this is what I love about this forum, real discussion and great knowledge! I read a bit on Kalman filters and a bit to try to recall my stats for Bayesian inference.. and that plus this discussion has been very enlightening or as the title says Aero sensors for dummies and I qualify for sure!. Thanks for taking the time to help!

Mathlab published some really easy to understand videos on youtube on how Kalman filters work.

Sorry, I should have explained myself better, Clocks are analog and digital is a analog clock less valid because it doesnt spit out a number. the same with the speedometer in my car its a needle pointing between two numbers, when my wife steps on the scales she is interested in wether her weight is going up or down not her specific weight. Have you ever heard any one say I have lost 432 grams?
BTW even with a 5 second averaging my powermeter varies so much plus or minus 5 watts is not an issue.

Ah. So, as I've probably alluded to before but may not have made clear, sometimes when we're looking at big changes in posture or equipment, we already have a good idea of what's higher or lower, so we're using the estimates we get more for affirmation than information. However, when we're interested in small differences where we really don't know which is better, we want to know not only which is higher or lower but also how confident we are in that answer. In that case, analog "higher" or "lower" doesn't make it easy to evaluate how close the estimates are and how confident we can be in that difference. In that case, actual numerical values make it easier to figure that out.

(And, as for your power meter, the analytical procedure we're talking about does a reasonably good job of handling that kind of variability in instantaneous power reporting.)

So I haven't gotten many opportunities to much testing between work and everything being on fire around my house (yes literally on fire).

I did a couple of testing sessions that may be of interest. The session was just looking at how consistent the calibration of the Notio is. After you mount it or whenever you change position, you are suppose to do a calibration ride. Looking at the image in the documentation, it appears to adjust the calibration of account for the high pressure wave in front of the bike and rider relative to the pito tube/wind sensor. I was curious how much variability there would be from one calibration ride to he next. Answer is not very variable. I did four back-to-back rides in my standard position and it was 1.51, 1.50, 1.51, 1.52. When I looked at the saved values it actually adds an extra digit, so the 1.52 was a 1.515.

It is probably a good idea to keep track of the calibration factor so if you switch back and forth between bikes you will be able to spot a potential error or bad calibration ride.

The second thing I did was download the rides from the Golden Cheetah Notio application to regular Golden Cheetah and redo the analysis using the same parameters. I did find that one of my "stinker runs" from before was from the speed sensor dropping from about 48kph to 20 kph. A little editing fixed most of the problems and now if I compare the value that the Notio software calculates with what I do with regular old GC the results are pretty darn comparable.

So I finally got around to doing a bit of testing where I was actually looking for improvements. The one thing I have noticed is that I think my road bike position may be too low. I have been riding what probably looks sort of like a typical ProTour position for 20+ years. The current bike is a 54 cm Madone with ~5" of drop from the saddle to the top of the handle bar and a 140mm stem. When I go hard I tend to ride on the tops of the hoods with my arms pretty flat, but I can't comfortably hold this position for really long periods of time. My FTP also falls by about 10-15 watts if I ride really low for too long, so if I am going going hard in the drops my forearms are no longer parallel to the ground (A more upright relaxed version of this is the position I have been using as my road bike baseline). In contrast I can ride with my forearms flat for long periods with the higher position on my gravel bike (~3 inches of drop from seat to bars). So I took my gravel bike out and swapped the Notio over. The CdA just riding comfortably in the drops jumped from my baseline of ~0.308 to ~0.343. About what I would expect with the 700x36 tire and very un-aero frame. Now I can ride a long time with flat forearms on the tops tops of the bars. Result, CdA of 0.298 on two runs.

I have always fallen somewhere between a 54 and 56 frame size, so I guess the next bike will be a 56 with at least some spacers under the bars.

I also sent some question to Notio about interpreting their graphs but haven't heard back yet. Nevertheless, I am starting to get to the point that I think I can start getting some usable results.

Thanks Mike, was really interested to read the update. And good that you at least have good enough air quality to ride outdoors still.

How stable have you found the calibration factor if you take the notio off the bike and then remount it between runs?

I'm getting fairly close to pulling the trigger on buying a unit.. you should be getting some sales comission if I do!

The calibration factor is not as position dependant as some people make it out to be. If you remove and put it back "almost" in the same position you will not see a difference

You will see more of a difference in variability of the calibration factor based on the algorithm and calibration protocol a vendor chooses to use for calibration.

There are also "bad places" to put it.

This does not just apply to a Notio but sticking any pitot tube up front.

It is surprisingly insensitive, though I don’t know how much variability constitutes “a lot”. I ran back to back calibration rides just to see how stable it was and it was pretty constant for my position at about 1.51-1.52. When you get in an look at the result it actually adds another significant digit and the last one I looked at was 1.515 so you can see how it would round up down quite a bit.

