1

I have built an electronic load. At the time I did not understand the SOA curves. Now I find the IRFZ44 datasheet does not have the DC curve. Any idea what the max current might be at 30V and 40V so as not to cause damage? Otherwise, I may have to change it all to a 2N3055 (and make a mess).

The electronic load does work @ (badly smoothed) 40V DC 3A. (I have tested it only for few seconds at this power). However, I have tested it a lot at 15-25V and 1A. Seems to hold up well. I have a large heatsink with a fan on it. Case temp does not go over 40C.

(The IRFZ44N and IRF540 seem to be the only power mosfets readily available locally, so any pointers about the DC capability of either will do.)

Added on 2018 APR 13: Well, the mosfet died. So to answer my own question... The IRFZ44 will happily work at slightly more than 20V (maybe 23-24) at close to 2A (I tested numerous times at 1850mA). I tested today at 42.5V, and it died shortly after reaching 400mA (maybe upto 500mA, I'm not sure). It was not even warm, maybe only at 35C or so. So that's that. In hindsight, I should have added a couple of resisters upstream of the drain to lower the voltage across the mosfet. It should probably tolerate a lot if the Vds is limited to only about 10V.

This also means the SOA curves in the datasheet are probably not all correct. Extrapolating the points where the mosfet worked and did not work (the imaginary DC operation line) does not give me a line parallel to the other lines at difference pulse widths. So, they too may in reality be curving downwards much more at higher voltages (i.e. a curve instead of a straight line in the datasheet).

So there... my 30 cents worth of IRFZ44N...

Added on 2018 APR 15: Another mosfet died. This one failed with only 2.5-3.5V at about 2.4-2.5A. So that probably explains the missing dc line in the SOA curve. This mosfet will fail at above 2A at all voltages below 20V. The dc curve is a flat line at about 2A till 20V, and falls sharply after that to 400mA at 40V.

Indraneel
  • 2,117
  • 1
  • 17
  • 24
  • 1
    I see a DC characteristic curve for the IRFZ44N so I don't know why the internet should be preventing you from doing that. – Andy aka Mar 13 '18 at 17:11
  • Which figure? I couldn't find anything at 0Hz linear operation. – Indraneel Mar 13 '18 at 17:38
  • 1
    25V*1A = 25W. 40V*3A = 120W. Big difference. And the absolute maximum power rating is based on a die temperature of 175ºC. Expecting reliability at any more than 30W (even with excellent cooling) is unrealistic. – Bruce Abbott Mar 13 '18 at 19:52

3 Answers3

5

The IRFZ44N is a hexFET designed for switching applications so using it in linear applications runs a risk of destroying it and you being left scratching your head as to why it went pop. It might not even be very warm at all. For linear applications you should consider using MOSFETs designed for avoiding thermal runaway. Yes, a MOSFET will go into thermal runaway if the gate-source voltage is below the zero-temperature-coefficient threshold.

This isn't some theoretical annoyance that doesn't really happen. I can vouch for seeing it on a design I was asked to look at. Take a look at some IXYS MOSFETs that are designed to handle "linear" applications.

On semi document on thermal instability

Infineon document on thermal instability

Ditto from Fairchild

Nasa document explaining it

Which figure? I couldn't find anything at 0Hz linear operation.

Figure 1 and figure 2 describe the DC operations despite this measurement being made with a 20 us pulse: -

enter image description here

It's made with a pulse to prevent self-heating and the remotest possibility of thermal runaway at lower gate voltages; note how a gate voltage of 4.5 volts (25 degC) produces a current of about 7 amps with 1 volt across drain to source; then note that as the silicon die warms (rapidly of course) the current increases to about 14 amps at 175 degC (figure 2).

