Measuring delivered RF power into an arbitrary impedance load

I want to measure the efficiency of a current mode class D (CMCD) RF amp when driving a resonant loop antenna (basically a series RLC circuit). I work with many different types of loop antennas, with frequencies from 50-150MHz, Q between 20 and 200, and resistance between 1 and 30 ohms. So far I've done this by using an RF current probe to measure the load current, and estimating the loop resistance by measurements with a VNA (either through measurement of Q and L to get R, or direct measurement of R), and just doing P=I^2*R. However, the impedance of the load changes significantly as modifications are done to the system, meaning, I have to re-characterize the system quite often, and even then I don't believe my results are very accurate. I was thinking that maybe it would be possible to create some sort of coupler which would give me a measurement of delivered power (meaning I*V, not just V or I) to my RLC load, whose accuracy is independent of the actual load impedance. The RF bandwidth is pretty narrow (less than 1MHz) though there is harmonic content that I'd like to reject. But it's important that the insertion of the coupler not significantly modify the impedance seen by the amplifier output.

Any ideas?

You can use a directional coupler, it measures forward and reflected wave. By nature, it has a finite transmission line length and a characteristic impedance. But if it's short related to wavelength, it won't cause large changes to the impedance matching, presumed the characteristic impedance is roughly in the same range as the source and load impedance.

Alterantively, you can measure instantaneous voltage and current at the interface terminals and calculate real and reactant power consumed by the load. This should be feasible in the intended frequency range.

yes you need high directional coupler. As you mention bandwidth is small so it possible to design it.
it has 4 ports one port connecting to input, 2nd port to load ( where 1 to 2 is through path), 3rd port is coupled port connected to detector and 4 th port is isolated port also connected to detector. you may use single detector provided maintaining 50 termination for the un connected ports of the coupler using switches.
coupled port gives the power delivered by the transistor. isolated port gives the reflected power. simple subtraction gives the power delivered to the load.

I should clarify that I'm not using any cable to connect the amplifier to the loop antenna, they are connected directly with just a few centimeters of copper tape. Like you say, very short transmission lines won't effect my load impedance, but at the same time how does one make a directional coupler with such a short line? Also the output of my amplifier is balanced, which makes most off the shelf couplers unusable.
In theory this is feasible, but getting the phase of the measurements to balance is a big challenge. Also my oscilloscopes can't do the math (difference between two voltage measurements multiplied by current measurement).

Directional couplers can be made small compared to the wavelength, as the devices used in the short wave band. They have a frequency proportional coupling factor for l < λ/10. The directivity can be optimized by adjusting the ratio of capacitive and inductive coupling.

Assuming symmetry, it would be sufficient to place the coupler at one side of the differential transmission line.

If you relate the directional coupler to the current probe mentioned in your initial post, it does nothing but adding respectively substracting a current proportional and a voltage proportional signal. So you may doubt if it can correctly measure the delivered power in case of strong impedance mismatch.