Genre: Software Driver      Creation Date: October 2015     Language: 6809 Assembly Language    System: Tandy Color Computer 3

This project was jointly created by John Kowalski, Robert Gault and Nickolas Marentes 


The Tandy Color Computer was introduced with two analogue joystick ports. The analogue joysticks provided a variable voltage to an internal Analogue-to-Digital converter (ADC) circuit which translates these voltages into a 6-bit value in the range of 0 to 63.

In 1984, a mouse was released and sold by Radio Shack. Like the joysticks, it also was an analogue device which mimicked the joystick's operation using two variable resistors for the X and Y axis. The resolution of the mouse was the same as the joystick since they were both tied to the same 6-bit ADC circuit within the Color Computer.

Unfortunately, the 6-bit precision of the ADC meant the joysticks and mice could only address a 64 x 64 area and this was too low to represent every pixel position of the original Color Computer's high resolution 256 x 192 screen. Furthermore, scaling the joystick to match the screen resolution produced a jumpy and innacurate mouse movement that was unable to effectively represent every screen pixel accurately.

When the Mac-like paint program CoCoMax was released in 1985, it came bundled with a rompak style cartridge containing an 8-bit ADC chip that gave a resolution of 256 points. This worked very well but unfortunately, CoCoMax was the only software that supported it.

In 1986, Tandy released the Color Computer 3 with a new 320 x 225 screen in 16 colors and a maximum 640 x 225 screen in 4 colors. The Color Computer's 6-bit ADC resolution was too low and another solution had to be found.

A new hardware interface was required that improved on the low 6-bit resolution of the ADC and so Radio Shack released the High- Resolution Interface in 1987. This device was designed by the famous Steve Bjork and was manufuctured and sold by Radio Shack. It utilized a different method of reading an analogue joystick that was also employed by the Apple 2 and IBM PC at the time.

Rather than convert the incoming voltage directly via the ADC to a digital value, it used a capacitor which was charged then discharged through the resistance provided by the joystick stick position. The time it took for the voltage to drop back to zero was monitored and the time of discharge measured then scaled to provide a value that matched the Color Computer 3's high resolution screens. It connected to both a joystick and cassette port, utilizing the cassette port to trigger the capacitor discharge and the joystick port (via the ADC) to sense when it reached full discharge.

A modified version of this interface was later released with the program CoCoMax3 by Colorware that operated the same way but did away with the cassette port triggering and only required the joystick port connection.

A third high resolution mouse option was later released by Diecom which utilized a PC style serial mouse and connected to the Color Computer's serial port. This was sold with their drawing and paint software, The Rat.

Common with all these high resolution options was that they needed a dedicated piece of hardware to replace the limitations of the Color Computer's built in ADC circuit.



"You built a Hi-Res Interface... out of a DeLorean!?"

Back in the early 90's, John Kowalski had an idea for a new way of reading the Color Computer 3 joystick ports that could extrapolate additional information  between the 6-bit sample points of the built-in ADC. He theorized that the voltage applied to the Digital-to-Analogue converter (DAC) which worked with the ADC didn't merely jump instantaneously between each of the 6-bit points but transitioned as a rapid voltage slide and that with very fast and syncronized software, extra readings could be had that would provide a higher level of precision.

But at that time, he didn't have any project planned to use such a feature and moved on to other groundbreaking software projects of which John has become famous for.

In 2007, John created the groundbreaking feat of  converting the Z-80 based Donkey Kong original arcade code to 6809 and emulating the functions of the original arcade hardware on a stock standard CoCo3. In 2015, he improved it by making it slightly faster and added 6 extra levels to the original game. On completion of the project,  his thoughts of the new hi-res software routines returned. He mentioned his idea to me and I immediately said "let's do it!".

I've worked with John in the past on some of his brilliant ideas. Firstly, he helped me to understand his groudbreaking Gloom 3D algorithm which I used in my game Gate Crasher, the first true solid 3D complete game for the Color Computer 3. He was very instrumental in explaining the math involved. As kryptonite is to Superman, math is to me. I hate math and his patience and clear explanations helped me to complete Gate Crasher in 2000.  LINK

We later worked on another of his ideas we titled DigiWiper which primarily involved the removal of the Macrovision encoding on commercial video tapes... for the purpose of creating backups of our original tape investment! As a bonus we were also able to add realtime video transitions between the dubbing of video tapes.   LINK

So, I had no hesitation in listening to his latest "crazy idea" and I began coding to see if it was indeed possible.

