Gather data from a set of coast-down sections into a single .tcx file. Your Garmin computer should only be recording data while you are actually coasting down. I recommend you set your Garmin to gather data every second, rather than using smart recording, so that you have plenty of data for the tool to analyze. For each coast-down section, follow this procedure:
If you have a power meter on your bike, you don't need to press start/stop for each coast-down section. Instead, the tool will automatically detect sections of your ride where you are not applying power. So, if you do have a power meter, start recording, accelerate to 20mph, coast down to 4mph, and repeat as many times as you like while continuing to record.
If you want, you can perform a number of different experiments on your aero position while gathering data. For each position, create a separate .tcx file, and gather a set of coast-down sections in that position within the .tcx file.
Once you've gathered your data, head home and upload your .tcx file to your computer. (If you performed multiple position experiments, you'll have multiple .tcx files.) Next, use weather data from the web to determine the temperature, air pressure, and dew point while you were gathering data, and use our air density calculator to estimate the air density rho. Finally, measure the total weight of you and your bike, including any equipment you were carrying during the data gathering, such as your helmet, clothes, water bottles, and bike tools.
Frolling (Newtons) = 9.8067 (m/s2) · W (kg) · Crrwhere Crr is the coefficient of rolling resistance and W is the combined weight of the cyclist and bike. Note that this force is constant, i.e., is independent of speed.
Fdrag (Newtons) = 0.5 · Cd · A (m2) · Rho (kg/m3) · (V (m/s))2where Cd·A is the drag coefficient Cd times the frontal area A of the cyclist and bike. Also, Rho is the air density, and V is the speed of the bike. Note that this force decreases as velocity decreases.
a (m/s2) = 9.8067 · Crr + V2 · ( (Cd · A) · Rho ) / (2 · W)or, simplifying for exposition:
a = c1 + c2 · V2During a coast-down test, your bike is gathering a number of data points. Each one contains a timestamp and a velocity, as well as other data. By looking at neighboring data points, the tool calculates the average deceleration over that time period, as well as the average velocity. In this way, the tool generates a set of (velocity2, acceleration) pairs based on your measured data in your .tcx file. Using the above equation, the tool uses linear regression to curve fit a straight line against the (velocity2, acceleration) pairs. The slope of this line is c2, and the intercept is c1. From these, you can calculate Crr and Cd·A.
This technique works as long as your data is clean. Calculating acceleration from velocity pairs can be noisy, so the Crr estimate is probably more accurate than the Cd·A estimate.