Yes, I recall your description of the number of runs above...but, what's missing is some sort of description of the equipment used and how the method was "validated". Here's a short list of things off the top of my head that could be helpful:

• How was power measured? What's the accuracy and precision of the equipment? Is the power measurement inclusive of drivetrain losses? If not, how is that handled (i.e. is gear selection fixed? Something else?)
• How well does the method detect a "known" change in Crr (i.e. the "Tom Compton Challenge"), such as what happens when mass is added to the setup? Does the Crr change in the expected way? What's the smallest change detectable reliably?
• What's the surface roughness which the test setup is attempting to duplicate? How well does it mimic that "specification"? (This is where the PSD data referenced earlier would come in handy to evaluate how well the test setup duplicates the intended surface).
• Was the same rider and bike used for all tests? If not, how do we know their "damping" is similar (for pressures above breakpoint)?
• How does the data compare (numerically) to what is found using simple roller tests and field tests? If it differs, then how...and what are the possible sources of the differences?
• How tightly is ambient temperature controlled in the test location? If it's not, is the temperature compensated for in the results (due to the somewhat important temperature effects of Crr)?

I believe all of the above have been reported and investigated for both roller and field test methodologies. I understand the desire of the Zipp Engineers to use their "rolling road" as a proxy for field testing. It eliminates translational aero drag AND puts to rest any arguments against validity from people who don't understand the "equivalence" of rollers to flat surfaces...in other words, it's slightly more "real world" for some consumers. That said, it is going to have it's own quirks, and being able to understand those quirks is the way we can put their results in context. It's just like understanding that roller testing (when done carefully) is extremely good for evaluating tire hysteresis properties (and thus Crr) for pressures below the "breakpoint" of the system. Again, it's all about context.

edit: I thought of one more (very important IMHO) piece of info on the treadmill:

• What is the thickness and material of the moving belt, and how is it supported? How does that compare to an actual road surface in terms of compliance and damping?

As mentioned above with the Bigham track discussion, surface compliance matters in these sorts of things. If the "rolling road" adds compliance and damping into the system, the results can be different than "real world".

I think the issue is exactly that there isn't enough transparency with this testing. All the claims they are trying to make are that their rolling road is a good simulation of real world with no data to back it up or even details on what type of road surface is being simulated especially as road surfaces are super variable.

Along the same note people even don't believe wind tunnel test and that's very well established with decades of validation and everyone is well aware of the assumptions and limitations involved in wind tunnel testing. Asking other engineers to believe an entirely novel setup without any validation or even pictures/description of the setup shown to them is going to be an uphill struggle. Listing out the number of runs, rider weights, how long they road for before testing isn't really useful in my opinion. That's small peanuts and is taken as a given that those things are controlled. The real important questions are as mentioned before behind the actual rolling road setup. As if that doens't simulate road riding or the representative surfaces I or others ride on, the results won't be meaningful no matter how much control there is of test setup. I want to believe that it simulates actual roads well, but having experience building test apparatus to simulate real life events, I know how difficult that is.

This is part of the reason the track "data point" I mentioned is so impactful to me. We know the test apratrus is good (the track), the main source of error (the rider) likely well controlled due to rider skill, and the test setup is likely good. Plus this is on very smooth surfaces which is a regime with sparse low pressure data. If much lower pressures are good on smooth surfaces, then it's a given that they will be good on rougher surfaces. Instead most of the testing thus far has been on quite rough surfaces which makes me believe that the benefits aren't there on smoother surfaces, and that this testing (if it is an accurate simulation of road riding) represents rougher surfaces to show a larger benefit.

Echoing what Tom said I'd like to know the following about the rolling road testing and will likely have more follow on questions once I see a diagram of the rolling road setup and a picture/cross-section of the surface. However I'll assume the Zipp engineers have the normal environmental, sensor setup, and sample preparation variables well under control.

• Do you have a diagram of the surface of the rolling road setup?
• How is each section/plank of the rolling road supported when rotating through the treadmill/how large is each section?
• What does the transition between sections/planks of the rolling road look like?
• Do you see periodic events correlating to the load/unloading of a new section/plank or the transition between sections?
• How is the micro/macrotexture on each section/plank of the rolling road applied?
• What surfaces are you trying to simulate?
• Do you have PSD's between the rolling road and the surfaces you are trying to simulate?

I know what it means philosophically. I don't know how the charts are trying to represent it.

