Hello All,

I made a PCB from 22 mm in diameter (circle) with 2 layers.
On this a frequency of 3 Mhz is made in a self oscillating configuration (and of course all the not wanted harmonics) with 2 FETS in half bridge configuration on 311 V (220V ac * 1,414)
When using the 2 layer PCB the FETS stay cool and temp raise is about 10 Degrees Celsius.
With this configuration I have about 10 dB to high level on signals to pass the tests

To reduce the EMI, I decided to order a 4 layer PCB with ground and VCC as two layers inside the PCB as suggested by many articles. Effect is that the FETS become very hot now (raise is more then 25 degrees).

I redesigned the PCB with 2 internal ground layers 5 mill from top and bottom and 47 mill between the 2 grounds to meet the 63 mill overall thickness.

I see many articles about impedance, length of traces and so on...it becomes too much for me to handle all

My question: 2 ground layers as proposed here will have the same effect as ground and power layer to lower the EMI?
I hope with only 2 ground layers that the FETS will not become so hot anymore, which is my goal

I also redesigned the PCB to add direct VCC and GND traces (on the other layers) to the parts needed and still have the 2 ground layers. Is that a good design? or do I face "bouncing" of signals.

Any advice would be welcome

 

Post schematic and board layout. Ground planes shouldn't affect heating that much, if at all. At 3MHz, you shouldn't have to worry about trace impedance; unless you're trying to match some impedance...

 

Ground planes shouldn't affect heating that much, if at all.

It most definitely will if all the dissipating surface area is now moved to an internal layer..

 

It most definitely will if all the dissipating surface area is now moved to an internal layer..

How can you be so certain? You don't know how the device was mounted. You have an insulator between the device (assuming it's using some copper on the top for heat dissipation). Ground/power planes are used for shielding, not power dissipation.

 

If the FETs were soldered to a large area of copper on the surface of the old board, they had a good heat sink from that. I've heard that when using PCB copper for heat sinking, it even helps significantly to have copper fill on the other side of the board. If you're putting any significant power through the FET(s), you may be forced to go with an actual heat sink, clipped or screwed onto the transistors.

If you've done any waveshaping on the signal to try to reduce harmonics, that might also be keeping the FET(s) in the linear region longer, resulting in higher power dissipation.

 

Yeah, we're in full guess mode because the OP didn't bother to post schematics and pictures. Just wants us to evaluate his design based on our overactive imaginations...

 

How can you be so certain? You don't know how the device was mounted. You have an insulator between the device (assuming it's using some copper on the top for heat dissipation). Ground/power planes are used for shielding, not power dissipation.

I was just making a general comment in case he did move a dissipating external to and internal.. Haven't seen the schem/board files as you already stated..
He could have had the FET tab tied to a nice external copper pour at ground potential and then eliminated/reduced it when he moved the ground internal.

 

If the FETs were soldered to a large area of copper on the surface of the old board, they had a good heat sink from that.

The two layer PCB had no extra copper for cooling beside only a very small heatsink on the FET's itself

If you've done any waveshaping on the signal to try to reduce harmonics, that might also be keeping the FET(s) in the linear region longer, resulting in higher power dissipation.

I did not changed anything to the design compare to the two layer. I think that indeed the two layers added capacitor values and the FETS stay longer in their linear region what course the high heating-up. With other words the gate signal is changed and rise time became less.

I don't think it is necessary to post PCB payout and schematics here, I don't want a discussion over the electronics or place of components on the PCB.
Intention to post it here is to get some more knowledge and maybe other technical guys did face the same experiences and found solutions.
Also become clear now that extra layers does not allways make things better and effects can be high see this one square inch board.