Data collected using onsumer-available power meters.

Somehow you've not convince me...

The difference falls within the noise of the data error. Let's assume the SRM and PowerTap have an accuracy error of 2.5%. That would be +/- 2.5W at 100W measured for each. Anything within +/- 5W @ 100W measured difference between the two should be considered noise, and inadmissible into evidence. That's good testing protocol. At 400W, you have a +/- 20W variance.

Basically, if you are not measuring, consistently, a greater difference than the inaccuracy between the two, then you don't have good data. Your graphs look very noisy.

You don't know what you're talking about.

C'mon guys, don't you DARE express a difference of opinion, or appear to have fun discussing a subject, or the mods will shut this thread down faster than an unlocked bike would be stolen at night in Detroit!

Oh, thank you for showing me the way. I never knew that noisy data that doesn't account for normal driveline losses, and measurement device error could be so clear. All my experience and education has been for nought. Your words have relieved a heavy burden from my shoulders.


Thank you


thank you


thank you!!!!

I hope this doesn't represent all your experience and education but, if so, yeah.

You're welcome.

Arguing test data with RChung

= very bold move.

Good luck.

More an argument over the testing devices' published error and no obvious account of it in his examples. My initial point was that consumer available power meters could not measure difference in chain line power outside of its published error.

I would like more insight into the conclusion. As I said, I can't see a difference between the 42x12 and the 54x15. It's certainly possible that one can be calculated. Statistics are funny that way. The only one that seems to show a clear advantage is 54x16 and I have to wonder if that isn't chainline rather than chainring size.

And the difference probably doesn't matter, but it's nice that we can have a civil discussion about it. I guess if it doesn't cut into the sales of steel bikes, certain elements (Fe?) don't care enough to derail the discussion.

I am the OP.
I am not an engineer and really don't understand the charts but from I am gleaning from this is that
there is no discernible difference between riding the big chainring or the small other than you might save some chain wear.
How to pro racers handle this... do the stay in the larger chainring so upshipfting is easier?

i watch a lot, maybe too much, pro cycling. the announcers often mention that so-and-so "is in the BIG chainring on this climb!". they just as often fail to mention the cog in back that he or she is using. in MTB racing they also fail to mention whether or not the racer is run 29" or 27.5" wheels.

it's really just a macho thing. there's a club around here that's called "53x11". haven't seen one called "34x28" yet. good thing is that it's even worse in the fixed gear set. oy vey!

Links to published articles on the subject of bicycle drive train efficiency.
Cycling Power Lab

Drivetrain efficiency of a modern bicycle (i.e. derailleur system with typical road gear range) peaks at about 98% in optimal conditions however variations of as much as 5% (down to 93%) are possible at realistic power outputs.
As power output increases efficiency increases because frictional losses become a smaller part of total input power. Typical best-case efficiency of a drivetrain in the 200 – 300 watt range is 96-97.5%. Above 300 watts typical best-case efficiency is 97-98%. Read on to understand what we mean by “best case”…
Efficiency is higher when using larger sprockets because the chain benefits from a less extreme radius of rotation – chain links going around corners cause greater frictional power losses. The take away here is that if you can achieve the same gear using a large chainring + larger sprocket combination than a small ring + smaller sprocket combination it is a worthwhile consideration. The efficiency difference between an equal gear that involves the 24 sprocket and the 13 sprocket can be worth 1-2 watts when riding in the 200-400 watt range.
“Cross Chaining” (riding with the chain at angles between the chainring and rear sprocket) really hurts efficiency – it’s pretty obvious that frictional losses increase in this scenario. You may intuitively avoid this scenario because it can be noisy and to avoid dropping a chain but efficiency of power transfer is another key consideration.
Efficiency falls in lower gears (riding at higher cadence) and improves with higher chain tensions (riding at lower cadence). Adusting ones cadence for this goal alone is a risky decision to take - the total efficiency of a rider and bike is more complicated than simple drivetrain efficiency. But have Tony Martin & Bradley Wiggins - two of the best time triallists at the time of writing - taken conscious decisions in this regard? This Bradley Wiggins Interview in which he talks about cadence and "something to do with rolling resistance and with the gears" suggests they have.
On a perfectly clean chain in a laboratory environment choice of lubricant makes little difference to efficiency. The real value of lubrication is to fill the gaps that would otherwise be filled by dirt and grime – things that do increase friction and decrease efficiency. Lower viscosity lubricants maximise efficiency. Friction-Facts suggest savings of 1-3 watts just by thoroughly cleaning and re-lubing new chains with thinner oil and sells “Ultrafast Chains” which have benefited from this process.
Brand new chains are less efficient than chains that have been run-in. The greatest efficiency gains are made in the first hour of use but gains continue throughout the first several hours of use. Don’t use a brand new chain for a key race unless you want to give up half a watt.
Cheap derailleur pulleys compared to high end equipment can cost 1 watt. Upgrades with ceramic bearings work but the marginal gain is tiny compared to the flagship offerings from Shimano and Campagnolo.
Cheap pedals with cheap bearings - technically a little higher up the power transfer chain than the drivetrain per-se - can also cost 1 watt.


https://www.ihpva.org/HParchive/PDF/hp50-2000.pdf

Amateur ex-racer here.
Know what the actual gear inches produced by each chainring/cog combination is: tape them to your stem.
Plan your shifts, if the terrain ahead is getting steep know when to bail on to the inside ring and what cog shifts will be necessary to maintain pace/cadence.
The inside ring is for climbing, use it on steep uphills.

