Monday, 1 April 2013

6809 LED output test

Now that I can reliably write data into the EEPROM, the next step is to build the simplest possible working 6809 computer, in order to test my knowledge of how a system built from a 6809 should function.

After looking at how Ben Ryvers built up his Z80 based computer, I decided to borrow some of his ideas for a trivial test.   The idea is that test data - a sequence of bytes indicating which LEDs should be on and off, held in ROM - will be outputted.  Once the sequence has ended it is restarted.  As simple as can be.

My test-bed 6809 computer consists of:
  • 6809 clocked by a 4Mhz crystal
  • The 28C256 EEPROM
  • 8 bit latch which drives a 10 LED bar graph
  • Glue logic
For this test, the computer has no RAM at all.  This, of course, makes writing even a trivial test program slightly challenging, but luckily the 6809 has enough registers that temporary storage is not a problem.

The memory map for this setup is completely trivial.  When started, the 6809 reads a 16bit startup vector from memory location $fffe.  Therefore ROM has to live at the end of the memory space.  In this case, 16KByte of ROM will be available at $c000 to $ffff.  The other half of the 32KByte EEPROM isn't mapped.

It is helpful to consider the memory map in terms of the high nibble, in binary:
  • 0xxx - RAM (in the future)
  • 10xx - IO
  • 11xx - EEPROM
Each "type" of device needs to be selected from the upper 2 address lines, which are fed into a 1 in 4 decoder.  Since there isn't any RAM yet, only two decoder output lines are needed.  A typical decoder, which I've used, is the 74139, which has a pair of 1 in 4 decoders.  One generally useful aspect of the 74139 is the outputs are inverted, which is nice because memory and similar devices nearly always have an active low Chip Select (or equivalent).

To avoid the risk of the 6809 modifying the content of the EEPROM, the R/W line on the EEPROM is tied high.   The /CS line for the EEPROM is derived from the address decoding, described above, whilst the /OE line is form by the negation of the R/W signal on the 6809 (alternatively it could be held low, but this will not be useful when the EEPROM is programmed in-place).

In this simple case, "IO" consists of a single 8 bit latch - I chose a 74HC574 for this purpose.  Attached to the outputs of the latch are 10 LEDs in a bar graph (the end 2 LEDs are unused).  Since the latch is clocked on the rising edge, it is necessary to invert the IO enable signal from the address decoder.

Two push buttons, one for /RESET and one for /HALT, and a 4MHz crystal with caps complete the circuit:


You can download a postscript version.

The code, in good old fashioned assembly - such as it is, is on github.  Note the $4000 delay loop.  This sets the speed of the pattern.  This is equates to a delay of about a half of a second with the 1Mhz clock.  This program is my first crack at assembly programming since the mid to late 1980s, on the ZX Spectrum!

To turn the assembly file into a ROM image, I used a09 (available here), along with hex2bin to form the 16Kbyte ROM image, which of course is written out to the EEPROM using the upload tool previously written.

I did make a video of the "computer" in action.  Unfortunately it seems I'm unable to upload videos to this blog.  You will have to make do with a picture instead:


As you can see, the breadboard is a mess - extending the busses is going to be difficult.  Power is provided by the same IDC header used by the EEPROM programmer - only the +5V and 0V lines are used.

Amazingly the circuit worked first time.

The next step is to decide what the next step is. ;-) I can either add some RAM and play with some more advanced testing towards making the computer "useable", or I could try to get an old UART working.

I suspect I will try my hand at the RAM, since a serial port is, after all, only a "peripheral"....

Update...  Finally figured out that the way to get videos on my blog is via youtube.  So here is that video:

2 comments:

  1. Very interested in 6809 married with TI. Looks like a great combo. What's best way to contact you for retro questions?

    ReplyDelete
    Replies
    1. Hey Carlos,

      I prefer good old fashioned email. aslak3 at gmail dot com. 6809 is a great proc.

      Delete