For instance, if the "delta" is the difference from the old wheel, then why does the first chart have values for the old wheel that aren't at zero on the y-axis? Shouldn't it be at zero difference from itself? The second chart does have the old wheel at zero, does this mean that the first chart is referenced to the 808 data? But even then, the 808 data doesn't cover the entire portion of the x-axis used by the other wheels, so what is the new-wheel data at those pressures in reference to?


I'm asking how the charts show the change from that underlying data to the new wheels.

the power was measured using a quarq power meter on the bike of the sole person on the only bike used for all 96 runs. accuracy? it's the accuracy of a quarq power meter.

because speed was constant for all runs. drivetrain losses are not material to because the actual power is not relevant (who cares how much power it takes a given rider to ride his bike at 20mph on a rolling road?) the only thing that's relevant is the delta in power between the set-ups.

the surface roughness has been asked and answered at least 3 times in this thread. what is relevant, that i don't know - also mentioned at some length in this thread - is the vibration in the entire system which could - just using my intuition - add to the "roughness" of the road. i think this is worth investigating.

was the same rider used for all the tests? yes. as i pointed out in posts twice above. how does the data differ from field tests? the data, per zipp, syncs well with their fields tests, which they perform at eagle creek park, on a circuit, and which i stated at least twice, both in this thread and i believe in the article on the front page.

how tightly is the ambient temp controlled in the test location? again, just my intuition, but in the 25,000 square foot factory where i built my bikes i hit on the idea of setting the thermostat on my HVAC. this has not been mentioned, explicitly, so for those who have this question: zipp's rolling road is inside its building, not outside.

OK then...and myself having had quite a long term experience with Quarq PMs (I was a beta tester for them at one time, after all), I also know that the data quality is only as good as the care is taken during their use. It's imperative that the torque slope is checked (and adjusted, if possible) and to "zero early, and zero often" ;-)



...unless drivetrain losses are changing across the runs, that is. Are we SURE gear selection was the same? You appear to assume it was because the speed was fixed, but is that the case? I don't know...there's no data showing one way or the other.

Additionally, how "fixed" was the speed, i.e. what's the +/- on the speed setting? Did they calculate Crr for each run (which would cancel out speed and power variation), or just compare watts? In other words, if you want to compare "delta power", you need to compare on an apples to apples basis. For roller testing, I don't try to tightly control the speed (in fact, I try to control cadence for a given gear combo), but then I use the speed, power, wheel load, and ambient temperature to calculate the equivalent Crr on the flat at 20C ambient. THAT is what gets compared, not the raw test values. That's the thing, if you have a good model, you don't necessarily need to control all the variables, just record their values and use the model.



Right...and it's been asked for good reason. You say they claim it's comparable...we're all asking "how comparable?" My intuition is telling me that based on the results, there may be some additional compliance and/or damping introduced in the running surface that's different than typical pavement. That's one way that continually lower pressures would lead to lower power requirements on surfaces consistent with typical pavement roughness.




Right...they say it "syncs well", and yet you said they aren't willing to reveal that data for some reason. It really should be the "golden data set", no?




Yeah...see, that's the thing...even HVAC systems can have quite a wide range (relatively speaking) of temperature swings throughout a day. I often run long-term thermal tests at my office and detect local temperature swings in particular areas of the office of a few degrees C. That's plenty of temperature change to affect Crr results, especially if one is looking to resolve them to within 1W (as their plots imply). So, if they aren't monitoring the temperature, and/or compensating for the temperature effects on Crr...well, that throws up some red flags to me.

Slowman wrote:

HVAC. i guess i'd need to see how the differences between 71°f air temp and 73.5°f air temp impact the Crr results. i'm not smart enough to know.

as to the field testing data. they're not ready to publish that yet. they are comfortable, as of now, that the field testing they've done syncs with the rolling road tests. but they're not ready to publish. it is my guess that field testing will become the norm for this brand. but let me tell you my takeaway. the fact that they say the field testing syncs means that there is some degree of validation of their rolling road testing. .

jcbesse wrote:

I would be quite easy for them to just show a picture of the 'rolling road' you'd think?

The do on their website

https://www.sram.com/en/life/stories/life-on-the-rollingroad


https://www.sram.com/en/zipp/campaigns/total-system-efficiency

jcbesse wrote:

I would be quite easy for them to just show a picture of the 'rolling road' you'd think?

The do on their website

https://www.sram.com/en/life/stories/life-on-the-rollingroad

https://www.sram.com/en/zipp/campaigns/total-system-efficiency

Quarq PM. it is my assumption that SRAM understands best practices using that PM. same gear throughout. you're right. i don't know for dead sure. but i'd be shocked if this were not the case. same speed throughout.

