MOSFET DC SOA collection

[Martinn]

TLDR: For linear mode operation like in electronic loads or linear power supplies, you need MOSFETs with specified DC SOA. As these are hard to find, here's a list (or ideas where to look).

DC SOA has been discussed here many times, so I'll add only a short intro.
Please search eevblog for any combination of "DC SOA MOSFET linear mode electronic load dc power supply IXYS". You can restrict google search to eevblog by entering "DC SOA site:eevblog.com". Try also searching on diyaudio.com.
Example hit from beginner section: https://www.eevblog.com/forum/beginners/mosets-safe-operating-area-inconsistencies/

Most MOSFETs these days are used for switching, so either turned fully on or off. However, if we want to build an electronic load, we want to operate
the MOSFET in resistive mode, not fully turned on. This is commonly called "linear mode", in contrast to switching mode.
Unfortunately for us, optimizing modern MOSFET famlies for switching had the invonvenient side effect of making them mostly unsuitable for linear mode operation.


Example 1: IXYS linear FET IXTK200N10L2
It specifies 4 V/150 A or 100 V 6.5 A. Note that TC is specified as 75°C - some diagrams are given for 25°C, which is useless for a constant high power application.
Of course you will need to kepp the case at 75°C at 100 V*6.5 A = 650 W, which probably also is near impossible to achive.


Example 2: Infineon IPW65R040CM8
DC SOA for TC=80°C is given as 4 V/45 A or 650 V 4.5 mA. Had you initially been impressed by the beefy TO-247 case or the 329 W max. power dissipation, the 650 W*4.5 mA = 2.9 W might disappoint.


But why this list? Primarily because none of the online selection tools (parametric searches of Mouser, Digikey etc) support filtering for DC SOA. As most modern MOSFETs are not linear mode capable (with this I mean they publish a decent DC SOA envelope, not something like 5 mA@650 V for a 100 A MOSFET) so finding appropriate ones is sheer luck.
I come across various candidates randomly and instead of noting them for me, I can put them here. Feel free to contribute.

In general, most elderly models were doing OK in terms of DC, as they had large dies that were later shrunk in order to reduce cost. Highly popular are the IRF540 series models, although I have not found a datasheet with DC SOA listed.
Also LCSC https://www.lcsc.com/products/MOSFETs_381.html?globalKeyword=irf540&s_z=n_irf540 lists 15 different (mostly Asian) manufacturers, will they all stick to the ancient and robust 1980s design?

Try filtering DigiKey search to the HEXFET family: https://www.digikey.ch/en/products/filter/transistors/fets-mosfets/single-fets-mosfets/278?s=N4IgjCBcoGwJxVAYygMwIYBsDOBTANCAPZQDaIALGGABxwDsIAuoQA4AuUIAyuwE4BLAHYBzEAF9CAVnoAGRCBSQMOAsTIgAzPTkUELEBy69BoiYRiyYCpSryESkctrpx9bTpB79hYyeDg6GzQse3UnEHkmfwBaACZgyH4AVzVHcilmWIpEuzSNMDhCxmj-a2gQAQATLmp6HI9jHzNCdgBPVlwuKuwUcXEgA
(HEXFETs were introduced by IRF in 1978)

If you can afford them, IXYS linear mode FETs are great, particularly those with the monstrous SOT-227 package.
IXYS linear mode FETs: https://www.littelfuse.com/products/power-semiconductors-control-ics/mosfets-si-sic/n-channel-linear
Example:
https://www.littelfuse.com/products/power-semiconductors-control-ics/mosfets-si-sic/n-channel-linear/l2/ixtk200n10l2

Infineon also has a linear series:
Infineon Optimos Linear FET https://www.infineon.com/cms/en/product/power/mosfet/n-channel/optimos-and-strongirfet-latest-family-selection-guide/optimos-linear-fet-100v-150v-200v-/

TBC

- Martin

EDIT: I realize that's a very short list. But hey! It's a start!

[mtwieg]

Interestingly, Onsemi's SuperFET series (which are superjunction devices) specify DC SOA, for example FCA47N60F. These are among the very few high voltage (like >=500V) devices I've found which specify DC SOA, and the SOA values are among the best I've found (like 100x the IPW65R040CM8 referenced above). But some of them have gone obsolete in the past few years, not sure how long the others will be around.

