The garmin speed and cadence sensor works great. It's pretty much all you need to keep track of distance, current speed and cadence. If you don't have a power meter, this is about the best you can do. I feel though that the data that you get from it is a little skewed and usually on the lower side compared to how it would be outside. But it gives you something to go off of besides heart rate only.

Any rear-wheel speed/cadence sensor + bike computer will work to display your speed, which is analogous to power.

See here for the speed to power conversion curve:
http://thebikegeek.blogspot.com/...tter-and-better.html

Another option which is brand new but seems cool (albeit with a monthly fee) is the Trainerroad.com website. You would need to purchase a $30ish ANT+USB stick for your computer but then the Garmin speed/cadence sensor (which is ANT+) will transmit data to your computer, and it'll display your power and other goodies on your computer screen like a Computrainer. Neat stuff.

In addition to the options mentioned above:

I haven't used it, but Golden Cheetah supports "virtual power" in realtime mode using ANT+.

If you don't want to run a computer with the trainer you can use SportTracks plus a plug-in to estimate power for post-ride analysis.

All of the options mentioned lack "calibration" so results for power estimation may vary. You will at least be able to track heart rate, distance and cadence.

Long time lurker first time poster.

Would it be possible to just read the speed data that the cadence/speed sensor is sending to the watch with the ANT+ USB stick that came with the watch?

If that is possible, then we can just plug the speed into that cubic equation on http://thebikegeek.blogspot.com/...tter-and-better.html and display it on the computer screen - instant and free power meter!

It would be awesome if someone with ANT+ expertise can chime in.

I have ANT expertise and I built something for you real quick: www.TrainerRoad.com :)

---

CEO at TrainerRoad
Co-host of the Ask a Cycling Coach Podcast

We find the power calculations to be pretty consistent if you tighten your trainer wheel down the same and keep your tire at the same pressure. The power curve offset moves a bit once the trainer warms up, but once you get going it's pretty consistent.

This is all dependent on the type of trainer you have though. I've been testing with the Kurt Kinetic Road Machine which is known to have a power curve that's not effected by heat.

---

CEO at TrainerRoad
Co-host of the Ask a Cycling Coach Podcast

Ummm...not really.

My experience with running a Fluid2 trainer in conjunction with a REAL power meter is that the load tends to vary significantly when the fluid heats up at certain power levels (IIRC, it was at ~250W and above).

Combine that with the uncertainty due to tire choice, pressure, roller force, etc....I wouldn't put too much stock in having anything remotely "accurate" about a power guesstimate based on speed alone.

Sorry to throw a wet blanket over the idea...

---

http://bikeblather.blogspot.com/

Buy at PowerTap hub.

---

Park City Condo Rentals
http://www.traverseridge.com

And quite a number of folks have had the exact opposite findings, finding that it's spot on dead-on accurate, with Powertaps. Even KK trusts theirs enough to place a speed-power converting head onto their KK.

My Fluid2 is dead-on reproducible. In a single week workout, when I'm targeting exact speed/power levels, my HR values are spot-on reproducible every time. I do check tire pressure, and make sure I lock it on the trainer similarly, but it's not hard at all. It has been extremely good for power-based training. And I can definitely notice the difference between 22 and 21mph quite readily (even more apparent at 23 to 22mph for me) so if there is any drift, it's way under 0.5mph, which outstanding precision.

Taking indoor numbers to outdoor numbers has its own challenges - I'll agree that you still should test outdoors to get the best outdoor numbers as conditions between both are different.

I do think it's exaggerating to say the Fluid2 power curve is not reproducible - there have been people who have posted their PT curves on from KK or Fluid2 and they've been dead-on.

To add what lightheir wrote, I've been testing the KK road machine power curve with a power tap and had my numbers line up really close. If I tighten down the wheel the same and have the same pressure I can use a consistent offset and get the avg watts for my intervals with 0-5 watts.

I have a fluid2 sitting in my garage waiting to be tested. I'll be blogging about those results once i've done a few rides.

---

CEO at TrainerRoad
Co-host of the Ask a Cycling Coach Podcast

Does anyone have any experience with consistency of the TacX trainers using speed as proxy for power (i.e. Satori or Booster)?

Well...quite a few number of folks have shared the exactly what I'm talking about right here on ST. Here was a thread just last year:

http://forum.slowtwitch.com/...ring=fluid2;#3039889

As far the KK goes, I don't know...but, we're talking about the Fluid2 here..



So...what you're saying is that you haven't actually checked it against a power meter. Got it.


