Amps
[Top] [All Lists]

Re: [Amps] SGC-Mini Lini

To: amps@contesting.com
Subject: Re: [Amps] SGC-Mini Lini
From: Manfred Mornhinweg <manfred@ludens.cl>
Date: Wed, 12 Aug 2009 01:55:15 +0000
List-post: <amps@contesting.com">mailto:amps@contesting.com>
When I joined this list a few years ago, it was in search of information 
and perhaps help regarding an envelope-restored amp I was trying to 
develop. It's funny that only now I learn about the Mini Lini of 2005!

I shelved my project several months after joining this list, even though 
I still have the prototype in a box somewhere. I ran into a stupid 
little problem I couldn't solve, and I have a feeling that SGC ran into 
the same problem, and that this was why they never sold (or even got FCC 
approval) for the Mini Lini.

In my prototype, I was using a linear modulator (accepting the 
inefficiency), as an interim step while testing the rest of the concept, 
and so the differential delay of phase and amplitude wasn't such a big 
problem. Nevertheless, I obtained a lousy IMD performance. The reason 
was an often overlooked characteristic of MOSFETs: The reverse transfer 
capacitance rises very much when the drain voltage lowers. With bipolar 
transistors much the same happens. And this creates a big problem: Since 
class E (or C) amps need to be driven into saturation, there is a 
relatively large drive power available, and when the drain voltage goes 
low, the capacitance goes high, the power capacitively coupled through 
the transistor wins over the amplified power! And since the two are in 
opposite phase, a funny looking distortion happens: Instead of the clean 
straight zero crossings on the envelope that are normal for a clean two 
tone signal, there are now TWO zero crossings, with a small sine between 
them, in which the RF has reversed phase. Spectrally, this manifests 
itself as very strong IMD, and when listening to the signal from this 
amp, it is audibly distorted.

Using transistors, I see no good way around this. Only applying some 
predistortion, so that when the original amplitude reaches zero, the 
drain voltage sreaches the level where coupled-through and amplified RF 
just cancel out. But this is frequency sensitive, load-sensitive, 
tuning-sensitive, and even heat-sensitive! It's a nightmare.
Perhaps someone cleverer than I am can see a better solution. If not, I 
will have to clever-up and invent some adaptive-corrective feedback system.

A Brazilian ham published an article in QEX some years ago, about such 
an amplifier, but he used tubes, which are a lot easier in this regard. 
Even so, he also must have had problems in the small-signal area, 
because he ran the amp in envelope restored class E only from a certain 
amplitude level upwards, and below that he froze the supply voltage and 
let the amp run in conventional linear mode. I don't trust the IMD can 
have been very brilliant that way.

So, I definitely think that the connector issue cited by SGC as the 
reason for indefinitely delaying the Mini Lini was just an excuse. The 
real problem lies deeper inside.

And to compensate for the delay of the switching modulator, well, let's 
just use a roll of coax cable as a delay line for the RF drive! With a 
250kHz switching frequency and all the filters and delays there, just a 
few kilometers of cable would be needed! There goes the "mini"!


So, my project of a high efficiency small solid state legal limit linear 
amp has reverted to the ages old class D concept. I just need some 
MOSFETs rated at several hundred volts, able to take 10 amperes or so, 
that can switch in about one nanosecond. Hey, semiconductor 
manufacturers, what are you waiting for?

Oh, well, I think I could settle for 5 nanoseconds! Just would have to 
drop 6 meter coverage, and accept some more spurs! ;-)

Manfred the optimist.

========================
Visit my hobby homepage!
http://ludens.cl
========================
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps

<Prev in Thread] Current Thread [Next in Thread>