> Would it be practical to sample the _touch_ of a 9' Steinway, and use
> feedback to a motor on each key of the synth keyboard to simulate that
> touch?  "Motor" in this case would be a voice coil, or some other
> solenoid form.

A touch-feedback system like this has been developed at Stanford
CCRMA (Center for Computer Research in Music and Acoustics).  The term
used for this kind of devices is "haptic interface".  I don't know whether
they have sold it to a company, but I wouldn't be very surprised to
hear that (Yamaha's two ground-breaking inventions (FM and physical
modeling) came actually from CCRMA).

You can find more info at:

   http://ccrma-www.stanford.edu/CCRMA/Overview/node49.html#SECTION0043500000000000000

(it really has those zeroes)

Zoli¶
---
  Zoltan Janosy, M.Sc.El.Eng. <zoltan.janosy@hu.bonus.com>
  Appaloosa Online Ltd. (formerly Novotrade Software Ltd.)
  H-1122 Budapest, Varosmajor u. 12-14., Hungary
  Tel:+361-213 0349 Fax:+361-155 2293

 [ Editor's Note:  Yes it does have all those zeros.  And for those
 [ of you who don't want to type all of those, here's a text excerpt
 [ from the Web page referenced above.
 [
 [ Jody

The Touchback Keyboard - Brent Gillespie

A virtual piano action (a mathematical model), a simulation algorithm,
and a set of 88 single degree-of-freedom haptic display devices
constitute a promising means to make a synthesizer keyboard feel like a
real grand piano.  Other keyboard instruments could be simulated with
the same device at the touch of a button.  We propose, then, to simulate
the feel of the grand piano by numerically integrating the equations of
motion of the multibody piano action in real-time in a
human-in-the-loop simulation scheme.  As the integration proceeds, the
finger-key interaction force is computed and generated by a haptic
interface.  In addition, the motion (needed for use in the integration)
is sampled from the haptic interface hardware.

We believe that haptic (tactile/kinesthetic) feedback from an
instrument, in addition to the aural feedback, plays a crucial role in
the process by which a keyboardist controls the tone and timing.
Further experiments in human-computer interaction can be made with
arbitrarily determined dynamical behaviors of the keyboard.

The behavior of the simulated piano action should appear passive to the
user even though it is realized with active devices.  We seek design
guidelines which will ensure that such dynamical behaviors
characteristic of active objects (chatter and stickiness) will not
arise.  Applications of both linear and non-linear analytical tools are
being applied to this end.