Dear Piclisters,

I have had my tries now ... and I'm fed up ;)
This is what I am trying to do:

I have to connect a pic16c84 to a pc.
"Connect" means that I have to send a 16Bit word (or two 8Bit
words) to the pc in irregular intervals (about 1 sec between the
events). As I already have a 14Bit A/D converter in my pc and my
"events" are rather rare pic-wise speaking I thought I'd convert
my two 8Bit words into two voltages which would then be re-digitized
in the pc.
The other possible options (just to mention it) I have ruled out,
rather: I have already two serial ports in action and wouldn't like
to have another one (using win95 ...). And I DO have a dig i/O card
but haven't dared to give it a try yet ;) (more problems like length
of cable expected).

So I got an LTC1454 (thanks to Linear Technology) which offers two
output voltages of 12Bit resolution. I just used the upper 8Bit of
each channel resp. to improve signal/noise ratio AND STILL I GET
DIGITIZING ERRORS !!! That is, after digitization I have values for
a given digital output of the pic that differ in the range of more
than 16, that is already more than the 4Bit I have forsaken to
avoid this trouble ... worst of all: It's not constantly there.
It seems to be better after a few hours of operation ...

My hardware layout is more than simple which is probably the problem.
I used the "typical application" LT offer in their data sheet
(can't find any traces of this chip on the webpage right now, which
makes me a bit nervous ;). I'll put the datasheet on my webpage for a
while:  http://www.mayn.de/users/hyla/14544li.pdf (ca.230KB).

What I did:
- connected the Din, /CS/LD, CLK to pic i/o pins directly
- connected the dacs REFout to both channels REFhi directly
- connected REFlo and X1/X2 for both channels to Gnd directly
- didn't forget the bypass cap Vcc - Gnd (used a tantal though !?)
- Dout not connected
- Vcc's and Gnd wired to a cheap power supply ;) (is that it?)
- and finally connected the two voltage outputs to my a/d card
  (again directly, what could I do). Result: see the above ...

Yes, I can measure a mysterious 1MHz signal wherever I like ;)

Piclist comments ( mailto:PICLIST@MITVMA.MIT.EDU ) and personal
e-mail ( mailto:hyla@mayn.de ) very warmly welcomed ;)

Sorry this got so lengthy ...

Christoph Klein

p.s.: and YES, I HAVE read Fr. Tom McGahee' elaboration on serial
registers with interest but I still can't part from my original
design ;)