The MDmesh M2 series from ST are newer devices which seem comparable to the SuperFET devices, but they only specify SOA down to 10ms, not DC....

[Martinn]

Interestingly, Onsemi's SuperFET series (which are superjunction devices) specify DC SOA, for example FCA47N60F. These are among the very few high voltage (like >=500V) devices I've found which specify DC SOA, and the SOA values are among the best I've found (like 100x the IPW65R040CM8 referenced above). But some of them have gone obsolete in the past few years, not sure how long the others will be around.

The MDmesh M2 series from ST are newer devices which seem comparable to the SuperFET devices, but they only specify SOA down to 10ms, not DC....

DK search:
https://www.digikey.ch/en/products/filter/transistors/fets-mosfets/single-fets-mosfets/278?s=N4IgjCBcoGwJxVAYygMwIYBsDOBTANCAPZQDaIALGGABxwDsIAuoQA4AuUIAyuwE4BLAHYBzEAF9CAVnoAGRCBSQMOAsTIgAzPTkUELEBy69BoiYRiyYCpSryESkctrpx9bTpB79hYyeDg6GzQse3UnEHkmfwBaACZgyH4AVzVHcilmWIpEuzSNMDhaTOj-a2gQAQATLmp6HI9jHzNCdgBPVlwuKuwUcXEgA

FCA47N60F (random pick from that list) has 600 V, 0.7 A, 25°C. Not bad!

Just searching if Superjunction technology has an impact on DC SOA and found this paper:
https://www.infineon.com/dgdl/Infineon-MOSFET_CoolMOS_7_linear_mode_at_high_voltage-ApplicationNotes-v02_00-EN.pdf?fileId=5546d46272e49d2a01730eef0c7529dd

[mtwieg]

Both the IPW65R040CM8 and FCA47N60F are referred to as being "Superjunction," but their SOA curves are different by an order of magnitude or more. Clearly there's something else going on besides the superjunction thing.

It should be noted that the SuperFET devices have SOA curves with a slopes of -1, which begs the question of how valid they are...

[Martinn]

It should be noted that the SuperFET devices have SOA curves with a slopes of -1, which begs the question of how valid they are...

https://www.onsemi.com/pdf/datasheet/fca47n60f-d.pdf
Seems to be the straight power limit, not accounting for any hotspot forming...

I recollect having read in this forum that DC SOA curves might be purely theoretical and untested. Particularly if the second kink in the curve is missing (like in the FCA47N60F one).
Also someone mentioned that the only way of reliably knowing DC SOA limits is to test them yourself. So I guess even when specified those limits (particularly straight line ones) are a guidance only.
At least noone would attempt to build a HV DC load with a IPW65R040CM8.

[PCB.Wiz]

This part is just listed from ROHM

https://www.rohm.com/products/mosfets/small-signal/single-nch/ry7p250bm-product

Has good hot-swap SOAR, main challenge for linear use would be practical cooling

[mtwieg]

https://www.onsemi.com/pdf/datasheet/fca47n60f-d.pdf
Seems to be the straight power limit, not accounting for any hotspot forming...

I recollect having read in this forum that DC SOA curves might be purely theoretical and untested. Particularly if the second kink in the curve is missing (like in the FCA47N60F one).

Yes, the DC contour is equivalent to a constant power contour of 417W, which is is the rated power dissipation at Tc=25C (which comes from the rated Rjc of 0.3 C/W). So the DC contour is certainly not based on actual testing.

For the other contours, one could compare them with the transient thermal impedance curves. For example, for a pulse width of 10ms Rjc is given at around 0.09 C/W. Based on that the maximum power would be 1389W. Looking at the SOA contour for 10ms shows a max power around 1200W, maybe that does account for hot spots somewhat...?

Typically I tend to give datasheet specs the benefit of the doubt (at least from reputable companies like Onsemi and IXYS). But this would definitely be a case of "trust, but verify" for me (just to be clear, I haven't done any testing on DC SOA for any of these devices).

