Which N Or P Power MOSFETs ?

It's ususlly easy to assign one opamp per fet, to ensure basically perfect matching. Opamps are cheaper than giant fets and giant heat sinks, and the source resistors can drop millivolts, not volts.

The claim I'm making is that those constant power curves at the right assume the temperature is uniform across the MOSFET die, as it's likely to be in a switching application where it's either fully off or each pulse is putting it solidly into the ohmic region.

Makes no sense. Fig 8 clearly has data at 200 volts and 5 amps.

I don't see any assumptions about temperature uniformity. The SOAR curve just specs what is safe.

(Not that I unreservedly believe them. Processes change more often than data sheets.)

There are several articles about the topic:

"Figure 1 shows the typical SOA curve included in most power MOSFET datasheets. The constant power curves, shown to the right of constant current line within the SOA boundary, are extracted from the thermal data with the assumption that the junction temperature is essentially uniform across the power MOSFET die. The dissipated power does not cause a catastrophic failure to the device, but brings its junction temperature up to the maximum guaranteed temperature when the applied power pulse is evenly distributed on the die surface."

Thermal gradients, blowin 'em up:

For a single pulse, yes? So, one pulse and you're done forever?

For switching applications the SOA curves are very useful tools for part _rejection_ they tell you unambiguously what a part can't do.

They're less useful for part-selection/optimization. And for linear operation the curves often look worse particularly for lower voltage/high transconductance devices, but why are they gonna show you the part look bad for an application 99% of people looking at the part aren't gonna use it for.

Most FETs are not good for using in linear mode, e.g Class AB due to the Spirito effect:

Nah many will work fine if you de-rate everything conservatively but you don't need a SOA curve to tell you this.

Sure. It's called a "plate modulator". ;)

Cheers

Phil Hobbs

The base of Tr6 connects to one end of R12. Is the other end of R12 supposed to connect to Tr10's drain?

What's the deal with the dashed line around Tr10, Tr11, L2, Sw1, etc? Is the dashed line supposed to denote a module?

Danke,

Klaus Kragelund fabricated :

============================

I bet you have zero experience dealing with commercial mosfet amplifiers.

..... Phil

Where would be the delay line in a plate modulator?

On Sunday, 22 November 2020 at 15:15:25 UTC-8, Don wrote: ...

Yes, I checked my copy of the original and the line is missing, but pretty certain that's where it goes.

If you follow the line around it is enclosing the low level circuitry that is constructed on strip-board (as described in the text after "Construction and Performance".

The power components are mounted on the chassis and heat sink.

kw

Oh My ! Process changes are changes that I have to worry about. Or used to have to at least. Back in the 1990s, in one of our power inverters, we used a certain Motorola (On Semi) FET in the circuit.

They worked fine untill they did a die shrink on that part.

Inverters have to surge and we rely on the large die area for transient thermal transfer. They said "but the RdsOn is the same !"

Well... They did not understand evidently how that can hurt the thermal impedance in the application.

As I remember, we changed to Harris Semicondutor (then Fairchild and On Semi again I think) and all was well.

Never had a problem like that again luckily.

"I think there is a class of RF power amps that modulate the power supplies of the final, to track the RF envelope."

Teasing friends is a bit of a way of life round here. ;)

Cheers

Phil Hobbs

I did a half-bridge motor driver and used two Moto fets. During braking, the substrate diodes conducted some. They had the ideal diffusion profile of a step-recovery diode, made enormous spikes, and blew out their own gates.

Does anyone know if that's still ever a problem? My new class-D amp has schottly diodes across each fet, just in case, but I may not need them. May well stuff them; they don't cost much.

Not needed for straight classic AM, but an EER amplifier handles more complex modulations by handling the phase and amplitude components separately. You may need delays to match the component timings at the final. The plate modulator is just the amplitude handler.

Aren't plates those red things in tubes?

The thing I was remembering was about using an ultra-linear amp to output a complex constellation, maybe with multiple carriers, where distortion would be expensive. So they delayed the signal long enough to have the power supplies get ready for the rare but big amplitude peaks.

It probably has a name. That sort of thing sells those million-dollar oscilloscopes.

Not usually - except in a ham contest.

There were a few tubes designed to run plates red at nominal dissipation. I recall Philips transmitting tube QE 08/200 which specified 'Anodes red hot, remperature 850 K).

You do not need anything more exotic than a cellular phone base station.