SuperJunction MOSFETs for RF

[T3sl4co1l]

STF6N60M2
Common source
Vds = 100V
Id = 185mA
Rsrc = 5.6Ω (50 ohm generator, 3:1 Guanella style transmission line transformer)
Rload = 50Ω



Overall build.  Bottom: 100V supply.  Left: reflectance bridge.  (Partial lie; this one isn't any good at high frequencies, so I changed to a Mini-Circuits ZFCD-10-1 which is reasonable at 200MHz.)  Right: 20dB power attenuator.  (Signal flow is left to right.)




Insertion gain (s21).  Ref for all of these is bottom grid line (measured by difference).  Not very impressive at high frequencies, but it's not optimized for any particular frequency.  With a tuned load and better matching, I imagine the gain at high frequencies is quite reasonable (without sacrificing too much bandwidth).




Input return loss (s11).  Coincidentally good match around 100MHz, implying Z_G ~ 5 ohms somewhere around there.  Input impedance is higher at lower frequencies (match improves with resistor connected in parallel), as expected for a MOSFET gate which typically has a lossy-capacitor characteristic.




Feedback (s12).  Quite low over most of the band, implying a reasonable margin towards maximum stable gain.

Note that this load is below the load line (i.e., 100V / 185mA = 540Ω), so the gain, as well as maximum power output, can be much higher than what's measured here.  Just that the bandwidth is then crap.

Speaking of power and bandwidth: these transistors should be just about useless at low voltages.  Coss is massive until about 30V.  They should be treated more like vacuum tubes, actually: a relatively high saturation voltage, so plan accordingly in terms of [full-bandwidth] power output and load resistance.

The main downside is, the availability of power dissipation versus V*I ratings.  These are switching transistors, of course.  They are usually capable of DC SOA (amazingly, despite the reduced die area).  But this transistor for example, 6A * 600V = 3600VA of switching capacity, but only 20W of power dissipation, so it's a long way off ideal.

This does make class AB PP amps a bit more favorable than class A.  More stuff to put together (two or three TLTs for each balun), but vastly improved power output and efficiency.

Tim

[boB]

I have wanted to play with mosfets for RF amplification myself.

I'm not sure exactly what you were showing with the spectrum analyzer ?  Were you sweeping with a tracking generator ?

[OwO]

Is there a reason for using this particular transistor rather than cell tower LDMOS transistors like the MRF5S9070N? Or is it just sourcing availability?

Also that's a pretty noisy trace I would get more suitable TE for the job.

[David Hess]

Is there a reason for using this particular transistor rather than cell tower LDMOS transistors like the MRF5S9070N? Or is it just sourcing availability?

I have often been asked why a ubiquitous and cheap switching power MOSFET cannot by used for linear RF amplification.  They can of course but for the reasons T3sl4co1l gave, they are difficult to use and not very suitable.

I wonder how well an inexpensive ring-emitter bipolar transistor like a D44VH10 would work.  They are common for audio power amplifiers and switching regulators before power MOSFETs became available.

[T3sl4co1l]

Is there a reason for using this particular transistor rather than cell tower LDMOS transistors like the MRF5S9070N? Or is it just sourcing availability?

Unless those are pennies on the dollar, they can't match the price of a switching transistor.

Sometimes you don't need the GHz capability, and taming it is a liability.  Seems these would be good for SW or low VHF work, which a lot of hams may find useful?  Better than the old fashioned IRF510 amps, which work okay up into SW bands, but still can't compare with modern FETs.

Also that's a pretty noisy trace I would get more suitable TE for the job.

Yes, noisy, it was done with a noise generator and averaging.  My spec doesn't have a TG output (it has LO output, but I don't have a mixer to make a TG channel).  I would ignore the small blips in the curve; at worst those are probably blips due to slight cable mismatch.

I have often been asked why a ubiquitous and cheap switching power MOSFET cannot by used for linear RF amplification.  They can of course but for the reasons T3sl4co1l gave, they are difficult to use and not very suitable.

I've also been curious about the other way around, using an RF transistor for switching.  Besides never giving Rds(on) (but you do get Vds(on) at Id, Vg(on), probably close enough), they're just too expensive to bother with.

I wonder how well an inexpensive ring-emitter bipolar transistor like a D44VH10 would work.  They are common for audio power amplifiers and switching regulators before power MOSFETs became available.

Not bad I think, but like the IRF510 amps, limited to SW, maybe low SW at that.  Most Sanken and On Semi audio/servo transistors also fit the bill.  Usually fT peaks around 60MHz at Ic = 1A or so, which isn't bad.  The feedback capacitance (Ccb) is killer.

Of course, you don't mind feedback so much in an emitter follower, but then you still need to generate the voltage swing, and you need almost as much drive current as output current.

I'm quite fond of low-Vce(sat) transistors for drivers; besides high hFE and high current ratings, they have quite reasonable fT.  Not much power dissipation though, which limits the opportunity to use them in the volt-amp stage in such an amp.  Maybe as complementary emitter followers to drive the output bases, maybe even from a bootstrap supply to reduce power dissipation?

Tim

[boB]

Right, not very stable, Vgs threshold wise.

You would most likely have to design in some bias technique to keep it the same with temperature.  Shouldn't be too hard ?

[T3sl4co1l]

I used 4V of source degeneration. Not bad. Still drifty. A Vbe multiplier or thermistor would do fine.

Tim

[David Hess]

I have often been asked why a ubiquitous and cheap switching power MOSFET cannot by used for linear RF amplification.  They can of course but for the reasons T3sl4co1l gave, they are difficult to use and not very suitable.

I've also been curious about the other way around, using an RF transistor for switching.  Besides never giving Rds(on) (but you do get Vds(on) at Id, Vg(on), probably close enough), they're just too expensive to bother with.

I have seen this done before when the highest possible switching performance was desired at any cost before SiC and GaN power MOSFETs became available.

I have also seen bipolar and MOSFET RF transistors used as cascode output stages in pulse generators.  The topology was essentially CML (current mode logic) but with greater output power, current, and voltage.

[T3sl4co1l]

Have also seen planar triodes used in similar ways. Can do some neat things when efficiency isn't a concern (and, often, lifetime either).

SiC are good enough that you can make ESD generators with them.

Tim

[OwO]

I can get the MRF5S9070N for under $1 which is not much more than the STF6N60M2. Also avoids having to deal with very high voltages.