long way x0.2-x0.3 for example, try modulating it with Env2, 0 attack, quick decay, set it on Src2, contol depth by AttVel) just to shape the attack. Sure it will sound less like a piano, but then again... YOU ARE NOT PLAYING A PIANO! :D
 
Why don't you try one of those MIDI-compatible pianos? I'm pretty sure they exist - a real piano with a MIDI in/out.

Or maybe I dreamt that...
 
Recap

Guys, first I'm not losing sleep over anything. My initial post said it all - I'm impressed with sampling, but it will never be close to true piano.

2nd, a couple of you seemed to have not read my post at all, in your praise of samples. Samples, cannot and do no create harmonics (overtones), etc. Period.

3rd, being a long time computer geek, computer power does not grow (i'm fairly sure) exponentially. It is linear. Maybe a steep curve, but not exponential. I could be wrong, but I don't think so. If you go back to the 8088 chip, followed by the 80286, the xx386, then the 486, then the first Pentium, onwards ho to the latest P4 and 64 bit chips... you map computer chip power in a few ways, but most telling is the mips (millions of instructions per second). It's a linear curve. Has it reached 8 MIPS, yet?

Lastly, physical modeling, or writing an algorithm to simulate all the harmonic frequencies that are heard in reality in a real piano, is of course, possible, but many a year off. Certainly more than 5 years. I just found this link which I think is pretty good, if you care to read a little more tech talk on the subject.
http://www.applied-acoustics.com/techtalk-physicalmodeling.htm
 
Toddskins said:
It's a linear curve.
Thank you for the kind correction.

I was speaking semi-poetically, not semi-technically when I used the term 'expotententially'.
 
Sounds like you're comparing apples and oranges. Saying samples don't sound real is almost like saying recordings don't sound real. It's not a valid comparison. It would be more fair to compare samples to a RECORDED live piano.

Do you have a steinway/whatever? Do you have DPA/scheops/whatever? If not then chances are the high-end grand libraries with a nice controller will kill you.

If you're only considering recording, and not performance, then there are no theoretical considerations: sampled sounds are every bit as good as the real thing, because they are the real thing. The problem lies in recording the performance via midi and translating that to samples. The limitations are practical not theoretical; i.e. how long is the sustain recorded, how many dynamic points are recorded etc.

Still, it makes sense that you may prefer a poor piano recorded live to "the best" samples, but it's because of the performance->midi->sound translation, not the lack of harmonics or any other timbre consideration.
 
ap said:
Still, it makes sense that you may prefer a poor piano recorded live to "the best" samples, but it's because of the performance->midi->sound translation, not the lack of harmonics or any other timbre consideration.

In the case of the piano, there are additional problems. As was stated in the first couple of posts, there are sympathetic vibrations in a real piano that occur and become part of the sound. You cannot capture those by sampling a piano one note at a time.

ap said:
sampled sounds are every bit as good as the real thing, because they are the real thing.

Not by a long shot.
 
What I think is suprising that nobody has yet mentioned such things as Native Instruments B4, or Apple Logic instrument models EVB3, EVD6 and EVP88.

These are all computer modeled (not sampled) instruments that have been painstakingly created to do what the real thing does. Surely with instruments such as these currently available, a modeled piano is on the way.
 
ap said:
Sounds like you're comparing apples and oranges. Saying samples don't sound real is almost like saying recordings don't sound real. It's not a valid comparison. It would be more fair to compare samples to a RECORDED live piano.

Do you have a steinway/whatever? Do you have DPA/scheops/whatever? If not then chances are the high-end grand libraries with a nice controller will kill you.

If you're only considering recording, and not performance, then there are no theoretical considerations: sampled sounds are every bit as good as the real thing, because they are the real thing. The problem lies in recording the performance via midi and translating that to samples. The limitations are practical not theoretical; i.e. how long is the sustain recorded, how many dynamic points are recorded etc.

Still, it makes sense that you may prefer a poor piano recorded live to "the best" samples, but it's because of the performance->midi->sound translation, not the lack of harmonics or any other timbre consideration.

Just found this thread and you beat me to this post.

+1

Carl
 
RezN8 said:
There is a Grand Piano patch for the NN-XT in Reason that is simply beautiful sounding -- and the closest to the real thing that I've heard so far.

It sounds amazing with just a little reverb and I often get asked how and where I miked that awesome sounding piano. It's definitely worth looking into.

Cheers, Rez

I use that one too...........even though I really don't know how to play piano, but it's fun trying. I usually use grand B with a little verb and it sounds pretty nice. When I listen to stuff with real piano though, Tom Waits for example, I can hear a huge difference. If I played piano it would probably bother me but I don't and it does'nt. :D

Ain't nothin like the real thang baby.
 
best piano samples

My piano background is to say the least, extensive. I know all about how pianos are sampled and why there is no substitute for the real deal but I stumbled along something that you might want to take a look at. www.synthogy.com
I have it and use it for midi piano. Without a doubt the best painos I've ever seen. Sampled at 10 velocities per key and many more features that add to the authentic piano sound. Check the site.

Peace,
Matt
 
You cannot capture those by sampling a piano one note at a time.

Ah, good point. I read that in the earlier posts, but it didn't register until now.
 
Back
Top