...changing the speed on the rolling road would be no different than changing wind speed in the wind tunnel. do we ask those performing wind tunnel tests that question?

Well...in the white paper I linked to just above, they state it's run at speeds "up to 20mph", so that's not exactly clear. But again, it doesn't matter if they control the speed tightly IF they measure the speed and combine it with the power and mass to calculate Crr. If they don't do that, then yes, the need to control it tightly, since higher speeds require higher power at the same Crr.



If the wind tunnel reports in drag force, but they don't say at what speed, then we ABSOLUTELY ask for the wind speed. The drag force value has no context. However, if they report in CdA, then test wind speed isn't as necessary (just want to make sure it's in a representative Re number regime). It's the same with my description of watts vs. Crr above.

Here is their data for good roads (Kona). Green 28mm, red 25mm, blue 23mm

A really cool triathlon site has the answer to EVERYTHING

https://forum.slowtwitch.com/...D_P7791676/#p7791676

Tom A. wrote:

The constantly lessening power requirements with reducing pressure, for even a 28c tire, still is hard for me to fathom...unless the road surface roughness is REALLY bad, and/or the surface has additional compliance/damping than typical pavement...

...or, unless it means that their wider internal widths widen the measured tire widths so far that even 30 psi is above the breakpoint pressure for those tires, speed, and road conditions? I'm not sure if that's exactly a good thing, if that's the case...for a variety of reasons.

Found another longer video with more details on the Rolling Road.




It looks like they're able to put different "belts" on the treadmill in an attempt to simulate different road surfaces. The ones shown in the videos on the SRAM website have some pretty nasty bumps in them (like riding in a rumble strip or on washboard dirt roads). Even if those were removed, The texture of the normal slats looks like they would provides a fairly "rough" surface, like riding on a boardwalk vs smooth asphalt. It's possible the plot you showed is from one of those really rough belts (maybe since they're showing the gravel king comparison?).

FWIW, there're 4 tris in Eagle Creek every year. The road surface inside the park is pretty good for for Indiana and it would make a good place for VE/Chung testing (lots of potential loops w car traffic). It's been a couple of years since I've raced it, but I don't remember any chip seal or the thunk-thunk of expansion cracking, but at the same time, it wasn't freshly paved asphalt either. It's what I'd call "normal roads" like Zipp and Dan have been saying they're trying to represent. The data to show that the Rolling Road is actually replicating that though.......

when i was at zipp a couple of months ago i had it on my schedule to see the rolling road. but a covid incident forestalled that plan. as i see it now, from this video, it's a much more sophisticated device than i thought. to cajer's point on a PSD plot, i wonder what's built into the software analysis zipp has when translating video from its camera. 200,000 frames per second.

So...a quick calc using the stipulated temp range above (which is only ~1.4C range, or ~1/2 of what I've seen in HVAC controlled office environments), and mass and speed claimed (85kg, 20mph), shows that temperature difference can cause the Crr results to vary by up ~0.5W (with a reasonably assumed Crr of .0030 for the application).

That doesn't seem like much...until you look at charts showing 1W differences, that is...and understand there are other sources of error contributing as well ;-)

Hmmm...seeing as the "planks" appear to be some type of polymer (Delrin, or Nylon?) material, and in the slow-speed video in the second link shows them deflecting somewhat, I think my suspicion of additional compliance/damping in their system is what is making their plots look the way they do. Unfortunately, that surface doesn't have the same properties as pavement.

Something that’s been nagging at me, and just because you brought it up: reporting measurement error doesn’t seem to be in the kit-bag for engineers (perhaps I should say industry engineers?). Recalling the Zipp data in question here has 4 runs for each setup, so it’s possible to show this.

Put another way, to you point (and BergHugi above?) – is all this just hair splitting if the test sensitivity is a relevant amount less than the size of the error? Dan argued above that Zipp (and other companies?) don’t show error bars because this is not up for peer review. That seems like strawman rationale if we’re seriously interested in teasing this apart.

That's likely because we are both looking at small deltas and the bike industry doesn't have that much money for development meaning they aren't able to do allot of runs to reduce errors. So it's best for them not to report them.

This is probably a silly question, but hey.....

Why doesn't ST set up a road test with a well accepted method (like Chung testing) to validate some of the data/claims made by the manufacturers ?

BTW, the TSE paper is actually really interesting. Not too many manufacturers release this amount of info. I still believe the P5 whitepaper was the reference of what a great report could/should be.

but you would probably have to reconsider making statements challenging the integrity of the company if you think you might like to be part of a team that ferrets this out. .