Bob, let's keep a level playing field. ;-) Ed Oxner is still working with fets. Siliconix is dead. However, he still works as a consultant for Interfet.
Ed says that it doesn't really matter whether the fets are V-mos or D-mos. He is also looking at really high frequency operation. It is not so good to make a big thing about bipolar transistors, yet we already have to cover the beta drive, so it becomes a tempest in a teapot.
Ed says that it doesn't really matter whether the fets are V-mos or D-mos. He is also looking at really high frequency operation. It is not so good to make a big thing about bipolar transistors, yet we already have to cover the beta drive, so it becomes a tempest in a teapot.
Nelson Pass said:
Of course, but with a BJT output stage, there would not be the
temptation to skip the driver stage. Thus only Mosfets suffer
the foolishness of being driven from the VAS.
None of you caught the butt of the joke (I think) because it is Nelson who in all of his 2 stage amplifiers, like the Alephs, drives the Output Mosfets straight off the Input Pair(s) and gets fantastic sonic performance. I presume that in this statement Nelson might have had his latest 2.5 stage designs in mind!!
john curl said:
Hi John.
Not sure what you mean by keeping a level playing field. As I said, I have a lot of respect for Ed, and I don't think I was disagreeing with anything he said in his paper. He himself said the parasitic NPN won't likely turn on. It was you who asserted that it does turn on. Read his paper more carefully.
He did speculate in his paper that the issue was likely the same for VMOS or DMOS, and I don't have an issue with that; however, he was referring to small MOSFETs made back in 1980. All I said was that geometry can strongly influence the relative magnitudes of what he is discussing, and we all know that geometry has changed a lot in the last 27 years.
After reading Ed's paper carefully, I think the matter of the parasitic NPN is not at all a problem in the context of audio amplifier design, so in that regard it is a tempest in a teapot; it was you who described it as a potentially serious problem.
Thanks again for sending me the paper, because things like this always make us think a little bit harder.
I'd still be very interested in what Ed has to say about the parasitic NPN issue in regard to today's MOSFETs in audio amplifier applications.
Cheers,
Bob
anatech said:
I've recently measured carefully the input capacitance of MOSFETs, and there is no rapid rise in Ciss as the device turns on, when measured properly. MOSFET input capacitances are no more nonlinear than BJT input capacitances.
I certainly agree that MOSFET output transistors should be buffered from the VAS, just as with BJT's, but also agree that good performance can be had from MOSFETs driven directly from a very beefy VAS.
Cheers,
Bob
I just spoke to Ed Oxner. Yes, the parasitic transistor is still very important and they even test for its potential effect at Siliconix.
When I say a 'level playing field': To emphasize potential bipolar transistor problems and to minimize potential mosfet problems is not being very unbiased. We all want to know the problems and the advantages of each and every design approach.
If you are of the opinion that the parasitic transistor is not very important, then you should show some evidence to prove it to be as you think, not just dismiss it.
You should talk to some real mosfet designers, like Ed Oxner, and get the inside info.
You never know, we all might learn something.
When I say a 'level playing field': To emphasize potential bipolar transistor problems and to minimize potential mosfet problems is not being very unbiased. We all want to know the problems and the advantages of each and every design approach.
If you are of the opinion that the parasitic transistor is not very important, then you should show some evidence to prove it to be as you think, not just dismiss it.
You should talk to some real mosfet designers, like Ed Oxner, and get the inside info.
You never know, we all might learn something.
john curl said:
I have tried very hard to be fair to both MOSFETs and bipolars, at least as much so as you, John. They both certainly have limitations. Based on my reading of Ed's own words, your characterization of the seriousness of the NPN parasitic transistor effect for audio amplifiers was a red herring. If it's really a problem, or a performance-limiter, I want to know about it as much as you.
I'll discuss directly with Ed about this, if possible. It may very well be "very important" in some context, but the relevance of that context for audio power amplifiers needs to be illuminated, something you have not done.
