David Harmon wrote:
> If you really want to build a programmer that can be used on a generic
> serial port with pins 2, 3, and 7, like say on a UNIX box, I think it has
> to have a baud rate generator, the functionality of a UART, and plenty of
> glue.  I would be delighted to learn that's not true.

David,

Although I could argue about the need for a baud rate generator (I use
a monostable to recover timing) I don't think that I can really refute
what you say.  If I wanted an RS232 hosted 8-bit bi-directional parallel
port I would use a UART without question.  I didn't; I wanted a simple
interface to bit-bang SPI/I2C using the serial port.  That's enough
functionality to program the 16C84 at least.  I admit the circuit I came
up with is not so very simple, but it is probably no more complex than
a UART based design.  Furthermore it has the advantages of smaller size
(it doesn't use a 40-pin chip), lower cost and can be powered directly
from the RS232 port in many cases.  I also admit it could be done
better by a micro-controller but that defeats my main objective.
I will even admit that the name Machine Independent Parallel Interface
is probably over the top for such a modest thing.

If you (or anybody else for that matter) is interested I can e-mail a
copy of my circuit and description and you would then be in a better
position to judge if the idea has any merit at all.

You quoted William Chops Westfield:
> > Creating a port capable of doing this in a general fashion, while
> > maintaining the ability to control timing of the bit-twiddling to a very
> > fine degree (which is probably why the parallel port was used in the first
> > place) is going to be a difficult task.
and wrote:
> Yep.  I would expect that things like the timing of a program pulse would
> have to be done with a one-shot or something.

One of the things I really like about the PIC16C84 is the fact that the
programming operation is self-timed (though you have to make sure you
don't ask it to program another location until 10ms or so has elapsed).
So I see a 16C84 as being the heart of my universal PIC programmer.  The
downside is that the 16C84 doesn't have enough I/O to program the 16C5X
PICs without the addition of some glue.  I guess that I could get the
complexity of the final programmer down to a bare minimum by a two-step
boostrapping process: use MIPI to program 'C84; use 'C84 to program 'C64
(say); use 'C64 in final design.  I think this is taking things a bit
too far.

As I said in one of my (too frequent) recent posts, I don't see members of
the PICLIST as being the primary audience for this project, so perhaps
I should tout my wares elsewhere.  Thanks to you and everyone else who
has contributed to the discussion.

David
--
david.tait@man.ac.uk