[Amps] SS Amps are expensive...why?

[Manfred Mornhinweg]

Hi Roger, and all,

> SS amps (including those in many high priced 100 and 200 W rigs) have
>  historically had poor IMD.

That's correct, unfortunately. And the reasons are various. Among them
the fact that bipolar transistors aren't complitely linear, MOSFETs even
less so, that in order to linearize them its necessary to use negative
feedback, but negative feedback requires surplus gain, and RF power
transistors, until very few years ago, were very short on gain. To this
we have to add poor circuit design, as almost every HF radio in the
world uses power amplifiers copied from early pioneering work done at
Motorola in the 1970s, where the goal was to get any RF power at all,
rather than getting clean RF power.

Add to this that it's mighty hard to find ways to correctly handle the
very low impedance of power transistors delivering 100W from a 13.8V
supply. As a result almost every of those final stages is only a very
poor approximation to what a true linear amplifier should be.

All this can be improved quite dramatically, but hams these days buy
radios instead of building them, and don't ask the manufacturers loudly
enough (that is, with their wallet and feet) to build better linearity
into the transmitters.

> To build a SS amp with a CLEAN output requires an amp capable of 
> running almost twice the desired output. So an amp for a clean 1500W 
> legal limit should be capable of close to 3 KW out. Adding the 
> REQUIRED low pass filters capable of 3KW may cost as much or more 
> than the rest of the >amp.

I have to state some corrections to this. First, what exactly is an
amplifier's power rating?  The fact is that one and the same amplifier
can carry several power ratings, even infinite power ratings, best
expressed in a set of graphs. These graphs indicate at what power level
the amplifier produces how much distortion, efficiency, and gain. As the
power gets higher, starting from zero, the efficiency improves, the
distortion first goes lower, then stabilizes, and then goes higher and
finally shoots through the roof, while the gain has small variations
throughout the low and mid power range and then drops more and more
toward high power levels.

Typically linear amplifiers are rated for the power at which the IMD
degrades to just the acceptable level. That could be anywhere from
-28dBc to -40dBc, depending on the quality desired, and the type of
signal fed through the amplifier.

For testing purposes, often an amplifier is rated for the power at which
the gain is 1dB lower than at mid power. That's the "1dB compression point".

Instead non-linear amplifiers are typically rated either for the maximum
power they can produce under practical conditions, or a certain power
level where the trade-off between efficiency, gain and power is good.
IMD is irrelevant, of course.

So, Roger, if you want a 3kW amplifier to use at 1.5kW producing a clean
signal, the 3kW is probably the saturated power output, and at 1.5kW the
amplifier might be capable of an IMD level of -35dBc or better. But in
this case, this being a LINEAR amplifier, the power rating would be
1.5kW, not 3kW!

Also neither the low pass filters nor the power supply need to be
dimensioned for 3kW operation, because the amplifier will actually never
run at more than 1.5kW - provided its control system is working
correctly, keeping it in its linear range.

The cooling system has to be dimensioned for the power level at which
the dissipation is greatest, not where the output power is greatest. The
amplifier mentioned in this example would probably reach peak
dissipation around 1kW output. Beyond that the dissipation starts going
down!

> This amp would pretty much require the latest state of the art 
> transistors. (Expensive and unlikely to be found at swaps)

It's indeed good to use state of the art transistors, because these have
extremely high gain, allowing the use of strong negative feedback and
thus better linearisation. Also they are less expensive, per watt, than
older ones.

These transistors are NOT expensive, compared to tubes. A transistor
(actually two transistors in one package) rated at 1.5kW output costs
$172 in single quantity, and less if you buy several:

https://www.digikey.com/product-detail/en/nxp-usa-inc/MRF1K50NR5/568-13079-1-ND/6562303

And if you get into more modern technology, using switchmode amplifiers
with envelope modulation, you can use gallium nitride transistors that
are even much cheaper than this.

What is true is that you won't usually find them at swap meets. But it's
quite possible that eventually somebody will buy 100 of them, or more,
at quantity discount, and sell them at swap meets cheaper than the
quantity for a single one at a distributor.

> Even then we are probably looking at two pallets with 2 transistors 
> each. IOW, 4 transistors, each capable of 1.5 KW (3KW out at 50% = 
> 6KW input) =  some big dollars plus combiners to combine the signals
>  from the two pallets.

No longer necessary. That used to be the case, but thanks to modern
devices, like the one linked, it is in principle possible to use  a
single one for 1.5kW output. Doing so in linear mode with a conventional
class AB amplifier would require extremely good cooling, though. A
well-designed, micro-machined copper water-cooled plate, along with a
radiator, high pressure water pump, and relatively quite fans. Instead
if you use two of these transistors, conventional air-cooling works, and
the whole thing is built as a single pallet. No combiners needed.

My own approach is using a single one, in a high efficiency amplifier.
If I write less and work more, maybe I might complete that project
before leaving this planet...

> The big limitation here is eliminating 3 KW worth of heat? from 4 
> physically small transistors.

The peak dissipation in a 1.5kW class AB linear amplifier using this
kind of transistor is only around 1kW, assuming the SWR is low and the
circuit design is good. To tolerate some higher SWR, more than 1kW
cooling capability is desirable. But 3kW cooling capacity is a large
overkill. Of course, this is still enough of a problem! Actually a very
tough one, if we want to use a single transistor.

