The Roll Your Own Split Gilbert Cell thing....
I have improved the 2N2222 phase mixer circuit and to my surprise
the FTT analysis of it shows that the 2cd, 3rd, and 5th harmonics of
which are the highest, are all down - 87 dB at a 10 MHz test frequency
fully modulated. Imagine a better transistor choice then? Not bad
though with these. And waveforms are all looking good.
Also, the Kelvin energy noise and low frequency as well as phase
noise in the circuit appear to down to -225 dB? Humh? I say appear,
providing I understand this "voltage dependent current source" test
devise set up in the LTspice circuit correctly. When clicked on, up
comes a plot of the noise in dB and Kelvins. This is with the outputs
of the circuit coupled to the devise for testing. Coupled by means of
a labeled output (spice directive). Which was a devise in the MC1496
test circuit set up supplied to me by Ray Anderson. Sometimes it
shows Kelvins and then sometimes it appears to show phase. All
relative to dB's. We will learn more about using this. Yet, if this
is the noise and ambient thermal energy levels then that is
interesting in itself. dB wise. {I am still learning the software here.}
The two transistors stacked on top of each other as the mixer cells
in the two halves of the mixers now have a compensation network idea I
borrowed from similar circuits like this used in audio amplifiers. It
works both as a feed back compensator and ratio balance network to
maintain balance between the two series mixer transistors and aids in
the mixing of the products. It shares "half the load current" and
forms a loop with the load absorbing some of the reflected waves from
the load. The later should be good on swr reflections from the load.
Well thats my added view on the reflected wave aspect in the loading
circuit. Spelling out a load compensation effect likewise. Apart
from the other attributes of this little added network, I do believe
it will aid with mismatch conditions if used over a wide bandwidth as
compared to use at a single frequency.
The little added network seems to also aid in the harmonic suppression.
So the idea borrowed from audio amps similar to this circuit seems to
have several attributes to it. But mostly it is a passive feed back
compensation and mixer balancing network in its intended use. And
offers some loading compensation.
I have been pushing the performance more within the 7 to 30 MHz range.
But with an adjustment in input levels for the mixer balance ratio,
you can use it down on 1 MHz or 3 MHz. And with audio as the I/Q
inputs for a transmitter anywheres in its spectrum.
It mixes I/Q phased audio with a carrier or demodulates rf inputs. So
it is used either way you want.
Them kind of figures for a circuit made of 2N222 type transistors is
surprising to me. I also figured out that if you use resistors with
a negative temperature coefficient in the collector load resistors and
positive coefficient in the upper bias voltage divider resistance you
can compensate the thermal change in the bias adjustment and keep the
circuit right on target if used in a full service mode as a phase
mixer for a broadcast transmitter. Not often however do you see
resistors with their +/_ temperature coefficients listed but sometimes
you do.
Better transistors can only help to make it better or else raise up
its output gain. Oh, I reduced the load impedance way down now so it
will work at output looking into more types of circuits.
As I mentioned in a earlier post, you can use dual gate mosfets in the
I/Q inputs so that you can control the I/Q modulation from software
with alc from the transmitter. The extra gate will be for the alc.
So that can kind of advances the circuit along them routes if needed.
It seems that this half baked idea is turning out to be an interesting
circuit for future two way communications uses. Just keep in mind
that it has part of Gilbert Cell hidden in it. Well its not really
Gilbert Cell anymore but we know where the idea comes from.
* I got the input impedance measurement in LTspice better figured with
using a rf generator set for the input amplitude but with no internal
series resistance. When clicked you can read the current and then do
the math on the rate of change in voltage divided by current at input.
And when the calculated Z is entered into the rf generator, the
output will not change. So that seems to be the most precise way to
measure the Z at input. And the old school way. So you should know
that this is the method I am using now. Some Internet articles on the
matter had me confused on the method used in LTspice.
Dan
