sorry! wrong user! muddled brain. please accept my apology. you're gold.

---

Dan Empfield
aka Slowman

No problem. Thanks

Yeah I'd like to see CRR and aero for both wheels with both tire sizes. Otherwise Zipp is assigning the RR gains that come from a larger tire to the new wheels but not for the old. What tire size they're "optimized" for is rather meaningless in this context. The old wheel might save the same 4W on RR with a 28mm tire.

And stating the obvious (not to you specifically)

this is why a Chung test should be done, old wheel with 25, old wheel with 28 and new wheel, on a "reasonable" road, representative of most triathlons. Ideally at a few pressures, pick the best number for each combination.

We will then see the real world difference. This is trivial to do if one has access to the wheels

You might be getting a bit hyperbolic with that last sentence, no? ;-)

Here's some alternative reasons why there may be a preference for roller testing, especially for TT/Tri purposes...at least these are the ones for me (I can't speak for others):

• Having done Crr evaluations by both "field testing" (Virtual Elevation, or Chung, method, Classic regression method, "Work per Lap" method [Adam Haile], to name a few way variations) AND small roller testing, I've found that the latter is significantly more time efficient, less problematic, more accurate and repeatable, and MATCHES closely the results of the field testing for pressures below breakpoint on pavement. This is especially so after having refined my protocol to reduce the effects of temperature.
• I've also studied some of the physical properties at play, and some of the relevant modeling of them, in order to better appreciate what each method of evaluation does or doesn't show. I've also done some testing on compliant surfaces, such as a typical modern running track, and directly saw how the additional compliance and material damping of the surface affects not only the Crr estimate, but how some of the non-linear responses of the surface can affect the CdA estimate as well. THAT is why I'm skeptical of the running surface material properties of the Zipp treadmill...especially since the pressure vs. Crr results from that device don't tend to match what is seen in many field tests, but also behaves in a manner that is consistent with the introduction of additional surface compliance and/or damping (i.e. continuous lowering of Crr with lowering air pressure, which doesn't match field testing OR roller testing results).
• Field testing can be good...but getting reliable results is not easy. Trust me, I know...I've made all the mistakes. There are many factors to take into account and many opportunities for introducing error. There's a reason wind tunnels and roller Crr testing are the established protocols used to evaluate transportation products. They work.

So...I think you can see why when data such as what is being discussed is presented and it isn't consistent with previous methods that are well understood, then there's going to be some skepticism about that data. You seem to be saying people don't like it just because it's new, or that there's "momentum" or "inertia", or some sort of "investment" in putting down this new methodology and it's demonstrably none of those things. It really comes down to the fact that the data coming from it appears to be an outlier over established methods for the surfaces and tires being compared...so there's going to be some explaining to do about why that is so? Is it actually revealing some sort of new information (and if so, what is it measuring that other methods don't, and why?) Or, is there something about the test setup that doesn't reflect the actual use case and is causing a systematic error? Don't forget, many repetitions of a measurement (and the resulting statistics) don't address systematic errors, only random errors.

Also...in regards to making opinions based on a Youtube video...well, when that's the best information available...just sayin' ;-)


It's not just about the roughness of the treadmill surface, but the material properties (i.e. compliance and damping) as well. This matters...see Josh's results comparing freshly laid asphalt vs. the same surface that has been allowed to harden/cure over a few weeks/months time. I have done direct comparisons of CdA/Crr estimates for the same setup on both asphalt and around a compliant running track, and the effects are quite noticeable and apparent.

---

http://bikeblather.blogspot.com/

Or...if one has wind tunnel data from the tire/wheel combos of interest, along with Crr data of the same combos from roller testing, it's actually possible to come up with an "overall estimate" that can be useful in evaluating which is faster, and when ;-) Such as here: https://bikeblather.blogspot.com/...t-3-after-party.html



Field testing of some of the combos of wheel/tire/pressure could be a good confirmation, but performing that testing isn't always the easiest to accomplish and is very time consuming, IMHO.

