EMRFD Message Archive 156

Message Date From Subject
156 2006-09-15 08:01:35 Stan transformer phase coupling question
I should know this one, but I can not find a reference in any book
either.

Assume we have two windings on a coupling transformer between
stages, i.e. osc to buffer, etc. Does it make any difference on
the polarity of with winding to the connection to the base of the
buffer?

Should we want the collector of the osc to go positive at the same
time the base of the buffer goes positive, etc.?

Thanks, Stan
159 2006-09-21 13:41:40 Wes Hayward Re: transformer phase coupling question
Hi Stan and group,

The question at hand regards the polarity of a link that is on the
output of a transformer. Let's design an experiment: Start with
a signal generator and a hand full of small signal transistors; what
ever type is on hand will do. I tend to use the ubiquitous
2N3904, but any similar thing will do just fine. Build a small
signal amplifier with the transistors biased for perhaps 5 or 10
mA. Use some RF emitter degeneration to aid stability and keep the
gain down and put a toroid tuned circuit in the output. Use a
powdered iron part so it will have some reasonable Q with a
collector winding of 25 or so turns. Grab a floor sweeping
capacitor to resonate it somewhere in the HF spectrum. The signal
generator should drive the input through a suitable blocking
capacitor. Put a link of 3 or 4 turns on the output. Ground one
side of the link and use the other side as the output. Then,
terminate the link in 50 Ohms in the form of a piece of coax
running off to your spectrum analyzer or 50 Ohm terminated
oscilloscope. Measure the small signal gain. Make sure that the
output is well below the maximum possible so that small signal
conditions are retained.

The gain is measured by first running the signal generator directly
to the analyzer or scope, maintaining the 50 Ohm environment.
Carefully note the output from the amplifier. Now break the coax
path from generator to analyzer and insert the new amplifier.
Measure the output again, taking care to stay in a linear range.
The gain, formally the "insertion gain," or "transducer gain" is
then the ratio of the two outputs.

Measure gain with the opposite polarity on the link. It should
not be any different.

Now build a second amplifier stage, operating from the same power
supply. You can make this another tuned amplifier or it can be a
feedback amplifier configured for a high gain. Put this on the
output of the first amplifier. Measure the gain of the cascade,
again taking care that the output stage is not being over driven.
Having done this, reverse the polarity of the output link between
stages and measure the gain again.

All of this should produce the conclusion that the link polarity
makes no difference to gain. Go ahead and increase the drive from
the signal generator to push the overall amplifier toward gain
compression and again confirm that polarity makes no difference.

To be really cavalier about it all, add a third stage of gain and
repeat the experiment. By now with 3 stages the cascade should be
up to over 50 dB of transducer gain. The conclusions should not
change.

But they might!

If you are doing measurements with an oscilloscope, use two channels
with one driven by a 10X probe that is attached to the base or the
collector of the first stage. Trigger the `scope from this
channel. Run the 50 Ohm terminator on the other channel and run
the output coax to that load. This will allow you to examine the
phase of the signal to the load. You will see it change (flip) as
you change the link polarity.

We have assumed that care was taken to decouple the stages. Each
stage should have a resistor coming from the power supply with a
bypass capacitor at the other end of each resistor. The bypass
caps will all use short leads to guarantee minimal extra inductance.

Now let's purposefully get sloppy. Get rid of those decoupling
resistors. Use .01 uF capacitors for the bypassing, taking care
now to pick the worst parts you can fine. If you have some old
paper caps with very long leads, use them. Or take high quality
parts, but leave as much lead length as you can on them. Tie all
three stages to the same supply. Just to be sure that the supply
is not helping with any good bypassing, put an RF choke in the lead
from the supply.

Now repeat the polarity flipping measurements. You might now see
a change in gain. If you are lucky you may even see an
oscillation. (Hey, LUCK in this experiment is in seeing the
problems, for we are trying to create an unstable situation.) Be
sure to try the experiments with both small signals and large
drive. You might even play with an output stage that operates in
Class C.

So, the bottom line is that polarity should make no difference.
But if it does, there is probably something wrong with the overall
design. Polarity then becomes an experimental tool that we can
use to detect the problems. Or it might be the factor that we
need to build an oscillator.

It's all about doing the experiments.

73, Wes
w7zoi