I actually question how much of a market, in the world of cycling and multi-sport, truly exists for this type of device. I'm fairly certain the golden age of aero has come and gone. Bikes aren't going to get much better, helmets (assuming you have a good position) are all testing very close to one another, what makes clothing fast is pretty well known at this point, and consumer knowledge of all these things is much better than 10 years ago. For positioning, this will be great. Once proven accurate, I'll use the hell out of these things. For everything else, there's not much low hanging fruit left out there.

I also don't believe it's about obtaining better data. We're already pretty darn good at predicting bike splits, so that tells me the data we're getting from the wind tunnel and velodrome is fairly accurate. If we can use this device as a cheaper alternative to those two, that would be great, but now we're talking about offering it as a service by people like myself who can combine it with at least some level of expertise. Once you get past the pointy end of the market, I just don't see wide adoption. How big a market is the pointy end? I don't know, but it will be no where near the power meter market. There was a time, but to quote Butch Cassidy & the Sundance Kid, "Those times are over."

Of course, I don't believe everyone entering this market is focused solely on cycling and/or multi-sport. ;-)

---

Jim Manton / ERO Sports

It's hard to know if we are going to get significantly faster bikes/wheels/tires/clothes/etc (probably not, right?) but there sure is a lot of room to improve (amateur) rider position. And the way different components interact with each other and the rider might lead to decent gains too. An easy way to tinker with this would be pretty cool.

I saw it and made a note to watch their progress but it appears they still have a ways to go before it's ready for prime time. I didn't think it was worth starting a thread or kicking up a fuss until we got closer to launch. That said the feature set is interesting. One thing that sort of bothers me about my mantis setup is that I've only tested it to 10 degrees in the tunnel and I don't know how it performs at higher yaws.

When I field test, I'm basically testing at zero yaw by default. I only test on still mornings and my whole route is heavily lined by oak and pine trees. Having a sensor that records yaw would definitely add an extra dimension and actually encourage one to test in the wind. I've also always been curious about the "real world" difference between something like a 303 front and an 808 front on a windy day. Does the extra butt-pucker really gain you what the tunnel says it should?

Hello,

This is John Buckley the co-founder of Velosense. Dr Barnaby Garrood and I started looking into measuring wind on a bicycle about five years ago as a pet project to try and get an estimate of our aerodynamic drag.
Both of us have been Aerodynamicists for our entire career, with the majority of the time spent in Formula One. Barney in particular had worked for several years as a track aerodynamicist, correlating real world conditions with the wind tunnel. The main difficulties of this job lay in measuring car forces (downforce and engine power) and capturing the wind direction while the car is cornering.

Measuring power dynamically on an F1 engine is extremely difficult as the environment is harsh (hot) and noisy (large fluctuations in torque). Modern bicycle power meters in their much more forgiving environment mean that now many riders have accurate and highly repeatable power meters. The force (coming from power) measurement is the starting point for resolving any aerodynamic force, so in cycling we already have a better foundation. However in F1 the speed works for you: pressures and forces are far higher so easier to resolve, and wind yaw angles are substantially lower. A strong background in aerodynamic instrumentation is therefore critical to getting an accurate aerodynamic device.

One of the biggest challenges for us has been creating a device which can measure wind yaw angle over a large range - in field testing with crosswinds, we often observe that the instantaneous wind angles are +/- 20 degrees of the average wind angle. So along a stretch of road, I may have an average crosswind angle of 10 degrees, but this is made of values from +30 to -10 degrees. During our career we have used multi-hole pitots which give a high angular range, but the results from these have always been poor, especially in the turbulent conditions commonly seen on bikes. We believe this is due to the way the air flows over these devices, separating over the measurement hole at larger yaw angles, requiring complex calibrations. So using a few tricks we had learned during our careers, we set out to develop something completely new.

Following 2 years of development and wind tunnel testing, in the spring of 2017 we made a few technical breakthroughs which resulted in a design not far removed from the Velosense probe (but much larger!) which is accurate up to 50° wind yaw angle. We have a patent pending on the shape and measurement system of the probe, and we may apply it to applications wider than the cycling market.

