How does drafting affect power readings?

[wellmt]

Hi

I've been trying to understand how power readings on the iBike would be affected by riding in a group (or train). Would the power readings be under reported or over reported and why? Would the Gen III also have any issues?

Thanks

[racerfern]

Power reading generally drop significantly in a draft or train, simply because generally speaking your efforts drop dramatically. I find that the iBike still shows accurate power with the exception of very high head winds or cross winds. Upwards of about 10mph headwinds or severe crosswinds the iBike seems to be "sheltered" somewhat and doesn't experience the air resistance that your upper body is up against. I feel wattage is under-reported in these high wind situations, however they usually only last a couple of minutes before you either turn in another direction or it's your turn to pull.

I have an iAero with an Edge 705. In long drafting situations, especially in high winds, I switch to the "Enviro" screen to have accurate effective wind display. I let the Edge display my speed and power. Then I will shift around looking tor the sweet spot usually slightly to one side or the other. Finding the lowest wind speed always yields the best spot in the echelon.

[coachboyd]

I have tons of data with drafting in it. Let me show you the crit that we recently did in Greenville. I will show you a screenshot of the last lap with the Quarq compared to the iBike.

To give you a story of what happened, my teammate and I got to the front a little too early (about 1.5 laps left to go). Fortunately, my teammate is one of the strongest cyclists I know and for that lap and a half was able to pull by himself and keep me in a good position to sprint. Unfortunately, there was a big headwind for the sprint and I waited a little too long to start my sprint and one guy got the better of me.

But you can see the data here for the last lap of the crit when I was just sitting behind one person. You can see a couple times the iBike reads a bit higher, and a couple times it reads a bit lower. I think for this section the iBike read about 15 watts lower than the Quarq, but that also includes me standing up and sprinting (which does change my CDA)



Now, because of the wind sensor on the iBike, you can analyze group rides and races in ways never done before. You can actually see how much work you were doing afterwards and whether you raced a smart race or not. We were showing a lot of people this graph at Interbike. This was a crit I did in Statesville, NC. I managed to get in a break pretty much right away, and I had one teammate there with me in the group of 5. You can see at the start of the race, I have circled in red everytime I was at the front pulling. We end up lapping the field and you can see now the wind speed dramatically drops as I am sitting in the pack.

The bad part of this race was that Rashaan Bahati (who would go onto win the criterium National Championships a couple weeks later) was also in our group of 4 (one person got dropped before we lapped the field). I know that I am not going to beat Rashaan in a sprint, so I have to try and get away from the field again. I gave it two more attempts to try and break away, but neither was successful. The field ends up all sprinting together and Rashaan wins the race.

But for athletes who want to see why they didn't feel great at the end of a ride, or coaches who want to analyze a race, this info is amazing to have. Not only can you look at how much watts you were putting out during the whole ride, you can see if you were sitting in and conserving, or at the front and doing (too much) work.

[rruff]

Do you have a comparison between the Quarq and iAero for a whole race... like maybe the last one shown above? How do they compare when you are in a large pack?

[coachboyd]

I'll show the race I had last weekend with the Gen 3. This was the data straight out of the unit as well. . .no analyzing afterwards or anything. The numbers you see on the screen are what was being shown on the computer head.

This is a race where I sat in the pack pretty much the entire day and had my teammate lead me out for the field sprint finish. As you can see with the power meter comparison, the numbers are almost identical, except the peaks (600+ plus) are a bit higher on the Quarq compared to the iBike, due to my CDA changing.

[coachboyd]

hmm, it might be a new file format with the new versions of the software. I didn't think about that.

Let me see if there's something I can do to get it to work with the software that's out right now. If not, I will do screenshots for everybody.

[Velocomp]

We have a new version of iBike 2 coming that works with the Gen III units. Older versions of iBike 2 won't be able to open Gen III files.

We'll have everything up in the next couple of weeks...

