Frequently asked questions and answers
Your bike needs a powermeter, speedometer and optionally a Notio aero device. AeroStar uses the 4Hz sensor measurements from the Notio aero device to compute your CdA & Crr.
If you don't have a Notio, you can instead use a FIT or TCX file from your bike computer. These files record data (e.g. power, velocity, elevation) at 1Hz but are missing some needed data (e.g. total mass, assumed Crr, density, airspeed) which you'll have to manually enter. I recommend a magnetic sensor speedometer since accelerometer-sensor speedometers report noisier velocities.
The Notio CdA algorithm is extremely sensitive to velocity; slower speeds result in giant increases in CdA. If you compare the Notio Virtual Elevation (VE) with measured elevation, you'll find that the VE is tens of meters higher = Notio CdA is far too low. So the Notio CdA bounces between being too low and too high. So Notio results vary a lot and are not accurate. Argh!
These problems irritated me so much that I developed the AeroStar algorithms to compute an accurate CdA. The AeroStar app plots all CdAs and VEs so you can easily compare them. Note that the AeroStar VE is usually spot on with measured elevation.
AeroStar analyzes every ride, so you don't need a special test ride. When starting out measuring your CdA, I find that its best to minimize environmental influences: so on a calm day ride on a flat stretch at about 10 m/s (36km/h, 22mph) for 1600 meters (1 mile), slow down by coasting, turn around then accelerate and ride back to your starting point. No braking! Ride a few laps with exactly the same position to get acclimated. Typically your position is changing lots on the first lap(s), so I ignore the CdA analysis for the first laps.
Riding on a loop (e.g. outdoor velodrome) let's you maintain your position and velocity making for simpler CdA interpretation.
Aero resistance and Rolling resistance are equal at 4 m/s (15 km/h, 9 mph), so if you ride faster then aero will dominate and you'll get accurate CdA values. If you go even faster the accuracy will further improve. If you're training for races than you should test at race speeds to check if you can hold your aero race position.
If you also want to estimate Crr, then your ride needs to include smooth acceleration from (very) low speed to high speed.
A Competitive can hold their position and consistently generate the same power. This results in very consistent CdA graphs making it easy to detect subtle changes. This makes it easy to do ABABAB testing of positions or equipment. Expect a CdA accuracy of about 0.002
A Weekend rider is not as strong and tires easily and their position is always changing slightly. This results in a CdA curve with appreciable differences for a 'constant' test ride. Expect a CdA accuracy of about 0.005.
A TT bike with aerobars 'locks' in the position of the rider much more than a road bike. On a road bike the rider can't really 'lock' the elbow angles leading to different aero drag resulting in a 'noiser' CdA curve. A road bike with aero bars will have more consistent CdA results than one without.
AeroStar uses an energy method similar to Virtual Elevation (VE), but with lots more math. VE uses an energy balance (Ewheel = Eaero + Eroll + Epe + Eke) that can be solved using Ordinary Least Squares (OLS), but that only gives you a single (static) solution -- typically an average CdA value for a lap. Using Recursive Least Squares (RLS) will provide a (assumed constant) CdA solution updated with every data point. RLS can be extended using sliding windows or forgetting factors to support changing (dynamic) parameters. A superset of RLS is Kalman Filters (KF) which supports dynamical systems and time-varying parameters. AeroStar uses KF which are further enhanced to provide robust solutions for multiple parameters (e.g. CdA, Crr).
AeroStar M2 solves for CdA, and AeroStar M4 solves for both CdA & Crr.
Solving for both CdA & Crr is very tricky, so I recommend a careful protocol to get good data. A flat test path should have the same surface, and allow 10 m/s through any turns. Start from a stop and accelerate slowly to +10 m/s. Ride a total of +1km (so probably several laps). At the end of your test coast down to a slow 3 m/s speed without braking. As an alternative, you can end your test with hard braking to a full stop, and then use AeroStar 'Solver Options' to remove the braking.
There are lots of ways to calculate a CdA for a ride -- some more accurate than others. To determine if a CdA is accurate, AeroStar compares the *measured* Energy, Power, Velocity and Elevation with *computed* EnergySum, Virtual Power, Virtual Velocity and Virtual Elevation. These curves will match if (and only if) the CdA is correct.
