[Coco] Sega Genesis Gamepad on CoCo

[J.P. Samson]

Here's some more reading I found regarding Genesis control pads:



============================================
= Sega Six Button Controller Hardware Info =
============================================

      Source: http://www.cs.cmu.edu/~chuck/infopg/segasix.txt
      Author: Charles Rosenberg, chuck at cs.cmu.edu
     Version: 1.0
Last Update: 9/9/96


==========
Background
==========

A couple of months ago I was looking for a cheap, robust, hand held
controller for a project I was working on.  The controller for the
Sega Genesis was the perfect choice because it was ready available and
had a standard DB-9 connector for interfacing instead of the
proprietary connectors used on the controllers for a lot of the newer
game systems.  I was able to find interface info for the 3 button
controller, but I could not find any information about accessing the
X, Y, and Z buttons on the newer 6 button controllers.  A little
reverse engineering uncovered the necessary info.  This document
describes how to interface to a 6 button Sega Genesis Controller.

============
Introduction
============

The controller for the Sega Genesis is a hand held controller useful
for many applications.  It has a joypad, a start button, and action
buttons.  Another big advantage is that it has a standard DB-9
connector for interfacing instead of a proprietary connector.  The
original version of the controller has three action buttons, labeled:
A, B, and C.  The original version was easy to interface to.  Newer
versions of controller have six buttons, the original three and three
new ones, labeled: X, Y, and Z.  The newer version is a little more
tricky.

======================
Three Button Interface
======================

The interface for the original three button controller is relatively
simple.  (I know that it has been described elsewhere, but I think
that this description is more straightforward.)  You can open up the
controller and see that a standard quad 2:1 multiplexer chip is
inside, a 74HC157.  A 2:1 multiplexer directs data from one of two
sources to a single input line based on the state of a select signal.
So, theoretically, this chip would allow 8 signals to be sent over 4
data lines.  The way the Genesis uses this chip, the data from the 8
switches in the controller is sent over 6 data lines, not quite the
most economical solution, but one which makes the controller backwards
compatible with Atari 2600 joysticks.

Inside the controller is a small PC board which contains the 74HC157,
some 10K pull-up resistors and some bypass capacitors.  The switches
are connected so when a switch is pressed (closed) it shorts its
output to ground, otherwise it's output is 5 volts (via the pull-up
resistors).  One additional connection is a pull-up resistor on the
select line.  If it is left unconnected it will be "pulled" high.

The following tables summarize how the signals are mapped to DB-9.
The output signal to a particular pin depends on the state of the
select line from the Genesis.  During a standard game, the processor
will set the select line high or low, depending on the which game
controller buttons it wants to read.

This table lists connections which are outputs from the Genesis to the
controller, the first column is the pin number of the DB-9:

Pin  Connection
---  ---------------
  5   Power: +5 Volts
  7   Select signal
  8   Power: Ground

This table lists connections which are outputs from the controller to
the Genesis and are dependent on the logic level to the select signal,
again the first column is the pin of number the DB-9:

Pin  Connection with select low  Connection with select high
---  --------------------------  ---------------------------
  1   joypad up signal            joypad up signal
  2   joypad down signal          joypad down signal
  3   logic low (ground)          joypad left signal
  4   logic low (ground)          joypad right signal
  6   Button A signal             Button B signal
  9   Start button signal         Button C signal

Notice that the signals on pins 1 and 2 are not affected by the select
signal, there is a direct connection from those switches to the
corresponding pins.  Also notice that signals on pins 3 and 4 are not
used to send switch info when select is low.

If you monitor the select signal to the controller on an oscilloscope
while playing a game on the Genesis, it will look something like this:

5V ---+ +-------------+ +-------------+ +---
       | |             | |             | |
0V    |_|             |_|             |_|

The low time of the signals is not drawn to scale with the period of
the signals.  However, notice that the signal is mostly in one state
(high), and then is low (inverted) for a short period.  A typical game
may sample the controller at 60 Hz, resulting in a period of 16.7
milliseconds.  The low time for a select signal pulse might be around
20 microseconds.  The timing can vary significantly from game to game.
The basic reason for this pulse pattern, is that every so often, the
game will call a routine which reads the state of the controller.
This routine will bring the select signal low and then record those
values and the bring it back to its default state.

That sums up the three button controller, now on to the new stuff.