> This means a LOT of work, skill, and knowledge in heat transfer.

Yes.

> The mating surfaces on the heat spreader and the transistors need to
>  be flat! Really flat! Lapping surfaces close to being close to 
> optically flat takes skill!? Sometimes hours of effort and a great 
> deal of mechanical aptitude.

Forget that. Solder down the transistors. Many modern transistors don't
even have bolt flanges. Transistors of this power level tend to buckle
up when hot, due to differential thermal expansion, and no amount of
lapping and greasing will keep excellent thermal contact. Also the
sources connect through the base metal, and absolutely need an extremely
low inductance path to the rest of the circuit. Soldering is the only
reasonable mounting method. And given that the thermal conductivity of
solder is dramatically better than that of the best thermal grease,
surface flatness is no longer as important. Of course, the surface
should still be pretty flat, but a mirror-like polish is definitely not
required. What IS required is a good soldering job, that results in a
thin solder layer, free of voids.

And the copper base needs to be soldered to the circuit board, to
complete the source circuit! Many, many homebrewers forget this.

> Many of these capabilities/requirements are outside the realm of the
>  sharpest people in electronics and old timers.? I know how, but no 
> longer have the physical abilities.

I feel like you... The older one gets, the more one knows, but the less
does the body help to turn knowledge into working equipment.

> SS amp control? It can be built in or done by computer. Neglecting 
> the software, we still need to get the required signals, in the 
> proper form to the computer which may require substantial hardware in
>  the SS amp adding still more cost.

The only control that's really required is TX/RX switching with proper
timing, low pass filter selection, and SWR/thermal protection. This
doesn't need any computer. On the other hand, considering that a
self-contained microcontroller containing several analog inputs digital
inputs and outputs, various internal specialized hardware, along with
copious memory and processor capacity, costs under $2, it's not such a
bad idea to use one to replace discrete circuitry!

Of course one can follow the modern trend and make the amplifier look
like a computer with an RF stage added in the back corner. But I don't
see the point in doing so.

> Now where I disagree on a couple of points. With today's licensing 
> structure (needed to build our numbers to justify our use of very 
> expensive frequencies) bringing in many younger and less educated 
> people with little? money or interest in electronics into ham radio, 
> lowering the average age of hams. Yes there are those few who are 
> expanding the envelope, bringing with them things like dynamic 
> predistortion. STILL the knowledge level (and abilities) of the 
> average new ham just like the average new college graduates, is 
> abysmal. (I taught at the university while working toward a masters 
> in CS and was dismayed at the abysmal intelligence in those classes)?
>  I can't ignore the inability of many old timers to adapt to the new 
> technology either.

Well, we don't really disagree a lot there... The lack of knowledge and
the superficiality of many young people is pretty obvious - but when I
look around old hams, it becomes obvious that those are in average not
much more knowledgable! They just have acquired more experience, over
their long lives. It seems to be a fact of life that most people are
ignorant, just live each day as it comes, while the ones advancing the
art - any art - are very few.

In my country the vast majority of new hams are so-called
"emergencists". They joined ham radio for the sole purpose of providing
emergency communications in the event of a catastrophe that renders
every other communication system inoperable. They meet daily in
emergency training nets on various HF and VHF frequencies, and have
absolutely no idea about, nor any interest in electronics, homebrewing,
IMD, DXing, contesting, CW, satellites, EME, and so on.

But were the old hams all technically savvy? No! Those were mostly
ragchewers, while a few were DXers and contesters. In the 1980s there
were also a lot of "radioneeders" here, masquerading as hams: People who
needed radio (mostly VHF) for family or business communication, and
obtained ham licenses as the easiest way to legally use radios. Most of
these disappeared when cellphones came around. The most numerous group
remaining today are fishermen illegally using our bands. And those don't
even bother to obtain ham licenses.

> At my age, I'm now a appliance operator.?

I fight becoming one! My main transceiver is an appliance, a TS-450SAT.
But it's modified, with phase-lock to a GPS standard, improved filtering
(Collins mechanical filters), and so on. The power supply is a homebrew
40A switcher. The station and home is powered with "homebrew
electricity". And I'm working (or at least pretending so!) on a high
power, high efficiency SDR transceiver project.

> Those large displays are often necessary to display many important
> functions in advanced SS amps.

Like which?  And are they really necessary?

Some rigs are expensive and over priced, but some bring new technology
with extremely clean signals.?

Like which? I'm not aware of commercial ham equipment having extremely 
good IMD performance. I would like to know of any!

We technically interested hams should take the lead in developing a type 
of modern legal-limit amplifier, that overcomes the problems making them 
so expensive now. For example, instead of having one lowpass filter per 
band, which is the reason for much of the bulk and cost of amplifiers, 
we should develop a transmitter that can work with a single lowpass 
filter. How so? Simple: Make use of the UHF capabilities of modern 
transistors to modulate the entire HF signal onto an UHF carrier, as a 
PWM signal or some other digital code, then pass the UHF signal through 
a high efficiency, high power switching stage, and then pass the output 
through a simple, single lowpass filter that strips the UHF signal and 
passes everything HF and lower. The harmonics of this amplifier would 
fall on UHF frequencies and be filtered out easily.

We should definitely NOT keep copying the class AB push-pull linear amp 
circuits developed in the 1970s, with all their problems and 
shortcomings, as if there could be no other way in the world to generate 
clean HF signals.

Manfred


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