Measuring air speed on a moving vehicle is always difficult due to the vehicles effect on the airflow, as Andrew Coggan's post above alludes to (here's that link again: http://www.hupi.org/HPeJ/0008/0008.htm). One solution to this is to move the probe a distance far enough away from the vehicle that the effect is negligible. Another solution is to place the sensor in a position where the airflow is affected by the vehicle body, but to calibrate the sensor to account for this effect. On an F1 car, the air flow sensor is on top of the nose, where it is affected by the volume of the car, the changes front wing angles, and even the height of the car above the ground. By testing and mapping these conditions, calibration factors can be determined to obtain very accurate results. The bicycle situation is complicated by different rider positions and bicycle configurations.

It is important that users understand the relationship between accuracy and repeatability. Repeatability allows users to make adjustments and observe positive or negative results, whereas accuracy is required when determining what the absolute time over a race will be. Our device will come with a pre-set factory calibration which can be used to take repeatable measurements straight out of the box. For users looking to measure to a high level of absolute accuracy, this pre-set calibration can be updated using a calibration routine.

Our aim has been to create an aerodynamic device which is as easy to use as a power meter. If elevation change, wind, air density, and rolling resistance can be accurately measured and accounted for, aerodynamic drag in differing conditions and locations can be compared. Aerodynamic devices will always be inherently more complex that power meters, but our calibration and data analysis methods will massively simplify the use of our probe. So for example, calibration can be as easy as riding up and down the same stretch of road, whilst our analysis methods can reduce the CdA to a simple number which takes account of the variability in yaw angle.

We are currently in Banff for the ANT+ symposium and will return to the UK next week to continue an intensive road testing program. We have followed the aerodynamic threads on slowtwitch and greatly appreciate all of the work that has been done by contributors to this forum, especially in the field of rolling resistance, which has been very valuable to us. In the coming weeks we can share some of our measurements such as wind angle variation from our road testing.

John Buckley

---

John Buckley
https://streamlines.aero

I would think that the ideal usability of VeloSense (and its competitors) would be less post-use analysis and more real-time control. In other words, you go to A2, you get data, you crunch it, you change your equipment/position and you're done. With VeloSense, if they can give you that "two diamond display" in real-time, you are able not only make those equipment/position updates grossly once every year or two, but to actively shift/correct your position throughout your race. Holding a position over 112 miles is tough, this could be used to tell you, "Hey! You're sub-optimal! Quit wasting watts!" Combined with power/speed/heartrate (maybe in place of power?) and it seems to me a really useful tool.

$1000 still seems steep, but I'm in the camp with slowman that tri does not require $20k in equipment to get you happily to the finish line!

How do you account for the fact that the shape of the"vehicle body" keeps changing?

For example, it seems quite likely that the flow field in the vicinity of your sensor would vary significantly based on the angle of the arms.

Hi John,
I take it this is the application? https://patents.google.com/patent/GB2558709A/

It's a bit unclear without the illustrations in the application, but as I understand the text, you have a series of pressure ports in the inner walls of the "ring" which give a pressure differential that is calibrated to yaw angle...is that the basic gist?

That sounds really cool...it's sort of like turning a mutiport pitot tube "inside-out" :-)

BTW, I'm glad you mentioned the corrected air speed calibration possibly changing with body positional changes...I think that's something which a lot of folks may not understand about how these devices work and it needs to get out there "early and often" for it to sink in ;-)

Thanks for the update!

---

http://bikeblather.blogspot.com/

Super excited to see this product, and more so the conversation around the testing that has been done to date.

I have seen several apropos designs, and none of them have addressed this higher yaw measurement capability. It has been my supposition that several of the wind yaw distribution studies have been flawed due to instrumentation shortcomings.

Some of the best yaw data I have gathered is with string and a GoPro, as the string doesn't have a limitation on the angles it can express.

My $0.02 on this issue is that the farther out from the rider the device is mounted the better. This has limits, however, as the longer the mount gets the more the device can "bounce" (oscillate up and down) when you go over a bump and this will introduce noise into the readings.

That's why back in the mid 1990s Martin et al. mounted their "whirlygig" device on a long, stiff boom extending from the head tube.

A question that I have been pondering: how long does the wind have to be coming from a particular yaw angle before it "counts"?

Obviously, an instrument with a very low temporal resolution will only provide a measure of the central tendency, whereas one with much higher temporal resolution will record greater variation. If, however, the wind comes from an angle for too brief of time to alter the overall drag, then such fluctuations aren't really relevant.

