The torque sensor calibration coefficient is derived from a single unit measured by Casainho.

Analysis of Torque Sensor design (in german) translate to english

Performance

• Interchangeability: Unknown
• Linearity and Interchangeability of TSDZ2 Sensors: Unknown
• Temperature Stability: unknown

Static Torque Calibration Methods

Spanner + Spring Balance

• Requires things you probably have directly on hand
• balance accuracy is needed
• difficult to get reproducible or accurate results

Casainhos Technique

Weight + Lever Arm

• Accurate
• Reproducable
• No special calibrated equipment needed.
• Need to make simple jig and get 5kg weight.

Make some hooks (or cable tie to the crank) so a wooden arm hangs off one pedal crank horizontally facing forward. Hang a gym weight off the end. (This is a 2 person job)

You can zero by either unhooking the arm, or (if cable tied) , rotate until the arm is vertical so exerting no torque.

Torque= 9.8*N/kg * (GymWeight.kg** distance-from-axis.m + Wood-weight*WoodLength/2)

5kg at 1m + 2kg wood ~=60N.m

 Power (W) = Torque (N.m) x Speed (RPM) 1000/ 9.5488

Power Measurement

Hill Climb

The hill climb seems a good way to measure power without needing anything to be calibrated. (as long as you live somewhere with hills, or a multi-storey carpark with tight ramps).

Energy(J)=mass(kg)*G(9.81)*height(m)
Power=E(J)/T(secs) 
PedalPower= Power/ efficiency. 

0.85 overall efficiency seems a reasonable guess for rolling resistance+powertrain loss+low speed wind resistance

So for my ride up the hills which is 514m climb, taking 60min,

  514m*118kg*9.81*3600secs = 198W overall. (and 680cals food @25% efficency)

Made some measurements during todays ride: 31m climb at the start = 361W. 183m climb at the end=254W.