for three devices: FQP33N10, RFP50N06 and IRFP4321PbF.
The third device is odd, a nice big TO247 power package, but lousy low frequency to DC SOA. Compared to other MOSFETs, this one has an extra, steep line added to the trace set.
What's the explanation?
Worth trying to see if these really can do better than 1A at 20V?
Why limited to 20W DC when they claim it's a 310W max power device?
150V at 78A (limited to 75A by package, hmm, not as bad as some IR datasheets).
That slopey part of the DC curve is about a constant 120 watts. Hmmm,
120 is nowhere 310, is it?
Lots of fets have a DC curve that slopes down more, in the sense that max allowed DC power falls as voltage increases. That's why switchmode type fets can blow up at fairly low power when used in linear mode.
The axis label are annoying. What's wrong with 1 10 100 ?
Ha, second breakdown. But MOSFETs aren't supposed to exhibit that phenomenon!
Vgs(th) has a negative tempco, which certainly should be capable of producing the effect, at least if transconductance and power density are high enough. Is that one difference between old school vertical DMOS and typical switching FETs, higher gain?
Tim
-- Deep Friar: a very philosophical monk. Website:
Ah Well: High voltage MOSFETS at low currents *can* exhibit secondary breakdown (as you say).
Some switching MOSFETS are not entirely MOSFETS, they are hybrids; They contain extra circuitry to improve switching. The characteristics may be "cranky" due to the gate drive enhancement circuitry. Large ISOTOP devices might be improved in this way.
This "feature" is often not explicitly documented (which is why my hair is now white from trying to get an analogue HV-amplifier to work with some of those hybrids).
IXYS makes FETS desighned for linear operation an explanation of there L2 FETS is in the above pdf. They arent giving them away cheapest is a T0220 $6 FROM DIGIKEY. But you would probably want a to-247 or TO-3P package and they start at $12.
Oh, OK... missed the other two graphs. That 3rd graph probably illustrates my point about voltage. The IRFP4321 can dissipate 300 watts at 3 volts, but only 20 watts at 20 volts. There's some equivalent to bipolar 2nd breakdown at higher voltages. This is an extreme case... looks like 5 watts at 50 volts!
I avoid IR fets in general.
We tested a bunch of 300 to 500 watt fets to destruction to find some that would survive 300 watts at high voltage for 100 milliseconds.
The maximum power is the maximum power at "absolute zero" that the package can dissipate. It is not the expected power dissipation from the package in standard operating conditions as that depends on too many factors for a simple number to describe.
It has to do with thermal resistanc calculations and is a sort of ideal power dissipation(all additional thermal resistances are 0).
Conceptually if you think of heat radiating from the junction to the outside package it has to move through the device packaging. That packaging has resistance which "slows down" the transport of heat energy from the junction. If the termal resistance is very large the junction temperature will rise to a very large number and of course will burn up.
If you add additional thermal resistances(heat sink, air, etc...) then it only makes matters worse as the heat energy has more barriers to cross. By giving a number that is independent of all those extra cases we get a pure number that depends only on the device package. That number is pretty consistent across all devices manufactured the same because of quality control. Now when we want to go calculate the total dissipation in our case we can do so because the thermal resistances will add as there will not be any co-dependence between the different thermal cases.
Another way to think about it is that in the ideal case, that is complete power absorbtion and removal from the device package, then the device package can dissipate it's maximum power dissipation withthe junction temperature being it's maximum.
Best to show an example I suppose.
2N7000 -
Maximum power dissipation = 0.4W @ STP Thermal resistance - 312.5 C/W
(Note there is no junction to case resistance because one generally uses these without any other thermal resistances. This should make sense as they realized no thermal calculations would probably be needed since the devices would almost all be used effectly the same. If you put this is liquid nitrogen then you might get more power out)
312.5*0.4 = 125C.
This tells us that if we run the device at 0.4W we will get a junction temperature of 125C, possibly it's maximum which will probably cause it to burn up. The data sheet suggests that the maximum junction temperature is
150C so it probably will be fine to run at 0.4W.
The total thermal resistance is R_JC + R_CA = R_JA = 312.5C/W.
Hence what they are telling us here is that the maximum power dissipation is pretty much what it says it to be. This is why many electrical enginners get confused about power dissipation numbers because they treat them all the same. In this case it so happens it is the same because they intended it to be the same. That is, the maximum power dissipation is the maximum operating power dissipation. If the ambient temperature changed then one could calculate the new maximum operating power dissipation.
Now for a high power device one generally adds in thermal resistances that effect the *maximum operating power dissipation*.
