Why did SMPS stop working? Chassis ground..

[cupoftea]

Hi,
We were doing a 300W push pull. (24vin 32vout, 300khz)
It had an all metal enclosure.
We had to wire in snubbers across the fets and diodes, as there were none.
The PCB was mounted in an all metal enclosure.
There were mounting holes with chassis copper net on them, so they connected to chassis with screws.

When the chassis net, and circuit_ground_net were soldered together (so that chassis and circuit ground were electrically connected), then the SMPS worked (it was mounted in the metal chassis)
But When circuit_ground_net and chassis were not connected , then the SMPS would not work when it was in the metal chassis. (it would work when out of the metal chassis)
To this point, there was no common mode filtering on the PCB.
So we then added Y caps and common mode chokes to input and output...(y caps from circuit ground to chassis)
After the common mode filtering was added, then the SMPS worked when it was in the metal enclosure...even though chassis and circuit_ground_net were not then directly soldered together.

What is the mechanism of the circuit not working here? (ie, when it was not working as explained above)

 

[Easy peasy]

If you can supply an A1 sized xtal ball with complete accurate wiring diagrams for the all the cases mentioned - including the case wiring connections and heat-sinking to the case of the various semis - then an answer could be surmised

without this one needs the more magical type of xtal ball

 

[cupoftea]

Thanks, I do have the xtal ball on order from Farnell......though the "magical" type was nil-stocked.........AYK, no semis were heatsunk directly to case.....all were bottom PCB mounted and gap padded to metal case.

Basically, when circuit_gnd_net and chassis were shorted, then AYK, that is like a massive y cap from chassis to circuit_gnd_net.

I mean, i had always known having circuit_gnd_net floating from metal chassis could cause EMC problems, but never saw it actually stop an SMPS from working.

The explanation, (i think?), is that when metal chassis and circuit_gnd_net are not connected (not even with y caps), then noise that escapes from the circuit cannot get back into the circuit, so it radiates like mad and even radiates so much it gets into the control of the SMPS.
Somehow, connecting circuit_gnd_net and chassis, stops the metal chassis from radiating so much.

 

[SWRmeter]

Hi,
We were doing a 300W push pull. (24vin 32vout, 300khz)
It had an all metal enclosure.
We had to wire in snubbers across the fets and diodes, as there were none.
The PCB was mounted in an all metal enclosure.
There were mounting holes with chassis copper net on them, so they connected to chassis with screws.

When the chassis net, and circuit_ground_net were soldered together (so that chassis and circuit ground were electrically connected), then the SMPS worked (it was mounted in the metal chassis)
But When circuit_ground_net and chassis were not connected , then the SMPS would not work when it was in the metal chassis. (it would work when out of the metal chassis)
To this point, there was no common mode filtering on the PCB.
So we then added Y caps and common mode chokes to input and output...(y caps from circuit ground to chassis)
After the common mode filtering was added, then the SMPS worked when it was in the metal enclosure...even though chassis and circuit_ground_net were not then directly soldered together.

What is the mechanism of the circuit not working here? (ie, when it was not working as explained above)

It seems that your connecting chassis & circuit directly, or indirectly through chokes, reduced the ground impedance between the two, and hence reduced some IR drop somewhere in your circuit. Your colleagues might be right.... it's a ground loop artifact !!

 

[cupoftea]

Thanks, and of course, connecting circuit ground and chassis ground at more than one place, can cause a ground loop...is this why its bad to connect cct ground to chassis ground in more than one place?

 

[FvM]

I think the discussion is missing the problem. If a PCB is designed with multiple metalized mounting holes, they are obviously intended to be connected to a metal chassis. That's e.g. the case with many RF modules or also PC mainboards. If chassis connection causes problems with your PCB, there must be a design fault.

--- Updated Feb 13, 2023 ---

If I should guess, it's probably not a simple ground loop problem.

But instead of guessing, the problem is asking for real measurements and debugging.

 

[cupoftea]

Thanks, i think its a general "rule of thumb" that (with PCBs in metal chassis) if you dont have cct ground connected to chassis ground via either y caps or hard shorting, then the noise situation in the PCB will be worse?

Question is...why?...whats the mechanism at play?