john curl said:
John, thanks for supplying this data. The 2SC3264 ring emitter transistor really is quite good, and the secondary breakdown effect appears to be quite small at 100V. A couple of observations, however. The complement is the 2SA1295, and its 10ms-100V SOA is only 3.5A. In comparison, the IRFP240 and its complement are rated by IR at 3A at 100V. The 200W ring emitter devices still slightly edge out the 150W MOSFETs, at least by the spec sheet.
In reviewing the IRF SOA curves, it is evident that they set the 10 ms SOA curve at a constant power of 300 watts, reflecting the fact that the 10 ms thermal impedance is about half the static junction-case thermal resistance. What this means is that their criteria for the safe area limit is a transient die temperature no greater than their nominal TJmax of 150C. This seems very conservative, since their reliability is such that they are rated to operate continuously at 150C. This made me think that their approach to defining safe area is very conservative.
So I decided to torture a few IRFP240s and destroy them. I applied 10 ms pulses and drove them to higher and higher pulse current until they blew, all with 80V drain to source (I couldn't quite get to 100V with my supply). They didn't pop until 9 amps, reflecting a power level of 9*80=720 watts, or slightly more than twice the rated value. Of course, we would expect some manufacturing margin, but this is still very good news. Devices from different lots years apart all blew around 9-10 amps at 80V for 10 ms.
Since the failure mechanisms may be a bit different, and since not all manufacturers spec with the same degree of conservatism, its hard to guess how much better than 3.5 amps a 2SA1295 might do in reality. Are you interested in giving it a try?
Two other things I noticed about these parts. Their ft drops like a rock above 6 amps, where it is already down to a typical of 40 MHz. Secondly, they have a lot of collector-base capacitance, 500 pF at Vcb=10V for the PNP.
Anyway, at least on paper these ring emitter transistors' SOA is as good as the MOSFETs', even though the MOSFETs don't suffer from second breakdown. But I don't think one can conclude that MOSFETs are more prone to break at high voltages. Thanks again for sharing this data.
Bob
Bob Cordell said:
Bob, it might be interesting to have a look at the lowly IRF640. According to IR, their 10ms/100V SOA is 4A, better than IRFP240, to which it is very similar. Of course, it's a smaller TO220 case and nearly 50% higher Rth from junction to heatsink, and costs very little, but that is what makes the SOA rating so surprising. That being said, there are numerous IRF640 'clones' out there with sometimes wildly varying dissipation and SOA ratings.
BTW, in your tests, how long did you take between applying 10ms pulses? I am looking forward to your tests re PMOS gm anomaly.
ilimzn said:
My guess is that the die inside the 640 is the same as the one in the 240, just different power package, so it's unclear why they would rate the SOA higher.
In the test that I did, the interval between pulses was 0.5 second, so the test was slightly more grueling than the single-pulse SOA data that IR publishes.
I re-did the test again this morning on another 244 sacrificed to the lofty goals of DIY science. This time I applied 125 volts, and it popped at 5.5A, for a pulse power value of 688 watts, reasonably close to the 720 watts I got in the earlier experiment, and still more than twice the rated 10 ms pulse power of 300 watts. Encouraging.
Bob
traderbam said:Bob,
We're just measuring speed in a different way. If you equate speed with ft then certainly VFETs are speedier than BJTs. And the source/emitter follower simulation you suggested is consistent with this.
I'm not just looking at speed in terms of current gain. The design approach I use, which I shall keep under my hat for now, causes me to look at it in a different way and this reduces the advantage of FET over BJT.
From what I have learned it does not surprise me that some have found more sonic success using bipolars than FETs. Many leading brands use bipolars. There are many brands that use FETs that sound mediocre. It is pointless to generalize about this.
Personally, I don't like either of them.The tube folk are blessed.
MikeBettinger wrote:
Sorry.Sound quality is hard to measure. That which is easy to measure tends to get the attention and tends to be "improved". Often improved much too far at the expense of that which is more important but hard to measure (or identify).