This is the thermal runaway I refer to.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • 1
    This is why Active load designs use current feedback – Tony Stewart EE75 Mar 13 '18 at 19:01
  • @TonyStewart.EEsince'75 ampage feedback won’t protect a mosfet if it’s cold and you apply a high gate voltage to get enough current flowing. Within about a milli second, a hot spot can develop in the hexfet and a localized temperature of over 600 centigrade results and blows a hole in it before the feedback loop can respond. – Andy aka Mar 13 '18 at 19:37
  • I was thinking even if shorted and 100ns delays 100kHz BW or even 10kHz loop BW is not too hard with V to I control to prevent that with 0.3J capacity on this device.. – Tony Stewart EE75 Mar 13 '18 at 19:48
  • The trouble is that it’s the centre 10 percent of the die that gets warm so the control loop closes down to keep running at the set point but all that current is flowing through the hot bit and that hot bit is about to boil off. – Andy aka Mar 13 '18 at 20:16
  • Thanks @Andyaka Now I know about Paolo Spirito and the recent Spirito Effects known in the last 10 yrs. and that sufficiently aggressive thermal clamping of the case can help prevent it with feedback faster than thermal time constant – Tony Stewart EE75 Mar 13 '18 at 20:16
  • OK, now I understand. So the 20us test pulse basically says "do not use linearly"? But I guess this will happen for every mosfet. I found this (old?) mosfet http://www.mouser.com/ds/2/149/FQP50N06-1009503.pdf and it has a DC curve in the SOA, thermal response curve with 5s pulse, and 250us test pulses. Anyway, it seems I should have an electronic fuse (with another mosfet) to protect my DUT, in case the power mosfet blows. I guess if a thermal runaway starts to happen, the mosfet will short and the gate will stop working, so any feedback loop on that mosfet will fail. – Indraneel Mar 13 '18 at 21:03
  • Still reading the very interesting NASA paper you linked to above! But it seems every mosfet if used in an electronic load may have this problem, unless it has been tested for DC. So maybe I really should switch to a 2N3055. At least it will work at 40V and about 2.75A (that SOA was also scary). – Indraneel Mar 13 '18 at 21:08
  • IXYS make ones specific for the task. If you want to recognise one, it’ll tell you on page 1 of the data sheet. Look for words like “extended SOA” OR “designed for apps such as eFuses or hot swap circuits”. – Andy aka Mar 13 '18 at 21:27
  • The mosfet datasheet I linked to above says it is a planar mosfet.... and now I know what planar mosfet means, and who Paolo Spirito is! The mosfet is also available on ebay @ 5/$. Planar mosfets seem to be still made for audio amplifiers, among other things. So maybe the datasheet is real and I can trust it? – Indraneel Mar 13 '18 at 21:40
  • No, this one looks bad. Look at figure 2 and think what happens when you apply a voltage on the gate at 25 degC to get 1 amp of drain current. Then imagine it warming up and taking 7 amps. This is the sort of thing to look for plus it says it’s designed for switching applications so it could fail. – Andy aka Mar 13 '18 at 21:52
  • 5.5A if I zoom in :) The IXYS ones look good and pricey, and likely difficult to find here. All temp curves show at least 1s pulse. But shouldn't being planar mitigate some problems for the FQP50N06? Doesn't that mean there are less mosfets on the die, so should be more stable and distribute heat better? [I'd hate to have to use a BJT for this] – Indraneel Mar 13 '18 at 22:31
  • I think you have all the facts. I can’t pat you on the back and say it will be alright. – Andy aka Mar 13 '18 at 23:32
  • As a quick test, I put a 500mA load and ran it for about 30 min with no fan. It warmed up a bit. The Vgs dropped from 2.70V to 2.66V. So a 40mV difference. Wonder what it means. At 1A load, Vgs was at 3.3V (while it was still warm). This was all at Vds=9V. But you're right, I have all the facts now. I'll accept your answer for the helpful links. – Indraneel Mar 13 '18 at 23:40
  • The mosfet died today with <500mA at 42V. – Indraneel Apr 13 '18 at 15:26
  • @Indraneel RIP dear MOSFET LOL. Hope you are not scratching your head about why. – Andy aka Apr 13 '18 at 15:34
  • Not any more! I already added comments to the original question at the top. I think the datasheet should also be interpreted with a pinch of salt. – Indraneel Apr 13 '18 at 16:31
  • @Indraneel quite a few data sheets are like this but the manufacturer's say on page 1 - it's a switching device! That is there poor get-out clause. – Andy aka Apr 13 '18 at 16:46
  • And oh, I forgot to mention.. I still don't have the electronic fuse I talked about in the comments above. So, after the mosfet blew, 2.5A started flowing through it and the SMPS voltage dropped to 6V. I would have expected the SMPS to have shut down. It was from an old EPSON printer and I was trying to find its max current. Luckily it seems to have survived. I'm quite glad nothing exploded. – Indraneel Apr 13 '18 at 18:56
  • How about an IGBT? They seem to come with the DC curve, if the SOA is shown. And about same price as a couple of 2N3055 (although I don't have any IGBT lying around). – Indraneel Apr 17 '18 at 08:09
  • So what mosfet to use? – piotr Aug 14 '21 at 23:55
  • @piotr I don't understand what your application is. My recommendation to Indraneel was to look at what IXYS offered. If that doesn't suit then, I recommend you ask a new question on this site that reveals what problems you are having and what circuits you have tried. – Andy aka Aug 15 '21 at 08:54
1

It's on page 1 of spec

Continuous Drain Current (Absolute max)

ID=50A @25'C
. = 36A @125'C, Vgs at 10 V ,

For a linear electronic load, the critical factor is the Rth thermal resistance from case to ambient \$\Delta T = (Vin-Vout)\cdot I\cdot d\cdot R_{th} \$ 'C above ambient, for some pulse wuth duty cycle d ( 0~1)

Use a flat greased heatsink Rjc=0.5 and a CPU cooler < Rca=0.5
Rth=Rjc+Rca ( Rca case to ambient with a fan is as low or lower than chip case to junction)

So for 50W you can have 1A at a 50V drop or 50A with a 1V drop.

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • The continuous current is when the mosfet is fully on. In the linear region, like in an electronic load where Vgs around 4V, the SOA curves are the guide. However, IRFZ44 has no curve for DC in the SOA curves. – Indraneel Mar 13 '18 at 17:44
  • DC is the SOA curve > 1 second. VGS must be greater than 4V to lower Ron. Usually 3x Vgs(th) – Tony Stewart EE75 Mar 13 '18 at 17:53
  • Can't find SOA curve > 1 second. Can you please link the datasheet? [In an electronic load, Vgs has to be low to control the current, so mosfet is never fully on, so linear operation] – Indraneel Mar 13 '18 at 18:27
  • Did you not understand my answer with Temp rise? that is the steady state DC ( from Ohms' law for P=VI) and spec Vgs = Vds , Id = 250µA ONLY so Vgs >>Vgs(th).. sorry forget about SOA – Tony Stewart EE75 Mar 13 '18 at 18:41
0

Damage could be caused by overheat in your case (DC load).
You have to work with power dissipation figures. It is huge difference between 3A @ 40V and 3A @ 5V.

Chupacabras
  • 5,394
  • 2
  • 23
  • 50