Mike, the current mode DC-DC converters DO NOT CARE THAT THE LOAD IS A RESISTOR OR A LED (sorry but i'm having a tough time getting my point across). Because the way the circuit works is by sensing a voltage drop across a very small resistor, the circuit doesn't care about whether it is driving a led or a resistor . As long as it senses the right drop across the resistor (Ohms law, V=IR) it will servo the loop to drive the error to zero, in this case the error is the difference between the reference in the converter and the drop across the sense resistor, typically a very low value resistor on the order of 100 milli ohms or so. Now what is critical here is that the voltage that is sensed is "smooth" and that's why monolithic DC-DC converters have an output filter network to smooth out the voltage that ultimately goes into the sense feedback circuit. This makes the converter operate *independently" of the type of load presented to it.
I did not measure using a LED because, as you pointed out, the LED is non-linear. Because of this non linearity I would have to:
1.) Infer the current through the LED by measuring the voltage. This is not accurate since you need the I-V characteristic of THAT PARTICULAR LED under test. I do not have the test equipment to characterize this parameter. Inferring current this way is very prone to error in your set up and measurement.
2.) I would have to use an ammeter in line with the LED. An ammeter also has a small series resistor so this would present a measurement error as the converter would have to compensate for the resistance of the ammeter. This can be a substantial error depending on the series resistance of the ammeter.
If you really want to do the measurement using a LED, I would use method #1, being very careful about setup error. In the old days, transistors I-V curves were characterized using a curve tracer. You need to use one of these curve tracers to characterize the LED's I-V curve. Also, since the curve is very steep once you turn on the diode, you need to have millivolt ABSOLUTE (not relative) accuracy in measuring the quiescent operating point of the diode in the circuit. This also presents another problem since the diode is driven by a switcher with tens of millivolts ripple. How are you gonna make the measurement accurately? Remember, a few millivolts change on the Vf of the diode amounts to an order of magnitude change in current through it. Are you getting the picture now? To get an accurate measurement using a LED is a lot more complicated after you think about all the gotchas that you need to account for. This is not worth the hassle when a more direct way (good 'ol ohms law) works with a lot less error. Even if I had a curve tracer, I would still use a resistor since I have the knowledge to understand that the assumptions I'm making are valid and that this is the elegant and accurate way to measure efficiency.
