On Fri, 08 Oct 2004 07:08:10 -0400, Dave Hall
wrote in :

On Thu, 07 Oct 2004 13:58:13 -0700, Frank Gilliland
wrote:

On Thu, 07 Oct 2004 07:06:26 -0400, Dave Hall
wrote in :

On Wed, 06 Oct 2004 11:47:29 -0700, Frank Gilliland
wrote:

snip
Compression -- a term usually applied to audio conditioning where the
amplification of a signal is varied inversely to it's input level. One
of the most common types of audio compressor is called "constant
volume amplifier".

That is but one type of compression. For a more detailed discussion of
RF amplifier compression, please refer to the following related link:

http://cp.literature.agilent.com/lit...5965-7710E.pdf

Generally speaking, amplifier compression occurs when the inpututput
power ratio no longer increases db for db.


You should know me by now -- I just -have- to disagree.

That's what makes discussion interesting. Unless, of course the other
party is nameless sociopath..... ;-)


Which one? There are so many.....


The use of the
term 'compression region' is really a misnomer. It's properly
described as a nonlinear region.

A device can be non-linear without going into compression. The way I
learned it, compression is a specific term that applies when a
normally linear device loses linearity as it approaches it's maximum
drive level.


The problem is the application; i.e, the operational limits of the
device. More below.


The reason is that different devices
behave differently, and while a few devices (some transistors, a few
tubes, incandescent lightbulbs) have nonlinear characteristics similar
to compression, the fact is that nothing is being 'compressed', and
the vast majority of other devices have nonlinear regions that do not
resemble compression.

Exactly! That's why compression refers to a specific condition, so as
to differentiate it from other forms of non-linerarity. Certain class
"C" "Modulator" type amplifiers, for example, deliberately run in the
non-linear region just above device cutoff, to take advantage of the
"swing" effect of the device.


......huh? That statement looks like a train-wreck. Assuming we're
talking Class C for both types, a device is usually pushed towards
-saturation-, and the result is clipping which forms a psuedo-square
wave that packs more energy than a sine wave, and making the stage
more efficient (a design that evolved into the Class E amplifier).
Also, Class C devices are -biased- above cutoff by definition. But
what is this "swing effect" you describe? I have never heard of a
-device- having such an effect unless it was a self-resonant device.
Last time I checked, most transistors don't tend to be self-resonant
below 30 MHz.....


It's non-linear, but it's not
compressed.


That's the point. It's -not- compressed. I see no reason to place that
label on a portion of a curve only because it loosly resembles one of
the characteristics of compression yet doesn't fall within its
definition. To me it's like using the word 'pill' to describe an RF
transistor, or 'swing' to describe modulation.


And whether that nonlinear region is gradual or
sharp, it's still clipping because it's a limitation of the device. It
also causes distortion that is characteristic of clipping. And yes,
saturation is within the nonlinear region.

I differentiate the term "clipping" from "compression" by the
application in which the terms are used. In an application such as a
broadband amplifier with a broad spectrum of carriers (Such as a CATV
amp), is applied to the input, the term compression applies as you
reach the point of non-linear gain and the incidents of composite
second order beats rises disproportionately.

I look at "clipping" as a momentary chopping off of an otherwise
linear signal reproduction, such as what you would encounter with AM
modulation peaks which run out of headroom in an amplifier, which
would otherwise be still in the linear range when unmodulated. You can
still "clip" while not being fully into compression.

Maybe I'm over-analyzing these terms, but that's me.....


I see the image in your brain -- "clipping" suggests a straight-line
cut from the top of the wave. And I understand your definition of
'compression' as the transitional zone between linearity and clipping.
But clipping, by definition, is caused when the signal exceeds the
limits of the device. And that's exactly what happens when the output
is pushed into the nonlinear region because it's operation is intended
to be limited to the linear region (otherwise it's not a linear amp).


I'll give him this one. This may be a matter of semantics, like noting
the difference between 'weight' and 'mass'. I had these things drilled
into my brain by my profs for the simple reason that using the wrong
term can cause a misunderstanding. And that's what I believe is
happening when people substitute the term 'compression' for when the
transistor has exceeded the limits of linearity and subsequently
clips, be it hard or soft.

But no matter what you call it, the result is distortion.

No argument there.


But the problem here is that the term 'compression' has been adopted
by voodoo techs as a euphamism for 'clipping', making it sound as if
the distortion-causing effect is not only benign, but sometimes
preferred. It is neither.

After reading the link, you might want to revise your definition.
There are a bunch of very talented engineers here who would be a bit
insulted to find out that you refer to them as "voodoo techs".


I'll stick with my definition. And while I may take issue with some of
the more liberal definitions used by engineers these days, I should
point out that I worked for HP (Agilent) several years ago and I'm not
a big fan of their engineering department.

While the link I provided, was put out by Agilent, it was one of many
examples. I don't work for them either, but I do work with some
talented RF engineers, and we use the term compression frequently to
describe the point where linear gain ceases, and distortion products
increase disproportionately.


I've known many talented engineers who learned switch terminology
backwards: for example, they would describe a 6-position rotory switch
as a 6-pole single-throw. I've known some talented engineers who
didn't know the difference between 'active' and 'passive'. I've even
known a couple talented engineers who thought electrons flowed from
positive to negative. I guess I'm just anal (like THAT'S a suprise!).


But you are right about one thing. a -33dbc harmonic rating from a
single carrier signal is pretty poor. Perhaps a chebychev lowpass
filter on the output will fix it up.....


That would be nothing more than a kludge. The fault is in the design.
The response isn't even close to linear. That may be due to the bias
class, the bias regulator, the choice of active device, or just crappy
engineering overall. I suspect it's a little of everything.

Usually, if it is a push-pull design, device matching plays an
important part. Bias is also important, as is impedance matching. But
even a "good" design should have a follow-up low pass filter to
minimize any harmonic content.


Absolutely. But even filters have limitations. Assuming Brian's amp
has 350 watt noise figures in the neighborhood of -18 to -24 dB (not
unrealistic), it would take a mighty stout filter to clean it up!

A 5 pole filter should bring it into line. All he would need is an
additional 30db of attenuation


Don't forget that as you increase the power you need to increase the
attenuation. This is because of the changes in the equipment rules
regarding harmonic emissions. The limit is now absolute whereas before
it was relative to signal strength. Regardless, a LP filter won't
filter the splatter.