Well...I'd question as to how exactly those power curves were generated. Here's a big reason why...check out the plot below. It's from a workout I did last January that I call the "Shortened Hour of Power" where the plan was to gradually increase the power every 3 minutes while throwing in a 15s "sprint" at the end of each 3 min. interval. I did the whole thing in a single gear. Take a look at the third interval where I started pedaling a bit more "evenly". The average power was 192W at 93rpm. As you can see, the average cadence (the green plot) stays basically constant throughout the 35 minute session, while the power (the yellow plot) gradually creeps up. In the last interval, which had the EXACT same average cadence (and thus same rear wheel speed) as the 3rd interval, the power was up to 223W for the interval. Thats a 29W difference!

---

http://bikeblather.blogspot.com/

This is interesting. Is cadence being measured with a sensor independent of the powertap in your trial?

Edit: i ask in reference to this post: anonymous.coward.free.fr/wattage/rosetta/rosetta_details.html, which refers to powertap cadence holes near your target cadence.

That plot is from a CinQo/Garmin 705 combo. Cadence from the CinQo.

---

http://bikeblather.blogspot.com/

Oh...and here's the plot of the constant cadence cassette "sweep" I was referring to in the post I linked to above. This is after warming up the trainer for 16 minutes. Compare the cadence and speed curves ( I used a Garmin speed/cadence sensor in this run) to the power for laps 4, 8, and 10...not too stable, huh?



edit: BTW, my intent with this run was to try to create a power curve for the Fluid2...actually, I was going to plot P/V vs. V^2 to get the "virtual CdA and Crr" (like I eventually did with the LeMond Revolution http://forum.slowtwitch.com/...rtual%20cda;#3178482 )...but, after seeing the above, I realized the effort was somewhat futile.

---

http://bikeblather.blogspot.com/

Good data, thanks. I suspect you're right about the fluid warmup, but I still don't think the variance is enough to make the Fluid2 not a good candidate for power-based training.

One critique for your data - you are using CADENCE vs POWER, which is not entirely the same as using speed. In fact, your graphs themselves show a much tighter correlation with SPEED vs POWER. I don't know why the speed/cadence is uncoupling, but your data is clearly showing it, whereas the speed - power correlation is much tighter.

I could theorize about why this might be, and it would just be guessing as to why the cadence/speed are uncoupling, but your data points are clearly showing both uncoupling of cadence vs speed as well as a much tighter correlation of speed vs power, to the point that I'd be very happy using it speed as the power correlate.

If you've got data with a flat speed profile and power, that would be helpful as well.

Either way, its still WAY better than HR.

Really? You don't think a variance of 29W over ~30 minutes at a constant cadence (and wheelspeed since it's all in one gear) isn't "enough"? Really??




Ummm...when in a single gear, cadence is directly proportional to speed...unless my chain is slipping, or something :-/




I think you might be getting fooled by the fact that cadence is reported in integer values (and therefore the plot looks more "chunky") as opposed to the finer resolution on the speed. However, if you really think that the speed power correlation is "tighter", I'm having a hard time seeing how you're getting that. I'll look into the individual correlations later if that will appease you...


I'm thinking your confirmation bias might be showing here...



It's right there in front of you with the first plot (cadence is proportional to speed)...also, the individual intervals in the second plot are at nearly constant cadence/speed as well...



In this case, I don't agree...

---

http://bikeblather.blogspot.com/

TomA - you're ignoring your very own data and just going on a presupposition.

I'll make it clear:

Compare the distance between the SPEED (blue) curve and the POWER curve in the 2nd graph. (You conveniently left out the blue curve in your first graph just to make your point about cadence.) The speed and power curves stay nearly the same distance apart during all intervals.

I admitted that I can't explain why your cadence doesn't couple with the power - I would expect it to, but as we're debating the merits of SPEED, you must look at the speed data and not 'infer' from your cadence data, which doesn't correlate well with either speed or power on your 2nd graph.

You can't cherry pick your data points. You might still be right, but at least with what you've posted, the data is supportive of good consistency for speed-power relations. If you reshow the 1st graph and include the blue speed curve overlaid on your power data, it may also further support speed-power.

I never use cadence as a surrogate for power, ever, and I don't know anyone who does as well.

Are you referring to the green curve?

---

http://bikeblather.blogspot.com/

Perhaps as the Fluid 2 "warms up" it provides more resistance at a given power output? Thus with the same gear and the same cadence you would produce an increased power output with increased duration?

---

__________________________________________________

"I'm gonna show you a hill that would choke a mule."

I'm surprised that the power load increased as you road. I'd think that the fluid would get thinner as you road and not take as much power to get through it.


We've seen big drops on some trainers/rollers. Even with the KK the wheel heating up changes the resistance. This usually flattens out after about 10 minutes though.