How the ADC works

The real work when converting any analogue voltage to a digital reading with the Color Computer is actually done using the DAC (Digital-to-Analogue) converter. To understand how the ADC (Analogue-to-Digital) circuit works, we must understand how the DAC works.

In simple terms, the DAC accepts data from the computer and converts it to an analogue voltage. Although the computer's data bus is 8-bit, the input to the DAC is only 6-bits. This represents a value from 0 to 63 or 64 individual voltage steps as output from the DAC. This is also how the 64 levels of volume is derived when the DAC output is diverted and used for sound generation.

In order to get a digital reading from the joystick port, the output of the DAC is fed into an input of a voltage comparator. At the same time, the voltage from the joystick or mouse is fed into a second input of the voltage comparator. This voltage comparator compares the two input voltages and sets an output high or low to indicate the difference. Software running on the Color Computer can read the comparator output and determine if the DAC voltage needs to be raised or lowered to attempt to match the joystick port voltage. This test is repeated until it locates the DAC voltage value which coincides with the incoming voltage. Through a process of succesive approximation, an accurate ADC value can be found.

This is the first stage of reading the joystick port that derives the normal 6-bit joystick port reading.

For the rest of the explantion of how we go beyond this 6-bit limitation, let's assume that we scanned the joystick port and found a value of 32 with a joystick at midpoint.

Deep scanning "the abyss"

Now to test John's theory and see if we can pull additional readings from points in between any of the 64 ADC points. In this example we want to scan between 31 and 32 as well as between 32 and 33.

The first step is to create several delay points that trigger from the moment we set a new DAC voltage level up and down 1 unit from the current. For example, with a 6-bit DAC reading of 32, we then need to switch the DAC voltage to 31 if we want to scan below. If we want to scan above 32, then switch to 33. The moment we set the DAC to this new value, we need to immediately begin our timings.

Multiple scans are then made at varying time intervals with the first interval starting at a 4 cycle delay from the moment the new voltage point is set. More voltage points were scanned at 1 CPU clock cycle iterations after the intial 4 cycles.

Why 4 cycles for the first delay point?

Because 4 cycles is the shortest number of CPU cycles to write and read the data to the DAC (with a stock 68B09 CPU set to 1.79Mhz). The remaining points then add to this. In the end we wound up with 5 time intervals down from the base voltage and another 5 time intervals up from the base voltage; i.e. 10 additional sampling points.

You can see from the graph on the right that the 4 cycle delay creates a period where we cannot obtain data.

Now to see what we CAN find!

We did many multiple scans, graphing the results and displaying them in realtime while we moved the joystick ever so slightly. The results looked promising and we discovered that the clearest and most accurate readings were to be found at the 4 and 5 cycle delay when the voltage was dropped down from the base setting. The weight of the readings varied the longer the delay from the base voltage but even the poor readings provide additional information that we could use.

I found that the readings when the voltage was raised from the base were not being detected as well as the readings when the voltage was lowered from the base. I suspect that a rising voltage happens much faster as more voltage is injected via the DAC, but a falling voltage is slower due to capacitance in the DAC circuits dampening the voltage decay. The faster shift of the rising voltage was too fast for the slow 6809 to detect properly and gives a very narrow window of data. These are not given very much weight but even these contribute to the total reading and help fill in a couple of the gaps.

Fuzzy Logic

After many days and weeks of experimenting and analysing, John applied his amazing skills of math and data analysis to average and weight all the results to eventually obtain a reasonably accurate 15-bit total joystick port reading. Quite amazing when you factor in the general noisiness of a 30 year old variable resistance device such as the Tandy joystick or mouse, not to mention the electromagnetic interferance from surrounding electronics. It's amazing to be able to piece together the little scattered bits of data and reassemble them into a combined whole reading. The noise in the readings also help make our multiple readings randomize a bit and by averaging out the roughness, we get a non-noisy reading with even higher resolution than otherwise possible.

A further massage of the obtained data to apply some extra smoothing and scaling to match the Color Computer's screen resolutions and we had a working Hi-Res joystick interface driver!

The Hi-Res interface you always had!

I created two programs to demonstrate the driver. The first DEMO320 places a mouse pointer on a 320 x 200 high resolution screen that can be moved smoothly across the entire area. I also added the function to scribble on the screen when the mouse or joystick button is held down while the mouse pointer is moved.

For fun, I placed a screenshot of Greg Miller's and Eric Gavriluk's insanely brilliant ColorMax Deluxe paint program as a background to highlight a potential use for such a driver.

Maybe this driver can be used to replace the Hi-Res Interface driver built into ColorMax Deluxe?

