Constant current load - MOSFET selection?

[Alex1]

Hi, I am going to make the constant current load that dave made in one of his youtube videos but have a question regarding MOSFET selection.

Since the MTP3055 here is very expensive I am having to use a substitute, but I don't really understand the safe operating area in the datasheet so it is making selecting a MOSFET hard (I have only ever used them as switches before).

Would these two MOSFET's be suitable?

http://www.rapidonline.com/Electronic-Components/Stp55nf06l-Mosfet-Logic-N-60v-55a-47-0550
http://www.rapidonline.com/Electronic-Components/Stp36nf06l-Mosfet-Logic-N-60v-30a-47-0552

Thanks.

[mariush]

I'm using a fqp50n06 in my constant current load : http://www.fairchildsemi.com/ds/FQ/FQP50N06.pdf

They're about 2$ in stores, about 4$ for 5 pieces on eBay : http://www.ebay.com/itm/5PCS-FQP50N06-50N06-TO-220-Transistors-NEW-CK29-/260905933998?pt=LH_DefaultDomain_0&hash=item3cbf349cae

[codeboy2k]

can't really answer your question about those two specific MOSFETs without knowing what your specifications will be for your load.

Maximum voltage is ??
Maximum current is ??

That defines the power.  The SOA requires that you stay within that power area of voltage and current.

Also, I didn't look at the datasheets for those two FETS, but a FET that is good for a switchmode PSU is not necessarily the best one for an electronic load. Just keep that in mind. A FET designed for a switchmode PSU is going to have the lowest R_ds(on) but a FET for an electronic load wants a higher R_ds(on) so  you can get some control over the resistance of the channel.

[codeboy2k]

Some additional information to help you understand more..

Here's a typical SOA graph from a datasheet.  This particular FET is about 2W max, 12A max. If you look at the DC  line, you see that for every point under that line, the combination of drain-source voltage X drain current (ohm's law: V X A = power) is always less than or equal to 2W.

That's the Safe Operating Area.  Any combination of drain-source voltage and drain current that is less than or equal to the maximum watts that the device can dissipate. Also note that the maximum watts are usually specified at junction temperature T_j = 25C , so you have to heat sink it to get the junction down.  You probably can't achieve a junction temp of 25C, so you have to de-rate the maximum watts down, to a safe wattage at whatever junction temp your heatsink can achieve.



The dotted lines are extended SOA, that can be used as long as the device is on for only that short pulse time as indicated at the line, i.e. 10s, 1s, 100ms. That is useful information when used as a switching device.  A DC electronic load operates at DC, constantly on, so you must stay below the DC line to keep all the magic smoke inside the device.

[Balaur]

... and then try to keep the input capacitance (Ciss (Qt)) small as possible.

In my opinion, this is the most important aspect, as usual opamps will have difficulties driving a high input capacitance transistor.

[fmaimon]

Use an IRF540 or similar. These are almost a jelly bean power mosfet.

[fmaimon]

Yea the fets cheap but you will need three or four times the heatsink compared to a beefier package with a lower jc thermal resistance. It depends though on what he wants power and voltage and current wise. For 30-40w 50-60V yea use 3 or 4 of those. If you want a couple hundred watts or more and wide voltage range use a different one.

The only beefier package with leads that I know than TO220 is the TO247. But for 30-40W, a single IRF540 will do. Any computer heat sink will be able to dissipate that easily with a single one. I've done at least 30W with a single IRF540 and a 403K heat sink.

[krish2487]

There is the to - 3p package that is a tad bigger than the to - 247 package. It is designed to dissipate more power than the latter . See the irfp90n20. It is a rugged mosfet. You could parallel a couple and dissipate a couple of 100 watts comfortably ..

Sent from my Chaser using Tapatalk 2

[Smokey]

SOA is actually pretty easy to understand.
The main thing you are trying to do is keep the die from overheating.  What makes the die overheat?  Power, watts.
The three things you need to be aware of (and who's relationship is shown on an SOA curve) are voltage, current, and temperature.
The thing that makes SOA confusing is that the device has a max value for each of those things, but that max value is given INDEPENDENT of the other values.  What this means is that, for some fictional device that has a max current of 10A and a max voltage of 100V, you can only realistically get 10A at really low voltages or run at 100V with really low currents.  You can't run at max current of 10A and max voltage of 100V at he same time.

So what the SOA curve is showing you is the relationship of the current and voltage you can run continuously AND keep the die temp within limits.  If you want more current, you have to drop the voltage.  More voltage, then drop the current.  It's all about the watts, because it's the watts that are heating up the die.

