Tradeoff between noise coupling from DC/DC to ADC and impedance of PDN

[gperoni]

I have what I consider to be a very silly question, but yet, I can't answer it myself.

I have some sensitive chips, like clock gen and ADCs. They are powered through multiple, properly filtered switching regulators. Here are my two tradeoffs:

1) Keep the switching regulators close to the sensitive chips. Like 1 cm / half an inch apart, separate from ground sticking vias and the shield made by the CNC-milled enclosure. Magnetic coupling between the two should be eliminated by the solid metal of the enclosure, and I get all of the advantages of having a super short power distribution network.

2) Meh the PDN, move the switching stuff far away from the sensitive circuitry. The CNC enclosure is still there, mostly for power dissipation reasons, but maybe it's not as critical, and maybe I can sell a cheaper version without it.

What would you do? Optimize for PDN inductance or for physical separation between switching regulators and ADCs? The main reason I'm doubting myself is the CNC enclosure, if it wasn't for that, I would go for maximum separation and add extra capacitors close to the sink.

[ejeffrey]

Option 2.  A short PDN isn't really needed.  Capacitors provide local bypass and ground/power planes or wide polygons provide a low impedance connection to the regulator output filter. 

The only real downsides to a longer PDN are higher DC resistance and the possibility of high frequency  parallel resonances involving the PDN inductance.  But you can have resonances from the capacitors regardless of the PDN length anyway as well and in either case they can be damped with RC snubbers if really necessary.

[David Hess]

The self inductance of a longer PDN solution helps filter out the high frequency switching spikes.

[donmr]

It used to be common to initially regulate to a slightly higher voltage and then use a linear LDO regulator near the sensitive chip for final filtering.

[David Hess]

It used to be common to initially regulate to a slightly higher voltage and then use a linear LDO regulator near the sensitive chip for final filtering.

That is required to reduce low frequency noise, but linear regulators are much less effective removing switching noise.

[jonpaul]

What is a PDN?

[brumbarchris]

Power Distribution Network