More Whining About Piano Sound

[lowerhodes]

Originally posted by Bridog6996:

"The question I'm surprised nobody asked is why you bought it if you were not satisfied with the sound. Did you not try it out before purchasing it?"

 

Yes, but as we all know, a music store is, ironically, not the optimal place to audition a musical instrument. That and the sales person is someone I have dealt with before and who will accomodate me here accordingly.

[Finale]

Four drunk Keyboard forumites having a conversation...

 

Guy A - My piano is much better than yours. It has an Orgasmic 33' Bechstein piano patch.

Guy B - No big deal, I have a Grand Ultimate 45' and Five Sixths on mine.

Guy C - Well, mine's got the new version of it called Grand Ecstatic 45' and Infinity.

Guy D - Wrong. The latest is called Grand Ecstatic 45' and Infinity, plus one. And it's only available on mine.

[marino]

Originally posted by soundscape:

I actually believe that this obsession with getting voices to sound "real" is the wrong way to go in a way. It's better to have a patch that doesn't sound real but sounds good than one with 127 sample layers which are just impossible to get all matched and playing together well and the end result is more "real" but has "rough edges." Perhaps it matters more with some styles of music, however.

Huh? Obsession? I don't get this. What the aim of a digital piano would be, if not to sound and feel like a piano? More real doesn't mean 'rough edges' - not necessarily, at least. And if 127 sample layers don't sound good, they aren't 'real' - just bad.

That said, I've noticed that often, the best dynamic response is not obtained with more layers, but rather with an accurate use of filters or other devices. In my opinion, one of the best piano sounds with regard to dynamics is the ProMega3; you can like its sound or not, but the dynamic control is excellent. They use a patented filter for this.

Also, I find The Grand 2 to have great dynamic control too, even though I think its basic sound is a bit harsh.

Overall, I still have to find the piano to replace the KeySolutions Steinway for the Kurzweil. Despite having only two dynamic layers, Mike Martin's excellent programming work on the Kurz filters makes it very playable and expressive. Plus, the piano which was sampled was simply stunning, and that makes all the difference in the world.

I guess that unless every last atom is modeled, the result is never going to be 100% "real".

I thought that's what I said.

[soundscape]

Originally posted by marino:

Originally posted by soundscape:

I actually believe that this obsession with getting voices to sound "real" is the wrong way to go in a way. It's better to have a patch that doesn't sound real but sounds good than one with 127 sample layers which are just impossible to get all matched and playing together well and the end result is more "real" but has "rough edges." Perhaps it matters more with some styles of music, however.

Huh? Obsession? I don't get this. What the aim of a digital piano would be, if not to sound and feel like a piano?

Sorry... to be completely clear, the jibe wasn't aimed at you... it's aimed at the way the industry pushes this week's 100Gb (or whatever) sample piano (or other instrument) that may well sound detailed but on close evaluation is sometimes of rather sloppy quality. It's not difficult to create big sample size libraries, it is difficult to create good ones though.

 

Originally posted by marino:

More real doesn't mean 'rough edges' - not necessarily, at least. And if 127 sample layers don't sound good, they aren't 'real' - just bad.

Quite... but that doesn't stop marketing claims from being made.

 

(For example, I know of one Mellotron-based sample library or software instrument that claims to be better because of the lack of noise reduction used in order not to suck out the life of the sound. Does that mean it sounds more like a "real" Mellotron, or the real sound of bunch of worn out Mellotron tapes, or simply that the developer was not able to get prestine Mellotron tapes or couldn't be bothered to go through all the careful processing to get rid of the gunk without sucking the life out of the sound? You get the idea.)

 

Originally posted by marino:

That said, I've noticed that often, the best dynamic response is not obtained with more layers, but rather with an accurate use of filters or other devices. In my opinion, one of the best piano sounds with regard to dynamics is the ProMega3; you can like its sound or not, but the dynamic control is excellent. They use a patented filter for this.

This is exactly the sort of detail that interests me, how to combine the synth engine/programming with the excellent sample set to get the best possible result.

 

Originally posted by marino:

Overall, I still have to find the piano to replace the KeySolutions Steinway for the Kurzweil. Despite having only two dynamic layers, Mike Martin's excellent programming work on the Kurz filters makes it very playable and expressive. Plus, the piano which was sampled was simply stunning, and that makes all the difference in the world.

I read the discussion between yourself and Mr. Coakley (ahem) on his site, and I noticed that you said it used a lot of VAST programming. I wish there were more audio samples of it online.

