Beggining PCB Design with Gallium Nitride GaN

Thread Starter

SiCEngineer

Joined May 22, 2019
318
Hi all,
Was just wondering about the feasability of me designing a PCB for use with GaN.
I am a PhD student, and at a stage where I now need to design a PCB board. My converter design has one half-bridge converter, and one fly-back converter (2 different supplies for 2 different load, w/ very different requirements - that's why I am keeping them seperate - but I digress). I am using the following demo board and its components to design my board https://www.ti.com/lit/ug/snou140a/...ps%3A%2F%2Fwww.ti.com%2Ftool%2FLMG3410-HB-EVM

However, as you can see, it is a half bridge board which would not be suitable for my flyback portion. So I am thinking of designing the board myself, using the schematic and layout from above as a general guide.

My question is, is this feasible for me to do, in say 2 months?
I know PCB design with GaN is very difficult, but I plan to use the design above as a guide.
How close could I get the performance of my board to this?
It looks very difficult and I was wondering whether it would be a suitable use of my time. But I don't see any alternative other than changing the board design completely. BTW the fly-back is a two-switch version for clamping the leakage. The half-bridge is a resonant LLC converter.

Let me know your thoughts on how feasible it is for me to re-design the board above. I've not much experience with PCB design at all, really, apart from a few project in my undergraduate.
 

Papabravo

Joined Feb 24, 2006
16,775
Hi all,
Was just wondering about the feasability of me designing a PCB for use with GaN.
I am a PhD student, and at a stage where I now need to design a PCB board. My converter design has one half-bridge converter, and one fly-back converter (2 different supplies for 2 different load, w/ very different requirements - that's why I am keeping them seperate - but I digress). I am using the following demo board and its components to design my board https://www.ti.com/lit/ug/snou140a/snou140a.pdf?ts=1605029763828&ref_url=https%3A%2F%2Fwww.ti.com%2Ftool%2FLMG3410-HB-EVM

However, as you can see, it is a half bridge board which would not be suitable for my flyback portion. So I am thinking of designing the board myself, using the schematic and layout from above as a general guide.

My question is, is this feasible for me to do, in say 2 months?
I know PCB design with GaN is very difficult, but I plan to use the design above as a guide.
How close could I get the performance of my board to this?
It looks very difficult and I was wondering whether it would be a suitable use of my time. But I don't see any alternative other than changing the board design completely. BTW the fly-back is a two-switch version for clamping the leakage. The half-bridge is a resonant LLC converter.

Let me know your thoughts on how feasible it is for me to re-design the board above. I've not much experience with PCB design at all, really, apart from a few project in my undergraduate.
If I were you I would prepare to throw at least the first one away. Regardless of any other consideration you're going to end up doing that. It is akin to writing a piece of software that compiles with no errors and works flawlessly without any need for debugging the first time. I've heard apocryphal stories of this happening, but I've never actually seen it.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
318
If I were you I would prepare to throw at least the first one away. Regardless of any other consideration you're going to end up doing that. It is akin to writing a piece of software that compiles with no errors and works flawlessly without any need for debugging the first time. I've heard apocryphal stories of this happening, but I've never actually seen it.
So it feasible to churn a design out, but expect that it won't work as I want and to have a re-design?
I am pretty sure the first design won't work, as you say. But the worry is then how long did it take me to design, and how long will it take me to re-design. I only have about 1 year left of my Phd to get this designed and tested.
 

Papabravo

Joined Feb 24, 2006
16,775
So it feasible to churn a design out, but expect that it won't work as I want and to have a re-design?
I am pretty sure the first design won't work, as you say. But the worry is then how long did it take me to design, and how long will it take me to re-design. I only have about 1 year left of my Phd to get this designed and tested.
I've known PhD candidates that were allowed 12+ years to get their degree.
 

ronsimpson

Joined Oct 7, 2019
1,456
I am using GaN. Keep the Gate drive close to the MOSFET. I read TI's .pdf and I think they are only running at 100khz which is slow for GaN.
I would look at several other companies .pdf and demo boards before making enything.
My first half bridge quasi resonant (maybe 25 years ago) ran very cool at 100khz. I have a hard time finding the right mix of Ls and Cs to get it to be happy. It needed a minimum load to be happy.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
318
I am using GaN. Keep the Gate drive close to the MOSFET. I read TI's .pdf and I think they are only running at 100khz which is slow for GaN.
I would look at several other companies .pdf and demo boards before making enything.
My first half bridge quasi resonant (maybe 25 years ago) ran very cool at 100khz. I have a hard time finding the right mix of Ls and Cs to get it to be happy. It needed a minimum load to be happy.
Hello Ron. You will find that TI are now offering an integrated gate driver solution with GaN in a QFN package. I am going to choose this to reduce the impact of poor gate driver design.. I can’t remember the name of the demo board exactly, but I’m sure if you google something along the lines of “LMG Texas Instruments integrates gate driver 600V demonstration board” it should flash up.
The only other company that is suitable for me would I believe be GaN Systems. As my input voltage is 243-297V. Take a look and see if you like.

Have you any advice in general, are you having trouble with GaN or do you find that if a tight loop of the power traces is present it performs well? In terms of time frame would you expect someone with not much pcb design experience to be able to get a first pass prototype within a month or two?
 

ronsimpson

Joined Oct 7, 2019
1,456
PCB should be fast if you can copy big pieces from demo boards.
My first board; I copied from a SiC board and the traces were too long for GaN.
This week I hope to get a new driver working. It has a pull up resistor I am setting to 10 ohms and a pull down resistor I am setting to 1 ohm. I am seeing Drain voltage rise time of 5 to 6nS on 400V supply. The D-G capacitance is pulling up on the Gate so I need to pull down hard. At this point I am driving a inductive load.
I like the resonant supplies where there are capacitors D-S across the MOSFETs. This limits the voltage rise time but allows the current to fall in 6nS.
You need a fast scope. I am having problems with 160mhz oscillation which is hard to see on some scopes.
 

Thread Starter

SiCEngineer

Joined May 22, 2019
318
PCB should be fast if you can copy big pieces from demo boards.
My first board; I copied from a SiC board and the traces were too long for GaN.
This week I hope to get a new driver working. It has a pull up resistor I am setting to 10 ohms and a pull down resistor I am setting to 1 ohm. I am seeing Drain voltage rise time of 5 to 6nS on 400V supply. The D-G capacitance is pulling up on the Gate so I need to pull down hard. At this point I am driving a inductive load.
I like the resonant supplies where there are capacitors D-S across the MOSFETs. This limits the voltage rise time but allows the current to fall in 6nS.
You need a fast scope. I am having problems with 160mhz oscillation which is hard to see on some scopes.
Do you know how to “copy” board from say a gerber file to a new pcb file? I was planning to just note the BOM down, crate the schematic exactly how it looks in the demonstration board PDF file instructions, and try to “by eye” place every component and trace where it should be. Maybe this is where I’m going wrong. Thats some interesting work. By pull up and pull down resistors are these the gate drive resistors? One for turn on and one for turn off, where the turn off one is usually smaller? I also like the drain source capacitance solution. I am attempting to drive at high frequency so this will probably have to be small. 6ns rise time however would be good enough for me I believe. Will you have a single gate drive signal with some logic gates for shoot through protection, or are you programming the gate drive signals separately? Are you using a 4 layer board? Do you have a recommendation for the oscilloscope bandwidth?
 
Top