Modifications to Casio keyboards from the 1980s

(Bigger version: Casio-M10-big.jpg)

There was a gig on 2009 February 13th: http://www.myspace.com/batterycovermissing  focusing on 1980 to 1985 Casios!

In 1980 I purchased a Casio M10, the first battery operated, portable, polyphonic digital keyboard.  In the next year or two I developed modifications for this, the MT-30 and the MT-40.  After producing two photocopied booklets, I published the modification iinstructions in an offset-printed booklet and sent them to 300 or more people around the world who wrote to me, with a small payment.  I got a very nice collection of appreciative letters and stamps!  Publicity was via "New Products" announcements in some music magazines, but I think the biggest response came from professional electronic magazines.  

Devo used a modified Casio - an MT-30 I guess.  Maybe they were used in famous pieces of music, such as Thriller ("froggie casio" notes     ).  They were used by lots of people having a lot of fun making their own music, often in places it had never been made before.  I took mine into a well on a farm, with a flanger and a transistor radio as an amplifier.  They could be swung around at arm's length, giving Doppler shift, while playing a chord.  (I have a key-ring on my VL1 and I think VL5 so they could be swung around on a piece of string.)

My M10 is still going strong - it is beautifully made, with a steel chassis.  All Devil Fishes are tested with this machine as an audio source.

The PDF of the 1981 booklet MODIFYING THE CASIOTONE INSTRUMENTS is here:  casio-mod-01.pdf   (31 Megabytes)

This is a PDF from TIF files which are scans of the offset printed booklet.  See the sys-admin directory of this site for how to make PDFs from TIFs. (The three update pages 14 to 16 are from a computer file.) The booklet was printed on 4 sheets of A3 and folded together.  The diagrams on page 13 were in the middle of the booklet, on an A3 sheet.  The artwork for the booklet was done with a 1959 Kenro process camera (sadly gone to scrap in 2006 ... I kept the lens) and the typesetting was produced in that camera (onto Agfa "bromide" Copyproof ) by reducing largish pieces of paper printed with an IBM golfball typewriter.  This was not an ordinary typewriter.  It used to be the console of an IBM 360 mainframe computer. (This was before CRTs - the 1960s.  A pair of 360s ran Houston Mission Control for the moonshots, with a clock rate of 500kHz, I think and 256 k bytes of RAM.)  The printer just had a bunch of solenoids to control the mechanism, which is entirely mechanical an based on the entirely mechanical IBM golfball typewriter.  I built hardware to drive the solenoids from a Z-80 PIO, and then some driver software to convert ASCII codes to the correct commands for how far to spin and tilt the ball.  It printed 10 characters a second, used a film ribbon if I wanted good quality.  All mechanics was driven by a 240 volt motor. The text was written on Wordstar 3.3 on a Z-80 machine.  This was on my home-built machine which ran Cromemco CDOS, rather than the more common Digital Research operating system.  I later ran a Big Board (1) with CPM.  Anyway, I picked up the IBM golfball printer from a computer service company in Sydney, along with a bunch of gold-bearing circuit board and connectors and a block of 64kbytes of magnetic core memory, on the pretext I wanted them for a stage-set for the band I was in (Equal Local).  The real reason for getting most of this was to extract the gold from the connectors and sell it for scrap - only a few people were onto it then.  Anyway, the golfball printer worked nicely from Wordstar, and there were various balls which could be put on it, with various typefaces.  There was no proportional spacing, as with a daisy-wheel, but I there were advanced IBM golfball typewriters which people used for low-cost typesetting. The fancy balls were made for those, but fitted the old 360 console printer.  

A moment's silence for the printers I owned and sadly, have no more:  A Creed 7B baudot teleprinter (1940s, once seen printing a telegram in a Humphrey Bogart movie), the IBM golfball and two Diablo Daisywheels  - beautiful devices with fiendish TTL-based digital servos for the carriage and daisywheel DC motors - I had one left, but it was destroyed in a flood.  (Googling indicates these extraordinary machines from the late 1970s are quite obscure now.) An old Tally 15" printer and a washing-machine sized Hitachi printer from the 1980s - too big to keep, it had to go.   I still have a piano-sized Benson plotter.