[Martinn]

OK, some more.
The Art of Electronics X-Chapters https://x.artofelectronics.net/ has two tables:

page 187 Table 3x.3 High-Voltage Amplifier MOSFETs
page 254 Table 3x.6 Representative Power-Amplifier Transistors

with the interesting chapter 3x.5.2 on SOA.

If you don't know them - both AoE https://artofelectronics.net/ and AoE-X are absolutely fantastic and a must-have IMO.

- Martin

[Martinn]

found in the forum: https://www.eevblog.com/forum/projects/igbt-vs-mosfet-for-dummy-load/msg1838819/#msg1838819
IGBT FGH50T65SQD-F155 , 650 V, 50A Field Stop 4 Trench

Check other FGH* devices https://www.onsemi.com/products/discrete-power-modules/igbts#products=fnN0YXR1c352YWx1ZX4zfiF+TGFzdCBTaGlwbWVudHN+IX5MaWZldGltZX4hfk9ic29sZXRlfjI1MDI0NzR+cmFuZ2V+Mn40MH44MH50ZXh0fnRleHR+MX5mZ2h+

I find this interesting, as IGBTs are commonly referred to as having miserable SOA.

- Martin

[w42i]

The problem of localized heating of a MOSFET die until thermal runaway is also called "Spirito" effect, named after professor Paolo Spirito. See for example, the references in (1). This application note also discusses the derating topic. There are also related documents from Infineon (2) and Toshiba (3).

Personally, I don't trust any SOA/datasheet of a MOSFET that is not fabricated in the "old" lateral way and does not have the "second kink" in the diagram and does not explicitly mention the suitability for linear operation.
Semiconductor manufacturers didn't have the problem on the radar for a period of time after they had transitioned to newer MOSFET structures. It has been getting better in the last years, but there are a lot of datasheets out there that show only a total power dissipation limit (without the second kink due to the Spirito effect) in the SOA diagram and are just useless for linear applications.

I had this (that the SOA diagram was wrong) even confirmed by an FAE of the manufacturer for a specific part I needed to be absolutely sure.

(1) https://assets.nexperia.com/documents/application-note/AN50006.pdf
(2) https://www.infineon.com/dgdl/Infineon-ApplicationNote_Linear_Mode_Operation_Safe_Operation_Diagram_MOSFETs-ApplicationNotes-v01_00-EN.pdf?fileId=db3a30433e30e4bf013e3646e9381200
(3) https://toshiba.semicon-storage.com/info/application_note_en_20180726_AKX00073.pdf?did=59471

[MathWizard]

Oh boy a single FCA47N60F is $12US on Digikey.ca

I'm working out how an Array3710/11 E-load works. It can do upto 300W, 360V max or 30A max. Unless I wanted to test a mains rectifier or boost converter (which may not be a good idea anyways at those typical frequencies), I can't think of any reason to go to such high voltage. So I'm just playing around with derating some aspects.

They used the 8x IRFP460, and also some other low voltage ones, to use as another bank of shunt resistors. Some day soon I'll have to get suitable heatsinks for all this, and low drift, precision resistors. I already have some nice, real, op-amps for it. I'll have to order some small transformers for the voltage rails. For now I'd get it working without the digital controls too. Making a case tho, will be another hurdle.

[Circlotron]

Just remember of course, parallel mosfets don't always load share very well in linear mode. The hottest one will as a consequence have a lower threshold voltage and that will cause it to dig in further...

When they are fully switched on that is a different story. The hottest one will have a higher RDSon and that will help it to back off.

In the early 80s there were a number of audio power amplifiers using paralleled mosfets e.g. Hitachi 2SJ49 and 2SK134 from memory and the current sharing idea was trumpeted as a virtue but didn't work all that well in linear mode.

[Martinn]

IRFP250 in BK8600 electronic load
Maybe as a reference on how the pros do it, they use ten of those on a fan heatsink to dissipate 150 W, that's 15 W per transistor.
You can also see nicely that they use individual current control on each MOSFET for paralleling, essentially having ten 15 W electronic loads in parallel.
https://www.eevblog.com/forum/testgear/bk-precision-8600-electronic-load-teardown-photosreview/

[Martinn]

Another DC SOA thread with more part options: https://www.eevblog.com/forum/projects/empirical-mosfet-dc-soa-testing/