The directional coupler used here was patented by Sontheimer and Fredrick in 1969 (see Fig.10). It also appeared in an article by William Spaulding, titled, "A Broadband Two-Port S-Parameter Test Set," in the November, 1984 issue of HP Journal. David Stockton, G4ZNQ, popularized this device among radio amateurs in his Winter 1989/90, SPRAT article, "A Bidirectional Inline Wattmeter." Wes Hayward, W7ZOI, presents a particularly clear explanation of the operation of this directional coupler on pages 156 to 158 of his book, Introduction to Radio Frequency Design.
"The ferrite binocular...can be used to realize both the T1 and T2 transformers on a single twin hole ferrite core. Low frequency response is dictated by the ferrite material characteristics. High frequency response is partially governed by total wire length, since the core effects are no longer dominant near the high frequency end. Interwinding capacitance, leakage inductance, copper losses and transformer coupling below unity (k <1) also degrade high-end performance. Small shunt capacitances to ground at the coupler ports can be used to improve match and directivity at the expense of bandwidth. At higher frequencies, lead length must be kept to a minimum to limit parasitic inductance. To achieve broadband performance, ground connection lengths must be minimized." Michael G. Ellis, RF Directional Couplers
The above quote is taken from an excellent, on-line tutorial by Michael G. Ellis, which may be found here. Mr. Ellis has also written a handy design program for directional couplers; which he has kindly made available here. Click-on "RF Coupler" for program details and click just to the right to download the program.
The turns-ratio of the two identical transformers is set by the desired coupling power ratio (the fraction of sampled power expressed in decibels). Given that we're using a one-turn primary winding, the coupling ratio is determined by 1/(N^2); where "N" is the number of turns on each of the two identical secondary windings. In a higher power meter one might choose to use a coupling ratio of -30dB (or less); where only one part in one-thousand of the input power would be dissipated in the power meter's load resistors. For this reason, a -20dB coupler will be more practical at very low power levels. A power ratio of -20dB is equivalent to 1/100; thus the number of turns required on each of the two identical secondaries is set at 10.
We need two secondary windings; each comprised of 10 turns; and each of which ought to produce a reactance of at least 10uH (XL of 10uH at 3.5MHz is ~ 4 x 50 Ohms). Sorting through my junk box, I turned-up a binocular core that I'd previously salvaged from a derelict Rhode & Schwartz television demodulator. I wound 10 turns on a side-leg (count the number of passes through the inside of the core) and measured 23uH. Good enough!
Elecraft's W1 power meter uses an identically-wound
binocular core. Please refer to page 6 of their construction
manual for the winding details. Here's a photo taken just after I'd
installed my coupler ("click-on" any of these photos for an enlarged
The two 51 Ohm load resistors, R1 and R2 (shown resting on the
copper-clad board in the above photo), ought to be closely matched and
My detector is based on a design by Wes Hayward, W7ZOI, that
appeared in the Spring, 1984, edition of SPRAT, titled;
"A Sensitive Microwatt Wattmeter." Wes observed that a sensitive
RF voltmeter may be made from ordinary Silicon signal diodes; provided
they are operated with a slight forward-bias. I matched the diodes by
measuring the voltage-drop across them while they were forward-biased
by a 9volt source in series with a 1MegOhm resistor. The collection of
1N4148's yielded by my junk box produced readings clustered around
0.38volts. The three diodes that I selected not only matched within a
resolution of 1mV on my DVM; but I also took care that all three diodes
bore identical markings of manufacture. Also, take care not to touch
the diode bodies during the matching process; as their forward
voltage-drop is a sensitive function of temperature.
The forward-biased diode detectors produce a square-law detection that goes a long way towards linearizing the resulting meter scale. I chose simply to list the calibration points on a sheet of cardboard. I calibrated my meter using an Agilent E4418B power meter owned by my employer.
The overall power consumption is low enough that most any wall-wart
producing 9 to 12vdc will suffice to drive it.