Sean RS Costall wrote:


>Just as a note to this: when the 'C74 was first released, we had
>preliminary data sheets that listed the reference voltage as 3.0V
>minimum.  Since this struck us as exceedingly stupid on a device that
>could run down to 2.5V, we called Mchip to find out what was up.
>
>We were told (at the time) that the preliminary specs were wrong, and
the
>reference voltage could be anything we wanted.  Fortunately for us we
>ended up using a 2.5V reference.
>
>Three years later we ended up trying to use a 1.2V reference.  Guess
what
>- it didn't work.  Mchip now says that the reference has to be at least
2.5V.
>
>Hope this clears up something for people who were similarily
misinformed.


I have used the ADC at references of 2.5V before, however based on the
message from Jason Bent I hooked up a variable reference to see what
happens. When I got down to Vref of about 1.65V the output of the ADC
started to act very non-linear. I expect this voltage will vary with
Vcc, temperature, and the batch. I always assumed the reason for the
higher reference was just to meet the linearity and absolute error
specs. It appears there are other issues also.

By the way the most recent (1997 dated) PIC16C7X data sheet still claims
a 3.0V minimum reference, which is not really possible running from a
regulated 3.0V (+/-5%)supply *even* if you use a Vcc reference (only
useful if you have ratiometric inputs).


Chip Weller.