So far, as very well suggested above, sometimes “EFFICIENCY” is not a top priority. What is most important is indeed the effective power that a certain AC/DC converter can draw from the AC side (generator) of a system and deliver to the DC side (load) at a certain speed.
To make it clear once again: “if one device gets an input of 3.0 W at 20 km/h and has an efficiency of 80% and the other device gets a input of 3.5 W at 20 km/h and has an efficiency of 75%, than the second device is the better device, in spite of its lower efficiency”.
To make evaluations on how good an AC/DC is, I personally do not even need very sophisticated laboratory instruments since I make all relevant measurements directly on DC side.
To do that I use a real generator, a real AC/DC converter (the one under test) and an adjustable ohmic load made-up of calibrated ceramic resistors.
The wheel/generator speed is controlled and finely tuned via an asynchronous motor powered by a 3-phase inverter. At this point the test method is extremely simple: I adjust (sink) the load till to get the maximum transferred DC power to the load taking care of course that the voltage doesn’t go below a certain “low limit”.
For instance for usb 5V voltage I’d set this limit to 4.75V.
The results seem to be pretty reliable so far.
The basic estimate criteria would be that an AC/DC converter is as better as bigger is the amount of ACTIVE POWER that it will let to flow from the generator to the load.
The test system I use is THIS ONE, home made but pretty well performing.
Here below a couple of sample plots based on a Shimano DH-3D32-QR hub dynamo and an e-werk respectively set at 4.9V (1st plot) and 5.6V (2nd plot).
As you can see at 20 km/h it doesn’t seem to be able to transfer more that 2.75W: