VSWR measuring, directional coupler

I don't know if it's common, but in contrast to Dave Tweed's answer, at least according to Wikipedia some SWR meters are built using directional couplers, just as you say.

Before we start, I'm not sure if you have a slight misconception about what's going on:

How are the forward and reflected waves separated out from the standing wave on the feeder?

The forward and reflected waves aren't distinct from the standing wave. The "standing wave" is just what we observe when we have a forward-travelling wave and a reverse-travelling wave superimposed on each other.

Now, for the benefit of future readers, what is the standing wave ratio or SWR? It's basically the ratio of the peak amplitude to the trough amplitude as you move a probe along a transmission line. It's given by

\$ VSWR = V_{max} / V_{min} \$

The max and min voltage amplitudes occur when the forward and reverse waves interfere with each other either constructively or destructively.

\$ V_{max} = V_f + V_r \$ and \$ V_{min} = V_f - V_r \$.

So

\$ VSWR = (V_f + V_r) / (V_f - V_r) \$,

or, to get closer to our usual formula for calculating the VSWR:

\$ VSWR = \frac{1 + V_r / V_f}{1 - V_r / V_f} \$.

So how can we extract the separate waves to compare them? The directional coupler is actually conceptually very simple. If we put two waveguides or microstrip lines (or PCB traces) carrying rf signals close enough to each other, some of the evanascent wave in the transverse direction (orthogonal to the line) will overlap with the propagating mode of the other line, and so some power will be transferred from the main signal to a parallel-travelling signal on the "coupled line".

(CC image by SpinningSpark at Wikipedia)

Unfortunately expressing this mathematically actually requires solving the 2-d transverse electromagnetic field problem for the transmission line structure, which can't even be done analytically for many important structures (like microstrip line).

Most of the power will simply continue to propagate in the same general direction it had been, so that naturally the forward-travelling wave is coupled to just one of the output ports, and the reverse-travelling wave is coupled to the other. Of course, there are also some reflections that happen at the discontinuities in this structure, so that the directional coupler can't have perfect directivity.

Also, if you look up directional couplers on Wikipedia, you'll see there are numerous other types beyond the simplest one I've shown here, and various tricks to enable the coupled-line coupler to operate over more than a small frequency band.

Taking your four points:

The diode is in its non-linear region. I ruled this out because the issue persists at high power and doesn't explain the high reflected voltage.

Your measurement at 5W looks like it suffers from diode non-linearity. At higher power, the SWR is consistent (but should be very near zero) so it looks like diode non-linearity is not an issue for these higher-power readings.

The issue is normal and I need to just calibrate for it in software. Though I've been unable to figure out how.

Have built more than a few, and all have satisfyingly-low SWR with no calibration. Your SWR results are quite far-off.

The hot glue I used that fills the center of the coils is lowering the Q of the coils significantly. This however, I don't think, would result in these symptoms.

You'll know if that hot-glue absorbs RF power if you plan to monitor more that a few hundred watts - it has a distinctive smell when it melts. Its melting point may indeed be a little low here (I've seen RF-sensing heads melt polyethylene coax under abnormal loads).

The two coils are coupling causing the high reflected voltage. However this didn't seem to be as big an issue in other people's circuits.

Possibly: Well-chosen ferrite should confine mag-fields. But those are big toroids, with a big area. Likely doesn't account for your large errors though.

When troubleshooting, a methodical approach breaks this circuit into its two components: voltage monitor and current monitor. Your Stockton circuit combines these two into a vector sum (having a very large error) that makes it difficult to see what's going on. Re-arrange the circuit to test each component on its own. I'll assume your RF signal source applies ten watts to a 50 ohm dummy load...

Voltage monitor

With 10 W into 50 ohms, RMS voltage into that 12:1 stepdown transformer should be 22.36 Vrms. Secondary voltage should be twelve times less: 1.86333 Vrms. The diode peak detector should yield nearly 2.635 Vdc. (not including diode losses). Test circuit at left:

^{simulate this circuit – Schematic created using CircuitLab}

Current Monitor

On the right, the current-monitoring transformer transforms R3 (51 ohms) down to 0.3541666 ohms on its primary side. With 0.4472 Arms flowing through 0.3541666 ohms, primary voltage of 0.15838 Vrms appears across its one turn. This voltage boosts 12 times to generate 1.9 Vrms across R3. This should yield 2.688 Vdc across C2, neglecting diode losses.

Note that this current-monitor DC voltage is very close to that of the voltage-monitor DC voltage. You could monitor these two DC voltages separately, but that would not properly yield SWR, because those DC voltages have no phase information. For SWR that gives a proper null only when the dummy load is 50 + j0, the RF voltage and RF current are combined before diode detection.

Once these separate voltage & current agree, proceed to try the combined circuit. If you still have problems, the problem is in the RF combining, not in your diode detector or transformers.