***             NTmail Dialup Gateway Service            ***
*** Unregistered - please obtain a trial key or register ***

>   I'm interested in using an Analog Devices ADXL05 accelerometer
> in a project with a PIC 16F84.  The accelerometer outputs a
> voltage in the 0-5v range.  At its default and most sensitive
> setting, the accelerometer outputs 200mV per g.  (It measures
> -5 to 5 g's.)  Although I don't know for sure (yet), the A/D converter
> may have to be able to detect voltage changes as small as 2-4mV.
> (Is this a problem?  How sensitive are most A/D converters?)
>  Of course, on top of all this, I'm also looking for something relatively
> inexpensive.  :-)

>    I'm very new to microcontrollers and electronics in general, so
> I don't know much about A/D converters.  I know the theory behind
> what they do and how they work, but that's it.  I am trying to find
> an A/D converter to use for this project, with no success.  I was
> hoping that someone here might have used the ADXL05 with an A/D
> converter that they might recommend.

I am using the ADXL05 myself, along with a 16C73 based system.
I do not use the internal AtoD since 8 bits would be inadequate.
I use the Motorola MC145051, a 10 bit serial device with 11 analog inputs.
(There is a version with fewer inputs, can't remember the part
number at the moment.)  Cost is less than $5.

Even though the '73 has built in serial support, it won't work
with the Motorola part... so I did the bit twiddling 'by hand'.
(If anyone wants the code, just ask.)  The 16C76 has improved
serial support which _should_ work (more choices of clock phase
and polarity).  BTW, you need 4 IO pins to drive this chip.
Clock, data in, data out and either 'End of Conversion' input
(which I use) or a chip select to the chip (I leave it permanently
selected for board layout reasons).

A little more on the ADXL05.  It's output is noisy.  You will need
some serious filtering to get .01g resolution, or accept a low
bandwidth.  I use the standard low pass configuration with a
cutoff of 8Hz and get approximately .01g resolution.  The datasheets
and application notes show the circuit and component values required.
I also scale the output for 1V per g, and offset for 0g at 2.5V,
also using the standard circuit.

With a voltage reference of 5V, .01g is (.01/5)*1023, about 2 counts
on the 10 bit AtoD.  This does look like I'm living on the edge,
but as I said, the sensor output noise level is about the same.
There is no need to go with a 12 bit AtoD.

BTW, this chip doesn't like any inputs over about 5.3V.  It will cause
errors for any input converted while another input is over 5.3V.
This can make input protection circuits interesting if you want
to be able to accept 0-5V inputs.  This is not a problem for the
ADXL05's output.

HTH,

Orin.