[Amps] Linearization problem

[Manfred Mornhinweg]

Peter,

> I use a bird sampler
> 
> http://www.birdrf.com/Products/Components/Variable-Signal-Samplers/4273-Seri 
> es_15-35-MHz-Variable-RF-Signal-Samplers.aspx

That one is non-directional. It's a plain simple sampler, that will produce an
output attenuated a certain amount from whatever is on the line. It will give
absolutely no rejection of the distortion introduced by the amplifier stage into
the drive signal!

You can do exactly as well by connecting a resistive or capacitive divider from
the drive line.


John,

> Maybe a band specific, power compliant isolator? 

That would be great, but the only idea I have about how to build isolators for 
HF results in something very impractical!

> A directional coupler could be used to detect the output of the
> transceiver before the isolator.

If an isolator is used, no directional coupler is needed. A simple divider 
connected to the line, between the driving transceiver and the isolator, will do 
fine.

The problem is just how to make that isolator!


Ron,

> Simple; Dougherty amplifier.  Two tubes one class "C" (positively modulated)
> the other class "A" (negatively modulated) many variations exist.

That has nothing to do at all with what I'm trying to do! A Dougherty amplifier 
is relatively complex, and it's hard to splice the two amplifiers together in a 
way that will result in low distortion. And the resulting efficiency isn't so 
great as to warrant all that complication. The approach I'm following is far 
simpler!

About amplifier classes, okay, let's then agree that some people call an 
amplifier "class B" when it is in fact class AB, but very close to B. In any 
case, when I talk about a solid state class AB amplifier, I mean one having just 
a small idling current. Just enough to get reasonable linearity in the 
cross-over region. Unlike tetrodes, which are often run with a huge idling 
current to linearize the mid-level amplitude region too, in solid state amps the 
mid region is more linear to start with, and the final linearization is done 
with negative feedback.


Gerald,

 > In the microwave world one would follow the coupler with a circulator and
 > the energy reflected from the amplifier would dump into a dummy load.

Yes, that would be the right thing to do at HF too, if there just was a 
circulator for HF!

 > Is there a circulator design made for HF?

  At HF it's surely simpler and cheaper to use several active buffer stages, 
than a circulator.I have never seen one, at least not one having a size and cost 
that makes sense. But I would love to get by without those buffer stages!


Bill,

 > Instead of driving the amp with RF, I would suggest doing it digitally.

In my opinion, that's certainly the way to go, for any RF power amplifier built 
into a modern, DSP-based radio. I think that the HF radios we should be using 
are basically a DSP box, with a full-spectrum, wide dynamic range A/D converter 
at the antenna input, and a 1.5kW, switching RF output stage, directly driven by 
properly predistorted digital phase and amplitude signals. The conventional 
concept of a big, bulky, heavy add-on amplifier, to provide the final 12dB of 
gain, looks so lame in comparison!

But the fact is that there are millions of conventional 100W HF transceivers out 
there, and most hams want to keep using them. So we still need those ugly, big, 
bulky, heavy, expensive add-on legal limit boxes. And what I'm trying to do is 
coming up with a way to make them significantly smaller, lighter, cheaper, more 
efficient, while at least maintaining decent IMD performance, or ideally 
improving it.

And for an add-on amplifier, digital driving just isn't very practical, because 
it means burning up the original analog 100W drive signal in a dummy load, 
dissecting a sample of it into its phase and amplitude components, then turn 
them into digital signals, put them into a DSP, and then built the whole 
digitally driven amplifier! It's hugely more complex than most homebrewers would 
want.


Steve,

 > > And a bonus question: Do you think that the phase distortion in
 > > such a class-A-AB-C-E amplifier will be bad enough to still
 > > cause poor IMD performance, even while the envelope linearity
 > > is excellent?
 > >

 > Yes. Around 1980 I prototyped an amplifier aimed at improving
 > efficiency in a military hf manpack. The output stage was a
 > wideband push pull affair which behaved like class B at low drive
 > levels, moving towards class D at full output. A 'polar loop'
 > (Petrovic and Gosling, Bath University) applied amplitude feedback
 > via the PA bias supply and phase feedback via the synth PLL. With
 > either not connected, the output spectrum was very wide -

Thanks for that info, even if it's bad news to me! But I suspected it.

Several years ago, while experimenting with pure EER and with combined class 
AB/F EER amplifiers, I ran into severe phase distortion problems. I finally gave 
up that line of development.

My present hope was that by keeping the supply voltage constant, and modulating 
the gain just through the bias, it might be possible to get a phase distortion 
low enough to live with it. But your experience tells otherwise.

So, it looks like I should better forget this whole stuff.

 > as best
 > I remember, both loop bandwidths needed to be some 10's kHz to get
 > acceptable results.

Correction loops will always correct the distortion products that fall within 
the bandwidth of the correction loop, and they will _worsen_ and __create_ 
distortion products outside their bandwidth! So, the correct way to use 
correction loops is to make the correction loop's bandwidth larger than that of 
the amplifier's output bandpass filter. In ham practice, this forces the use of 
tuned amplifiers, and with pretty high Q networks.

Now I have to think if I can come up with a reasonably simply phase modulator in 
the drive signal path, that would allow implementing a phase correction loop in 
addition to the bias-controlling amplitude correction loop.

It all is starting to look unreasonably complex. I mean, just to get 12dB more 
TX signal...


Last item in this post, for all those who suggested directional couplers to get 
my envelope signal: I did the theoretical analysis this morning. The result is 
that a directional coupler could indeed be used to provide a correct, 
undistorted envelope signal from the drive signal, even while the amplifier 
stage introduces a varying load and thus some distortion, but _only_ if the 
driving radio behaves like a perfect voltage source with a fixed series 
resistance! This is how signal sources are normally modeled, but unfortunately 
HF radios don't behave like that, at all! In practice, when the amplifier puts a 
  non-constant load on the drive signal, the radio's output will vary in a way 
that is not how a perfect signal source with fixed internal resistance would do. 
As a result, the drive signal would get distorted, at the output, input, and 
side ports of the directional coupler!

Sure, an attenuator would reduce the distortion, at the cost of some drive 
power, but only by a few dB.

Do you want to see the effect yourself? Take any good SWR meter. They are 
basically directional couplers with built-in detectors. Connect the SWR meter 
between your radio and an antenna tuner. Load the tuner with a dummy load. 
Adjust for 1:1 SWR. Then set the radio for a safe output power, say 10W, and now 
detune the tuner while watching the forward power indication. It will vary 
strongly! This indicates that the radio is not performing as a true 50 ohm 
signal source.

Putting any kind of directional coupler in place of that SWR meter, and my 
non-linear amp stage in place of that tuner and dummy load, would be the same 
situation. As my amp varies its input impedance depending on drive and bias, the 
sampled envelope would become untrue.

Okay. Project shelved. It was a nice dream.

Manfred


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