---

http://bikeblather.blogspot.com/

i don't disagree with the anything you wrote above. i just know that when i have broached the subject of rolldown testing, regardless of the protocol, i've gotten a toilet flushed on my head for the very valid reasons you post above.

as to the treadmill, again you're right. healthy skepticism is in order. but it's not a running treadmill. it's not a woodway. the segments are not designed - or necessarily designed - with compliance in mind. so, i would just hesitate to make any assumptions on material density, sag, compliance, damping. that established, i would also not make any assumptions that the entire system isn't subject to vibration beyond what's in the belt or track. for example, i wouldn't mount a camera or photosensor on the frame of the unit to measure vibration. i'd mount it on the ground.

all that established, i like the rolling road for the same reasons you like drum testing: ease of use, precision, repeatability. i'm not the scientist you guys are, but my instinct tells me if you could (say) put an accelerometer on the bike frame, ride it on the rolling road, and then find the road surface that most closely reproduces vibration of a similar amplitude, maybe you could make a statement about the road surface the rolling road (without any features installed) mimics.

---

Dan Empfield
aka Slowman

  
As you know, prying apart CRR and CDA improvements in the field is possible but not trivial.

Measuring "total system improvements" (in watts), at a variety of pressures is pretty trivial and can be used to confirm their findings using other methods.

Put it this way. If you had those wheels, you could in a few hours, confirm the savings published in those graphs. Correct ?

Yes...rolldown tests are a subset of the field testing protocols (most field testing procedures are really just evaluating the same equations of motion for a bike+rider system, just from different "angles"...at least they should be.) The main advantage of roll-down methodology is that a power meter isn't required, and thus power meter accuracy and precision doesn't come into play. But, to do it correctly, one WILL need accurate speed recording (GPS speed isn't going to cut it, and multiple magnets for a magnetic speed sensor are an advantage...although the same is true for other field test methods). Then, of course, tire rollout measurement will be critical as well as part of that speed measurement (again, same as other field test methods). If one want's to attempt to "pry" CdA and Crr, then that will require having entry speeds for the start of the roll-down segment vary as well.

All that is saying...if one proposes to do roll-down testing like some manufacturers have (i.e. hand-timing through a segment after a roll-down), then one is rightly getting a "toilet flushed on their head", because doing it in that manner doesn't allow one the sensitivity to reliably state that one setup is different from another...no matter how many runs are performed, nor what statistics are applied. I've seen claims made from that type of testing as "there's no difference", when in reality it's just that the method isn't capable of detecting the differences that are there. This is one of the reasons a "Tom Compton Challenge" type test is a good first start for any method.

I say that if one is going to do field testing and one DOES have access to a good power meter (and the late, great PT hubs are ideal for the purpose), then it's far better and more time efficient to do one of Virtual Elevation variations, such as "Half-pipe" out-and-backs, Loops with varying speed (i.e. Shen method), or even the classic "regression" methods of out and back runs at varying speeds on a flat course (and again, they're all just looking at the same math in different directions).




You are right that I don't have detailed knowledge of the treadmill structure...but, knowing what I know about polymer materials that are apparently being used, and being able to visually SEE the planks moving/deflecting in the low speed video (even across the "flat" plank sections) I'm pretty confident in saying that someone is going to have to SHOW me that aside from the surface roughness and vibration amplitude, that the material behavior of that surface and structure is consistent with a paved road. The fact that the data shown looks more like what one would expect on a compliant surface rather than a rigid paved road points to the idea that there's some of that behavior influencing the results. That's just my engineering opinion though...I welcome information that shows it's not so.



The "rolling road" could be a great tool, with the caveat being that it still needs to be shown that it reflects road properties (especially for smooth pavement applications). One big advantage it has is that it side-steps the inability of a lot of folks to be able to wrap their heads around roller testing being a good proxy for on-road performance. Another is that for rougher conditions, it brings in the damping of the rider body into the system....so, just like with rough roller testing using a load support with representative damping could do, the testing can actually demonstrate the breakpoint pressures of the systems, like one sees in field testing. That could be cool.

However, even if the vibration amplitude and frequency content at the rider is the same for the rolling road and representative pavement, if ADDITIONAL damping in the rolling surface is introduced, then the results are NOT going to reflect the pavement performance. More energy is going to have to be input to the system to get the PSDs to match in that case, because additional losses are occurring in the rolling road. In fact the results will end up more like what has long been measured for MTB rolling resistance, where no matter the tire, the lower the pressure the faster it rolls...and then pressure selection just becomes "what's the lowest pressure you can run without pinch flatting for the tire you've chosen"...which is EXACTLY what their power vs. pressure plots from the rolling road look like in this wheel introduction. It's another clue that there's something amiss at that interface that doesn't reflect actual pavement. But, like I said, that's just my semi-educated engineering speculation.

---

http://bikeblather.blogspot.com/

Probably...it would all depend on having the right course and the right conditions.