Tom, what do you think about a roll down test to account for this sort of drift on trainers that are affected by heat? We're trying to work that out now. ANT doesn't send us enough speed messages though so we are having to interpolate some of the values. I'm not convinced that we will be able to do it.

The idea would be that if you could do a quick roll down from 20 to 10 mph while you worked out that would take in account for the sort of drift you were seeing and adjust the offset for the power curve. On some trainers, you might have to do multiple roll down tests during your workout to correct for drift. Kind of a PITA but if you did it right it wouldn't be too intrusive.

---

CEO at TrainerRoad
Co-host of the Ask a Cycling Coach Podcast

Nice try to play dumb.

As per the axes on your very own data:
Green - cadence
Yellow - power
Blue - speed

Blue and yellow curves correlate well. Nobody correlates cadence to power as you are doing.

The behavior is consistent with what was reported about the "levels" in the thread I linked to above...i.e. increasing temps correlate to increasing load. Although one would expect things to get "easier" with higher temps (as the fluid viscosity drops), that would only be true in an "open" sytem. In a closed sytem like the Fluid2, my suspicion is that the increasing load is pressure related.




Right...and in that case it's due to the rolling resistance of the particular tire decreasing as it heats up. In some Crr testing I've done on some 4.5" Kreitler rollers lately, I find that there's a bit of hysteresis in the power values based on what the tire speed was prior to the particular interval. In other words, it doesn't necessarly warm up to a constant temp...it's dependent on the wheel speed (and thus, on rollers, the power).



That might work...you might want to try not having the speed sensor trigger off of the rear wheel, but glue a magnet to the roller drum and then mount the sensor there. That way you'll get more "triggers" per recording event...and presumably a more accurate speed calculation. I've done this for my roller Crr testing since I didn't want to have to compensate for tire diameter when testing various tires. You'll have to remember to enter the circumference of the drum instead of the tire for the rollout setting in your computer, though ;-)

---

http://bikeblather.blogspot.com/

Not playing dumb, just trying to give you a chance to correct your mistake. I just thought that you were misreading the plot. After all, you stated that the speed and power curves kept a constant spacing between them within each interval. If you really think that the speed curve keeps a more consistent gap to the power curve than to the cadence curve in those intervals...well, then I don't know what to say. Especially if you look at Lap 4 and Lap 8.

Since you don't seem to be recognizing your mistake yet, I'll now address your previous statements:


Again...look at Lap 4 and Lap 8 (the higher power intervals) in particular, and make sure you're looking at the correct gap. BTW, I didn't "conviently" leave the speed trace off of the first plot to make a point. It just doesn't exist. There was no separate speed sensor on board in that run, and with a cadence sensor and knowing the gear ratios, there's no need for one. Wheel speed and cadence are the same "signal", just taken at different parts of the mechanism. As long as the chain doesn't "slip", or something crazy like that, the ratio between cadence and wheelspeed is fixed.



Wheel speed and cadence are the same "signal", just taken at different parts of the mechanism. As long as the chain doesn't "slip", or something crazy like that, the ratio between cadence and wheelspeed is fixed.



Frank? Is that you? ;-)



That doesn't mean that using cadence as a surrogate for wheelspeed isn't correct. It's simple math...and recognizing that it's a chain drive with fixed ratios. Holding cadence constant is a LOT easier IME than trying to hold wheel speed constant when one is doing testing like I was describing above.

BTW, are you aware that RChung consistently looks at speed data and cadence and is able to back-calculate the gear ratios? That wouldn't be possible if one couldn't calculate speed knowing cadence and gear ratio.

http://anonymous.coward.free.fr/...ng/testprotocol.html

http://anonymous.coward.free.fr/...ents/components.html

---

http://bikeblather.blogspot.com/

Well, you've proven to me with your last post that you are intent on ignoring your very own data just to prove to yourself your own imaginary points.

It doesn't take a genius to see that in your 2nd graph, the green (cadence) curve and speed (blue) curve do not correlate, so any bets you're making on saying cadence = speed are totally off on your graphs (duh).

You can still coast on a trainer as well - if you stop pedaling, speed does not go to zero. Dead spots in the pedal stroke could account for some of the variance, although I freely admit that I'm surprised you have so much variance of speed vs cadence in your graphs, which I wouldn't expect. Regardless, it's there, and obviously contradicts any such statements you make about locking cadence to speed (or power).

I'm not the one making stuff up here - in fact, I'm agreeing with everything your graphs show.

That's a good idea to put it on the fly wheel. We're trying to develop a commercial product so we wouldn't be able to have everyone set up their unit like that, but it might be a more accurate way to collect the data for the roll down offsets and then interpolate the points for end users. Maybe have them do 3 roll downs or something.