Cheers,
Bob
Bob Cordell said:
Bob,
The NPN parasitic (N source- P body- N epitaxial layer) in vertical MOSFET's has the base shorted to the emitter by the source metallization which overlaps both the source implantation/diffusion and the P body. Also, around the metallization contact, the P body is extra heavily doped to provide a good ohmic contact. This effectively kills the parasitic bipolar beta. The price payed for killing the the NPN parasitic is the formation of the body diode, with all the troubles regarding the stored charge and switching times on reactive loads. If there's something that potentially may affect the HF behaviour of a linear amp driving a complex reactive load then it's the body diode and not the parasitic NPN.
Now, the base emitter short is constrained to the source contact, so, given the non zero resistivity of the P body, there's still a 3D effect near the MOSFET channel. High commutation dv/dt causes high current density of minority carriers (positive carriers, or holes) in the body region, which can build up enough voltage across the body resistance (from the channel area to the source contact) to turn on the parasitic BJT. This is the reason for the peak commutating (body diode recovery) dv/dt limit in the datasheet. Peak commutating dv/dt is higher for a vertical MOSFET vs. lateral MOSFET, because of reduced minority carrier lifetime in the P body.
I fully agree that for linear audio application the parasitic NPN is a non issue. Under any circumstances, the parasitic NPN is a reability risk rather than having an impact on the amp performance.
syn08 said:
Hi Sy,
Thanks for this explanation. It pretty much parallels what my understanding was of what Ed Oxner was saying in his paper. In particular, that it would take a very high dv/dt to create enough voltage drop to turn on the parastic NPN (much more than, say, 300 V/us).
With regard to the body diode, my understanding is that it is actually quite a fast-recovery diode in MOSFETs like the IRFP240, largely as a result of it needing to be fast when acting as a commutating diode in switching supplies. The IRFP240 has a typical reverse recovery time of 250 ns at 18 amps, which places it in the industry definition of fast recovery diodes (Trr<500 ns). Also, I believe that in a properly designed audio amplifier, this diode will never be turned on other than in a fault condition, perhaps, for example, where I-V limiting comes into play when driving an inductive load.
Finally, the capacitance of this diode is included in the drain-source capacitance of the device, which is about 350 pF for the IRFP240.
Thanks,
Bob
lumanauw said:
The IRF9540 is spec'd at 5000V/us max dv/dt. It is unlikely that a reasonable power class D stage (say, 200 watts) running at 500 KHz would reach such a rate.
I'm not an expert in switching amps, but I do know that when a dv/dt limit is approached in any power switching circuit, then snubber circuitry should be used to protect the power devices.
Sounds good to me, but it does also seem to limit the real gain bandwidth of mosfets as well.
The point that I am trying to bring out is that we have to 'worst-case' real devices, if we are to truly understand them.
The same thing happened to me, decades ago with worst case dV/dt in audio circuitry.
People often used the wrong phono cartridge, microphone, or other input source, and would then establish a slightly to ridiculously low slew rate criterion. For example, in the '70's Dr. Lipshitz et al used a Shure V15 (of the time) to measure slew rate. I proved in my IEEE paper that this was ridiculous, and I had given him a copy of my paper, so there was really no excuse. Oh well! Some folks might use a bandwidth limited transformer with an MC cartridge and get less than worst case results.
Much of the criticism of Otala was based on these half baked efforts to find true worst case for all practical conditions.
I would hope that if bipolar transistors are extensively analyzed, that mosfets are analyzed with the same intensity, and that nothing is brushed off, just because it is inconvenient to look further.
The point that I am trying to bring out is that we have to 'worst-case' real devices, if we are to truly understand them.
The same thing happened to me, decades ago with worst case dV/dt in audio circuitry.
People often used the wrong phono cartridge, microphone, or other input source, and would then establish a slightly to ridiculously low slew rate criterion. For example, in the '70's Dr. Lipshitz et al used a Shure V15 (of the time) to measure slew rate. I proved in my IEEE paper that this was ridiculous, and I had given him a copy of my paper, so there was really no excuse. Oh well! Some folks might use a bandwidth limited transformer with an MC cartridge and get less than worst case results.
Much of the criticism of Otala was based on these half baked efforts to find true worst case for all practical conditions.
I would hope that if bipolar transistors are extensively analyzed, that mosfets are analyzed with the same intensity, and that nothing is brushed off, just because it is inconvenient to look further.