Oh no it isn't. Size matters!

 

It is not really the silicon size as a limit - although for switching speed that is the factor. Here is an Infenion 3600A single - http://www.infineon.com/cms/en/prod...?productType=db3a30443679872b0136ab191c2d1ec1

The naming of the IGBTs is done similarly to MOSFET - however this is typically at a chip temp of 25C ( older types) and often 80C today and this "name" as I prefer to call it is determined based on the number and size of chips used ( 3600A probably is 18 x 200A IGBT chips ) this is roughly a DC current IF you hold the chips at the rated temp, i.e. 80C - this can be difficult in the real world.... Vce @ 3600A ~ 1.8 V ... that is 6480 Watts of heat being generated IN the module - and that does not include the switching losses yet....

IGBTs have much higher switching losses than MOSFET - and the module design, reduces the thermal efficiency, but the finished product is typically better thermally so the design of the system is typically more important than the datasheet Rth of the device & but the thermal resistance is typically more consistent over time than discrete types - so there are trade offs - etc. For a typical application , AC output of an IGBT RMS current may be 1/2 to 1/5 of the module size (name) -

Basically IGBTs can carrty a lot of current - as long as you can cool them, and they have the ability to interrupt VERY high current (faults)- again - as long as their Maximum Junction temperature is not reached.

As far as making bigger ones - like I stated previously I am sure they have been built in the lab, but for any practical application the use of industrial modules and paralleling them is the way to go, because of the cost to develop anything bigger for the low quantity ( as in one piece for a lab ) - get to be prohibitive, and is typically much less reliable.

 

As far as making bigger ones - like I stated previously I am sure they have been built in the lab, but for any practical application the use of industrial modules and paralleling them is the way to go, because of the cost to develop anything bigger for the low quantity ( as in one piece for a lab ) - get to be prohibitive, and is typically much less reliable.

When you say paralleling, you mean paralleling the modules or paralleling the chips inside the modules? Or both? The things I have read seem to indicate that paralleling IGBTs becomes exponentially more complicated for each device in parallel above 2 in parallel. This is because they have a negative temperature coefficient and whichever one gets hotter faster will hog the bulk of the load and get even hotter even faster. Thermal runaway. I believe this is the reason for the modules; the chips inside are matched and balanced, thermally linked, to prevent thermal runaway. But I would assume that when you start putting the modules in parallel, you run into the same problem again. So it would make sense for the modules to be made in even larger sizes, into the tens of kilo-amps and above? But then as more chips are piled into closer proximity, the more problems you have removing the heat, so maybe that's why we can't find any >4700A? Maybe for very high power applications, a mondo MOSFET module would be better?

 

I once met a nun that used to be a priest.
I thought she was cross-dressing, but no.
She was a trans sister.

 

The silicon die is a limiting factor in thermal dissipation. Kind of a critical mass issue (surface area to volume ration for heat removal). The next big thing in high power is in the early stages today but growing quickly. Silicon Carbide (SiC), silicon Germanium (SiGe) and gallium nitride (GaN) compound semiconductors with high electron mobility. These allow higher current because of lower on-resistance.

An added benefit of each is also higher switching speeds with less heat. SiGe and GaN seem to be the next generation material for CPUs that could allow leaps in processor speeds to 60-100 GHz in the next 3-5 years.

GaN has been the material of choice for high-brightness LEDs for many years. Building process tools and substrates for use in ICs is newer. EPC corporation and international rectifier are just releasing power transistors based on GaN. RFMicrodevices have been making them for RF amplifiers for a number of years.

In the end, I don't have a photo of a big transistor (yet) but look forward to some soon. These would be used in inverters for solar farms and motor controllers in Diesel Electric vehicles and even eVehicles.

 

have a look at this
http://www.roadsideamerica.com/story/8180

Dieter

Wow! It is really, really big and it is water cooled as well.

 

Wow! It is really, really big and it is water cooled as well.

Unfortunately, it just looks like a transistor. It doesn't act like one.