Prev Next
marcag wrote:
If you are make changes that affect the calibration of the air speed sensor yes a new calibration needs to be figured out. But that recalibration can be part of the test. It's not make change, calibrate, do test. Calibrate and do test can be combined.

I am aware of this... but it actually doesn't work well because the calibration adds a large amount of unnecessary error.

When high precision is needed the rider is doing A-B-A-B runs, switching between configurations. In this scenario, it is not so important that the calibration be accurate, so long as it is not different for A vs B. If the airspeed is off say 2% from reality, it will be off 2% for both. If on the other hand the change in configuration necessitates recalibration, calibration accuracy becomes extremely important, since it will directly affect the results.

I assume that calibration is performed by making the assumption that the wind vector relative to the ground averages for out-back runs? In other words you say that the wind does not have a bias. But how many runs do you need to perform in order to have confidence in this assumption? I can tell you that on a windy day, the wind varies a lot in direction and intensity. On a single out-back you could easily have 2mph net headwind or tailwind. How many runs would you need to "cancel" it out to a reasonable level of confidence where it doesn't impact your result? We are typically interested in resolving to CdA differences of 1% or less. The calibration error would need to be much less than this to be "acceptable". Maybe 0.25%? The wind error that would result in a CdA error of 0.25% is only ~0.1%. If you are testing at 30mph, this is 0.03mph. That ain't happening! Even if you are out there all day going back and forth you are not going to reach that level of confidence in the "wind canceling" assumption.

Regarding the "wind canceling", I'm already setting a wind vector when I do my analysis, with the assumption that the wind cancels out. I'm currently varying it with each run; wind cancels for each run and I average the CdA results. It would be better to have a global constraint of wind canceling for the whole test rather than each run, but I kinda suck at spreadsheets and am even worse at programming, so I haven't bothered. But if I did that I would acheive nearly as good results as the Notio... without measuring wind at all.

So, no... recalibrating won't work. The calibration needs to be stable for any changes you make in configuration.
rruff wrote:
marcag wrote:
If you are make changes that affect the calibration of the air speed sensor yes a new calibration needs to be figured out. But that recalibration can be part of the test. It's not make change, calibrate, do test. Calibrate and do test can be combined.

I am aware of this... but it actually doesn't work well because the calibration adds a large amount of unnecessary error.

When high precision is needed the rider is doing A-B-A-B runs, switching between configurations. In this scenario, it is not so important that the calibration be accurate, so long as it is not different for A vs B. If the airspeed is off say 2% from reality, it will be off 2% for both. If on the other hand the change in configuration necessitates recalibration, calibration accuracy becomes extremely important, since it will directly affect the results.

I assume that calibration is performed by making the assumption that the wind vector relative to the ground averages for out-back runs? In other words you say that the wind does not have a bias. But how many runs do you need to perform in order to have confidence in this assumption? I can tell you that on a windy day, the wind varies a lot in direction and intensity. On a single out-back you could easily have 2mph net headwind or tailwind. How many runs would you need to "cancel" it out to a reasonable level of confidence where it doesn't impact your result? We are typically interested in resolving to CdA differences of 1% or less. The calibration error would need to be much less than this to be "acceptable". Maybe 0.25%? The wind error that would result in a CdA error of 0.25% is only ~0.1%. If you are testing at 30mph, this is 0.03mph. That ain't happening! Even if you are out there all day going back and forth you are not going to reach that level of confidence in the "wind canceling" assumption.

Regarding the "wind canceling", I'm already setting a wind vector when I do my analysis, with the assumption that the wind cancels out. I'm currently varying it with each run; wind cancels for each run and I average the CdA results. It would be better to have a global constraint of wind canceling for the whole test rather than each run, but I kinda suck at spreadsheets and am even worse at programming, so I haven't bothered. But if I did that I would acheive nearly as good results as the Notio... without measuring wind at all.

So, no... recalibrating won't work. The calibration needs to be stable for any changes you make in configuration.

I am not sue I understand everything you are saying but I believe we both agree that cancelling wind in the out and back is probably the simplest way but there are better ways.
What other ways are there?

What I was trying to get at, is if you need to recalibrate when you change configurations, you're screwed. The sensor needs to be located in a place where the changes you are making won't affect it. If *that* is true, then calibrating by assuming the wind cancels is fine. You just need to have the same calibration for both configurations. Calibration accuracy is not so important then. Your absolute CdA value will still be off, but there won't be any bias favoring one of the setups.

If I calibrate for configuration A and it's +1% off (airspeed says 50.5km/hr when it's really 50), and I calibrate for configuration B and it's -1%, I've just introduced a 2% airspeed bias in favor of A. That's about a 5% CdA bias in favor of A! See the problem? B could actually be 4% better than A, but you'd erroneously think A was better.
Last edited by: rruff: Feb 27, 19 17:05
rruff wrote:
What other ways are there?

What I was trying to get at, is if you need to recalibrate when you change configurations, you're screwed. The sensor needs to be located in a place where the changes you are making won't affect it. If *that* is true, then calibrating by assuming the wind cancels is fine. You just need to have the same calibration for both configurations. Calibration accuracy is not so important then. Your absolute CdA value will still be off, but there won't be any bias favoring one of the setups.

If I calibrate for configuration A and it's +1% off (airspeed says 50.5km/hr when it's really 50), and I calibrate for configuration B and it's -1%, I've just introduced a 2% airspeed bias in favor of A. That's about a 5% CdA bias in favor of A! See the problem? B could actually be 4% better than A, but you'd erroneously think A was better.

Are you saying no device can work properly if it's placed in the area below extensions, above top pof wheel, before head tube, behind front of wheel ?
marcag wrote:
Are you saying no device can work properly if it's placed in the area below extensions, above top pof wheel, before head tube, behind front of wheel ?

Pretty much. If the device is located any place where adjusting the equipment or rider's position necessitates recalibration, then the device will have limited utility. I very much doubt the Notio mount location will work. At any rate the manufacturer needs to demonstrate the utility of their device and the expected error. Not just for this bias, but overall.

An "acceptable" bias error in the airspeed measurement is in the <0.2% range (0.1km/hr at 50km/hr), for a CdA error of ~0.5% (or for instance 10cm^2 out of 2000cm^2). If changes in position or equipment result in a calibration error greater than this, then it's failed IMO.

EDIT: I think I erred regarding the relationship between airspeed errors and CdA errors. It's 2x, not 2.5x.
Last edited by: rruff: Mar 6, 19 10:29

Prev Next