[Velocomp]

Here are screen shots from Boyd's ride. This is his whole ride; you can see exactly where he was pulling (spikes of positive wind)...

[Velocomp]

Here is his big 1000W sprint at the end of the ride

[Velocomp]

Here are the overall stats from the ride. Remember, this is straight from the Gen III unit--no post-ride analysis!

[Velocomp]

Here is the tilt curve for the ride. Remember that we're all used to seeing elevation changes in the hundreds of feet; the vertical scale has a divisions of 10 feet and the horizontal scale is in miles.. Average pre-analysis tilt for the ride was 0.01%

[Russ]

Wow, this is a huge improvement. Looking forward to my gen III unit!

I notice in Boyd's 1000watt sprint power curves, it looks like a phase lag on the higher power, say 300w and up, parts of the graphs. I mean, the dpfm rise and fall lead the gen III, resulting in (if it really is a lag in reporting) a greater difference in the difference graph than was real if the lag was removed. But there is no visible lag in the lower power peaks ??? I would not think the lag, if it is there or not, would affect the average power and all that so it seems, for our purposes, mostly a cosmetic thing. Any comments from the guys in the know on this?

Thanks,
Russ

[rruff]

The chart below was made from Boyd's file, with a 30 sec rolling average. It is Quarq power - iBike power and is typical of the whole race. The first vertical axis line is 100W, so what you are seeing is a +- variation of up to 100W. There is a greater fluctuation between the two than I expected, but oddly enough they cancel out, with the iBike on average reporting 7W less than the Quarq. It looks like reports of the iBike being unable to function in a drafting environment are greatly exaggerated. Or is this something that has changed between Gen II and III?

[coachboyd]

Well, keep in mind that I was still on Beta firmware. And there have been lots of tweaks to the firmware trying to get all the features to work so it'll be accurate in all conditions, drafting, flats roads, bumpy, climbing. There is always going to be variation between any two power meters in a race type situation.

I wish I would have had a power tap wheel on for this race as well, I am sure you would see similar results with the Power Tap compared to the Quarq. Unfortunately, there are no more crits until the end of next March. I will be able to get some good group ride data in about a month though with a power tap, quarq, and Gen3.

I will post those results here.

[wellmt]

Thanks for the replies and useful info guys. It's amazing how much extra info the enviromental sensors on the iBike can actually give you. A big part of learning to use the iBike effectively is undertstanding this stuff.

[travispape]

Ok, I'm back on the boards after a 2 month hiatus. I didn't mean to be away so long, but it was the only thing I could do.

The chart below was made from Boyd's file, with a 30 sec rolling average. It is Quarq power - iBike power and is typical of the whole race. The first vertical axis line is 100W, so what you are seeing is a +- variation of up to 100W. There is a greater fluctuation between the two than I expected, but oddly enough they cancel out, with the iBike on average reporting 7W less than the Quarq. It looks like reports of the iBike being unable to function in a drafting environment are greatly exaggerated. Or is this something that has changed between Gen II and III?

I know you had to open Boyd's ride in Excel because of the new CSV file format, but note that you can view the exact same thing in iBike3 in the Power Meter Comparison window by selecting the "Power Diff" plot and then turn off filtering. (By the way, have people noticed that you can type any filter size in that control that you want to?)

In a nutshell, what you are seeing is that iBike is a different technology than power meters that directly measure rider force. You said, "but oddly enough [the fluctuations] cancel out", but you yourself have posted how well the iBike works over longer time scales and how tolerant it is of different wind conditions. Remember that the iBike technology is to measure the forces of resistance to your motion and the sum yields the rider power; but in order to do that, the FW has to process and clean up each of the signals that it measures. Over time scales of a few seconds and longer, it works really well as we have shown over and over again and you have seen yourself; however, on a time scale of 1 second (roughly 1 to 1 1/2 revolutions of the crank) there is no way to expect the iBike's technology to get the sum to be exactly the same as a DFPM.