(In wind tunnel design and validation, a distinction is made between longer-term fluctuations in overall wind speed and shorter-term variations in turbulence, with the cut-off between the two based on how long it takes the air to flow past an object... I need to go revisit how those calculations are made )

So you have image processing to calculate the angle represented by the string?

Actually, it's about how much it turns relative to the apparent wind ;-)

But, what value would you term as "wild"? 10 degrees? 20??

I only ask because in my own (admittedly somewhat limited) playing around with an Alphamantis Aerostick (mounted on handlebars), the only time I saw what I would term "wild" yaw swings (i.e. on the order of 10-20deg larger than the nominal) was when performing a low speed, 180 degree turn, in which there would be a large yaw swing in one direction at the start of the turn, followed by another in the opposite direction as the bars returned to center. While riding along in a straight line, the sample by sample variation was much smaller.

For example, here's a data section with one such 180 degree turn:

---

http://bikeblather.blogspot.com/

What would be cool is to overlay stats taken from a solid-state yaw sensor onto the video of the string to see how well they correlate :-)

---

http://bikeblather.blogspot.com/

please bear in mind i have no idea what i'm talking about. that stipulated...

what i don't see on your graph is the time increments along the x axis. so, each triangle is representative of - what? - maybe 1sec, or 2sec? (wild guesses.)

now, you see these oscillations, and it seems like what you're getting is a pair of yaws that are 4° from each other, or 2° off of centerline in each direction. what i think you're reading from john above is maybe 4x that sort of swing. he can correct me if i'm wrong.

this is what i referred to. this is what got me to stop as i was walking by the booth. that chart, printed on an 8.5"x11" paper, that looked just like yours, except yours looked like a 3 on the richter scale and his looked like an 8. and then i thought of your good friend and mine hambini. whom i have not alerted to the existence of this thread yet. because i thought we could have some quiet time with john first.

but maybe i'm not interpreting all this properly.

---

Dan Empfield
aka Slowman

Ok I have been following this thread all day and can't help myself. I too have thought about this and basically apply a cutoff based on the length scale of turbulence that is on the scale of the rider. The temporal resolution required to resolve such a scale can be roughly be calculated from the convective speed of the turbulent fluctuation as it passes through the cyclist. For a turbulent fluctuation that is 1 meter in length scale and a relative speed of 10 m/s, you are looking at 0.1 seconds for that fluctuation to pass through. Apply Nyquist criterion, and you need at least 20Hz sample rate. We sample at 200[Hz] currently in our little wind sensor thingy (AeroLab Tech Sensor). *Note: this is a very rough approximation, and this theory would need to actually be tested for its applicability and influence on overall drag*

---

Chris Morton, PhD
Associate Professor, Mechanical Engineering
co-Founder and inventor of AeroLab Tech
For updates see Instagram

I do, because google has some nice image processing libraries. Really not difficult if you have good color contrast.

Honestly pretty interesting to just watch at 5x speed or so, especially if the GoPro can see the computer as well.

I am not trying to measure anything anymore, just trying to understand some things from a different perspective.

For years I have chased frames and wheels with great zero/low or high yaw performance, and then chose the "right" one on the morning, but more and more I see that it's so much more complex.

I think the answer is that all time "counts" as far as energy expenditure is concerned thus the higher the sampling rate, however "noisy", the better. Given the shape of most yaw distributions we have and the fact that most samples have been taken at ~1hz (iirc) I think we can safely conclude that the resulting yaw distribution skews low. Maybe not much, but at least a bit.

I'll throw this out there: Josh Portner (Silca/Zipp) recently let an interesting tidbit slip. Zipp did some testing where they went from 30 degrees to 0 degrees to see when the flow would reattach and they found that reattachment did not occur at the same angle as detachment for the rims they tested. Instead they found that flow didn't reattach until a lower yaw for most rims. He also hinted that shapes that performed better at higher yaws saw their flow reattach sooner when going from 30 to 0 degrees. For most modern rim/tire combinations we see today flow seems to detach around 12-15 degrees. For all the modeling that's been done (that I'm aware of) it has been assumed that a given wheel's performance from 0-20 degrees is the same as its performance from 20-0 degrees. Separation is admittedly rare for most riders but, again, this is a consideration that might be biasing us toward low yaws in design and equipment selection.

