Disklavier

The URL I provided below seems to be dead.  Here is an alternative URL 
for the same report.

http://ftp.ai.univie.ac.at/papers/oefai-tr-2001-27.pdf

Bob Welcyng

Robert Welcyng wrote:
> James,
> 
> If you are considering buying a DKV, my suggestion would be to spend 
> some time in a dealer's showroom.  Sit down and play one (or a have a 
> pianist friend play it) and evaluate by ear how well it reproduces the 
> original performance.  Try listening to some PianoSoft diskettes with 
> full dynamic range and decide if you like what you hear.  Try a 
> GranTouch, as well.
> 
> Acoustic measurements that I have made reveal that my Mark II does not 
> perfectly reproduce what was played.  You can also check 
> www.iua.upf.es/mtg/mosart/papers/p35.pdf for a university paper which 
> discusses the DKV's limits of accuracy.  However, despite those 
> limitations, I, and I'm sure many other others, are more than pleased 
> with their instrument.
> 
> While examining specs and pondering their effect is certainly 
> appropriate for anyone to do, the question really comes down to how much 
> the shortcomings are going to interfere with one's use and enjoyment of 
> the instrument.  You could set down ideal specifications for a 
> reproducing piano, but, given the few brands available to choose from, 
> what can you do if none fully satisfies your demands?
> 
> Finally, recognize that recordings made on DKVs are the ideal 
> performances for playing on DKVs.  Expect to have to make adjustments to 
> files that were recorded from other sources that had different 
> characteristics from a DKV.  Some files, such as you find on the Net can 
> be made to play well, and others are horrid--worth exactly what you paid 
> for them.
> 
> 
> 
> James Fry wrote:
> 
>>On Sat, 8 Mar 2003, Carol Beigel wrote:
>>
>>
>>>I probably should have explained a little more about the "escapement"
>>>mechanism of a piano action being different from just a laser shutter
>>>breaking a beam measuring  an electronic value.  An electronic key can
>>>measure all numbers, 0 to 128 for velocity.  All data recorded on an
>>>electronic keyboard with a tone generator that is well designed should
>>>record and playback as accurately as recorded.
>>>Hammer sensors and key sensors measure the travel of the piano you are
>>>recording on and generate a set of numbers.  Take those numbers and play a
>>>different piano, you are dealing with different amounts of friction and
>>>piano action adjustment.  Also, the position of the laser shutters on the
>>>hammers is BEFORE the hammer reaches let-off.  No matter what that number is
>>>it is overwritten by the mechanical let-off ajustment - which is usually set
>>>between 1/8 and 1/4 inches away from the string.
>>
>>
>>The escapement mechanism is largely irrelavent here though; what the
>>optical sensors measure is the speed that the hammer is flying into the
>>strings before it hits. Even if thats before "let-off" it is pretty darned
>>close to the final velocity of the hammer.
>>
>>Anything recorded on one system and then played back on another is going
>>to sound system. This applies to cassettes, CD's, MP3's, samplers, digital
>>pianos, and pianos, but the piano should have compensation for that in the
>>electronics (essentially statistically normalising the data based on known
>>parameters of the device).
>>
>>
>>
>>>Bob is right in that a mechanical piano action plays as low as 20.  You
>>>certainly would not want it to play harder than 100 (and I think that is too
>>>loud)because the solenoids will pound the keys much harder than they are
>>>designed to be played. In other words, the dynamic range of a fine piano is
>>>not 0 to 128, but more like 20 to 100.  The velocity measurement in MIDI is
>>>not affected by friction of a leather knuckle or the escapement of the piano
>>>action.
>>
>>
>>Sure, its a different way of measuring velocity than a digital
>>keyboard, but there is no reason at all that it could not (and it most
>>certainly _should_) output a full range of velocities in exactly the same
>>way.
>>
>>What you are saying concerns me as I am about to buy a disklavier - I can
>>see what you are getting at, but the possible dynamic range of piano
>>should bear no resemblence to MIDI velocities.
>>
>>If the piano limits MIDI velocities 20-100 in the way that you say, it is
>>significantly reducing the resolution of any recording and introducing
>>very significant quantisation errors. The human ear is a very sensitive
>>organ - try listening to 8bit digital audio (256 values) and 16 bit
>>digital audio such as CD (65,536 different values). Similarly, a pianist
>>can reproduce many more than the 80 different volume levels that you
>>suggest (and many more than the MIDI ideal of 127).
>>
>>Ideally the output from the piano on the very, very, very quietest notes
>>should be MIDI velocity 1, and the very loudest you could possibly get out
>>of it should be 127. In an ideal world it wouldn't be stuck in 1983 and
>>7-bit MIDI values - we'd be using 14/16/24-bit velocities. Note velocities
>>are a continuous function, and MIDI velocities are discrete and are always
>>going to be an approximation, but you do a lot to reduce the inaccuracy.
>>Limiting to little over 6-bit resolution is a bad idea.
>>
>>Your argument for playback is also confusing me - surely MIDI velocity 127
>>should be set to the loudest that the solenoids can play without damage
>>occuring, not 100, and MIDI velocity 1 should be set to the quietest
>>audible, rather than 20?
>>
>>
>>
>>>This Disklavier Pro is the best thing out there for recording live piano
>>>performances.  That it will not record MIDI numbers between 0 and 20, and
>>>100 and 128 is really of no consequence.
>>
>>
>>It is bordering on a tragedy because some of the subtleties will have been
>>lost which make the performance sound live. A friend owns a Yamaha P80
>>digital stage piano, and that is similar in that it wont output notes
>>below around 20 and above 100, which he finds extremely annoying when
>>trying to use it as a controller with Steinbergs "The Grand" etc. Perhaps
>>it is a design "feature" of Yamaha's ?
>>
>>Regards,
>>
>>James
>>
>>
>>To Post a message to the group, send it to:   disklavier@...
>>
>>To Post a private message to Todd Muncy, the group's founder and moderator, send it to:
>>disklavier-owner@...
>>
>>To reach our group's web site go to:
>>http://Yahoogroups.com/group/disklavier
>>
>>Todd's family web site was completely updated 01/15/03.  It contains some fun disklavier content and links to midi sites among many other things, The url is:
>>http://MuncyFamily.com 
>>
>>THINKING OF LEAVING THE GROUP?
>>If you are thinking of unsubcribing because you are getting too much mail, go the the web site and change your email delivery option instead.  That will fix the problem, while maintaining your access to the group.  If you insist on leaving us completely send a blank email to:
>>disklavier-unsubscribe@... 
>>
>>Know someone who wants to join?  Have them send a blank email to:
>>disklavier-subscribe@... or give them this link:
>>http://Yahoogroups.com/group/disklavier/join 
>> 
>>
>>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ 
>>
>>
>>
> 
> 
> 


-- 
Robert Welcyng
Anchorage, Alaska