Optical Scanner Design
By John McClelland
This is a short collection of thoughts about Spencer Chase's questions about using a hand scanner to read piano rolls.
The major issue that comes to mind when thinking about making an optical reader is to hold the device steady enough so that there is no movement or vibration to interfere with the reading device, so the notes (holes) don't move. The other issue is to insure sufficient resolution so that the notes you capture are actually where you believe they are.
So much for the theoretical aspects; I tried using a CCTV camera to capture the image, much like a series of photo slides. Scan an image, index the roll and line up on some known reference point, then scan the next image. One of the problems is the time it takes for the image to be captured, the time it takes to move the roll and mostly, the time it takes to find the reference point. There is then no assurance that the reference point is absolutely in the right place. The software that it would take, 'fuzzy logic', is beyond my ken & interest. In short, too many variables to control.
I have also seen a much reduced photograph of a piano roll (a 30 foot roll reduced to about 20 inches long). The accuracy of the roll, and the notes on the roll was faithfully reproduced. Sort of a micro roll, for a micro piano.
I spoke last year with Mike Ames, who has a device set up to read a roll one line at a time. There was something said about cracking the code on the microprocessor that controls the unit that was difficult, and took a 'geewhiz' computer guy to figure it out. I didn't really explore as to whether Mike would share the technology or license it, since I feel sure that he has some time and money invested. That also might be an avenue.
I don't believe that a hand scanner would do the job. 1600 pixels sounds like a lot, but spread over 11.25 inches it comes out to a pixel every .0070 inches. For safety sake, I would suggest making the field of vision 11.5 inches, to accommodate wavy and damaged edges of the roll. (Which also begs the question of what does the compuer software do with the damaged edges?) This then places a pixel looking at every .0072 inches. With a hole diameter of approximately .075 (on average), then there will be about 10 pixels per hole.
Another issue about the hand scanner is somehow focusing the roll image on the pickup: if it's out of focus then the holes won't show up or will not be in the correct place.
Then there is the issue of having a tick or time pulse that happens at some regular interval and ties the event (or non-event) to the tick. Is the pulse generated by the computer, the scanner or where? And the most formidable obstacle about the hand scanner is its advantage in all other applications: it's portability. Again, I think it is critical that the relationship between the optical device and the roll be absolutely rigid and unmoving, otherwise its the old 'garbage in, garbage out' scenario.
No real easy solutions, Spencer, but I would welcome working with you and someone who has some programming skills; I think I can handle the hardware end of it.
John McClelland
[ Thanks for sharing your experience with us, John. I've asked Wayne
[ Stahnke and Mike Ames for their input, too. -- Robbie
|
(Message sent Mon 17 Mar 1997, 04:21:39 GMT, from time zone GMT-0800.) |
|
|
Mechanical
Music Digest™ Archives