Take the IRFZ40 TO-220
Maximum continuous power dissipation @ STP = 140W Maximum junction temperature = 175C
R_JC = 1C/W
Hence R_JC*140W = 140C
BUT THIS IS THE IDEAL CASE. This assumes that the heat dissipating on the surface of the case is completely and instantaneously absorbed into the ambient. If they could get the device into an absolute zero atmosphere then they could run the device at 140W and it would produce a junction temperature of 140C.
IN REALITY, we have additional resistances involved. If we do not use a heat sink then
R_JA = 62.5C/W
and if we ran this at 140W then the unction temperature would be (62.5 +
1)*140W = 8890C!!!!! To make sure the device survives we can only run it at
140C/(62.5 + 1) = 2.2W.
Why is it much less than the idea situation? Because the air cannot draw out the heat from the case quick enough so the junction temperature will increase faster. 2.2W will produce a junction temperature of 140C if the package is in ambient. Probably ambient with no convetion. If you used water then you would get a different thermal resistance and could run it at a different power level.
Suppose you needed to run it at 5W, then
140C = 5*(1C/W + X) ==> X = 27
This means you'll need to find something with a thermal resistance of 27 to get that 5W power dissipation.
As you hopefully realize now, the maximum power dissipation is that which brings the junction temperature up to it's maximum. It depends on all the thermal resistances inbetween. By taking the ideal case, that is R_JA = 0, they can get R_JC. This means you can do computations by simply adding the thermal resistances.
In the 2N7000 they didn't need to an ideal case since the device was meant to be used only in ambient air. For power devices one is meant to use a heat sink and that adds an extra dimension. This requires a calculation and without R_JC one cannot do it.
Ask yourself, what is the "maximum power dissipation of a device"? Can you answer that? The answer is no. To many factors involved. Unfortunately in the devices that are ment to run ambient they generally mean "maximum operating power at STP an in ambient" and in power devices they mean "maximum power dissipation at absolute zero". Hence the confusion when they are using the same term with two different definitions.
In fact I don't even know why they need to give the maximum power dissipation for powe devices at all since you never will use it in thermal calculations. Of course They must give you either the ideal maximum power dissipation or the maximum junction temperature and they will be related by T = R_JC*P.
What you can do is know that if two devices or two different packages have two different power ratings that the largest one will let you run it "hotter". The difference between a 100W BJT and a 200W BJT effectively means that the 200W BJT's package has 1/2 the thermal resistance than the 100W BJT. This doesn't mean too much as your total thermal resistance will be much larger than that of the package. That is, in your calculations you'll have stuff like "small number + larger number" ~= large number. The small number is R_JC. If you start using very efficient heat sink's and other advanced cooling methods then it might start mattering.
If you look at most power devices in TO-220 their R_JA is approximately
60C/W. This is because the thermal resistance of air is much greater than that of the case. R_JA = R_JC + R_CA ~= R_CA != 60C/W. it depends on the surface area of the package so it can change significantly with the package shape. This means most power devices can dissipate around 2W in air without additional heat sinking.
Absolute zero[1]? The datasheet states Tc = 25C. They enforce that quite literally and stringently, making the *entire exposed surface of the case* held to exactly 25C. At any rate, they provide the corresponding thermal resistance.
What the OP is talking about is an apparent reduction in power dissipation at high voltages, shown on the SOA. This is nothing you can fix at the case level, it's an internal thing.
One reply stated a curve as low as 20W. I trust even the smallest power devices (e.g., DPAK, assuming suitable attachment) to handle that much, let alone a TO-247. This goes well beyond simple heatsinking -- besides, the most a manufacturer ever specifies in terms of heatsinking is "greased heatsink thermal resistance".
[1] Curiously, lots of diodes, resistors and other devices have a flat power, current or voltage limit up to a limiting temperature. GBPC35xx rectifiers are 35A up to 65C, derating linearly to 175C or so. I don't think I've seen this type of rating on transistors, though, so there may be additional power dissipation available from, say, a liquid nitrogen cooled heatsink. Assuming you don't cool it so far the silicon stops working (which specifically precludes "absolute zero").
Tim
--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
The temperature used for this is by industry standard 25 C. The device can safely dissipate "full power" if every square micrometer of its surface is cooled to 25 C. If seconary breakdown is avoided, that is.
If the device has a heatsinkable surface, cooling every square milimeter of that to 25 C is supposed to be good enough for the device to get away with dissipating full power.
Maximum current of some power MOSFETs is another story. I remember a small number of parts mentioned before in S.E.D. having current ratings that their leads appear to me to have trouble handling, unless slightly unconventional connection means are used. Such as maybe beefing up the leads with copper soldered-on externally, with close fit to avoid problems of solder having higher resistivity than copper. Or soldering the ends of rectangular copper "wires" to top and bottom surfaces of the leads.