I have a very similar position on my gravel bike and I switched to the sensor over. I came up with a 1.53 and 1.54 using a slightly different position. I haven’t moved it over to the TT bike yet. Location and position will be significantly different, so I would expect more change there.

One thing I have done is put together a spreadsheet to track results. I have a column for calibration factor. It would probably be useful if the results start to look odd and to highlight situations where recalibrating after moving the device between bikes didn’t go well.

Good test to do! How is the calibration performed, and what were the conditions?

What has happened to all the companies that were teasing a product two years ago, any of them still planning to enter the market? Here are some that I remember.

Garmin
Aerolab Technology
Red is Faster
Swiss Side

---

BoulderCyclingCoach.com

Aerolab was on a recent podcast and they sound like they are close but will be going through fitters/testers rather than general public. Smart move

Swiss side once said they did not want to go consumer, but later said they would, but didn't.

Garmin is a mystery. Maybe they see this is a lot of support and they can make more money off head units/watches.....

Red is Faster is on ST so maybe he will respond


What happened to the post from flocrest about his challenges to get stable numbers from his device ?

I was going to pick up a Notio a few days ago but... still no Android support :/

---

My Blog - http://leegoocrap.blogspot.com

I bought a used iPhone 5 just for Notio. Not many months later they made their updated app that only works with iPhone 5S and up. D'oh... I'm still using the old app :p

I should get back to bugging them again about making it Android compatible. I think there's quite a big group of tinkerers on Android who are not feeling like getting an iPhone just to get Notio working. Either you buy an old iPhone, which is still costing you something, and then you have to lug another device around and you don't know how long it will be before it's not supported anymore. Or you buy a newer one, almost doubling the actual cost of getting a Notio, and you also have to switch your digital life on your phone over to a new device and ecosystem. It's just not a smooth solution either way.

Especially a guy like you doesn't need the bell, whistles and hand holding of the app. Maybe a configuration, a diagnostic....

I don't use a Notio anymore but at the end I would use it to just switch configurations or upload....

Yeah, I just use the iPhone 5 for configuration (changing wheel size, connecting sensors, powermeter ZO, etc.) and transferring the files from Notio to Cloud/computer. Nothing fancy, could even be done with the Garmin I guess, but probably faster and easier with a phone app since you usually have a phone on you anyway... Just not with the right OS :p

Biggest hurdle with the old app is it works through Apple's TestFlight so I have to get it manually re-activated by the Notio guys every 2-3 months. And I need to remember to check that there's charge on the iPhone before I go test. So 'just' annoyances, not any impact on actual testing.

Until when do you test in your neck of the woods ?

Lately, pretty much never :/

I used be good about getting up before the sun came up and getting some runs in, but my lack of a TT bike (and motivation) hasn't helped this year.
Was hoping to start doing some testing at the track soon...ish...maybe :D

---

My Blog - http://leegoocrap.blogspot.com

Depending on the day, I go back and forth on if I should get one of these devices. Right now I'm generally pretty decent at performing my test runs on a nice circuit near me and doing the analysis in GC, but I feel like my testing is still pretty limited to a defined ABBA per session. The thing I would love to see is a "live" result so I know that B is better than A, which would then allow me to do a second BCCB test etc. I can do the analysis in GC afterwards to see what the CdA value is more precisely, but the big game changer would be the ability to more time efficiently test out different setups.

Doing ABBA to find out B is better than A, then going right to BCCB to find out B is better than C, then going to BDDB to find out D is best is so much better than: ABBA, ACCA, CBBC, DAAD, BDDB, etc. etc. etc.

And reading what I've seen in this forum, that does seem to be where we are at, correct? These sensors with their "live" data are good enough to say if A is better than B, but not necessarily what the CdA of A or B is with any real certainty until you do the detailed analysis yourself? If that's the case, then I think I am sold and will bite the bullet and go for it.

I don't know how Notio works, but CdACrr gives you a CdA for each lap in the field.

Determining if A is better than B is not a simple task when there is little difference between them, and you have the random wind factor. Measuring airspeed accurately enough to parse small differences isn't that easy either. If you can test without wind, you don't need an aero sensor.