There are some riding conditions that are going to cause the iBike technology to diverge if you zoom all the way in to 1 s resolution and riding in the dirty air of a crit is one such instance. There is a lot of wind turbulence in a crit. The good news is that the iBike's wind measurement behaves very linearly and it only takes a few seconds of integration to get an accurate reading that compares well with any DFPM. The bottom line is that the iBike technology works extremely well over very large range of riding conditions for time scales of only a few seconds and larger.

And keep in mind that the DFPM power meters have their own issues at time scales near 1 s. For example, due to the PowerTap's 1.26 s sampling rate, there are certain cadences that make the PT power signal fluctuate more than reality. The SRM measures power per revolution, so if your cadence is not exactly 60 rpm, what is it recording at the 1 s recording rate?

Anyway, if you keep in mind that the iBike is significantly different technology, there are no surprises here. You can spend a lot more money, insert something mechanical in your drive train and with a direct measurement of drive force and get a good measurement of speed, power, and maybe cadence down to 2 or 3 s time intervals; or you can get an iBike and get a good measurement of slope, speed, cadence, wind, elevation, and power down to time scales of about 10 s (or even shorter time scales in favourable conditions). Make it an iAero and you can measure your CdA down to time scales of 30 s so that you can learn how to ride faster.

Travis

[Tom_Anhalt]

And keep in mind that the DFPM power meters have their own issues at time scales near 1 s. For example, due to the PowerTap's 1.26 s sampling rate, there are certain cadences that make the PT power signal fluctuate more than reality. The SRM measures power per revolution, so if your cadence is not exactly 60 rpm, what is it recording at the 1 s recording rate?

Actually, Travis, the PT signal at those "special" cadences DOES reflect reality...it's just that those cadences happen to correspond to conditions where one of the fixed time samples has 1 more downstroke in it than the preceding and following samples. The data stored in those samples is merely the average of the torque and the wheelspeed over that 1.26s and thus it IS reality (or pretty darn close ) It's just not how us humans "expect" to see the data appear But yes, fixed time rate averaging/sampling gives a more "spikey" appearance to the data stream.

Also, I believe the SRM stores the average of all the completed pedal strokes in the selected recording time period. Obviously, if the cadence is less than 60, there are going to be some repeated sample values. Also, if the recording time period is increased, it STILL averages ALL of the completed pedal cycles over the whole time period. Which, IMO, is the correct way of handling things (I hate downsampling)

Of course, all of this brings up the reasons why I don't particular like comparing the output of PMs on a "point by point" basis. Due to the differences in sensors, sensor technologies, support electronics, algorithms employed, etc., there's bound to be differences in the "timing" of how the power is displayed and recorded. Even though the iBike2 software does an excellent job of "aligning" things...one can still see the "leads and lags" in the comparison plots...and I wouldn't put much stock into what that all means, just so long as the "content" ends up being as close as possible for each "power event" in the end.

Now...the question I have for you is: Since the CinQo outputs a data packet every 1/4s that has the values from the last complete pedal cycle, how does the iAero sample and record that data stream? I've been told that the Garmin merely stores at each sample period whatever data packet happens to be "presented" at that sample time (i.e. downsamples the data stream). Does the iAero do the same? Of course, IMHO all 4 data packets should be averaged together for each 1s sample, but I understand that sometimes there are reasons something like that can't be done.

[travispape]

Actually, Travis, the PT signal at those "special" cadences DOES reflect reality...it's just that those cadences happen to correspond to conditions where one of the fixed time samples has 1 more downstroke in it than the preceding and following samples. The data stored in those samples is merely the average of the torque and the wheelspeed over that 1.26s and thus it IS reality (or pretty darn close ) It's just not how us humans "expect" to see the data appear But yes, fixed time rate averaging/sampling gives a more "spikey" appearance to the data stream.