That said, using certain methodologies, it may be able to easily pry apart the CdA/Crr pairs in a reasonable time. In fact, I mentioned doing some testing on a High School running track at one time. Below is a portion of the data from the test session (I was mostly using VE for the aero brake evaluations, and it's data from the last 2 VE runs that are listed as runs #1 and #2). I was trying out this evaluation method of Adam Haile's at the time, and so using final bike configuration from the VE testing, I did some laps on the track on the way home from the halfpipe testing venue. As you can see, the method does a good job of separating out the apparent CdA and Crr for each road surfaces and setups, and one can also see that some of the non-linear loss performance of the "soft" track surface ended up being evaluated as a CdA increase in addition to increases in Crr. (Because, as you well know, the model assumes all non-linear effects are CdA related, and linear effects are Crr).

Pretty neat, huh? I've long had a task for myself of adding a "wind" column to this spreadsheet, but am usually discourage by my lack of "Excel-fu" to be able to pull it off <LOL>

---

http://bikeblather.blogspot.com/

I was looking back at your article again, and through the thread, and I don't think this plot shows what you are asserting above. Remember, it's a delta, or difference plot...in other words, it's just the relative difference from a data set selected as the thing to compare against. The reference curve could be nearly any shape, and all that these plots show is how the relative difference between the reference curve and these other data sets vary, or don't. Additionally, since the old 808 w/25 plot is flat and centered on zero, it appears that the original data curve for that combo is chosen as the reference...and thus, the flat delta curve for the old 808 w/25 makes sense, if not being redundant. It's NOT showing that they are impervious to pressure changes. It's just saying the relative difference to the reference plot is constant across the pressure range. Without that reference plot though, there's no way to determine how impervious (or "pervious"?) the setups are to pressure.

Now then, what does that say about the other 2 data sets (for the 28 and 30)? It only says that the relative difference is basically a fixed amount (within 1W actually) from the reference curve. The curves could be any shape, they're just offset by a fixed bias.

Here's the thing that this type of data presentation DOESN'T reveal though...let's say that the raw Crr vs. pressure curve actually follows they typical decreasing Crr with increasing pressure shape seen with other test methods. If we were to look at those curves, we might find that at the pressures one would typically run for each tire size, that the Crr is actually almost identical. This is very possible, even with a fixed offset of the curves, and what it means in a practical sense is that there's no speed difference from Crr when run with the proper pressure for the tire size. Presenting the data as delta plots only doesn't allow for that insight.

Of course what adds even more confusion is that 858 NSW plot, which has no obvious reference data set...unless it's also relative the old 808 w/25? No way to know with what's been presented.

---

http://bikeblather.blogspot.com/

This thready is getting super nerdy (which is great)

Here is my less nerdy assessment of where we are

- Zipp created an environment called the rolling road so they can simulate and better control road conditions.
- Like any simulation environment it can have quirks not necessarily representative of the real world. We don't understand those quirks although you can take some pretty educated guesses. There are ways to measure how close/far an simulated environment is. But not everyone appreciates a SPD.
- The numbers they produced show data that goes against previously observed behavior, both in the real world and other simulated environments.
- We are not sure of what conditions they were trying to simulate. Was it Paris Roubaix or Kona ? Maybe this explains what we think are quirks.
- There is missing data, for example aero performance of 28mm tires to get a full picture
- While simulated environments are great, real world data, in real world conditions for a triathlete is what we need to see. They have this but have not released it yet.
- We have no idea if this total system gets us 1, 3 or 5 watts at Kona, IMNZ or Challenge Daytona.

My own opinion here : There is only one person that has released more data than Zipp, it's Tom A. I mean that as a compliment to both.

Can you not simply take an RMS of the orange line to get a good baseline calculation?

For now I'll say I don't know

We have a ton of accelerometer data from good and bad roads
We see very measurable degradation of "performance" between good and bad roads
Manually it is easy to measure this degradation (as per my charts), and it's easy to see where the roads go from good to bad in the data

What isn't clear is how to quantify the vibration in a way we can say "it increases CRR by x". Of course I can analyze it and in that one case say it's 10watts, but how do I quantify a base value from say BRR then adjust it using the vibration data ?

Maybe we can't. Maybe the best we can do is say "the optimal pressure for this surface is y".

We have done a fair bit in terms of CRR measurement but it's more prying apart CDA/Crr on a relatively uniform road (good or bad). Kind of what Tom did above but in a more "ad/hoc", less structured test. We have some pretty interesting stuff.

But now that I got a Guinnes powered-supernerd doing PSD plots in Matlab I suspect we will learn something over the next few months. So following along.....

