RF phase shift network options for the R2/T2 are many and varied. Here is my favorite, and an elegant setup to bandswitch a group of them..
It has a wide bandwidth, small number of components, takes care of splitting AND quadrature in one fell swoop. And, you can bandswitch by just opening and closing the A, B, and C ports onto common busses, making bandswitching relatively easy.
(In-phase is the +45 port, quadrature or -90 degrees is the -45 port.)
Assuming all ports are at 50 ohms,
VARIATION 1, as detailed in the schematic:
L1: bifilar twisted pair on toroid, reactance 50 ohms at center frequency.
C1, C2: 100 ohms reactance each at center frequency
Port D: terminate in a 50 ohm resistor.
You obviously have to calculate your values from the reactance/frequency formulas found in your ARRL Handbook or other references.
VARIATION 2 - above, but eliminate C2, and make your C1 twice the value (i.e., 50 ohms reactance at the chosen frequency). This is what I'm using lately to save variable capacitors!
This gives you 90 degree phase shift over a broad range. However, the amplitude ratio between the two outputs will vary as you depart from the design frequency. No big deal - just tweak your amplitude balance pot on the R2.
In addition, for a perfect null, I make the capacitance adjustable with tiny 1/4 inch trimmer caps I found at a hamfest.
Now, in the real world, not everything is 50 ohms; and for various reasons, the audio phase shift networks in the R2 and T2 are not exactly 90 degrees, either. So things may depart from the theoretical. With my setup, I need much more capacitance, about 2-4 times the theoretical value. When I find out why, I'll let you know!
Why? Some hints as I pursue perfection... June 2001
I've had Glen
Leinweber's R2a writings for some time, waiting for the right time
to implement the ideas or pieces of them. He makes his audio phase shift
network tweakable, and provides a simple
schematic of an accurate audio quadrature generator to adjust it right
on the nose.
In addition, I started playing with PSpice (as provided in the free OrCad Lite v. 9.2 from Cadence Design Systems - see my simulation primer), modelling everything in sight to find out the whys and wherefores of how things work, what happens if I tweak "widget X", etc. In modelling, I found (with perfect parts) that:
I built up his audio quadrature generator, in preparation for building the R2a phasing network... and thought maybe I should check my current implementation of the original R2 network. Sure enough - over the communications audio band, there was quite a bit of audible difference in rejection, sometimes quite good, sometimes not so good. And, of course, the typical CW listening tone of 750 Hz was one of those not-so-perfect points!
So, clearly, I had been working hard to get a perfect null at that frequency, tweaking networks, adding capacitors, switching coil windings, and on and on... little knowing that my pursuit of the perfect null led to compromises in the total picture - like good null over the audio bandpass, and closer-to-perfect nulling over a wider RF tuning range.
With this discovery, combining computer and shop experimentation, I'm ready to pursue perfection. Stay tuned...
Meanwhile, here's something that works fairly well...
Components are a mix of surface-mount and leaded components. Don't shy
away from surface mount - with a little care, a good magnifier, and some
silver-bearing solder (even Radio Shack has the silver solder now), it's
a compact and flexible way to do homebrewing. Paralleling components to
improvise the right value is a matter of just stacking your chips! (And
I don't mean at the casino...) A hamfest purchase of a whole reel of 3.9k
resistors, plus little packs of 500 chip capacitors (180 pF and .1 uF),
a kit of assorted NP0 chip caps from DigiKey, and miscellaneous other hamfest
purchases make it possible to come up with the necessary values.
The LO input goes through an attenuating pot, then a 2N5109 amplifier
stage (not shown on schematic), to provide a uniform source impedance and
adequate drive power. This feeds the center microstrip.
The networks do their thing, then send their signals to the outer microstrips. A single PIN diode in each of the A, B, and C legs of each network makes or breaks the connections to the microstrips. I use the MPN3700, but the lower-voltage MPN3404 would do fine. (These are former Motorola parts, now made by ON Semiconductor, and available from Allied Electronics). Forward bias on the diodes (about 6 mA each) switches them on, and isolation is good enough for this application (about 40 dB at HF) by simply removing the bias. (Reverse bias would increase the isolation, but is probably overkill except for more critical applications like front-end filter switching.)
I did, just for fun, try garden-variety 1N4148 diodes in one network. With the values above, there was definitely an audible loss, so I didn't research it further. My guess is they'd need a lot more forward current to turn them on better. Even if that helped when "on," I don't know if the isolation would be adequate when "off." It's probably worth shelling out your shekels for the real thing.
The two output striplines feed the I and Q RF inputs on the R2. I tried two CMOS buffers after the phasing networks (74HC04 or 74AC04 - see this link for details). I heard various noises suggesting instability, however, as I tuned up the phasing networks before them, and some noise specific to some tuning frequencies, so I abaondoned this attempt at a nicety. Two short pieces of .141-inch semi-rigid coax are soldered to the bottom of the board, their center conductors terminating at the microstrip outputs. They connect to the R2 board via SMA connectors, making a nice rigid and shielded mounting for the board.
Drawings later... Bye for now.