 

Originally posted by marino:

I guess that unless every last atom is modeled, the result is never going to be 100% "real".

I thought that's what I said.

Yes, it looks like we are in agreement over many other aspects as well.

[niacin]

Originally posted by marino:

I wish that digital piano makers would refrain from calling their patches "Superb Grand", "Real Bosendorfer", or "Heavenly Acoustic". They just ain't.

You're expecting "dodgy digital grand #1", "best we could do Bosendorfer" or "not really an acoustic"? I remember the ad for GEM's Pro2. It had quotes from various players including Keith Emerson and Herbie Hancock. Hancock said something like "well it's really not too good, but it's the best of what's available". I had to laugh.

[dp2]

Originally posted by niacin:

Originally posted by marino:

I wish that digital piano makers would refrain from calling their patches "Superb Grand", "Real Bosendorfer", or "Heavenly Acoustic". They just ain't.

You're expecting "dodgy digital grand #1", "best we could do Bosendorfer" or "not really an acoustic"? I remember the ad for GEM's Pro2. It had quotes from various players including Keith Emerson and Herbie Hancock. Hancock said something like "well it's really not too good, but it's the best of what's available". I had to laugh.

lol

[burningbusch]

Carlo, I have to disagree with the idea that filters can be used as successfully (or more) than multiple velocity samples. If you listen individually to the layers on let say and 8 or 10 velocity instrument you will hear much more changing within the sound than can be approximated with filters. As you know, the instrument isn't simply getting brighter; there are some very complex things happening. Our best hope is probably modeling with using sample-based input to generate the waveform.

 

The only pianos that I've found that approximate the dynamic range of an acoustic are Ivory and Akoustik Piano. They become addictive to me BECAUSE of their dynamic range. But I was using Akoustik Piano in Receptor the other night at a gig and it crapped out. I need to work with it to see if it's usable.

 

We seem to be inching forward.

 

Busch.

[dp2]

Originally posted by soundscape:

Originally posted by marino:

Originally posted by soundscape:

I actually believe that this obsession with getting voices to sound "real" is the wrong way to go in a way. It's better to have a patch that doesn't sound real but sounds good than one with 127 sample layers which are just impossible to get all matched and playing together well and the end result is more "real" but has "rough edges." Perhaps it matters more with some styles of music, however.

Huh? Obsession? I don't get this. What the aim of a digital piano would be, if not to sound and feel like a piano?

Sorry... to be completely clear, the jibe wasn't aimed at you... it's aimed at the way the industry pushes this week's 100Gb (or whatever) sample piano (or other instrument) that may well sound detailed but on close evaluation is sometimes of rather sloppy quality. It's not difficult to create big sample size libraries, it is difficult to create good ones though.

If the vendors weren't chasing the "how real can I get this sound on that patch dreams? . . ." They'd do something similar to what many computer manufacturers (who, when they realized that processors have reached the theoretical speed limits): begin a new arms race. They'll probably add more effects and layers to help us to make those "real" sounding patches sound out of this world (euphemism for 'faker').

[soundscape]

Originally posted by burningbusch:

Carlo, I have to disagree with the idea that filters can be used as successfully (or more) than multiple velocity samples. If you listen individually to the layers on let say and 8 or 10 velocity instrument you will hear much more changing within the sound than can be approximated with filters. As you know, the instrument isn't simply getting brighter; there are some very complex things happening. Our best hope is probably modeling with using sample-based input to generate the waveform.

As mentioned the GeneralMusic have a patented filter ("FADE".) I've been reading through the blurb on their site and watched the video where it's explained, and it's "supposed" to change the partials in the same way that a real piano would, not just open/close a low-pass filter.

[marino]

Originally posted by burningbusch:

Carlo, I have to disagree with the idea that filters can be used as successfully (or more) than multiple velocity samples. If you listen individually to the layers on let say and 8 or 10 velocity instrument you will hear much more changing within the sound than can be approximated with filters. As you know, the instrument isn't simply getting brighter; there are some very complex things happening. Our best hope is probably modeling with using sample-based input to generate the waveform.

 

The only pianos that I've found that approximate the dynamic range of an acoustic are Ivory and Akoustik Piano. They become addictive to me BECAUSE of their dynamic range. But I was using Akoustik Piano in Receptor the other night at a gig and it crapped out. I need to work with it to see if it's usable.

 

We seem to be inching forward.

 

Busch.

I'll try to better explain myself.

