They measure the effort that the retarder has to put in to maintain the rpm that the r/r has asked for (using a strain gauge attached to the retarder). That could be a constant ramp if it is "shootout mode" or it could be a sustained load point.

So the r/r measures the effort the retarder has to put in to maintain roller RPM as requested.

As roller RPM increases, the effort required to turn the rollers (just themselves with no retarder attached) increases.

Therefore with a basic car that has around 100 bhp and say 85 lb ft. If the car is run up in say 2nd gear (with rollers turning at 55mph e.g.) X amount of effort will need to be used by the retarder to maintain roller rpm (stop the car accelerating). If you shifted into 3rd, the rollers will be going around quicker, and the rollers themselves, take more effort to turn, so the retarder will need to use LESS than X effort to maintain rpm (stop the car accelerating) as the rollers themselves absorb more effort to turn at higher rpm. They are like a fan in the way that 1kw = 100rpm, 2kw does NOT equal 200rpm, due to increasing losses of efficiency, it will be more like 3kw = 200rpm.

SOOOOO - what that means, to u, is that the r/r measures the effort put in by the retarder to maintain set roller rpm, and therefore it does NOT measure torque at wheels as u suggest. A calculation is then completed that turns the effort used by the retarder, (initially measured using a strain gauge iirc) into a torque figure. Therefore, how can u be sure ur initial ATW figure is "accurate" in ur terms? It will just become smaller and smaller as roller RPMs rise. SO u need to take a big power, reliable standard car and run it up once in a while to use as a benchmark / calibration in each gear.

That is my take on it anyway and goes to show there is more to it than meets the eye.

RW