[Amps] High voltage MOSFETs

[Manfred Mornhinweg]

Bill,

> One intriguing possibility with these is direct coupling (through an
> LP filter and DC isolation of course) to 50 ohm coax. 

Forget the low Q lowpass filter. This transistor is rated for 300pF 
output capacitance. At 30MHz that's a reactance of 17.7 ohm. But the 
load resistance for 300V and 750W operation is roughly 60 ohm. This 
means that you cannot possibly absorb the high capacitance of this 
device into a low Q lowpass filter. Instead you have to use a medium Q 
circuit between the device and the load - exactly like is the case with 
tubes.

Instead of the Pi circuit practical with tubes, a T or LCC circuit is 
better suited in this case. And while tubes typically require a Q factor 
around 10 or 12, and often even higher on the 10m band, this transistor 
could work with a significantly lower Q, like 6 or so.

Broadband operation would be possible to around 10MHz only. Better than 
tubes, but not good enough for full HF coverage.

 > No transformer needed.

That's right. And a single-ended amplifier is entirely practical. This 
eliminates the main cause for "typical" low efficiency of semiconductor 
amps. And the "conventional" layout of a single-ended stage with a tuned 
output network might even appeal to people used to tubes!

> Something to think about. 

I have been thinking about this for years. I hope to be DOING something 
about it soon - but I make no promise. It's just a good intention.


Joe,

> Only on 160 meters for 750 Watts output.  By the time one gets to 27
> MHz the optimum output impedance is down around 5 Ohms ... 

You are misinterpreting the datasheet. It says that the conjugate of the 
optimal output load impedance at 27MHz is roughly 5-j15 ohm. That means 
a series connection of a 5 ohm resistor and a 393pF capacitor, but the 
actual transistor is NOT such a series circuit! The impedance values are 
given as series impedances only because that facilitates the design of 
matching networks. The actual behavior of the transistor is like a 
current source having a capacitance IN PARALLEL, not in series. If you 
convert the given values of series-expressed impedance to their parallel 
expressions, you will find that

- the equivalent source resistance is roughly constant over the 
frequency range;

- and the output capacitance is also roughly constant over the frequency 
range!

That makes life much easier, when you are considering a broadband circuit.

> it would be a real trick to design a broadband output. 

As far as my understanding goes, it would be impossible to design a 
broadband output circuit that covers to 30MHz. Broadband output circuits 
are possible only up to the frequency where the device's output 
capacitance, plus strays, can be absorbed into a lowpass network having 
a Q of 1. That would be roughly 10MHz in this case. If anybody here 
knows otherwise, please teach me!

 > The best approach might
> be impedance matching in the output filters for each "band".

Yes. Either using fixed-tuned networks per band, switched by relays just 
as if they were lowpass filters; Or using a single, tunable network, 
which could be manually or automatically tuned.

The switched network approach is no-tune, but requires lots of high 
current, high voltage capacitors and coils - much more so than lowpass 
filters, due to the higher Q. It would probably prove too expensive and 
bulky. Which leaves us with a single, tunable network, very much as in a 
tube amp. Using variable caps, tapped/switched inductors, etc.


Steve,

> they were 
> originally designed for narrow band saturated cw use and one 
> common theme is the problems of trying to use them as linear amps. 
> Bias control is a problem and reduced supply voltage is 
> recommended, even then linearity is poor. One 'linear' design 
> quotes IMD as -25dB ref pep or -19dBc!
> 
> I dare say they are capable of better than that, but it won't be 
> simple and straightforward.

And I would say that instead of trying to force them to behave well 
enough in a linear circuit, we would do better using them in the way 
they like: Saturated, biased well below the conduction threshold, 
running at high efficiency (there I go again!), and then we can achieve 
the linearity by other means: Power supply modulation, along with 
predistortion.

In other words, I'm proposing to use devices like this in a tuned 
amplifier, running class C, E or perhaps F or inverse F, in an EER 
amplifier driven by an SDR.


Peter,

> Tokyo High Power used a pair of ARF1500 in their HL2.5KFX.
> 
> http://www.ab4oj.com/dl/thp/hl2_5kfx/manual_hl2_5kfx_complete_schematics.pdf
> 
> 
> Two amps with those transistors were built in Germany:
> 
> https://www.yumpu.com/de/document/view/21269897/kw-pa-arf1500-mods-hamcom

All of these use the ARF1500, rather than the ARF1505. That's the lower 
voltage version, to be run from 100-150V supplies. The output 
capacitance is nearly the same, but at the required 4 times lower load 
resistance the resulting bandwidth for broadband operation is 
correspondingly larger. So these transistors do allow broadband 
operation to 30MHz, unlike the ARF1505.

But they need transformers in a push-pull arrangement, and thus suffer 
from all the troubles that come from insufficient drain-drain coupling. 
Mainly low efficiency and high IMD, both of them a result of significant 
common-mode voltage swing, causing somewhat peaky drain voltage waveforms.

Manfred


========================
Visit my hobby homepage!
http://ludens.cl
========================