Tailor your gearing to your fitness & terrain.
Select the low/low that will get you up the steepest local pitches w/o undue stress and the hi/hi that you are willing to run out on the downhill.
Cram as many cogs in between as practical and have at it.
Don't forget to printout the GI pattern, you'll need to know where you are to determine where to go next.



-Bandera

That assumes the error in the device is random. If the power meters are consistently under or overestimating power then the conclusions drawn by RCHUNG are still valid.

The cassette was custom assembled from individual cogs to minimize cross-chaining. Spacers were used so the 12 and 13 cogs were "centered" on the 42 ring, the 15 and 16 cogs were outboard and "centered" on the 54 ring.

Those dotted red lines are roughly 8 to 10 minutes apart so each segment represents around 500 to 600 paired observations. But, in a larger sense, the reason why accuracy isn't an issue is that we weren't interested so much in the exact amounts of power the SRM and PT were reporting -- we were interested in whether any difference varied with the chain speed and chain tension. In fact, if you look closely, you'll be able to see that the mean difference between the SRM and PT is *negative* for one of the intervals. No one in their right mind thinks that the PT was *adding* power to the drive train. In fact, although both power meters were torque-zeroed prior to the test and (as you can see) there was a long warm-up period to bring both power meters up to operating temperature, and both power meters had had their calibration checked prior to the experiment and both were within spec, the PT was reading slightly high within spec and the SRM was reading low within spec. But, once again, since we were only interested in the difference, the absolute accuracy didn't matter. And, there's an obvious change in the difference between recorded power from the SRM and PT that occurs exactly at the points where gears were changed. Silvesx80 claims it can't possibly be there. Who do you believe, Silversx80's experience and education, or your lying eyes?

We were interested in testing this because of work by Spicer (for example, here where he talks both about cog size and chain tension and their effects on drive train efficiency). Spicer claims that larger sprockets are more efficient but also that higher chain tension is more efficient. As I pointed out above, at a given power, higher chain tension occurs when the chain moves more slowly, and for a given power and gear ratio, the chain moves more slowly when the chain ring is smaller. But when we ran the test we either got close to no difference or else a difference in the other direction -- higher chain speed and lower chain tesnsion gave lower losses and higher efficiency.

BTW, while we know that side-loading on chains is bad, Spicer also found that in bicycle applications, chainline matters less in terms of losses and efficiency than sprocket size and chain speed/tension. Of course, maybe you'll take that finding with a grain of salt now.

Yeah, this is also written by Spicer, and makes the same claims as I mentioned above. These are the claims we were testing.

Some pretty severe assumptions, falling almost into a straw-man argument. You're either not understanding my point, or misrepresenting it on purpose.

You cannot guarantee accuracy of those testing devices beyond the published variance. Only PowerTap and SRM can do that, and I would like to believe they'd be keen to update their marketing material if they could. The fact still remains that you can only report the degree of accuracy in your test of the equipment used, no matter how extensive the calibration. With the proper calibration, those units are still only good for (and I checked the latest devices) +/- 1% for the SRM and +/- 1.5% for the PowerTap G3. This must be reported.

A static calibration to the manufacturer spec guarantees the manufacturer accuracy in a dynamic situation. If you were to calibrate against more accurate equipment under dynamic load, then you can report better accuracy.

The posted ihpva report, on the other hand, appears to have much more precise measuring equipment... 15 years ago.

“The important takeaway,” says Friction Facts founder Jason Smith, “[is that] it's not good to ride the big ring-big cog combination, but it's much worse to ride the small ring-small cog combination.”... LINK to Bike Radar article

They are equally efficient. Every revolution moves the bike the same amount.

Crank deflection would show slightly more power at the crank in the smaller chain ring with a crank based meter.

A hub based pm would show slightly more power at the hub on the big chain ring.

The power difference would however be effectively zero.

I took this chart to mean that it makes sense to go to the small ring if you are inside of the 3rd or 4th biggest cog depending on the cassette. IME this matches up with the noise generated by the chain getting louder in those combinations.

Small ring has more friction pretty much everywhere else.

My lying eyes see no strong correlation between power loss through the chain and the big chain ring variable. What I see is one single chainring-cog combination that resulted in a different chain loss. That tells me there is some other variable unaccounted for in the design of the experiment.

My gut tells me that the experiment would have been more conclusive with an electric motor as a power source.

There are power losses through the chain, they vary (so the drivetrain is not equally efficient in every gear), and they dwarf any error introduced by the different chainring deflections. But thanks for playing.

They stay in whatever gear makes sense for the race situation at that moment. Tactics, terrain etc...

How can chain tension vary depending on chainring size for the same gear inch ?
If riding at 90 rpm, 80 gear inches, putting out 300 watts, isn't the chain under the same load regardless of chainring/cog combination ?

This question was answered almost a century ago, and is common knowledge in the chain drive world. Larger sprockets increase efficiency for a number of reasons. Chief among them are reduced chain tension at the same input/output torques, and reduced movement as the chain winds onto and off the sprockets.

Lower bearing load at the chains pins, and less movement translate to higher efficiency.