If you use 1 BCD per A/D channel the tolerences are not a problem. If you
wanted to do two per channel then that would be pushing it and not
manufacturable. One client is using this technique without problem in a
consumer product at very high volumes.


At 12:40 PM 8/22/97 +0100, you wrote:
>>
>>this could be done with one i/o line and about $3.75 USD in components
>> OR three i/o lines and about $1.00 USD in components
>> OR seven i/o lines and about .10 USD in components
>> OR twelve i/o lines and no additional components
>
>
>Thanks for all the suggestions:
>
>I've seen the three pin A-D solution, with resistor networks, but it used
>1% tolerance resistors in various odd values - as someone's pointed out
>recently this is unhelpful froma manufacturing point of view.
>
>I hadn't thought of the shift register idea though. I suppose that actually
>a '554 or '61 (with 13 i/o pins) could read twelve pins and clock the data
>out of the thirteenth in some kind of slow async. serial protocol - and it
>can sit and do this 100% of the time. So only 1 pin is needed on the
>primary PIC. And it might as well do the BCD-binary conversion while it
>sits there.
>
>And while I sit here writing this, it's just occured to me that there's a
>remote-control IC (HS12E I think) that takes 12 binary inputs and clocks
>them out in an odd but decipherable PWM format. That might be the cheapest
>single pin solution.
>
>
Larry G. Nelson Sr.
L.Nelson@ieee.org
http://www.ultranet.com/~nr