Hi Robbie, Thanks for your interest in my system, details of which
I'm please to give.  Here goes...

I've never told this story, which I hope readers find interesting.
It's about how I solved a major dilemma: getting lots of music for
my Ampico without spending much money.

In 1976 I met Denis Condon, a local collector who got me into the
hobby.  The next year I bought an unrestored Ampico A, (1923 Knabe,
5'4"), and with lots of help, restored it to playing condition.
It's here the problem arose; great piano, no music!

Being a lecturer in electronics, I decided to attempt developing
a means of recording Ampico rolls for playback on my Ampico.  But which
first: the recording machine or the playback system?  After all, I knew
of no one else doing this sort of thing, and I had no idea where to
start.  As well, I was tightly constrained financially, due to young
children, mortgage and so on.

However, by 1979 I had a prototype system working, which I demonstrated
to an amazed group of local collectors.  From then on, virtually every
Ampico owner who knew me wanted my system, usually attracted by the
huge library (eventually over 1500 recordings).  I hand built each
system, and installed it for a total cost of A$600.  What little money
I made came from selling the recordings.  Eventually others took over,
manufacturing and distributing the system, and I think ultimately
around 15 or so systems were built.

Of interest is how I did it.  In brief, I first developed the serial
data transfer format, devising what I guess would now be called a form
of FSK (frequency shift keying).  The scheme involved 101 data bits,
with three frame bits and 98 Ampico bits.  A data 0 is represented by
one cycle of a 6 kHz sine wave, a data 1 by one cycle of a 3 kHz sine
wave, with a 0.3 millisecond gap between frames.

This format was compatible with standard C60 cassettes, and gave a scan
rate of around 55 scans per seconds.  The electronics to decode the
data comprised a few 555 timer ICs and two digital chips which supplied
all the signals needed to clock the data through 13 8-bit shift
registers, and to time a number of operations.

The valves were very simple, comprising a coil of 4000 turns around a
3/8" mild steel rod, cut to 1.25" in length.  When energised, the coil
attracted a 1/2" mild steel disc that would uncover a hole.  (That's
about 1/2 million turns of wire per system, allowing for discards, yet
I wound them all!)

There's a photo of the recording machine on my web site
(http://members.optushome.com.au/eleced/index.htm), but it's a bit more
sophisticated than it looks.  Like Wayne Stahnke, I used a system of
electro-pneumatic sensors, having decided that optical reading gave too
many problems.

In principle, my sensors consisted of a pouch of Perflex (it's still
fine!), with a 1/2" metal disc sitting on top of the pouch.  The
disc formed one plate of a capacitor, so when lifted by the pouch,
sufficient radio frequency energy would be passed to trigger an SCR.
The underside of each pouch was connected to a tracker bar, held in
a pressurised spool box (around 2.5" WG).

This meant I could adjust each "switch" to give optimum repetition by
adjusting the size of a bleed to atmosphere.  It took me some time to
get everything fine-tuned, as the throw of the solenoid valves was as
critical as the size of the bleed in each switch.  I developed a means
of accurately adjusting the valves by using a manometer.

Like Wayne, I also used an Ampico B spool box, or at least one built
to emulate this box.  My electric roll drive was a VW windscreen wiper
motor, operated through a speed control circuit that gave a digital
readout of the motor speed.  So I know the rolls were recorded at the
correct speed.

The system behaved like a remote spool box, in that I could put a
roll on and listen to it on the Ampico.  I started by recording onto
reel-to-reel tape (Revox tape deck), monitoring the recording off tape
during the process.

The main problem I had with systems sold to others was the quality of
the tape player being used.  So rather than do the Superscope trick of
doubling the tape speed and supplying a special tape player, I decided
to go computer, as by now I had purchased an Apple II clone.

It was quite an ask to make a 64K Apple fit a roll recording, basic
DOS and the program to play the data, but with some tight assembler
programming involving a complex data compression routine, I was able to
achieve this will all but a few recordings.  These files required the
use of a memory expansion card (128K card).  I eventually burnt all the
programs into the Apple ROM (16K), which meant the user had to merely
turn on the computer for it to be ready to accept a disk (108 of them).

As far as I know, I'm probably the first guy to actually interface
a mechanical piano and a computer, and to make it marketable.  Once
computerised, I was able to increase the scan rate of the recording
machine, and to also test all functions of the Ampico from the
computer.  I included my test program in the ROM-based machines.

I last sold a system in the late 1980s, although they have now become
quite sought after.  These days I play the Ampico from CDs, although
the trusty Apple is still working, mainly for the Ampico test program.

But a real legacy of this work is the library.  All files have now been
converted to MIDI, and I mainly listen to the roll recordings on my
Disklavier.  But I can also play the same recording on the Ampico,
which gives a good comparison.  I can now look at the files in Cakewalk,
and for the most part they are clean and ready to go as MIDI files that
sound terrific.

So, it all worked out in the end; I got the music and sold enough
systems to pay for it all.  Now, with all that history and experience,
I feel poised to go on and develop a better system (as described in MMD
030616), but one that still operates from an audio signal.  I love
computers, but I also love the simplicity of inserting a CD and pushing
play to hear my favourite music on the Ampico.

Peter Phillips
(Australia)
http://members.optushome.com.au/eleced/index.htm

PS: I have another story about an electro-pneumatic Duo-Art Vorsetzer
I built that was used to record the Grainger performance of the Greig
Piano Concerto with the Sydney Symphony Orchestra in 1978.  The record
won awards in the US.  But I've probably taken up too much space
already.