Tilt plot- getting strange slope in ride

[GWPOS]

Just had to have my unit replaced. I took my new IPro and did the whole process of cal rides and such to get my new wind scaling, tilt etc. Now on my new unit I get weird power readings after going from uphill back down. I looked at my tilt plot and I can see why. I went and looked at a couple of my rides that were done on this new unit and they all show the same behavior. So out of curiuosity I looked at some rides from my old unit, and although not perfect they did not have the drastice tilt changes that this new unit has. If anyone can look at this example ride file and give me some ideas of what is causing it I would really appreciate it. If I use the route analyze feature it corrects whatever is wrong with the ride. However that does no good while I am actually riding. Thanks for any help or insight that anyone may have.

iBike_12_23_2011_1601_17_Miles_RAW.csv

(1003.89 KiB) Downloaded 689 times

[Velocomp]

We will take a look at this.

[GWPOS]

Thanks in advance. All my other data is looking good as of the last couple of rides. This tilt thing is causing me to see zero watts no matter how hard I am going. I know it eventually comes around but as you can see in my file it shows zero for awhile and then mysteriously goes the other way and shows 300 and 400 watts when just going easy.

[GWPOS]

Just did same ride as above and got about the same results. Here is the ride file for tonights ride. I have been doing these out and back rides on recovery days to check my calibration. For the most part it is pretty close except for the tilt plot??? Runnung the analyze ride feature in the software fixes the ride file, but its the actual ride that I need to be worried about. Thanks

iBike_12_26_2011_1732_16_Miles_RAW.csv

(1.04 MiB) Downloaded 681 times

[GWPOS]

And yet one more ride file from a different ride. Anyone have any ideas yet. I have tried everything that I know to do and nothing is getting any better. My old unit was never like this, although it did do some strange things once in a while. I really like the IBike but getting it set up this time is turning out to be almost impossible. Will go and try my neighbors unit and see how it behaves??? Like I said its the not showing any power while ging downhill...no matter how hard I go that bothers me.

iBike_12_27_2011_1726_20_Miles.csv

(1.34 MiB) Downloaded 609 times

[travispape]

Hi GWPOS,

Looking at these rides, it looks like on-the-road slope_offset algorithm is getting a stiff workout when you descend. Remember that the slope_offset algorithm has a 5 minute reponse time, and since the descents are only about 7 minutes long, they are almost over before the algorithm has a chance to completely respond. So I think the the main question is what has changed recently that might account for the iBike's tilt to change slightly as you descend.
Riders have a natural tendency to tuck when descending, and if your ride position changes it might change the flex of the bike or bars some. It doesn't take much torque on the bike or bars to become visible in the tilt plot.

Also do you have any ride files from your previous unit on the same route as the 23rd and 26th that you can send to me at tpape at velocomp?

Travis

[GWPOS]

Thanks Travis. I remember our discussion from before about this tilt lag. On my old unit it did the opposite. My numbers where always higher after turning around and decending where as the new unit show nothing for a while. I still have my old unit that I need to ship back to you guys. I could do the exact route with it and then see how the tilt looks with it. It had a wind port problem but should still do ok for this little test. My neighbor also has on e and I was thinking of using his and seeing what happens with it as well. I do know one thing, my old unit was never this stange though. I have attached a ride from my old unit ( a different route than posted before) and my new unit. They are the same route for the most part just one less interval. The Old IBike file was from before I had problems with the unit.

Old_IBIKE.csv

(1.77 MiB) Downloaded 691 times

iBike_12_27_2011_1726_20_Miles.csv

(1.38 MiB) Downloaded 673 times

[GWPOS]

Travis, forgot to mention that I tried very hard to keep the same position for the entire ride. I did this for 2 reasons, 1) So that I could verify my profile was correct 2) because I wanted to see what was happening to the tilt while riding. I did not stand or anything. Kept nice even pedal strokes etc. Basically tried to go at about 200 watts for the entire ride which I did not do perfectly but I did try. I do realize that even the slightest move could still have cause some kind of change in tilt no matter how hard I tried to not change my position. That being said I would not have thought that it would throw it off that far. Thanks

Clint

[GWPOS]

Here is the exact same ride as the new unit I have been using like you requested. I did it at lunch. Thanks again Oh and sorry the date and time are wrong. Forgot to set it after putting in the battery.

iBike_05_03_2011_1828_16_Miles.csv

(826.85 KiB) Downloaded 710 times

[Zoltan]

Hi GWPOS,

Looking at these rides, it looks like on-the-road slope_offset algorithm is getting a stiff workout when you descend. Remember that the slope_offset algorithm has a 5 minute reponse time, and since the descents are only about 7 minutes long, they are almost over before the algorithm has a chance to completely respond. So I think the the main question is what has changed recently that might account for the iBike's tilt to change slightly as you descend.
Riders have a natural tendency to tuck when descending, and if your ride position changes it might change the flex of the bike or bars some. It doesn't take much torque on the bike or bars to become visible in the tilt plot.