Also, I believe the SRM stores the average of all the completed pedal strokes in the selected recording time period. Obviously, if the cadence is less than 60, there are going to be some repeated sample values. Also, if the recording time period is increased, it STILL averages ALL of the completed pedal cycles over the whole time period. Which, IMO, is the correct way of handling things (I hate downsampling)

Of course, all of this brings up the reasons why I don't particular like comparing the output of PMs on a "point by point" basis. Due to the differences in sensors, sensor technologies, support electronics, algorithms employed, etc., there's bound to be differences in the "timing" of how the power is displayed and recorded. Even though the iBike2 software does an excellent job of "aligning" things...one can still see the "leads and lags" in the comparison plots...and I wouldn't put much stock into what that all means, just so long as the "content" ends up being as close as possible for each "power event" in the end.

Ok, I think we agree that once you start getting close to the time scales of pedal cadance that each power meter is going to record a power number that is a mix of the actual power output of the rider and the details of each power meter's respective technology. Unless you have a research grant to get a SRM Science model--it is interesting to see the forces on a ms time scale if you can afford to add an extra zero to the check. With the lessor SRM models and all of the other power meters, the reality that you record at a 1 s recording rate has variety of technical details convoluted in. It only takes a couple of seconds for the DFPM models to get to a power number they can agree on. It typically takes a couple more seconds of integration for the iBike power to have enough data to agree also.

John said he contacted you about your other question.

[rruff]

In a nutshell, what you are seeing is that iBike is a different technology than power meters that directly measure rider force. You said, "but oddly enough [the fluctuations] cancel out", but you yourself have posted how well the iBike works over longer time scales and how tolerant it is of different wind conditions.

The only odd part was that there was a large fluctuation (+-100W) between the two *and* they ended up canceling out. Note that the graph above uses a 30 second rolling avg, so it is already significantly smoothed. I'm puzzled by what could be causing a such a large fluctuation on that time scale... a few seconds I understand, but not 30 sec. At any rate, I'm happy that they cancel out over the whole race.

[travispape]

The only odd part was that there was a large fluctuation (+-100W) between the two *and* they ended up canceling out. Note that the graph above uses a 30 second rolling avg, so it is already significantly smoothed. I'm puzzled by what could be causing a such a large fluctuation on that time scale... a few seconds I understand, but not 30 sec. At any rate, I'm happy that they cancel out over the whole race.

A big chunk of the explanation is the time lag of the iBike power. The following graph shows both the power and the power difference for 30 s averaging.

crit power difference.png (105.25 KiB) Viewed 13809 times

For whatever reason, there is enough time delay on the iBike power signal that the power difference looks bigger than it needs to look. This also explains why the difference ends up canceling out.

Laps for this crit were only 0.7 mi, and Boyd may have been in and out of his seat several times during that 5 mi window. Also, there is a lot of wind turbulence in a crit. In an intense race like that, Boyd is probably pulling and torquing on the bars quite a bit causing quick, offsetting changes of tilt. With laps that short, he is almost alway leaning in corners. In short, this is a challenging environment for each of the iBike's sensors; however, despite that Boyd regularly gets good aggregate power results even in crits. The power difference you see at 30 s is magnified by the time offset between the two power signals.

Travis

[Russ]

In my effort to 'restate' in simple terms what I see with the graph here is another try:

I notice that (by eyeball) in the above comparison graph, the iBike seems to read higher power at the bottom of the valleys and lower at the peaks. If correct, this suggest, to me, a higher degree of smoothing.

I also notice regarding the 'lag' or if you will the way the iBike curve seems to trail the DFPM,
that if you grab values on the slope, because of the lag then the difference is exaggerated and obviously caused by the lag. I say this because if you were to move the iBike curve a little to the left, those differences would largely disappear, which I think is what Travis is saying.

Russ

[travispape]

I also notice regarding the 'lag' or if you will the way the iBike curve seems to trail the DFPM,
that if you grab values on the slope, because of the lag then the difference is exaggerated and obviously caused by the lag. I say this because if you were to move the iBike curve a little to the left, those differences would largely disappear, which I think is what Travis is saying.

Yep, that's the main point.