Replacing IGBT with MOSFET - what to compare in datasheets?

[okw]

I need a large quantity of IGBTs for repairing motor controls. They're quite expensive, compared to MOSFETs with same specifications, so I want to try MOSFETs.
There are six IGBTs in the circuit, four of them are controlled by two gate driver ICs. So I guess two for each rotation direction, and two for breaking(?).
I guess MOSFETs are generally faster, so switching speed wouldn't be a problem.

What do I look for when I choose a new MOSFET? I understand every application has differences, but on a general basis.
The obvious, N-channel and same package.
And:
VDS MOSFET >= the IGBT (600V)
ID MOSFET >= the IGBT (4A)
VGS = VGE (+/- 20V.... should it be equal or can it be lower?)

Anything else?

[MariuszD]

Calculate the power losses in the IGBT and select the MOSFET's Rds(on) so that they are no greater. You'll find that a comparable MOSFET is more expensive.

An IGBT with a maximum current of 4A can operate at 4A in a real application.  The MOSFET's maximum current is a marketing gimmick, it doesn't correspond to reasonably chosen currents in the target application.

For example, the STGD4H60DF 4A 600V IGBT will have a 1.5V voltage drop and 6W static losses at 4A.

The "4A" 600V STP4NK60Z MOSFET will have a voltage drop of approximately 16V and static losses of 64W at 4A. This is because the cold Rds(on) is 2 Ohms, at high temperature it is even twice as large.

[wraper]

You need to consider gate charge/capacitance too. Chances are gate drive circuit is too weak to drive the vast majority of MOSFETs with RDS(on) power loss similar to VCE(sat).

The MOSFET's maximum current is a marketing gimmick

Yeah, it's a spec that's barely worth looking at because you don't have an infinite heatsink in real devices.

[ejeffrey]

The "4A" 600V STP4NK60Z MOSFET will have a voltage drop of approximately 16V and static losses of 64W at 4A. This is because the cold Rds(on) is 2 Ohms, at high temperature it is even twice as large.

8V and 32 watts, but the point is still correct.  And as you say, higher given the unavoidable temperature rise. 

[Salitronic]

Yeah, MOSFETs only start to make sense when the bus voltage is low, once you’re up around 400–600 V, IGBTs generally win. The Rds(on) of a 600 V MOSFET is just too high to compete with the ~1.5 V drop of an IGBT at the same current. You’ll end up with way higher conduction losses unless you pick some massive (and expensive) part.

Also keep in mind your gate driver. It was designed for IGBTs, probably swinging ±15 V or so. Many MOSFETs aren’t rated for that, and the driver might not have enough grunt to charge/discharge the much larger gate capacitance of a low-Rds(on) part.

[viper]

No clue on actual application other than I have decades in industrial automation and if OEM is IGBT, that is for good reason.  With that said, if older gear, MOSFETs have improved but you already know you need margins. 

As I recall, one of the perks of IGBT was the failure modes.  trying not to take the driven motor with them.  If working fanuc, there are usually aftermarket FET packs that are reasonable. 

[jbb]

It sounds like you’ve got a H Bridge motor driver (or maybe a three-phase driver, to which the following still applies). In this case, the MOSFET body diode reverse recovery can be a major problem.

Basically, it can generate huge current spikes during switching events and blow up devices.

I would suggest that you stick with the IGBTs if possible.

[xvr]

For high-voltage MOSFETs, SiC MOSFETs can be used—they have high Vds and low Rds. However, they may be more expensive than IGBTs.

[okw]

Thanks for all the advice. So I'll go for IGBTs.

What properties should I try to match as close as possible, and which are OK to go over (and by how much), or go under?
(Except the obvious: Collector-Emitter Breakdown Voltage (Vces) / Current - Collector(Ic) / Pd - Power Dissipation.)
Also, my IBGT is a 4A part, but many of the options I found are 10-30A. So the required saturation values could probably be less when replacing a 4A part?

For my part (also attached datasheet), these are a selection of values (maybe others are needed too?):

Property	min	typ	max
Vce Saturation(VCE(sat))	1.7V	2.0V	2.4V
Gate-Emitter Threshold Voltage (Vge(th)@Ic)	3V	4V	5V
Gate Charge(Qg)	x	24nC@15V	31nC@15V
Input Capacitance(Cies)	x	264pF	317pF
Td(on)	x	22ns	26ns
Td(off)	x	237ns	284ns
Turn-On Energy (Eon)	x	0.070mJ	0.081mJ
Switching Energy(Eoff)	x	0.061mJ	0.079mJ
Reverse Transfer Capacitance (Cres)	x	17pF	20pF

Also, the symbol in the datasheet doesn't show a body diode, but the text says:
"Energy losses include “tail” and diode reverse recovery."
Can I assume it has a body diode? Since many of my part options show a diode in the IGBT symbol in the datasheet.

[ahbushnell]

Yes body diode can be a problem.  Back in my youth we ran into that.

SiC mosfets don't have that problem. 

[okw]

Yes body diode can be a problem.  Back in my youth we ran into that.

SiC mosfets don't have that problem.

But the datasheet for the original IGBT says:
"Energy losses include “tail” and diode reverse recovery."
Can I assume it has a body diode? Even though it's not drawn in the symbol?

[ahbushnell]

Body diode is for the MOSFET.

[okw]

I'm not getting any wiser about the diode. Most of you say IGBT doesn't have a body diode, however, more often than not, I see IGBTs shows a diode in the symbol.
See attachments.

However, the original part in the design does not show a diode in the symbol, but datasheet mentions diode several times:
Graph: E(on) and E(ts) include losses due to diode recovery.
Table: Energy losses include “tail” and diode reverse recovery.

Will this say the original IGBT has a diode?

[harerod]

Check out fig.E in the d05n60 datasheet. Shows the "body" diode and explains how switching losses are determined.
Edit: put "body" in quotes, thanks to phoenix for nitpicking

[Phoenix]

I'm not getting any wiser about the diode. Most of you say IGBT doesn't have a body diode, however, more often than not, I see IGBTs shows a diode in the symbol.

A single discrete MOSFET has a parasitic/accidental diode already in the silicon structure. This is called the "body diode" and will generally have very poor switching characteristics as it can not be optimised.

An IGBT (NPN style construction) does not have this parasitic diode. When a component shows a diode, the package has 2 silicon dies in it, one IGBT and one diode, this can be called co-packaged. This diode is called an "anti-parallel diode". It can be optimised for switching and will perform far better than the MOSFET body diode.