The second demonstration DEMO640 demonstrates the driver operating in 640 x 225 and I have placed a 'mockup only' background of an SDC Disk Management utility that I was toying with. It utilizes the highest resolution graphics mode the CoCo3 can produce and was an ideal candidate to see how well the new driver can operate at this resolution. I even provide the scribble mode from DEMO320 to see the fine linework that can be produced.

Surprisingly, the mouse precision is quite smooth with just a few hints of imprecision noticeable when drawing. As for mouse positioning, it's just about perfect and quite amazing that it's operating without any hardware Hi-res Interface... using only the CoCo3's built in 6-bit ADC/DAC!

You can try both of these demos for yourself by downloading the distribution disk in DSK format at the bottom of this page. To start the demos, type RUN "DEMO320" or RUN"DEMO640".

This driver code works on a real CoCo3. Software emulators such as VCC and MAME do not emulate the original ADC/DAC characteristics required for this to work properly at full precision. They will work but only as a 64 x 64 resolution mouse. You will be amazed by how well the averaging and smoothing software is working to give the appearance of more precision.

This demo clearly demonstrates the success of this project and is yet another entry that John Kowalski can cross off the list of things that are impossible to do on the Tandy Color Computer.

Use it in your BASIC programs!

Robert Gault has joined us and added support for BASIC programs by adding additional scaling options supporting various screen resolutions as well as optimisations in speed. It's now possible to use a PMODE 4 (256 x 192), HSCREEN 2 (320 x 192/200/225) and HSCREEN 4 (640 x 192/200/225) screen resolution.

He has created a demonstration in BASIC that calls the high resolution joystick subroutine and places a mouse pointer over the screen. His program is available on the distribution DSK and can be started by entering RUN"DEMOBAS".

To interface your program to the NEWJOY subroutine
Clear memory above &H7000 and load the NEWJOY subroutine. NEWJOY loads from location &H7000
Set the desired screen scaling from BASIC by POKE'ing to locations &H7004 and &H7005.

XSCALE:  POKE &H7004,number    (1=256  2=320  3=640)
YSCALE:  POKE &H7005,number    (1=192  2=200  3=225)

Call the NEWJOY subroutine. EXEC &H7006
Read the X-value (XVAL) and Y-value (YVAL) returned by PEEK'ing at locations &H7000 to &H7003.

XVAL:  PEEK(&H7000)*256+PEEK(&H7001)
YVAL:  PEEK(&H7002)*256+PEEK(&H7003)

As an added bonus, Robert's program also patches BASIC to operate in graphics screens of 200 and 225 lines. Yet another undocumented feature that was possible back in 1986. I will be providing details of this on a future web page.

Use it in your Assembly Language programs!

For anyone interested in implementing this software-only Hi-Res joystick driver into their application or even patching an existing program to remove the reliance on the Hi-Res Interface hardware, the source code is provided on the distribution disk in EDTASM6309 format. A text file is also provided for importing into other assemblers (see below).

Prior to calling the routine, you must select the desired joystick port and axis by setting the appropriate bits in the Color Computer's Analog Multiplexer.









Then, if you wish to sample the X axis of the joystick or mouse, load the location AXIS with a 0 and load register Y to point to #DACX prior to calling the routine (READ). This will return a 16 bit value in XVAL, scaled to the XSCALE setting you chose.

If you wish to sample the Y axis of the joystick or mouse, load the location AXIS with a 1 and load register Y to point to #DACY prior to calling the routine (READ). This will return a 16 bit value in YVAL, scaled to the YSCALE setting you chose.

You will be calling the routine once for each axis for each joystick or mouse you wish to read. Below is the code I used to call the high resolution routine in the demo.

      LDA   $FF01 ;SET JOYSTICK X 
      ANDA  #247  ;RESET BIT3

      STA   $FF01






      ORA   #8    ;SET BIT3

      STA   $FF01





You can delete the scaling parts from the NEWJOY source code that you don't require to make the routine occupy less memory.


We've achieved what we had hoped to achieve... to prove that the Color Computer 3 always could read the joystick in high resolution.  We wish that this had been done back in 1986 when the Color Computer 3 was first introduced then maybe there would have been more mouse driven programs released. Hopefully, we will see new programs utilising this driver and even some old favorites patched to use it.


View the NEWJOY subroutine SOURCE CODE

NOTE: The demos provided on the Distribution Disk only work correctly on real CoCo3 hardware. The VCC and MESS emulators do not emulate the voltage characteristics obtained from the DAC/ADC circuits found on real CoCo3 hardware.