Now another thing you have to be aware of is the temperature of the package.  Remember that what you are really interested in is the temperature of the die, which is always going to change temp faster than the temp of the package.  If the package is already hot, then you have less room for more heat so you can't have as many watts going into it.  What this means is that you need to derate the SOA curve based on the temperature of the device, or lower the running temperature.

That's pretty much it.  The SOA curve is just letting you know the combination of amps, volts, and temp that will keep the device from overheating.  One more thing to watch out for though is that you won't get much of a warning if you exceed the SOA curve.  The device will just overheat and blow up.  Because of that, it's a good idea not to try to run right on the edge of the given SOA.  Give yourself some headroom.

If you are looking for a good device to run in a linear application like this(as opposed to switching which is what most FETs are made for), a good place to start looking is the Audio industry.  They are the guys that are primarily pushing the high SOA linear operation devices at reasonable prices.

[fmaimon]

Well you’re pushing that FET. Its only rated for 88W max power dissipation at 25C case temperature. You have to derate 0.57W/C  so assuming you can keep the case at 100C that leaves you with 45W. Then given the high junction to case resistance if your case is at 100C the junction is likely at max; with isolating material its exceeded maximum. I prefer to keep the junction  at 100-110C max better long term reliability.

88W? The datasheet says 130W. Actually I don't use this rated power, so I did my calculations. The thermal resistance, from the junction to the sink, is 1.15 + 0.50 = 1.65 C/W. Let's use 2 C/W just to be sure. At 30W, I'll have 2 * 30 = 60 C temp rise. My heat sink claims that it's temperature rises 55 C without a fan, so I have 115 C temperature rise. With an ambient temperature of 40 C, the junction temperature will be 155 C, still 20 C below the maximun junction temperature. Hot, yes, but still within limits and some margin. If you use the values from the datasheets, the junction temperature will be around 140 C without a fan and 115 C with a fan. Pretty good for me.

[Rufus]

SOA is actually pretty easy to understand.

That's pretty much it.  The SOA curve is just letting you know the combination of amps, volts, and temp that will keep the device from overheating.

It isn't that simple. If it was just about watts a power rating and junction to case thermal resistance would tell you all you need to know.

Bipolar transistors and to some extent MOSFETs operating out of saturation have a temperature dependency which allows them to pass more current with increasing temperature. That means a local hot spot in the silicon will pass more current and and so generate more heat resulting in thermal runaway and rapid catastrophic failure at that spot.

An SOA plot is bounded by the device maximum current, voltage, and dissipation. Within those bounds the shape of the SOA plot is determined by stuff like the uniformity of the silicon and doping, thermal resistance across the die and silicon bulk resistance. Pulsed ratings are often provided because they are much higher, local hot spots taking time to develop.

 

[Smokey]

True.. True.. I should have said "the basics of SOA is pretty easy to understand."

Since the thread is about MOSFETs, here is a quick app note about about how, depending on the device and current levels, they can have a positive thermal coefficient and are less susceptible to thermal runaway than bipolars.  They are pushing their devices with on-die thermal sensors a bit at the end. 
http://www.onsemi.com/pub_link/Collateral/AND8199-D.PDF

[peter.mitchell]

Another package that is good for high power is TO-264, even though i have used all 3 packages (-247, -3P and -264) I can never remember which is biggest; it appears the TO-264 is.

Here are some package outlines;
TO-247
http://www.vishay.com/docs/95223/to247.pdf
TO-264
http://smartdata.usbid.com/datasheets/usbid/2000/2000-q4/to264.pdf
TO-3P
http://www.icbank.com/icbank_data/semi_package/to3p.pdf

[SeanB]

I worked with some equipment that used bipolars at both high current and with marginal heatsinking. They would run so hot that the TO3 packages would unsolder themselves from the wiring, and then they would either blow up or stop working, depending which lead was off first. Base meant blow up as then the bipolar was so hot it would go further into thermal runaway and eventually blow both the lamp ( invariably 2 as there was a spare lamp and of course it would be selected and poof) and the fuse. Emitter and just no go. Eventually we selected the base drivers for VceSAT at 1A using a curve tracer, as this was more reliable than any other method. Provided the pass device had a gain over 15 at 10A no other parameter was important. The driver was the important device.

No, to change the design would involve 2 years, a mountain of paperwork and ulcers all round. A simple form that stated - use additional pages per section if needed, and had 25 sections to be filled in in detail. Then await the real forms........ Was cheaper to just order case lots of 2N2219A's and 2N3773's, along with a lot of BUX23's. then again this board had unijunction transistors on it as well, those are pure unobtanium.