Also, a text file which describes how to give commands to the upD931 chip, as found in the Casio MT-65. 

Casio-931-2006-06-17.txt

If you have anything to add to this, or any relevant pages to link to, please let me know.  Please do not republish these files elsewhere.  It is best that I maintain a single copy here, which  saves various out-of-date versions lying around causing confusion.

DAC improvements

Here is a rough outline of improving the quality of Casiotone DACs.  I have found some detailed notes on modifying the CT-202 which explain this better.  Please keep an eye on this page - I will turn them into a PDF and put them here.


To get a really good sound quality from these 14 bit DACs (as used in the original M10 etc. series and the MT-65 series), it is important to rebuild the circuitry in a new way, and to drive the 12 bit R2R network (which needs to be removed from the old circuit, since you can't buy these things) and another two bits of matching discrete resistors for the two least significant bits, from latches, not from inverters etc. driven by the LSI chip's outputs.

The timing of the 14 bits coming out of the LSI chip is not ideal.  These DACs have a sample rate of 567,086 Hz (for the M10 series - I think the MT-65 series is similar or the same).  The DAC has no sample-and-hold.  The output of the R-2R network goes straight to the audio amp.  Use a nice op-amp and don't use the dull RC filters the Casios have - or maybe use a pot to fade between the raw, bright, DAC sound and the filtered sound.

The problem with the way Casio build the DAC is primarily that the output of the LSI has different slew rates going positive and negative, and that the CMOS buffers used to drive the R-2R network also have skewed slew rates.  This means that on a clock transition in which the signal goes from 01111111111111 to 10000000000000 there can be a brief time when the R-2R network gets 00000000000000!  This only needs to be a few nsec and you have a nasty glitch, since this is such a big deviation from the two, almost identical, voltages which should result from the first two numbers.  This would be if the downwards slew rate was faster than the upwards.

Anyway, the trick is to get the 14 bits and latch them into several 74C174s.  Nowadays, we would use 74HC174s. Clock them from the LSI chip's clock signal.  These latch chips (in my experience) have essentially symmetrical timings for rising and falling edges of their outputs.  So there are no such glitches.

You need to use multiple sections of a 74HC174 or whatever for the most significant bit, and maybe the next few bits which are not quite as significant.  This is because it is very important to have the correct resistance in the R2R network for the MSB - bit 13, the one which is worth 2^13 = 8192 steps in the full 16384 range of the DAC.  We can assume the R-2R network has nearly the right value, but when added to the output impedance of a single 74HC174 or whatever, this adds some low resistance (tens or hundreds of ohms) to the drive for the resistor array, and then it is no longer an R-2R network.  So gang about 4 sections of a 74HC174 in parallel to drive the MSB.  Maybe use two or so for bit 12.

Two more things are required to get this right.

First, you need a trimpot to finely adjust the weighting of bit 13.  Get two latches and drive one's input from the LSI's bit 13 and the other from the inverse of the LSI's bit 13.  Now string a 50k pot between the outputs of these latch stages and from there, say a 1M or so to the output of the R-2R network where it drives the audio amp.  This enables you to tweak the trimpot and compensate for whatever error there is in the weighting of bit 13.  The best way is to have a decaying sound with a waveform which goes positive and negative - say a piano sound.  Tweak the trimpot for minimal distortion at the end of the decay.

Finally, in all this work, there is going to be a different capacitive coupling from bit 13 compared to all the other bits put together.  This results in glitches when going from 01111111111111 to 10000000000000 and back, even if you have the DC situation completely adjusted, as with the just-described trimpot.  This capacitive glitch removal requires careful design, layout and tweaking.  Maybe use the outputs of the normal and inverted bit 13 latches and just a tiny bits of bent wire (or did I use insulated wire twisted around another wire to just the right amount), to form a sub-picofarad trimmer cap more to the normal or the inverted latch output.  Couple this directly to the summing point where the R-2R network drives the audio amp. This can zero out a nasty glitch and the sound can be really pure and sweet.

Links