3 ms·
Just because I was doing my own hobbyist research on power switching, and encountered the IGBT vs power MOSFET tradeoff: The current path for an IGBT looks lar
by aidenn0 7y ago
Just because I was doing my own hobbyist research on power switching, and encountered the IGBT vs power MOSFET tradeoff:
The current path for an IGBT looks largely like a diode; it has a fixed voltage drop (0.6V in silicon), so the losses in the "On" state are linear with current. A MOSFET, however looks largely like a resistor in the "On" state, so the losses are quadratic with current.
There are also switching losses, which are a bit more complicated, but largely scale quadratically with frequency (for IGBTs there usually aren't actual formulae but just a curve from the datasheet; MOSFETs are a bit more amenable to calculation). For relatively low-frequency and high-current switching like a high-power permanent magnet motor, you will tend try to keep switching losses low relative to conduction losses; in any event IGBTs are low-frequency devices so if IGBTs are in discussion typically switching losses will be low.
As anyone who has studied asymptotic performance will know, for low-frequency applications, as current goes towards infinity, IGBTs are better. At low currents, however, MOSFETs trounce IGBTs; A resistor that drops 0.6V at 1A would have 600mOhms of resistance, and a $0.10 Power MOSFET can beat that in low voltage (<120V) applications, with on resistances on the order of 1mOhm. At 100A there are similar conductive losses.
As you design power MOSFETs for higher voltage handling though, the on resistance goes up. There are ways to design power MOSFETs with lower on-resistance, but they will make them much harder to switch and increase the switching losses to the point where that bit I said earlier about conduction losses dominating becomes not true. For switching 900V you will want a MOSFET rated for at least 1200V and now you are looking at 100s of mOhm of on resistance, so even at say 2A, the IGBT is looking quite good.
Enter SiC. I know absolutely nothing about semiconductor design, but the way it was explained to me is that the design parameters of an SiC MOSFET are such that it can block a larger voltage than a silicon MOSFET, so that the envelope shifts "up" for working voltage at any given amount of on-resistance ant gate-charge (which relates to switching losses).
Lastly, if you aren't a racecar driver, most of the time the power draw will be far below the max, so even if an IGBT beats an SiC MOSFET handily under hard acceleration (it almost certainly does), for highway cruising there is a good chance the MOSFET will win.
- gens 7y agoFrom what i remember when i cared about such things: It's the gate capacitance (and gates resistance) that leads to switching losses. Usually the biggest part of the switching losses is because the transistor is not fully "on" or "off" for the whole time, but is in between when in the process of switching. The gate capacitance is probably one of the biggest problems (more for FET, but also for bipolar), that influences the speed of the transistor (presented as frequency or as slew rate). Otherwise you could just put a bunch of MOSFET-s in parallel and get much lower resistance, thus greater efficiency (and it works, to a point). Another thing to consider is that FET-s let the current go back and a motor is an inductive load, while bipolar transistors do not. I forgot how much that matters, but it probably does plenty. Thankfully electronics is one of those fields of engineering where you can get all the information you need from a datasheet. Unlike from those damned headphone manufacturers (embarrassments to engineering, 99.9% of them). PS "..you will tend try to keep switching losses low relative to conduction losses.." I think you meant the other way around.
- Kirby64 7y agoAlthough FETs have built-in body diodes (to let current free-wheel back), power IGBT packages usually include built-in/added free-wheeling diodes. From a power handling perspective, power IGBTs and FETs act similar just with their own challenges.
- aidenn0 7y ago> PS "..you will tend try to keep switching losses low relative to conduction losses.." I think you meant the other way around. A 6 pole motor running at 10000 RPM has a switching frequency of 1kHz. Any application where you are considering an IGBT is probably under 20kHz. Cruising on the freeway uses about 15kW. Conduction losses are probably going to dominate.
- gens 7y agoHence you will try to keep the conduction losses low relative to switching losses. Yea, 1kHz is nothing for modern transistors. FET, IGBT or bipolar. Not that it's not worth considering them, ofc.