Question for Win Hill

Rolling pin.

...Jim Thompson

I propose only to do a MOSFET amplifier design better than that done by the proponents, NOT to show that MOSFET amplifiers are better, because they're NOT.

I place people who espouse the benefits of MOSFET power amplifiers in the same category as those who think tooobz are wonderful... almost as bad as those who insist that BJTs are voltage-controlled ;-)

...Jim Thompson

Usually we just design something, lay out a proper production-quality pc board, release the documents, let manufacturing build us a first article, and we test that. You're going to have to do that sooner or later, and sooner gets the project that much further along. Multilayer boards and parts are a lot cheaper nowadays than an engineer's time, and breadboarding complex surface-mount stuff just isn't feasible any more. We seem to simulate only tiny snippets of circuits, often just passive r-l-c gidgets, and not whole functional circuits.

I see too many people wasting time Spicing and not thinking, and fiddling with the simulation (and the part models!) until they think it works. That teaches some very bad habits.

I did a tour of the Cornell EE department; PC screens outnumber oscilloscope screens by maybe 20:1.

John

I like mosfets in power amps because they're so rugged - no second breakdown - not to mention fast and easy to drive. I like to assign one opamp to drive each fet, making all sorts of drive, sharing, and thermal problems just go away. In some cases, especially high-power pulsed stuff, the fets wind up being cheaper than bipolars because they don't mind being pushed hard.

Haven't used a bipolar power stage in years. One of my guys designed a simple zener-emitter follower preregulator using bipolars. It kept blowing up. He quit and I replaced the bips with fets, and it just works now.

John

I'm curious about that configuration. Can you show me.

Sounds like load-line and/or inductive loads. I've not lost a BJT in the field (and I used to design ignitions); not to say I haven't killed many in the lab while designing snubbers ;-)

...Jim Thompson

Simple...

----+------ | ------------+ d out---+------g +--- - | s | c | +-------------+--r1 --+ | r2 or whatever | ------+------

which makes sort of a perfect follower: high Zin, millivolts of offset, and you don't give a damn about fet matching. The C-R1 thing can sometmes be deleted, depending. LM8261 is a nice driver amp.

Of course, when the fet blows, the opamp goes too, but opamps are cheap.

I thing turnon surges got it sometimes, but more likely the base zenered at turnoff. As long as you don't overdo the gate voltage, fets are very tough.

John

I read in sci.electronics.design that Jim Thompson wrote (in ) about 'Question for Win Hill/ Athlon64', on Fri, 31 Dec 2004:

Provided you design the loudspeakers WITH the lamp cord leads, everything will be OK.

I've lost plenty in final application; they are in fact fragile beasts, and the SOA is truly a killer.

Geez, it's an idea. I only do full designs for money. But I will suggest that precise followers need precise biasing circuits.

John

That's barely a switch... show me a push-pull version ;-)

...Jim Thompson

Howinthehell are you gonna get the oxygen out of the lamp cord?

I have the first three of his articles in a stack and expect the fourth final article to arrive soon. I've glanced at them, but clearly some concentrated study is in order. Hopefully he has addressed two of my four issues above (the two most important to audio amplifier designers). I have recommended his articles to others (hoping they'll check them out). Such as yourself, Tony?

BJT power amplifiers are so easily destroyed by reactive loads.

Win. Are you following/understanding Cyril Bateman's articles on modelling power MOSFETs in EW, and what are your opinions on his approach?

I think the concept has been, all along, is that engineers learn theory in school and practise on the job. That actually makes sense, because the theory is what sticks with you all your life, as the technology keeps changing. What many schools miss is the fact that concurrent hands-on work greatly enhances the understanding of the theory. If the theory is just equations and stuff, without physical insight, it's soon forgotten.

Engineering education used to be very practical. After WWII, various forces conspired to make it more "academically respectable" at some cost to true engineering, which is to design.

John

The mosfet model I'm often concerned the most about is the thermal model; that's what determines whether the part will blow up. I haven't had a lot of luck reconciling the transient thermal resistance curves on the datasheets with real life, so we test them ourselves, either by measuring Tj versus time, or just by pushing them to destruction and backing off by some margin for the real design.

Some of my amps digitize everything - heatsink temp, power rails, Vout, Iout - and simulate junction temps in real time, shutting down when Tj get scairy. For that you need a practical time-domain thermal model of the real-world, heatsunk mosfet.

I wish people would publish thermal models that were similar in detail to the electrical ones.

John

Agreed, but maybe 12 isn't too late. It's like tennis, skiing, driving, languages, music: if you wait till you're almost grown up, you're unlikely to be really good at it.

John

But I've had not one fail driving speakers.

...Jim Thompson

Playing devil's advocate for a minute, the nature of engineering has changeddramatically since WWII, and I'm not talking tubes vs. semiconductors or PCs (well, not directly, anyway.)

When I went through school, in a somewhat different line of engineering, the principles were all known, the equations were pretty well known, but there were almost no tools available to do the math!! As a result of this, the engineering "art" was largely one of estimating answers, because finding the"correct" answer was either impossible or impractical. A lot of what you did was based on tabulated data or historical info, based on what usually worked, or something that somebody put together and measured.

A whole generation of what I will call "Radio Engineers" grew up, who could use tables to slap together a resonant circuit, figure out its Q, make allowances for external effects and move on to the next problem. There was really no "engineering" involved, just patching together bits and pieces of what had been done before. These people absolutely, positively had the "hands on" experience that seems to be missing today, but I am ill at ease to call them engineers. Think about how much of the consumer electronics stuff of the 1950's worked "by accident," changing lead dress would throw it into oscillation, etc. Hell, think about all the truly rotten test equipment of the 1950's. This stuff was largely designed by the aforementioned "radio engineers." All experience and intuition, no real analysis or synthesis.

Today we are at the opposite end, largely due to the fact that simulation lets us instantaneously perform calculations that were impossible years ago. The student doesn't develop a sense for component values and functions because exact calculations are painless and instantaneous, no estimates or guesstimates are necessary. It's hard to force one's self to learn a hard or inaccurate way when the exact way is at you fingertip, and that is where we are today; hands-on is the hard inaccurate way, simulation is the (arguably) exact way.

Think about the chatter here a while back about the "decade boxes' being offered by another poster. A perfect tool for gaining experience, a lousy way to do engineering.

It's an uphill battle to push engineering schools or students back into "hands-on," but, as another poster said, if you aren't into it by the age of