Mike, actually I'm using a 12-Bit A/D with 4096 steps. I should have
worded things differently. I did prototype a version with a lot of analog
`glue' that provided an output of 0V at 28 InHg and 4V at 32 InHg. I would
still need at least a 10-Bit A/D and it required a few trim pots. I decided
to go with a 12-Bit A/D with 1mV/step and do the rest in software.

   - Tom

At 08:03 AM 6/9/99 +0100, Michael Rigby-Jones wrote:
>> Tom Handley wrote:
>> [snip]
>> > In my design, I wanted a range from 28 to 32 InHg and a resolution of
>> > 0.01 InHg. This required an A/D that could resolve 400 steps and would
>> > require a typical 10-Bit A/D but I already had an 8-channel, 12-Bit A/D
>> > (MAX186), in my design with a resolution of 1mV/step. I selected the
>> Sensym
>> > SCX15ANC transducer....
>>
>        Wagner replied:
>
>> ... if the full scale is 32 InHg and the resolution is 0.01 InHg, it
>> will demand 3,200 x 0.01 InHg to go to full scale, so 400 steps is not
>> enough, 4096 is, so the 12 bit a/d.  By my experience, all the
>> calculations, scale corrections and temperature linearity math requires
>> a better a/d resolution, to give you the required slack to deal with
>> math fractions and rounding.
>>
>        <snip>
>
>But he wants to start at 28 InHg i.e. the sensor/signal conditioning is
>arranged so that 28 InHg gives an input of zero volts to the ADC.  (32-28) /
>0.01 = 400.  Just under 9 bits are needed for the resolution Tom requires,
>obviously not a commonly available device :o)
>
>Regards
>
>Mike Rigby-Jones