I respect your desire to keep your proprietary design approach under your hat for now, but I don't understand why that prevents you from explaining your definition of speed as it applies to the source-follower (emitter follower) output stage that 90% of us use.
Moreover, I'm not just equating speed with ft. As I pointed out, large-signal turn-off speed is just as important, and the MOSFETs are faster in that regard as well.
What other speed criteria is there that is applicable to the source follower (emitter follower) stage that you can tell us about? How else are we to evaluate your assertion and even possibly agree with it?
Bob
Bob Cordell said:
That's what I thought too, but apparently the jury is still out on that. What I do know, is that the IRF240 and IRFP240 share(d) the die. It is entirely possible that from some point on the IRF640 and IRFP240 have the same die. Diagrams look the same, although some numbers point to process tweaks and cannot be accounted for by the mere change of case - like 30pF more Cds for the IRF640, hardly accounted for by the REDUCTION of the case size, just like increased SOA from a smaller mass of copper slug. The sim models are different too (but we all know this should be taken with a BIG grain of salt...). Perhaps a question to IR would shed light on this, or, failing that, I may have to take a hammer to a P240 and 640 and see for myself
Ultima Thule said:But if runing it in CS it will be more noticeable, but we then also have much more voltage gain to fix the transconductance anomalies by means of NFB.
I'm not sure my conclusion is right but do I guess right that your comments regarding IR's P-ch transconductance anomalies and the differences depending on CD or CS is valid only in a NoFB amplifier, or should I say more noticeable?
With no Source resistance and in Common-Source operation,
you get something like a 4 db variation in gain from bottom to
upper mid-band. This is reduced when there is Source resistance.
So when you run CS with no feedback, you have a problem.
Enlosed in a feedback loop, you contend with a 4 dB variation
in open loop gain, and with a generous feedback figure, this
is not much of an issue, but with low feedback designs, it
becomes more prominent.
Bob Cordell said:
I decided to blow up another IRFP-240, this time seeing how well it would do at elevated case temperature, which is more reflective of the real world.
The test was as before, with 120V Vds and 10 ms pulses spaced 0.5 seconds apart. This time, however, I mounted a second 240 on the same heat sink and used it as a heater to heat the heat sink to 60 degrees C.
This time the IRFP-240 was able to survive 5-Amp pulses at 120V at a case temperature of 60C, for a pulse power of 600 Watts. This is very encouraging, in that the device still exhibits a SOA of twice its rated room temperature value while actually at 60C.
I am now very curious to see if the BJTs have as much margin against their published SOA limits.
Bob
nuvistor said:
And another article, http://www.microsemi.com/micnotes/APT0002.pdf
"Power MOSFETs are generally designed as switches. Total power dissipated is the sum of "on"-state losses plus losses generated during the very short switching intervals. Desirable characteristics include the lowest possible ON resistance, high breakdown voltage, very high gain (g or Gm ), minimum switching losses, and a low gate threshold voltage VGS(th)
In those applications requiring operation in the linear region, these characteristics are not ideal. Firstly, Gm is too high and, secondly VGS(th)has a negative temperature coefficient which makes it impossible to maintain constant drain current without negative feedback. Finally, and most dangerous of all, large switchmode MOSFETs exhibit a phenomenon known as "hot spotting" or current tunnelling."
Usually I consider that as a good thing, somewhat less snake-oil businessBob Cordell said:
Link that Nuvistor gave is better reading than APT's somewhat shorty explanation:
http://powerelectronics.com/mag/511PET24.pdf
(related stuff, lateral vs. trench-mos)
http://powerelectronics.com/mag/Ely January 2004 PET.pdf
Workhorse said:
The 10 ms safe area line on this transistor is at 2500 watts - very impressive. This would appear to be 8 times the 300-watt rating for the IRFP 240. But I don't know anything else about this transistor, for example, how big is the die, how big are the capacitances, how much does it cost, does it have a p-channel complement?
Bob