Take any modern, REALLY well-recorded solo piano performance and the sound of the piano--that is a real piano--blows away any sampled sound out there: Steinway B, Gigapiano, Coakley\'s Perfect Piano, ILIO\'s product, Tractman\'s piano, the Bos, etc... But it doesn\'t have to be that way. My Steinway B sample (Truan) CAN sound very close to an AVERAGE solo piano recording (for example, my Bach at MP3.com); but it\'s just not competative with the very best. The fidelity just isn\'t there. What\'s needed is better sampling: more dynamics, better mic placement, better mics, better recording equip., better room acoustics, etc. IS IT OUT THERE??? HAS IT BEEN DONE YET??? ANYBODY????
John, my guess is that the best sampled pianos are using the same quality of equipment and recording techniques as the better (possibly best recordings) recordings. All of this has improved so much in the last ten years or so. I do think capturing more aspects of the instrument will help, but I don\'t think it will get us all the way. I think the problem is that it is nearly impossible to capture the instrument as a whole, as it\'s being played (if that makes sense). Sampling separate C, E and G notes from a piano, then playing them back sounds different than playing a C triad on the instrument and sampling it, to my ear. In solo work, this becomes very apparent. I think sampled pianos can sound great in a mix. The Steinway B with a good reverb (I use the TC Electronics M3000) fits beautifully into some of the mixes I\'ve done. I really have no complaints. But solo, it\'s a different animal. Eventhough I think the Steinway B is an overall better sample, I\'ve been going back to the GigaPiano when I just want to play solo (bang out some tunes). It is fuller and sounds more like you\'re sitting in front of a real piano. I just wish it was more consistent.
All of this has gotten me to look more seriously at an acoustic alternative to the sampled instruments. In the next six months I hope to add a Yamaha 5\'3\" MIDIGrand to my studio. This seems like a great way to use MIDI to capture/perfect your performance, but get the real sound of an acounstic piano. But at $25K+, they\'re not for everyone.
As a side note, the other thing that\'s gotten me excited about an acoustic instrument is the action. I\'ve been playing electronic instruments exclusively for so many years, I can\'t play a real piano any more at the same level of performance. I use the Kurzweil K2500 which has weighted keys, but no where near the weight of a grand. I\'ve been fooling my self for years thinking my \"piano\" playing was a a certain level.
Good topic though. I\'d like to hear other thoughts.
John I\'m listening to your Bach playing now, very nice! I assume this is the Giga-specific Steinway sample from east_west (the big one) right? Do you have any impressions of this one you could share? I assume you also have the Trachtman Steinway gig, care to share impressions of that one?
As to why sampled pianos are so different from the real thing, I\'ve got some ideas since I\'ve been programming \'pianos\' for a while. The biggest one is all the sympathetic resonances. Every note you play is going to excite all the other open strings, and change their overtones very dynamically, while every note on a sampled piano always stands alone and plays the same every time. Another thing is that when you sample a piano, you capture a single overtone series for that strike velocity, and then you can just lowpass filter it for lower velocities, but lowpassing doesn\'t at all capture the way the overtone series really progresses with strike velocity, plus you don\'t capture degrees of sound from the transient of the hammer strike, which is the loudest part of the sound. These big multistrike samples are getting us closer, but on a real piano you\'re not at all just picking from 3 or 4 strikes.
There\'s lots of other things that make me feel like I\'m playing a real piano, like how the sound comes from \'everywhere\' rather than 2 speakers, and how the instrument vibrates under your fingers, but you don\'t capture those in a recording, though you certainly play or groove to them and I wouldn\'t underestimate the importance of that to your performance.
Finally there\'s the miking itself. You usually pick relatively bright mics for sampling and put them right on the strings; since you\'re going to lowpass it you usually want to capture the brightest signal possible. But if I\'m recording a classical piano, I\'ll be choosing the mic and positioning it more for beauty than brightness, though this sound might be really dull if you tried to lowpass it.
John- I don\'t think the engineering side like mics, preamps, recording, or even the piano is a problem for most sets. Some things that the pianos could use is increased number of velocity levels, chromatic sampling, careful level editing, and extra \'dimensions\'- soft pedal as well as more standard ones like damper pedal and note release resonance. Last I knew Warren Trachtman was considering a very large Steinway piano sample with lots of velocities and more which should be good. There are also some interesting things that might help make the next \'ultimate piano\", like using the room reverb as part of the release trigger, or recording simultaneously with maybe 5 pairs of stereo mics in the room so that there are seperate samples for classical, jazz and pop contexts. That one I think would be particularly nice, and even if it made the set huge, the editing would be at least be the same for each pair of mics.
