is Chip Select a trap for low noise systems?

[SArepairman]

I have a DAC, ADC and some digipots on my analog board. I thought it would be a good idea to save on isolators (everything is isolated) and use chip select. However, now I am thinking if i should give a separate network for the active SPI stuff then the static/calibration stuff.

How much isolation can I expect from a chip select? Is it a crap shoot based on the part? Are there rules of thumb?

[T3sl4co1l]

What are you isolating?

[SArepairman]

What are you isolating?

The MCU.

[jeremy]

I'd be interested in the answer to this too.

Although my strength is not low noise analog, I would have thought that most of your gains would be through just isolating the bus from the micro (is it galvanic? transformer? optical?) but I don't even know if this is necessary. Presumably, if they are mixed signal parts they are already designed for low noise applications? So you just need to be careful that you use a satisfactory power supply topology. I mean, my DSO is full of digital and analog stuff, and they're even on the same board with no isolation. I've only ever seen serious isolation when a) one side of a connector can't be trusted or b) the instrument is supposed to connect to a person

edit: also c) when one side of the board has a stupendous voltage on it

[David Hess]

Slowing the edge rate of the digital signals down at their sources with an RC filter will help a lot.

[Bud]

I worked with an 24-bit ADC and tested bus connection via RC-filters and direct and found no detectable difference in noise floor, even with other devices operating on that shared bus during ADC conversion period.

[David Hess]

I worked with an 24-bit ADC and tested bus connection via RC-filters and direct and found no detectable difference in noise floor, even with other devices operating on that shared bus during ADC conversion period.

Many converters deliberately arrange to start conversions while the bus is likely to be quite but even without that, integrating converters have high inherent noise immunity.  Digital noise is a much larger problem with sampling converters and in single ended designs where the analog ground gets pulled around.