Also do you have any ride files from your previous unit on the same route as the 23rd and 26th that you can send to me at tpape at velocomp?

Travis

Hi Travis,

Is there anything against to decrease the 5 min response time of slope-offset to let's say 2 minutes? Why 5 minute averaging is really needed? (I guess we are speaking about simple, non time weighted averaging). Would it cause more problems than benefit?

Let me repeat here too that I am about to buy a Newton just for getting a reliable, very precise and accurate enough gadget for checking environment data including slope. Thus I will definitely continue the habit to change my riding positions when ascending and descending and do not care any rider profiles, because the power will come from DFPM. But wind speed, temperature and slope are holy data for me. So this is why I am asking whether there is any benefit in using 5 min response time and not 2, or let me be agressive 1 minute. The tilt is changing at once when I sit up or bend forward, that is one needs a relatively fast response from iBike to correct for the change in the tilt of the bike.

[rruff]

Not Travis, but I guess 5 minutes of data is used (vs 1min) so that small errors in barometric altitude do not lead to wide gyrations in the slope determination.

I wouldn't worry about your position changing the slope very much. Back in the day (Gen2) I noticed ~+.1% going up steep climbs and ~-.1% going down.

If your roads are rough, that can have a significant effect. After the fact, the software will do a decent job of fixing that, though.

[Zoltan]

Do you mean that slope_offset (or Auto Tilt) was invented and are applied just for the rough road effects?

All the other factors of tilt errors, which Travis has listed*, are just for theoretical discussions?

I am quite surprised to hear that if I correctly calibrate the tilt on my road bike, during the ride on some average-like quality asphalt there would be just a minor error of +/-0.1% in the slope accuracy, which is targeted to eliminate by the slope_offset.

I still did not get a clear answer why an accelerating bike does not cause a huge error in the instantenous slope accuracy.

* from Travis: frame flex, fork flex, head tube flex, play in the headset bearings, stem/headset interface, stem flex, stem/handlebar interface, handlebar flex, interface of the iBike mount, unequal tire loading (downhill vs. up), rider position & weight distribution, rider applying torque to handlebars or aerobars (can be a large effect depending on mount position).

[rruff]

The auto correction for tilt was to take care of a large number of people who had difficulty setting their tilt correctly. But the only times I noticed a significant problem on the Gen2 was when I was on rough pavement. I actually prefer to *not* have auto tilt most of the time, but there is no way to turn it off. Ideally being able to push a button and do a tilt cal while riding, and then turn it off, would be ideal for me.

All the things mentioned cause *brief* tilt errors, and there is no good way to account for them. One reason I liked using the fixed tilt was because I could have confidence in the numbers when the conditions were "normal". With auto tilt, the accumulated errors of the last 5 minutes are going to influence your current reading. For instance, if you ride on alternating smooth and rough patches of road then your numbers will always be off. Previously at least they'd be good in the smooth patches. Plus, if your ride includes long climbs the baro altitude measurement seems to have significant errors.

Oh... about accelerations causing slope errors... you have a wheel sensor that is used to calculate accelerations.

[Zoltan]

"With auto tilt, the accumulated errors of the last 5 minutes are going to influence your current reading. For instance, if you ride on alternating smooth and rough patches of road then your numbers will always be off. Previously at least they'd be good in the smooth patches. Plus, if your ride includes long climbs the baro altitude measurement seems to have significant errors."

Agreed. Really. If bike acceleration effect is filtered out using wheel sensor I must say no auto tilt is what a road biker wishes. Of course MTBers must love auto tilt.

[GWPOS]

Mountain bikers do not love the auto tilt. The 5 min required for it to self correct has already passed and is gone when you start your next climb or descent. It then has to keep recalculating but never catches up with itself. I noticed the same on the road while going through rolling hills, but I know they are working on tilt offset algorithym or something like that to address these issues.

[Zoltan]

Thanks for the correction.