You're data above though is disheartening. It looks like that even with a roll down test the Fluid 2 would vary a ton in just a short interval. With the KK it looks like after a warm up period it's pretty stable.

Like I said before, I have a fluid 2 sitting in my garage waiting to be ridden on. I'll probably post some excel files back to thread after I've done a few rides. I suspect I'll have the same conclusion that you do - fluid2 sucks for speed to power :).

We're thinking of making our own "Crank Armstrong" that KK has for their tests. We would just hook a motor up to the bottom bracket and set the speed/power output via power taps. That way we could output 300 watts for a few hours and see what happens to the power curve. Same thing at 200 watts, 100, ect. Maybe we could get more data about those "levels" in that other thread.

---

CEO at TrainerRoad
Co-host of the Ask a Cycling Coach Podcast

I think what is trying to be explained is that the gearing was different between the intervals, but stayed the same during each interval. Not that the gearing was the same for all the intervals.

If you take any individual interval the speed/cadence are identical. However you can see the power requirements (watts) particularly during interval #4 for example climb throughout the interval, even though the speed/cadence remain the same.

Huh? Are you missing that in the 2nd graph, the run was a constant cadence "sweep" across various cogs? As I explained before, the idea was to vary power and speed to then be able to plot P/V vs. V^2 to derive the virtual CdA and Crr coefficients. Each of the even numbered "laps" was done at a constant cadence in a given gear...and within each of those laps, the gear was held constant and you can easily see that the blue and green lines track exactly the same with a constant offset between them on the graph. The power, however, tended to rise within the higher power level "laps" as compared to BOTH the speed AND the cadence lines.


Strawman. I explicitly said the intervals were constant cadence through each. That means no coasting.


Not likely...especially when you realize that the cadence is triggered on a once per revolution basis.



Well then, here's more graphs for you to agree with...the first being a comparison for cadence vs. speed and power vs. speed for "lap 2" (one of the more "well-behaved" intervals) and the second graph being the same thing for "lap 4". I don't know about you, but judging by the R^2 values of the fits, it sure appears to me that speed correlates better to cadence than it does to power when looking at it on a point by point basis (which I actually am not particularly fond of doing...but, I digress...).

BTW, don't be distracted by the "banding" in the cadence vs. speed plots. Just remember that cadence is reported and recorded in integer values and you'll understand...of course, even WITH that artifact the correlation is significantly better than power vs. speed in BOTH cases.

Lap 2


Lap 4

---

http://bikeblather.blogspot.com/

The two different graphs were with 2 different bikes. One geared one not geared

---

Styrrell

Yes...they were actually 2 different bikes (not that it mattered), but both had multiple gears. The main difference between the 2 plots was that in the first one I didn't shift (So yes, effectively "ungeared"). Well, mostly anyway...I think I shifted after either the 1st or 2nd interval, but then didn't need to shift the rest of the way (Funny thing...I could probably figure out where I shifted by comparing the speed vs. cadence...funny how that works ;-). And the average load of each 3 minute interval increased by nearly just the right amount I intended to do (~1-2W every 3 minutes) and I was able to just keep pedaling at ~93 rpm (my preferred cadence anyway). It's almost like the load unit was designed for this workout ;-)

In the second plot, I was keeping cadence constant and switching between gears to get large variations in speed and power in an attempt to characterize the Fluid2's load curve. Obviously, if it can't be even remotely consistent within a single 2-3 minute interval, that's going to make characterizing the load curve quite difficult (at least not without a temp sensor on the casing as well! ;-)

---

http://bikeblather.blogspot.com/

We are trying to make a "Crank Armstrong" like the one Kurt Kinetic has: http://www.kurtkinetic.com/testrobot.php

We plan on getting a 1HP motor with a speed controller. We would then attach a cog to it and put it behind the bike. We'd put a chain from the motor to our power tap which would be on a bike (in a trainer).

We'd then regulate the power output via the speed controller while looking at the watts outputted by the power tap. By recording the data we should be able to quantify the drift that you and others are experiencing with the Fluid2 (and every other trainer we can get our hands on).

Do you see any problems or have any improvements on this idea? I'd rather attach the motor to the bottom bracket, but that doesn't seem as easy as just putting a cog on the motor. We'd also use a Continental trainer tire to help guard against tire wear that could throw off crr.

Edit: We'd also need a gear reducer. We'll have to do some math about what sort of ratio we need based on the cog we put on the motor and how many RPMS it's at.

---

CEO at TrainerRoad
Co-host of the Ask a Cycling Coach Podcast