I just realised that I never shared the graph of my power output from the Turbo Trainer method I described earlier this year. Here is one now:
The protocol for this test was invented on the spot and intended to approximate the effort-profile of a criterium. I had a 25 minute warm-up in 50×17 to determine my comfort cadence; then 60 seconds at the same cadence in each of 50×16, 50×15, 50×14, 50×13, and finally 50×12 before cooling down in 50×16.
I am working on an app to make turbo trainer rides more engaging; it will do interesting things with the data from your sensors as you ride. Because some of my customers will have power sensors I have had to buy one for testing. When it arrives I will be able to attempt the same (or a very similar) test with both sets of sensors enabled and then calculate the correlation between its measurements and the values calculated by this method.