comparing SiC and GaN for a specific application

nazardo

Member
2020-01-17 9:23 pm
Hello
I would like to develop a class D with a switching frequency around 400 or 500kHz, a bus voltage of 200V (or dual 100V), with a pulsed max current of about 10A (no continuous high current needed)

Initially I was inspired by TIDA-01605 which uses UCC21530 gate driver, and I selected the SiC mosfet C3M0065090D.
Advantage of this design is that there are transformers which allow for a virtually rail-to-rail configuration, and the design is already done so it is more easy to avoid mistakes.

However, lately I came across LMG3411R070 which are integrated modules with a GaN mosfet inside.
In theory those modules should be more efficient, I am not sure which of the two design is best. For sure the second has the advantage of using much less space.

Now, the idea could be to apply the concept of the TIDA-01605 design to these modules, but I am not sure how easy it is and I am not fully confident on the comparison that I did which for the moment is in favor of the second design.


What is your opinion?
 
The LMG3411R070 is only a single GaNFET - you want a half-H-bridge as a module, ie 2 GaNFETs and a single high voltage driver that does the bootstrapping and dead-time for you. Any risk of shoot-through will toast the FETs, its best to have that tightly under control, and GaNFETs are super fast, beyond most cheap 'scopes.


I've got some of these 80V GaNFET epc9203 modules: https://epc-co.com/epc/Portals/0/epc/documents/guides/EPC9201_qsg.pd which is a neat half-H-bridge in a surface mount module, EPC are one of the companies coming up with many GaNFET devices of all ratings.
 

nazardo

Member
2020-01-17 9:23 pm
The LMG3411R070 is only a single GaNFET - you want a half-H-bridge as a module, ie 2 GaNFETs and a single high voltage driver that does the bootstrapping and dead-time for you. Any risk of shoot-through will toast the FETs, its best to have that tightly under control, and GaNFETs are super fast, beyond most cheap 'scopes.


I've got some of these 80V GaNFET epc9203 modules: https://epc-co.com/epc/Portals/0/epc/documents/guides/EPC9201_qsg.pd which is a neat half-H-bridge in a surface mount module, EPC are one of the companies coming up with many GaNFET devices of all ratings.
Well, the epc9203 is interesting indeed. However, the voltage is too low for my application.

I know about the risk you are talking about, in fact the driver you mentioned uses a NAND and an AND to avoid simultaneous conduction. What is not clear to me is how to handle the dead time: how do you adjust them?



And, I would prefer not to use the bootstrapping diode, but some transformer or similar trick (there are others methods too) because of the rail to rail performance and also for other reasons (I have to switch this thing on and off many times so I need to avoid the click at the startup, this is why I need to keep the high side driver always ready).

My thought is that I could do the same, applying a similar circuit to the LMG3411R07, using the transformers for the high side power, but I have not very clear in mind how to handle the dead times.
I have also seen that EPC has other options, but they are not detailed. I will drop them an email
 

Leonelf

Member
2020-01-29 10:41 am
As you can see in the attached image (taken from the quickstart guide of the devboard), the outputs of the gates are coupled with the gatedriver through RC-element. When turning on the output, the RC-element works as a turn-on delay, when turning off the output, the resistor is short-circuited by the diode and discharges the capacitor immediately.
So the deadtime is the time the capacitor needs to charge to the gate-driver's input voltage threshold (for a rising edge, since it uses schmitt-trigger inputs).
 

Attachments

  • SumatraPDF_1SpQtM78hi.png
    SumatraPDF_1SpQtM78hi.png
    36.6 KB · Views: 102
For a bus voltage of 200V a 600V GaN is overkill, besides LMG3411R070 datasheet interestingly omits any information related to current ON/OFF A/ns slopes.

A normal class D PCB does not support more than 1000A/ns, regardless of type of semiconductors. This is 30ns for switching 30A. Do I care about a 1ns or 5ns crossover time? Do I even care about body diode reverse recovery, when 200V MOSFET is so low in Rds-on shunting body diode?

We are talking about 300%~500% price increase for 20% loss reduction, as a rough estimation.