GaN Mosfet --- Arlo?
- Thread starter athlon
- Start date
liveforphysics
100 TW
They turn on and off with RF level speeds. This sounds like a nice idea, but makes them a bit of a challenge to control in practice.
as far I know the driver part is the same of regular mosfet , also the load sharing is the same of the mosfet
For sure I know that 3 Formula E team already use this kind of mosfet but form online store looks like now is possible also for "normal" people to use it
I'm not an expert electronic but maybe can be possible to ask Marcos to adapt the VESC and see how it works
For sure I know that 3 Formula E team already use this kind of mosfet but form online store looks like now is possible also for "normal" people to use it
I'm not an expert electronic but maybe can be possible to ask Marcos to adapt the VESC and see how it works
liveforphysics
100 TW
Formula E teams are running SiC not GaN (I was just in Montreal formula E and had lots of controller related discussions with team crewchiefs) . No argument GaN will replace mosfets and SiCfets soon enough.
If you swapped them onto a VESC today it would be good for about 20-50amps for a few hundred bucks in GaNfets alone, and that assumes it didn't just explode from from switch time rise rates inducing noise and control issues (which is as far as the friends of mine got with them.)
If you swapped them onto a VESC today it would be good for about 20-50amps for a few hundred bucks in GaNfets alone, and that assumes it didn't just explode from from switch time rise rates inducing noise and control issues (which is as far as the friends of mine got with them.)
marcos
1 kW
- Joined
- Nov 19, 2016
- Messages
- 348
Actually... I'm working hard with GaN, controlling them with a similar microcontroller that vesc uses. You would only need a different gate driver and A LOT less deadtime.
For motor drives I'm using 1us deadtime. With GaN its 15nsec, so imagine how critical the layout is and how difficult is to parallel those devices. I lost an unbelivable amount of time telling the fab what I needed.
Also, the gate driver needs to be special, the gate input cant exceed 5.5v.
Its a ton of work and money, requires a scope I still can't afford.
On the other hand, you can get 99+% efficiency which means much easier thermal requirements, ie smaller heatsinks. I'm still struggling with the instruments calibration because the damn thing will measure >100% efficiency with any environment change.
You can also switch at MHz and use much smaller components, but you dont care about that in a motor control application.
Of course you can ask me to implement that, and I'll gently provide a quote
For motor drives I'm using 1us deadtime. With GaN its 15nsec, so imagine how critical the layout is and how difficult is to parallel those devices. I lost an unbelivable amount of time telling the fab what I needed.
Also, the gate driver needs to be special, the gate input cant exceed 5.5v.
Its a ton of work and money, requires a scope I still can't afford.
On the other hand, you can get 99+% efficiency which means much easier thermal requirements, ie smaller heatsinks. I'm still struggling with the instruments calibration because the damn thing will measure >100% efficiency with any environment change.
You can also switch at MHz and use much smaller components, but you dont care about that in a motor control application.
Of course you can ask me to implement that, and I'll gently provide a quote
KalSteve
1 W
Pulling a Kayne here...
I just gotta say, thanks Athlon, Luke and Marcos. These discussions tickle my brain. I am not clever enough to understand most of it but the language is at a level that I can learn a bit. Awesome stuff.
And once again, thanks Justin for keeping this forum going so this bleeding edge info is made available to the masses. It will be interesting to one day look back and see how much this sharing of information drove the electric transport revolution.
I just gotta say, thanks Athlon, Luke and Marcos. These discussions tickle my brain. I am not clever enough to understand most of it but the language is at a level that I can learn a bit. Awesome stuff.
And once again, thanks Justin for keeping this forum going so this bleeding edge info is made available to the masses. It will be interesting to one day look back and see how much this sharing of information drove the electric transport revolution.
athlon said:marcos said:Of course you can ask me to implement that, and I'll gently provide a quote
OK , let's do it ! send PM to me
don't want to miss the opportunity to have the first GaN e-bike
something like the VESC-X bu with with this Mosfet http://eu.mouser.com/ds/2/692/GS66516B%20DS%20Rev%20170321-1107590.pdf , Low current , high voltage , very efficient.
Also it will work good on my solar car
zombiess
10 MW
liveforphysics said:
I went to an IEEE presentation on High Electron Mobility Transistors (GaN) devices a while back. After the presentation I was speaking with one of the people that does the material sciences for them and he was telling me how they become difficult to work with due to layout inductance. The latest and greatest transistors are capable of very high switch freq (way into GHz range on leading edge) which makes the layout critical. He showed a picture of 7/8 dies configured on a laminated bus and said they can't go higher due to diminishing returns from unequal current sharing + layout inductance. I asked him if they matched Vgs(th) and he was a bit surprised by my question, but said that they did and it played a big part in paralleling them. The other issue is they don't have any high current / high voltage capabilities yet, but just 10yrs ago GaN devices were mostly lab experiments since spec's weren't good for most real world applications, so we've already come a long way. I don't fully understand their physics (something about setting up a quantum well which constrains things in one dimension forming a plane), but they are using properties of quantum mechanics. Pretty amazing tech.
marcos
1 kW
- Joined
- Nov 19, 2016
- Messages
- 348
^^ this.zombiess said:
And this -> http://epc-co.com/epc/Portals/0/epc/documents/papers/eGaN%20FET%20Drivers%20and%20Layout%20Considerations.pdf page 5
Those are a bunch of 0.15mm vias inside the pads... plus a lot of small but relevant design considerations. It really pushes the pcb fab limits and it takes some time to wrap your brain around its requirements.
Amazing devices.
Similar threads
