Electronics > Projects, Designs, and Technical Stuff
GaN transistor in buck converter turning itslef back on?(problem)
(1/1)
relu:
Hello dear EEVblog forum members,
I have a problem with a buck converter.
It is designed using the GS61008 GaN driven by the LMG1205 pre-driver.
The problem that appears is that during turn-off of the HS transistor, the transistor turns itslef back on.
Please see the attached pictures.
The converter, and the measurement points:
Layout (GaN devices are on the other side of the board)
Scope measurement, where you can see the turn-on:
Phisical board
Do any of you have any experience with this?
Would appreciate any help with this.
Thank you and best regards,
Relu
julian1:
Since noone else is commenting..
What is the purpose of the 27p cap. Is it a misplaced boostrap capacitor? It looks like it would just AC couple switching transients onto the HS Fet gate. Also the ringing/ transients look large, does it need a snubber to dampen?
relu:
Thanks julian1 for the question. I have used the 27pF capacitor to try to slow-down the transistor turn-on/off. (togheter with the turn-on / turn-off resistors).
Regarding the ringing - you are right I have to place a snubber to dampen these transients, but I had this problem with the turn-on first.
In the meantime I found the issue with this - it is in my sub-optimal layout.
I was used to consider a GND plane to have the same potential all over, but for the speeds the GaN transistors are switching, this is not the case anymore.
I have placed a capacitor right on top of the swiching cell, and the turn-on effect dissapeared. see the images attached.
uer166:
Looks like some serious speed.. So turns out adding more decoupling capacitance at the input fixed it, although in your first traces the Vgs voltage was below threshold, so it shouldn't have turned on the FET anyway. I guess the probing wasn't 100% correct then and the "real" Vgs was larger than shown in the trace.
T3sl4co1l:
This is two layer PCB? Seems dangerously optimistic...
With 4 layers, you can make the switching loop vertically oriented, i.e. within the PCB layers, so the current path is very wide and shallow. Here, it's a relatively large and thin (flat) loop, going along the transistors, then running beside them through neighboring ground plane, up to the bypass cap. (Note that there's no direct ground path under or over the switching node, only beside it.) The field is only able to peek through the gaps between copper polys, but that's still enough to total a few nH.
Tim
Navigation
[0] Message Index
Go to full version