I have an application where I need to emulate a conventional 5K linear 
pot.
The result should be indistinguishable from a mechanical pot when 
viewed through the eyes of an A2D converter which is controlled by 
embedded software which is designed (it seems) to make life hard. A 
microcontroller is used. Incremental cost should approach zero.

_____________

A controller that replaces a unit with a motor driven feedback pot 
interfaces to "a computer" that expects a pot. The computer provides 
3v supply and ground and expects to see the pot wiper. An electronic 
replacement for the pot should be a slow cost as possible while 
working well. eg $US1 is most undesirable. Adding an opamp would 
assist accuracy but add cost.

I have implemented a pot replacement using a PWM drive to a transistor 
buffer. This produces acceptable voltage but has essentially constant 
output resistance across its range, unlike a pot and is not explicitly 
ratiometric without software feedback.

The computer to pot excitation signal can and does sag and rise quite 
widely. As it is read by an ADC using ratiometric conversion (pot top 
supply = reference) these changes are relatively invisible to the 
computer. To make the voltage track the excitation sag I can read the 
excitation voltage and alter the output voltage to track. PWM 
filtering then needs to exclude PWM ripple but to be fast enough to 
track pot excitation variations. The software in the computer is fixed 
and unavailable and after having made an "endpoint-reached" decision 
it stops sampling until it decides another change is needed. Pot 
emulation must therefore track as fast and as well as a mechanical pot 
does to avoid the computer erroneously deciding the pot value is 
correct. (eg there is a delay in the PWM filter such that the endpoint 
is reached but the output is still settling. The target computer stops 
sampling but the output varies somewhat due to the delay.)

I could use a digital pot, but the cost is annoying. I could add an 
active filter to the PWM allowing crisper response - adds more cost.

I will definitely be able to make the result "good enough" using a 
transistor or two and some passives plus software feedback. But maybe 
someone else has done this sort of thing already and can comment 
accordingly. The killer is the need to meet the embedded software's 
expectations re pot wiper settling tine.




            Russell McMahon

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist