Design and build

The main design concept is to use a tandem match directional coupler, and then use highly sensitive AD8307 log amplifier to measure the forward and the reflected power. Alan Wolke, W2AEW, has an explanation of how this coupler works. The AD8307 log amplifier I have used also as part of my stand-alone lab power meter (still need to fix some software bugs).

Schematic of coupler.

The schematics of the coupler is not complicated, but I only feel I understand it fully in short glimpses. The core idea is to sense both current and voltage in the line from A to B, and that can be measured over C and D to give us the forward and the reflected power. The full design then use 50 Ohm terminations and an AD8307 each at C and D for the RF measurements. The two transformers are \(1:N\) turns (with the one turn being a coax through the center of ring ferrite cores that have the \(N\) windings). There is also an electrostatic shield connected to ground.

I built it up on one of the board I had ordered. Everything looks very good on the nano VNA - fully terminated, the return loss at the transmitter port is 35 dB or more.

Picture of coupler and VNA.

Measurements

For measurements, I used the TinySA spectrum analyzer. I put -10dBm into the transmitter port at 50MHz, varied the load impedance, and measured the power in the forward and the reflected port (taking care that the measurement port that wasn’t connected was terminated with 50 Ohm).

Here are the measurements, Power is measured in dBm and the calculated return loss (RL) in dB:

Note that R presents this with a bit more apparent precision that is called for.

To the results:

1. With proper termination, we measure 22 dB down in the forward direction. Since 1/12 in voltage correspond to \(-20 \log_{10}(12)=-21.6\) dB, that is reasonable. A return loss of 24 is quite good, corresponding to a SWR of about 1.1.
2. With an open load, all power is reflected, as expected.
3. The short was improvised with a pair of tweezer in the BNC jack. At 50MHz probably imperfect. So a tiny bit of return loss.
4. With two 50 Ohm loads in parallel (on a T adapter), the return loss was 11 dB. and the short (implemented with a pair of tweezers) was probably not a good RF short. Theoretically, the return loss should be \(-20 \log_{10} \Gamma\), with \(\Gamma\) being the reflection coefficient. With a purely resistive load of 25 Ohm, \(\Gamma = |25 - 50|/|25 + 50| = 1/3\), and the calculated return loss should be 9.5. Instead I measure more than one dB more. But with the return loss being the difference between two imprecise measurements, I don’t think this is a something to worry about.

I also kept the input constant at -10 dBm and varied the frequency from 1 to 90 MHz:

Respecting the accuracy of the measurements, it seems that the the coupler behaves well up to 60-70 MHz, and the gross deviations from linearity only happens above 80 MHz. I am quite happy about this.

Design considerations going forward

The important question is how turns should be wound on the transformers. DJ0ABR design with 24 turns, for a 2kW capacity. I have no planned need to measure that large amount of power, and there are two costs to having too many turns: First, there is limited dynamic range in the measurement system. Second, more turns will degrade performance of the transformers at higher frequencies.

One place to start is with the capacity of the AD8307. It is specified to measure up to 17 dBm and down to -75 dBm. I would also like to have accurate measure of return loss with QRP rigs running perhaps only 1W. I try to calculate some scenarios.

First, let’s vary the number of turns from 10 to 30. Then, I would like to measure the reflected power for a good match with a return loss of 25 dB. I calculate what forward and reflected powers I’ll need to have the sensor read for different powers from the transmitter and see how this range fits within the dynamic range of AD8307.

Note that there is also a voltage divider in front of the AD8307 in DJ0ABR’s design. This also involves a coupler of DC blocker capacitors, but these are large and basically shorts at RF. The internal impedance of AD8307 is, according to the datasheet, about 1k1 Ohm. In the design, this is paralleled with an 180R resistor, giving a parallel resistance of 155. This is the lower leg of a divider with 470R in the upper leg, so the voltage division is by a factor of around 4, or -12.1 dB.

I don’t completely understand how DJ0ABR can use this design to measure up to 2 kW with only 24 turns. Maybe the resistor network in front of the chip could be adjusted, or maybe a modification is called for.

If I were to change anything about the design, I think that I also would like to create a bit more space for the transformers. The 1-turn of coax through the ferrite cores makes it difficult to solder the very short piece of the shield without melting the coax. Perhaps it is my Ultraflex coax that doesn’t take heat well, but putting things together would be easier with more space. Maybe that would compromise the coupler at VHF. But if I could run with fewer turns on the transformer, I think that would make more of a difference at 50 (and maybe 70) Mhz.