If you are using a device on your own and without a coach or trained fitter (e.g., purchasing a Notio or other consumer direct product), I would suggest you start out by doing AAAAAAAAAA. Then do the same thing the next 3x you go out (same socks, shoes, tire pressure, tires, helmet jersey, body position, same shaved legs, calibrate your power meter, etc etc). Record the weather conditions on a piece of paper or in a note on your phone, including if it was sunny vs cloudy, air temp range, wind (steady, gusty, calm), and your own fatigue levels. There is a bunch more here I would recommend but unfortunately I don't have loads of time to write it all out!
I believe this would be about 30-40km of riding each day, total of about 100km over 3 days doing the exact same equipment with 10 trials each day.
Write down anything unusual that happened on any of the trials (cars passing, having to use brakes, needing to pass another cyclist, getting passed by a cyclist, excessive winds, possible brake rub, etc.)
If the wind is higher than 5kph and you are using one of the current consumer products on the market, I would probably just do a free ride and test on another day, or collect 10 more trials of data understanding that this particular data may not be suitable for direct comparison with other days.

Take a look at how much variance (2.5 times the standard deviation of your measurements as an approximation for 95% confidence in a test with a low number of samples) you have in your tests on a given day which were considered "clean", and throw away any tests that contained "unusual" events (even if the CdA number looks good in your view, throw it out anyway!)
Once you know the CdA variance you have on a given day, then you know how much change in CdA is required for you to statistically state that B is faster than A (with 95% confidence).

Then, take a look at how much change you have in CdA from day 1 to day 2 to day 3.
Let me know when you make it to day 3 and if you are not yet ready to throw the device at a wall :)


Edit: my distances are based on the understanding that the Notio has something like a 3km out and back recommendation. Perhaps that has changed but this is just my understanding. As a side note, the AeroLab sensor we have developed currently uses a protocol for aerodynamics testing that lets you set the distance. We currently use 400 [m] out/back. So 10 trials is a reasonable 8km of riding. I am also kidding about the throwing of the device. They were frustrating to use in the early days, but if you become an expert user, they are extremely useful.

---

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

You make it sound SOOO exciting... LOL

---

Sr. Salitre

Maybe I didn't state really well what excites me about these and what I'm looking to do with them.

I'm NOT looking to have a device that outputs a CdA value on my head unit that I take as a meaningful number in any sense towards what my real CdA is.

What I'm looking for is direction when I'm out doing my normal Chung Method testing for what seems to be lowering my CdA so I don't waste time with something that quite possibly less aero than my base position. Then once I'm back home and upload my data into GC, I'll do my normal analysis and get a better picture of the CdA using Aerolab. (Not your company, the Golden Cheetah plugin)

If the headunit via say the Notio told me position A was 0.267 and position B with hands raised 1cm was 0.263, well I'm going to tweak position B a little more and do a test with my hands 2cm up as a position C test. I understand when I get back home I might very well find out position A was closer to 0.245, position B was closer to 0.248, and position C was even worse and a total waste of time. That's fine and I can deal with that and hope that's not what happens every time...well maybe if it was always wrong I could just do the opposite in that case...so hopefully it's better than 50% accurate :-)

And I get it that I won't be nearly as good as going to a pro with lots of experience in a super-controlled environment like a velodrome or wind tunnel, but if I ever get to that point, hopefully I've been able at least get somewhat close riding around my neighborhood course and analyzing via the Chung Method in Golden Cheetah. If you don't think that's even something that's worth-while for us to do on our own, I'd imagine that's a whole other conversation?

+1 on this. Also try a few different courses until you find one where you get very consistent outcomes. My tests are about 7 mins in length and I consistently get +/- 0.001 on CdA without a wind sensor, but it took a fair amount of testing to refine the technique and a bad wind/cars day still throws things off. Also to get this level of consistency I have to temp-correct the CRR.

Aero Testing...

---

Jim Manton / ERO Sports

I would say that the Notio will do some, but not all, of what you want. The phone app will analyze your ride and provide an estimated CdA. It produces a graph of distance vs speed and then it overlays the parts of the graph that it is using to calculate the CdA. I was looking back at one of my runs that is a high-to-low speed looping circuit and there we lots of gaps where the app seemed to indicate that it wasn't able to estimate the CdA. After the ride you can download the data to the GC Notio software and it will recalculate a new CdA and provide some diagnostics. I also have been exporting the data and reading it back into standard Golden Cheetah and using Aerolab. So far the GC Notio and Aerolab estimates are very similar.

I have asked Notio support if some of the diagnostics are a virtual elevation profile and the answer was all the elevation graphs are derived from the device and not a VE profile, so taking this along with the sporadic chunks of the ride used for estimating CdA, you can kind of surmise that they are using a different estimation strategy. Physics is physics, so you would be solving the same equations, they just use a different approach to solving the problem.

So long story short, you can use the phone app to get an estimate in the field, but you really should go home and use the GC Notio app and maybe Aerolab to refine and check the estimates.

As an aside, I did get some suggestions on simple methods for error checking with the device. The one I like is looking at wind speed on an out and back loop. If the average speeds differ by more than 1 or 2 mph, something is off and you should check your calibration.

I hope to get back to some more testing if the fires don't kick up again and will update when I have something intersting