Hello together,

this is a more generic topic regarding software design of placement of filter algorithms within the application architecture.

Assume the application samples a raw input, maybe directly the result of an ADC conversion.
This value is later interpreted as a physical value like a temperature or pressure and likely to be imprecise and has some jitter. Thus a filter is needed.

Which way would you choose?
a) filter the raw ADC value first, convert it later to a physical representation on the application layer
b) first convert to physical representation on the application layer and filter it afterwards

(Physical representation here means a fixedpoint representation such as 1/10 degrees celsius per bit)

I personally feel a bit undecided which one to choose. It probably depends on the actual hardware in use.

Are there any comments or general design guidelines on this topic?

Regards,
Matthias