Back to back testing between my Powertap and SRM showed that the SRM was reporting higher wattages which could partly be explained by drive train losses. By lowering the SRM slope setting from the factory setting of 17.5 to 16.9 Hz/Nm I improved parity at my functional threshold. However, it still didn't provide the perfect solution as it was apparent that the rate of rise for the Powertap was higher than for the SRM. In other words, whilst I got good agreement at around 300W, the Powertap reports higher power than the SRM for wattages above that value. Subsequently ride TSS is always a few points higher for the Powertap.

I weighed a number of gym weights from 10kg to 35kg and made a spindle to allow them to be hung onto the pedal of my bike similar to this. I then mounted my bike on my turbo trainer and followed the aforementioned SRM calibration protocol. However, instead of using a fixed horizontal crank position, I rotated the rear wheel of the bike backwards to lift the weights off the floor. I then continued to lift the weights whilst observing the power meter's frequency. The highest value occurred at maximum torque when the weights were hanging at right angles to the crank. I calculated the torque applied for 6 different weights along with the increase in strain gauge frequency. The results were plotted on the following graph.


I used the trendline feature of Excel to calculate "m" for the graph which is the value of the SRM slope. I obtained a value of 16.6 Hz/Nm. This was interesting because it was close to my empirically chosen value following my Powertap comparison. Further searching on the Wattage forum revealed that SRM factory calibrations are often set too high.
I was then curious to know how the weights I used compared to real life pedalling forces. I exported an SRM file to Excel and plotted the following graph to compare pedal torque against power output.

My minimum calibration weight of 9.65kg applied a torque of 17.4Nm and typically this would produce power around 160W or level 1. My maximum weight of 34.55kg applied 59.3Nm and would typically produce power around 600W. I am satisfied that I have checked the calibration over a representative "real life" range of torques and that the power meter responds linearly.
I want to repeat the same experiment on the Powertap, but will have to factor in the gear ratios in order to compare with the torque measured at the hub. I can then establish whether the slope of the Powertap is indeed higher than the SRM.

I have two powertaps, an older pro and a new 2.4. One training one racing. I wish i had just got an srm to start with. I was wondering why your still using both? Given the variation you will probably always get with two different meters.

I use the Powertap on my training/winter bike. It has taken me a while to get used to not having a rolling average constantly on display with the SRM, but I am getting more used to it now. I do like the Powertap and the ease of changing batteries is good. For workouts up to threshold I get good agreement, but at the moment I'm not sure about higher efforts.