Rtheta(jc), thermal resistance from junction to case in degrees C per watt.
Rtheta of any heatsinking compound and insulators, plus Rtheta of te heatsink itself (normally including that of the air around it). These are thermal resistances in degrees C per watt.
In other words, a transistor can take full power if it's in Niagara Falls and the water temperature is not exceeding 25 C.
Whoa! That's hot!
I have heard of silicon taking that and even 200 C, but for what life expectancy? 10,000 or 1,000 hours?
Wanna run silicon conservatively, especially in plastic packages, I feel better with 125 C.
So, the device should be able to "get away with" 140W when the heatsinkable surface of the device is cooled to (175-140) degrees C, which is 35 C. For "a little margin of safety" they probably say 140W when cooled to 25 C.
Absolute zero is not zero C. To the nearest degree C, absolute zero is
-273 C.
BUT, I find heatsinks tend to get warm to the touch - meaning usually exceeding 35 C. Often, when they are of sizes used to make products reasonably compact, they get fairly warm - 45 C.
One designing a product with a heatsinkable device and a heatsink should know the thermal resistance of the heatsink (including the air around it), plus that of anything between the device and the heatsink (thermal compound joint and any insulator).
Otherwise, figure that the heatsinkable surface of the device often gets to 50 C even without an insulator. (If you use one, know its thermal resistance.)
That is in 25 C ambient. But somebody is likely to use your product in outdoor air or in an unairconditioned building in midafternoon of a minimally-hotter-than-average early July day in Phoenix AZ, when the air temperature is 45 C - 20 degrees hotter than 25 C. So, unless you know good cause to depend on a more favorable temperature (such as on basis of known thermal resistance of heatsink, etc. and power dissipation and your worst case air temperature saying better), I would expect that the heatsinkable surface of the device could go as high as 70 C if the heatsink "normally feels only warm to the touch").
Heatsinkable surface of the device may go as high as 70, and for "conservative operation" I like the junction to not get past 125 C. That means 55 C between junction and heatsinkable surface. In the above example with Rtheta(jc) being 1 degree C per watt, this means 55 watts is the most that I am usually comfortable dissipating into a 140W plastic case device in products that get operated in large numbers away from my supervision. A little more is allowable if the heatsink is so big that its thermal resistance (including that of the air around it) is so small (and known) that more power will not push the junction temp. past 125 C in a 45 C ambient. Or if your product will only be operated in more reasonable ambient temperatures, such as if it is a personal project.
Where did that +1 come from? R_JA is thermal resistance from junction to ambient, not from case to ambient. Do not add to this the junction-to-case thermal resistance.
2.24W is max without a heatsink pushing the limits in a 35 C ambient.
I would be afraid that the 62.5 degree C per watt Rtheta(JA) may assume some typical soldering onto some typical length, width and thickness of PCB traces on a PCB that is in free air rather than air stagnated by a small enclosure around the PCB.
But what if the air easily reaches 35 C and I want the junction to not exceed 125 C so that reliability is improved? That means 5 watts through a temperature difference of 90 C from junction to ambient.
Thermal resistance then is 18 degrees C per watt. Subtract from that the 1 degree C per watt R_JC of the device, and that leaves 17 degrees C per watt for the heatsink (which normally includes that of the air around it), any heatsink insulator and/or thermal compound, and any enclosure around the heatsink including any air inside the enclosure made more stagnant by the enclosure than it would be without the enclosure.
Make that 15 instead of 17 if the device is going to be used by any old Joe where the air temperature may hit 45 C rather than 35.
My impression is people look at the top chart, rather than the IR MOSFET SOA curves in the bottom chart.
That's my point, I've got little I-Pak MOSFETs claiming they'll handle
50 or 60W at or near DC, yet this IR beastie claims 310W peak power, but can only let an amp through at 20V D-S? The other device SOA curves are there as the control, these are what 'normal' MOSFETs look like.
The IR MOSFET has some dismal effect that implies that huge TO-247 package is wasted on near DC operation.
Thanks, Grant.
Well the odd things here is there knee turning southeastsouth below about 35A.
Normal MOSFETs have the lines going southeast or eastsoutheast without the knee this IR beastie has in its SOA chart.
I don't think I've seen 2nd breakdown in any of them. A lot of the time, they don't even provide the DC/pulsed curves -- just some crazy, almost square, unlabeled thing.
If you test them without safety glasses, you might be able to win some money *and* teach them to write better datasheets! ;-)
Tim
--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.