Disklavier

While varying levels of dynamics are important, I think far more 
important than increasing the number of levels is the timing accuracy. 
As mark says, the action wasn't designed with this application in mind, 
and I find it hard to believe that

Studies have been done that show playback accuracy of player pianos is 
not as good as you would hope - there was a paper somewhere (possibly in 
the archives?) which compared a mark II disklavier to at least one other 
system and found that there were great timing inaccuracies when the 
velocity was varied. Given how important small timing changes are in 
other areas of audio, I wouldn't mind betting that it makes quite a 
difference with pianos too. It would be interesting to see how the Mark 
IV compares to previous versions of the disklavier.

Regards,

James
Mark Fontana wrote
On Mon, 13 Nov 2006, Carl Youngblood wrote:
>
>
>
> I'm sticking to the "mechanically impossible" statement, adding the
> reasonable assumptions that the piano is not in a vacuum and uses a
> standard grand escapement action, which just wasn't designed to achieve
> this level of precision.
>
> Yamaha's own admission that they can control the solenoids with 1024
> levels but realistically get only around 256 distinguishable levels
> already leads one to extrapolate that achieving beyond ten bits of
> playback resolution is unlikely... adding more input bits isn't the answer
> to overcoming mechanical limitations.
>
> PianoDisc mentioned at the 2006 NAMM show that they have a high-res
> playback system in development, and it uses ten bit expression levels.
>
> Wayne Stahnke's Boesendorfer SE system is ten-bit.
>
> Boesendorfer's new CEUS system is ten-bit too.
>
> Ten bits appears to be the practical upper limit for this application.
>
> Regarding the amplitude resolution of human hearing, these links are
> interesting (though more casual than scientific):
>
> http://www.ethanwiner.com/BitsTest.html 
> <http://www.ethanwiner.com/BitsTest.html>
> http://www.pcavtech.com/test_data/ <http://www.pcavtech.com/test_data/>
>
> Mark Fontana
>