We also have a setup inspired by this
https://bikeblather.blogspot.com/...ollers-chartand.html
It took us a while to tune the setup but we get very repeatable and predictable measurements for a smooth surface.

We have a way of texturing the rollers using strips of wire. We use wire guage as our method to control amplitude and the number of strips to control frequency of the vibration. We have all kinds of data but never really did much with it. Again, how do we relate it back to the road.

Since Zipp has rolling road, I think I am going to call it "suicide street" because every time I over did the vibration, the rider (not the best on rollers) crashed :-). But after seeing Zipps setup a little flash occurred and I ordered a harness from our local climbing store.

I predict a fun winter in Canada

I'm glad that was some help. It was my whole intent with writing that...i.e. putting a means to evaluate tires "in the hands of the people" :-) It still amazes my how few people who are PAID to give tire evaluations even attempt to do this...<I'm looking at YOU, "cycling journalists">



Sounds similar to something Al Morrison attempted way back when HE was doing all of the roller tests and I merely was a guy who figured out the equations to attempt to relate the results to a flat surface. He had taped wires onto his plastic rollers. But, every time he took data with that setup, it basically just resulted in a relatively fixed bias when looking at Crr vs. pressure, despite the fact that his roller setup, as is mine, uses a live rider and mass, albeit with a front fork mount. This was true even for setups that Al commented were quite rough to ride on.

We didn't go too far in trying to figure out why that would be, and why there wasn't an apparent breakpoint observation as seen in field test data. One thing I've tried to wrap my head around is the fact that the effect of flexing in the contact patch is "amplified" when using the rollers, and for rollers the size I use (4.5" Kreitler aluminum) the raw power value is basically 3X what one would expect for the same roughness surface on the flat, due to the additional flexing due to curvature at the contact patch. This is great, because it's part of the reason why rollers can "tease out" the differences between tires, especially since all of the other losses (i.e. bearings, etc.) are also divided by 3 in the results.

That got me thinking though that when adding roughness, that's going to be requiring an additional energy input that has nothing to do with the contact patch curvature, and thus that roughness contribution would be divided down as well if just lumped with the rest of the power requirement (when it probably shouldn't)...I guess that means that for my roller setup, I need energy input to be 3X what one would need on the road? Or, does that just mean you need to do a smooth roller evaluation, and then consequently look at the absolute power requirement increase over that baseline for a given roughness? (Great application of a delta plot in that case :-) I'm sure there's a way to figure that out...I just haven't had the time to be honest...and, if I did have the time, I'd probably put the effort into putting together an automated roller testing machine with a damping representative of a floppy human instead. Doing roller tests with a live human is time consuming (but significantly less so than field testing, to be fair).

Additionally...did the use of a fork mount inordinately change the amount of damping in the system since ~1/2 the vibration input was being left out?

Again...lots of questions to figure out when attempting to create a "lab testing rig" that accurately represents actual outdoors riding...as we've seen with the subject of this entire thread ;-)




Ha! Yeah...that's part of the reason I went with the fork mount approach. Less chance of an "off"...

---

http://bikeblather.blogspot.com/

I know this post may set this discussion off in a different direction, but so far all the data provided compares the Zipp 808/858 against the last generation of 808/858. However, for me, the new 808 (in particular) has some appeal as a crit racer. In the past, I would not have considered the 808, as I would have felt that the added weight of the 808 over some of the competitive 50mm wheels out on the market, would offset any aero gains on the deeper profile.

The 808 firecrests appear to be in the same weight range as many 50mm/60mm on the market (Roval CL’s that come on the Tarmac SL7 Pro/Ultegra build, old 404’s, HED RC6, etc.)

I’m more curious if two wheels are the same weight, what are the aero gains from an 80mm rim over a 50-60mm rim, given the same tire/tube/pressure setup?

Is it a meaningful upgrade in crit racing, say, or is it going to just be a watt or two gained, and maybe a little more weight towards the edge of the wheel, making it a bit slower to accellerate out of corners?

You could go through some of the aero weenies or other data, make some ass-u-me-ptions and figure out the difference in drag between different depths of the same generation wheel from a given brand. It’s probably less than 5 watts at crit yaw angles. Aeroweenies.com/data.html


I’d encourage you to do the math on that power difference. It’s pretty easy to consider the kinetic energy of the rider/bike/wheel system at say 20 and 30 mph, then add a couple hundred grams to each wheel and see how that changes the energy of the system. Power is energy over time.