Something that has always bothered me, personally, is that we test in the tunnel with the front wheel perfectly aligned with the frame when, in reality, the wheel is constantly moving a few degrees back and forth from zero. If I had the time, I'd figure out a way to hook up a servo encoder to my headset to gather data.

...rambling here....

If we think about what happens when a side wind hits a rider, the wind tilts the rider/bike away from the direction of the wind which causes the front wheel to turn back into the wind. The rider then counteracts this and resets the system. In this instance, if you had two different yaw sensors attached to the bike at two different places (the handlebars and, say, the top tube) you'd get two different readings for two different parts of the system. If I'm thinking about this correctly the frame would see a higher yaw than the front wheel. The unknown for most of us is how the front wheel down tube system works when the two aren't aligned. It would seem to me the most pragmatic approach would be to include the steering deviation (1, 2, maybe 3 degrees) from center and add that to the yaw distribution that the wheel "sees" as the goal is for the flow to stay attached to the front wheel and then transition smoothly to the down tube.

Wrapping up my ramble here...

In light of the forgoing, perhaps we should be looking more closely at the performance of our equipment at higher yaws.

Edit: this might also make one reconsider tire selection as some fast rolling tires really hinder the performance of a wheel at higher yaws. The Turbo Cotton and GP TT come to mind.

You seem to be forgetting the very high correlation between power demand as predicted from wind tunnel measurements and actual power required under very windy conditions in the field.

"In God we trust - everyone else must bring data." - W. Edwards Deming

Here is the sort of thing I am thinking about/the sort of data that needs to be considered when deciding just how frequently yaw should be measured:

https://aip.scitation.org/...1969?journalCode=phf

Oh, I know. I've seen your work on the subject and others. My own field testing correlated well with my trip to A2. We're chasing really marginal stuff here. The last step of the optimization process is for a manufacturer to optimize a tire/wheel/frame system to eek out a watt or two... or maybe five. Tops. Specialized was the one brand I had pegged to take on that development path with the new Shiv but, sadly, that appears to not be the case.

I suppose the final frontier for TT bikes is comfort.

Yeah...each point is a 1s reading from the WASP utility app (I read out the Aerostick through a WASP-n ANT+ bridge, which is then connected to an iPhone 5 running the WASP utility app). Now then...as I understand it, the Alphamantis Aerostick is broadcasting ANT+ packets at either 4 or 8Hz, but what I DON'T know is if each of those packets is an average or a downsampled value...and furthter, I don't know if the WASP utility is then further averaging or just downsampling for the 1Hz recordings.

Anyway...what I need to do is try to determine the "noise" in the reading. In other words, given a known non-varying yaw input at a non-zero wind speed, how much does the device just naturally vary? You can't just cap off the end (I tried this) and see how it does, because the yaw reading is based on a pressure differential, and if the pressure (i.e.wind speed) signal is small, well...then the angle estimation can vary all over the place just from instrument noise...make sense? Since I don't have a small wind tunnel to calibrate that, maybe I'll try mounting it on a car mirror and see what happens :-)

---

http://bikeblather.blogspot.com/

There's a reason you're sampling at 200Hz though, and it's because you are not getting 200 accurate measurements per second out of your ICs!

Curiously, that’s exactly how I started out 6 years ago! I then moved to using a vane as it meant I had more freedom positioning the camera and measurement bit. Then I just added a pickup on the end of the vane shaft to get direct measurements with higher resolution. It also meant I could up the sample rate. I’m aware that there is some damping with a vane, but the frequency response is pretty good. Using a vane also has side benefits of low speed performance and relatively straightforward calibration/ datum setting.

---

Developing aero, fit and other fun stuff at Red is Faster

Ease of use.

As I understand you need for the Velosense, as for the Aeropod, a speed sensor.
And that is a step back for me. After all the hassle with spokemagnets and interference from overland power lines and so I am very glad since a couple of years now to have finally rather a stable speedometer thanks to GPS.

This is what the Aeropod and Velosense should have: internal GPS for speed. (I guess it is not possible to get the speed from another GPS device: which is in fact the reason that use must be made of the archaic speed sensor).

One should also not forget that the Velosense and Aeropod do not show a true CdA: if you go from smooth tarmac to rough tarmac the measured CdA changes (as I understand the devices) although the real CdA stays the same of course.

Nevertheless great stuff, although I probably would rather spend my time on training (I do not even have a PM upto yet).