For instance consider the energy balance Ewheel=Eaero+Eroll+Epe+Eke. Several partial sums can be formed, for example lets isolate Epe=Ewheel-Eaero-Eroll-Eke. The left-hand-side (LHS) is the potential energy and the right-hand-side (RHS) is the Virtual Epe. Divide both sides by m*g so the LHS is elevation and the RHS is Virtual Elevation. Plot both and compare. Are they super close? Then the CdA is accurate. Are they far apart? Then the CdA is wrong. Note that the measured elevation is NOT included in the VE. AeroStar performs these validations automatically to show you that the AeroStar CdA is accurate.
The AeroStar computed Virtual Power (VP) matches measured power quite well, and can even detect spurious data from SRM power meters from hard starts. (Look at the 'Power' graph for Demo3 and you'll see spurious *measured* +500W 'spikes' that are constant for 3-5 seconds but the velocity doesn't change. In contrast the AeroStar Virtual Power shows no spikes.)
You need to find and hold a consistent position! A 'noisy' CdA curve implies lots of position variations. Try testing at a slower speed (lower power) while holding a particular position. Very subtle position changes will change your CdA.
That's a complex question with three simple answers: Position, Position, and Position. Avoid cylinders and spheres perpendicular to the wind, so forearms parallel to the ground, head tucked down. Try ABABAB tests with small position changes. For example try A='head down eyes looking at your front hub', and B='head looking down the road' to measure the penalty of looking where you are going.
A consistent position will give you a 'flat' CdA curve with low variation. If you stop and switch equipment (e.g. helmets) and ride the same laps again, then you can compare the CdA for all laps. If the CdA is consistently lower than you've identified more Aero equipment.
Note: this is much easier if you are a competitive or Pro rider since the resolution of your CdA curve may be ~0.001. I recommend comparing equipment or positions with AB-AB-AB testing on a 400m loop using 3 sets of 6 laps of A followed by 6 laps of B (12 laps per set). The total test will be 12+12+12=36 laps. Note that at 12 m/s (43 km/h) a CdA change of 0.001 is 1W. Yes it's a lot of work to identify repeatable 1W changes!
The AeroStar algorithms run on flat or hilly terrain, constant or varying speed, constant or varying power.
Note that if you brake this is an unmodeled dissipative force -- so the CdA and Crr values will increase resulting in garbage results. Don't brake during testing!
Braking is a dissipative force that is not modeled, so the AeroStar algorithm thinks your aero drag resistance just got lots bigger! Sometimes braking shows up on the Virtual Power validation graph as negative power spikes. 'Negative Power' -- what's that you ask? Simple, pedaling is Positive Power and braking is Negative Power. If you want a smooth flat CdA curve than avoid braking and keep the same position.
Login on the 'User' page. Optionally subscribe to AeroStar to enable the AeroStar analyses. On the 'Load' page click 'User File' to load and analyze your Notio JSON file. In the upper-right corner is an 'Options' button which overrides several settings such as weight, crr, density, elevation, airspeed, ... Press Sensors to see the basic sensor data. Check that your airspeed and elevation looks ok. Press 'CdA & More' to see the Aerostar computed aero coefficient and compare with the Notio coefficient. Press Validation to see Virtual Elevation, Velocity, Power and Energy and compare with measured values. Toggle between AeroStar and Notio results to see which one is consistent with all the inputs. Hint: it's Aerostar :)
Login on the 'User' page. On the 'Load' page click User to load your TCX or FIT file. Note that a TCX or FIT file is missing lots of data that is needed for an aero analysis, so use the 'Options' button in the upper-right corner to set weight, crr, density, elevation, airspeed, ... Press Sensors to see the basic sensor data. Press CdA to see the Aerostar computed aero coefficient. Press Validation to see Virtual Elevation, Velocity, Power and Energy and compare with measured values.
Most graphs have a 'Camera' button to add the current graph to a report list. On the 'Report' page select each item and add annotations as html-formatted text. Press the checkmark to save. When done with all items press the 'PDF' button to save the report as a PDF file.
On the 'Load' page there is an 'Export' button which saves your ride data in a JSON file.