Hacker News Reader

Comment by grogenaut | Hacker News Reader

Are the power switching components similar to MOSFETs not a relatively advanced and recent device?

Peoe usually think of advanced as Nvidia chips, but there are other parts.

MOSFETs (both Si/SiC and more recent GaN) switching transistors may be made on special (Epi or Sapphire) wafers, but the Fabs are very old school (e.g. 2-6in vs 12in for CMOS) and the process geometries are practically neolithic (e.g. 90s). Processing equipment is more limited by the lack of desire to build or support something so old. Die sizes are also quite reasonable (e.g. 4kW/cm2) However, there is still enormous capacity for processing them around the world (both in US/EU, and around Asia).

Perhaps, for the new GaN on Sapphire wafers there could be a significant manufacturing limit? Mostly, those are used for Blue LEDs today.

Are GaN2 and 3 more exotic, r are they just newer iterations of GaN?

Those appear to marketing names applied to DC Chargers. I suspect that the GaN2&3 refer to higher frequency operation. This may be due to lower parasitic capacitances, and would allow smaller components/inductors to be used at comparable efficiencies and lower cost. Certainly, the wafers and materials seem to be the same.

High power MOSFETs are definitely advanced devices... but not produced on the same process as say a CPU or GPU chip.

Perhaps, but my guess is they don't need to operate at high frequencies, which lowers the bar for complexity.

the slower your power MOSFET, the more power you waste and heat you generate.

I am not an professional EE, but a person who professionally works on circuits from time to time.

That's absolutely not the case. The higher is the switching frequency, they higher are the switching losses in a MOSFET.

MOSFET is an electrostatic device, the charge needs to both go in, and out from the gate. The more switching, the more current goes through the gate back-and-forth.

You're talking about different things. A slow FET has a slow rise time. It spends more time crossing through the inefficient region between fully off and fully on. Switching frequency is how often you cross in one direction.

Up to a point, you can decrease total converter loss by trading conduction loss for switching loss by increasing frequency and resizing components. After that point, you start losing efficiency again. There's a sweet spot.

Yes, I see your nickname. That's the case if we go into details.

I think he specifically means high-end, high-frequency power FETs, since all the rage today is about them.

I would add that there is no real need for super-high-end switches in EVs, and component size for them is not a huge premium. There is no shortage of smart inverter topologies which achieve very good size, and efficiency without using any kind of fancy GaN of SiC switches. They will not be as small as them on the lower end, but around 100kw there is virtually no difference.

It's not a detail. It's a more complete picture. Both of your comments are different slices of the same pie.

I don't see anything written above that pertains specifically to any type of FET.

Your last paragraph is a new topic. It's true that Si can be as efficient as SiC and GaN, despite what the marketing usually says. But there are other benefits to new materials. They already support higher voltages with planar technology, and 3D will increase that further. Gate driving will get easier and more rugged.

> Your last paragraph is a new topic. It's true that Si can be as efficient as SiC and GaN, despite what the marketing usually says. But there are other benefits to new materials. They already support higher voltages with planar technology, and 3D will increase that further. Gate driving will get easier and more rugged.

There are 1000V+ silicon switches today which are not as fast, but have for, for example, very good RDSon. 1000V I think is an upper limit one would go for passenger car EV motors.

True, but even GaN will eventually exceed that, and there are other applications that will benefit.

thanks for addressing things with far more patience than i would have.

This is true, but if you don't switch very often, then it doesn't matter as much :)

The do need high frequencies. It's orders of magnitude slower than digital, but the amount of power being switched presents entirely new problems. You could flip your statement around and say that digital circuits are less complex because they're low power. But that would be equally meaningless.

Digital FETs are CMOS made with fins or whatever to switch as fast as possible. Analog FETs are something else to make clean predictable signals from mature manufacturing processes. Power FETs are big arrays of trenches to not destroy themselves. You can't really compare them. That would be like comparing a racecar to a passenger van to a rock hauler. They're all vehicles, but they have different purposes and success metrics.

Reply on news.ycombinator.com