Shielding the switching node from sensitive signals

[cupoftea]

Hi,
We are doing 4-layer 60W BuckBoost with Vin = 10-30V to vout = 5v to 20V
Fsw = 420kHz.
The enclosure is very tight, we had to have the power inductor on the opposite side
of the board to some sensitive digital signals which affect the SMPS operation.
Also, the inductor overlaps some of the SMPS control chip.
However, there was no choice in this due to space constraint.
We do have on layer 2, just under the inductor, a pretty solid ground plane which we hope will shield the switching node from the sensitive signals. However, the SMPS is working
intermittently, sometimes works, sometimes doesnt.
We cant adjust the FET drain rise/fall time as all 4 fets and their drives, are inside the chip. Only the inductor and an output current sense resistor is outside the chip.

Do you think a thin layer of internal plane copper is enough to shield the switching node
from sensitive signals? The SMPS controller is especially sensitive as it contains
a charger controller micro aswell.

Pretty much the entire power stage and input/output caps are on a 4-layer board which is some 20mm by 20mm.

The digital signals which go into the chip set its vout. Part of the noise operation we are seeing is that the vout of
the chip changes from regulated 20v (say), to regulated 5V....as if noise has somehow got into the digital
signal lines. This often happens as we manually raise reduce the load resistance (increase i(out).

 

[Easy peasy]

There is no information supplied on de-coupling of chips, or smaller caps to gnd on current sensing pins - or other sensitive pins,

no information on whether the inductor is a shielded type or not - no pictures, no schematics, no information about any ( small ) snubbers fitted ( or not fitted )

can the entire chip be shielded ?

 

[KlausST]

Hi,

impossible to answer without seeing the PCB layout.

In my opinion .. I´d rather move other - non critical parts - to the other side.

Klaus

 

[std_match]

Some small capacitors to ground from the digital signals?
Maybe combined with a series resistor in the driving end to reduce the current spikes.

 

[hiphei]

You have no possibility to make a 'power plane' from quieter of the supplies in addition to gnd plane in between the sensitive layers? That just might give some extra dB of shielding..And have You checked and double checked on how the switch currents return..and that you have no ground loops..? Every conductor (on pcb too) has some L and C..also the ground plane..so sometimes a gap on gnd plane can *guide* the noisy switch return currents away from sensitive parts.. ..also emi in *common mode* may creep along input and output wiring and couple back..maybe try some filtering there as well.. Appears it's not an easy task you're facing..