But here\'s the real problem.. as Busch and Sam mention, making an true accurate piano reproduction isn\'t quite possible with the tools. Sampling just can\'t capture some aspects of a piano\'s sound. The most prominent missing piece is the effect string coupling(the sympathetic resonances between strings) can have over time on any note played. When two (or more) strings are undamped and just one string is played, the coupling of strings through the bridge/soundboard can change the tone of certain harmonics (I believe the attack is softer and decay longer- I\'ll get details if anyone\'s interested) as compared to only the one string undamped. When the damper pedal is down(all strings undamped) that\'s an extreme case of this, but tones are different all the time without pedal just when more than one note is played at once. Considering all possible combinations for damped/undamped strings when even playing a single note, making samples that reflect this isn\'t really possible. With sampling, it\'s also problematic to reproduce effects like partial damping after a string is played, or repeated strikes to an undamped string. These last things are smaller details, but it adds up, and it\'s part of the reason it\'s so hard to get samples to sound as good as the real thing.
The perfect piano reproduction(or the next evolution) might come from a combination of sampling and physical modelling. Physical modelling of instrument/string behavior might be able to address the problems, but unfortunately I don\'t think it\'s a tool available to many people, at least not in sampling packages now. Of course it\'s hard to say how much it would help even if it was available. Still, I\'m very interested in what it might accomplish.
[This message has been edited by hurchalla (edited 01-01-2000).]
Yes David and Busch, there\'s nothing like a real acoustic piano! I pity the poor folks that live and learn with a digital piano, never knowing the world of difference an acoustic makes.
And the action alone... ah..
I grew up playing a Kawai semi-grand; but I don’t expect THAT kind of equivalent. My concern is the RECORDED sound, that is, the piano as heard through the very imperfect medium of speakers.
When the K1000 grand piano sound hit the market, years ago, an equivalent to the RECORDED piano already seemed within reach. Of course, the K1000 benchmark has long since been surpassed. The Steinway B for Gigasampler, carefully manipulated, is INDISTINGUISHABLE from medium and low-end recordings of acoustic pianos. Thus only the very best ear is able to distinguish my Bach on MP3.com from the “real thing.” Embarrassing, perhaps, but true. Joe public has to be reminded that they are not listening to a real piano recording.
The advances since the K1000 grand piano benchmark were made possible by vast increases in memory over the last 10-15 years. GeneralMusic, and one or two others, have already incorporated physical modeling into their samples, with some pretty good results. I don’t think we’re very far off at all from an utterly convincing HIGH FIDELITY piano sample. My guess is that the only barrier to bigger and better samples (yes, “bigger” is better—contrary to Coakley) is the apparent absence of a strong market for the product. Memory is no longer a factor.
I, for one, am willing to shell out a grand or two for a super-piano. Truan spent, I think, 25 hours sampling to create the basis for the Steinway B sample. A little more time and memory and he could have added
1. A true pp and ppp layer, which the piano now completely lacks
2. Some physical modeling, as in the Gigapiano
3. More, that is, MUCH more bass clarity.
4. More layers generally, to create a more seemless transition from ppp to fff.
I have many classical recordings of solo piano that are indistinguishable from Truan’s Steinway B. My pet peeve are the recordings that are head and shoulders above it. Listening to them tells me that all that is lacking, initially, is “fullness” of sound The sampled sound is just a bit “THIN.”
To use a simple example (sorry, it’s classical), take the C major Prelude from the Well Tempered by Bach. That’s a piece that’s played with NO pedal, one note at a time, except for the last chord. The BOS, the Steinway B, Gigapiano, ALL sound a wee bit thin next to a high end recording of a good acoustic piano. That doesn’t have to be the case. I’m convinced the technology already is at hand to go one better. Just as it quickly became evident that a mass produced sample like the Grand Piano of the K1000 could easily be improved upon; so I think it is now the case that the Steinway B (currently the best on the market in my view) is due for a drubbing.
>>I use the Kurzweil K2500 which has
>>weighted keys, but no where near the
>>weight of a grand
Have you tried a Yamaha GranTouch? Either the GT3 or GT7 has a (well, almost) real grand action. I don\'t care for the sound, but as a controller that feels like playing a grand, it might be worth the $7K or so...