====================
Six Button Interface
====================

The three button controller description was just a preamble for the
six button controller.  When you open up the six button controller,
you notice a bunch of differences in the circuit besides the three
new buttons.  The simple one is the addition of a fourth button which
acts as a mode select and tells the controller to be compatible with
three button controller games.  The ugly one is that the nice standard
HC157 has been replaced with a custom chip (a microcontroller, a GAL?)
and most of the pull-up resistors are gone.  So instead of being able
to trace out the circuit, you have to resort to the oscilloscope (or
logic analyzer) and see what the signals are doing while playing a
game that uses the additional three buttons.  My test subject was MK3.

So why did Sega go to all of this trouble of replacing a nice simple
logic chip with a custom IC?  Did they have too much time and money on
their hands?  (Of course that is a possibility.)  No, the reason seems
to be backwards compatibility.  As the descriptions following I hope
will show, the new controller is designed to be compatible with almost
all three button games, without the use of the mode select switch.

Some quick probing on the circuit board reveals that the pull-up
resistors for the switches have moved inside the chip and all of the
switches still output a logic low when pressed.  The pull-up for the
select signal still exists as a discrete resistor.  There are also
some additional capacitors for power supply filtering and generating
the clock signal for the custom chip.

If you monitor the select signal generated during a game, you will see
something like the following:

5V ---+-+-+-+-------------+-+-+-+-------------+-+-+-+---
       | | | |             | | | |             | | | |
0V    | | | |             | | | |             | | | |

The amount of time between the start one of these pulse groups and the
next is about 16.6 milliseconds or 60 Hz.  In my interfacing
experiments I have successfully used 33.3 milliseconds or 30 Hz.

A closer examination of the pulse groups on the select line:

5V ---+ +-+ +-+ +-+ +-----------------------------------
       | | | | | | | |
0V    |_| |_| |_| |_|

The period of these signals for MK3 was about 27.6 microseconds.  That
is the high or low time for the pulses was 13.8 microseconds.  In my
interfacing experiments I successfully used a signal with a 200
microsecond period, 100 microsecond high or low time.  I suspect that
there is some maximum period length associated with a time-out in the
custom chip, but I have not investigated what that is.

Interesting things reveal themselves if you monitor pin 3 of the DB-9
with the X button pressed:

5V ---+ +-+ +-+   +-------------------------------------
       | | | | |   |
0V    |_| |_| |___|

And if you monitor pin 3 of the DB-9 with the X button NOT pressed:

5V ---+ +-+ +-+ +---------------------------------------
       | | | | | |
0V    |_| |_| |_|

Now what Sega has done can be deciphered.  To obtain the new button
data, a series of pulses needs to be sent to the controller on the
select line.  For the first two pulses, the controller responds in the
same way as a three button controller.  However, immediately after the
third pulse goes from low to high, the controller will output new data
on existing data lines.  You can think of this as a new value for the
select line: Low, High, and what I will call Pulse-3.  The fourth
pulse resets the controller to compatible mode.

The updated table with the third select state is as follows, the first
two columns are the same as the three button case:

Pin  Select: low  Select: high  Select: pulse-3
--- ------------  ------------  ---------------
  1  joypad up     joypad up     button Z
  2  joypad down   joypad down   button Y
  3  logic low     joypad left   button X
  4  logic low     joypad right
  6  button A      button B
  9  start button  button C

This table does not change from the three button controller case:

Pin  Connection
---  ---------------
  5   Power: +5 Volts
  7   Select signal
  8   Power: Ground

So what about compatibility?  Under normal circumstances this system
maintains compatibility.  So what's the use of the "compatibility"
mode switch?  Obviously the controller could be fooled if some games
do strange things with the select line.  Some game code might not poll
and store the state of the controller, if game code accesses the
controller throughout its code, then unpredictable pulse trains could
be sent to the controller which put it in "pulse-3" mode.  And how
about games that aren't compatible no matter what?  My guess is that
that has to do with the speed of the custom chip.  Because the three
button version uses a discrete logic chip, the outputs of the
multiplexer can react to the state of select line in a matter of
nanoseconds.  However a little microprocessor or a slowly clocked
state machine made from custom logic would not be able to respond as
quickly, causing incompatibility problems if code was written to
assume instantaneous response on the select line.

==========
Conclusion
==========

So interfacing is not that difficult after all if you are interfacing
the controller to a microprocessor of some sort.  In my case it was a
68332 running Lisp.  I highly recommend the controller it packs a lot
of punch in a small price and the DB-9 interface makes it easily
replaceable if it goes bad.  Have fun interfacing.