 

First: One thing is dynamic range, another thing is dynamic control. 12 layers? Ha. For some psychoacoustic reason that I find hard to describe in words, those multi-giga pianos sound even more artificial to my ears, than a few digital or sampled pianos with a good implementation of dynamics. Even 128 MIDI gradations aren't enough. My guess is, they sampled those 12 layers quite well, but then they didn't work enough on making the transitions among those layers sound smooth.

The result of this is that very often, you have to treat dynamics in a sampled piano a lot more roughly, and with less subleties, that you can do on a real grand.

 

Second: I'm not talking about linear filters. To my understanding, for example, the unique filter on the ProMega is designed to have a continuously variable response thru the whole scale of velocities. In short, this means that it brings out certain groups of harmonics at some velocities, and other groups at different velocities. A traditional "cutoff+resonance" filter would be the wrong tool for the job.

Also, about the Keysolution piano: Mike did a stunning programming job with the VAST filters, fine-tuning the response of every aspect of timbre. Sometimes, he built nine or ten VAST layers for a single program. VAST layers aren't different sets of samples; rather, split and layered "zones", each one customized to give the best overall response. I learned a lot about sound programming by trying to reverse-engineering his piano patches. I don't see such an attention to detail in those large-memory pianos; their main concern is to provide large, clean samples. But that's not enough.

 

Third: Other factors, like pedal response, sympathetic string resonance, sounboard resonance, and the resonance of metal parts, aren't receiving enough attention. Yes, some manufacturer claims to have 'sympathetic resonance' or 'half pedaling', but those are rather crude implementations; still a long way to go! I wouldn't attempt a detailed discussion here... I have a concert to practice!

 

Fourth: The basic quality of the strument itself which is sampled. I've heard sampled pianos from recordings of instruments which sound clear and clean, but *not* warm and full. All pianos are different, and a really exceptional one is rare. They should put more work in the search for an excellent piano to sample. It makes, obviously, a big difference.

[marino]

Originally posted by soundscape:

As mentioned the GeneralMusic have a patented filter ("FADE".) I've been reading through the blurb on their site and watched the video where it's explained, and it's "supposed" to change the partials in the same way that a real piano would, not just open/close a low-pass filter.

Thanks Soundscape - you have anticipated my response.

[burningbusch]

Ivory, for example, uses an extrapolation technology. If you have sample A at one velocity level and sample B at another, the change in the waveform from A to B are extrapolated and recreated in realtime. Obviously this takes some significant CPU. I think this is very useful technology, not just for pianos but any sampled instruments. I hope was see it more in the future, likely in soft samplers. I think Ivory is fairly successful here, but it is first generation. I don't know exactly what NI is doing in Akoustik Piano. I do know that the full scripting language built into Kontakt 2 (the AP engine) is way beyond anything else out there, hard or soft.

 

I agree with you regarding warmth. I've yet to ANYTHING that captures the warm of a quality grand in the mid section of the piano.

 

There are so many compromises and issues. You have to determine what bugs you the least and learn to live with it. I, for example, can't stand many Yamaha and the Kurzweil Triple-Stike pianos. When you brighten them up even a little, the looping which kicks in after 1-2 seconds becomes very apparent and makes them sound, to me, like an RMI electronic piano. Other people can't stand anomolies of the Rolands, Kawai, etc.

 

I was working with Lounge Lizard 3 recently and found that they had done a very good job recreating the complex tonal change that can happen when you do rapid key strikes at high velocity, sustain pedal on. LL3 is completely modeled. Tonally it's not an exact recreation of the Rhodes but how it responds and the changes in timbre possible are FAR closer than ANY sampled Rhodes. In the end, I'm afraid sampling is just the wrong technology. A Rhodes is a vastly simpler instrument than a piano. The amount of CPU required to accurately model the complexities of a piano is likely beyond what we have available today in a standard computer. Plus the necessary software hasn't been developed or is in a rudimentary state. As it stands today you have to grab which ever one bugs you the least and try to not lose sleep over it.

 

Busch.

[Bobsk8]

I wish I had a nickle for every acoustic piano that I have played that sounded like crap......

[lowerhodes]

Originally posted by Burning Busch: "As it stands today you have to grab which ever one bugs you the least and try to not lose sleep over it."

 

 

This reminded me of something I learned a long time ago as a kid playing bars and high school dances (as a guitarist). Your musical identity isn't defined by what tool you use, it's how you use it. The day I stopped obsessing about gear I became a better player. I guess I haven't learned that yet regarding keys. But there's hope. I'm just going to move forward with the 700SX, which is a pretty good do-it-all board and start obsessing about music . What a concept! Thanks to everyone who responded.