Monday, November 16, 2009

The Chung Method is No Joke

I did TT tests yesterday at my Brookhollow course. When messing with the data last night, I came to the conclusion that my test hill must be four feet higher than the 83.1 vertical feet it was designed and staked out to be. Increasing the vertical drop by 4 feet was the only way I could get everything to work out with reasonable Crr and CdA numbers.

I couldn't stand the suspense, so I took a survey instrument (my trusty Topcon GPT-3000 W, shown at left) out there this morning to check it out. Guess what? The contractor built the road with 3.51' more vertical drop than it was supposed to have according to my design (I guess he wanted to save some cash on fill dirt).

The Chung Method is definitely no joke if you use it right. I don't know what Garmin Slipstream pays for tunnel time, but I think I have a comparable tool now for free.

I moved each of my aerobars inward by 3/8" yesterday, and apparently I was able to discern the difference in CdAs from that tiny change (which, by the way, would mean 11 seconds in a 40k TT). If I can see that type of change, this process is going to be fun (and fruitful).

More details to come.


Colin Griffiths said...

Ah, so there IS another person in this world who understands why I wander up and down little lanes, GPS in hand, making notes in a little book and looking highly suspect! We've had windy conditions here, so I've not been able to do any more tests :(

Robert said...

Holy cow.

Robert Jordan said...

I don't want to get too fired up about the results until I prove to myself that they are repeatable on different days in different temperatures et cetera. But I'd bet good money that they will be.

Robert said...

In general, you need an additional constraint in order to pry apart (Crr, CdA) pairs -- and knowing total elevation change is one of the possible constraints that let you do it. A year or so ago I posted some data on the wattage list from two runs up the same hill at different speeds. Since different combinations of (Crr, CdA) imply different virtual elevation change it was possible to separate the estimates of Crr and CdA. However, in that case I only knew the elevation delta to about a meter so it's really cool that you have the tools to measure the elevation change so precisely.

I thought that another way to pry apart the (Crr, CdA) pairs was with delta mass (do one run with empty water bottles and another with gravel-filled watter bottles) but I think I either needed bigger bottles or to fill them with lead.

Robert Jordan said...

I am able to separate Crr from CdA with pretty high confidence, I think, because my test hill is rather steep - about 8%. So particularly at low power outputs there is very little air drag effect against the rider on the climb; and Crr plays a rather small role in the force against the rider on the descent.

So I modify Crr until the climb height is about right, then adjust CdA to lengthen or shorten the descent to the appropriate length. Only a very little bit of iteration is required.

That's how I knew my 'design height' hill was too short: I had to bump the Crr up to .0075 to make it work out, and that resulted in CdAs that were much lower than I expected them to be (all that caused by just a 3.51' vertical construciton error!). It turns out that I was correct in assuming that number was not realistic. Once I plugged in the real hill height, Crr popped out at 0.0053 and CdA at about 0.230, if I remember correctly.

Robert Jordan said...

Actually, I think I said that backwards: I think I had to use a Crr that was very low, like 0.0015 to make it work the first time. You get the idea, anyway.

Robert Kendrick said...

its an expensive head unit, but the iAero gives good CDA.

Chung defintely works -- have fun getting slippery!

Alex Simmons said...

Well as a fellow user of this analysis method, I have found remarkable things when looking at data from an indoor track.

But the thing I'm most impressed with is how anyone can maintain the same position when going up an 8% gradient!

Rich W said...

Do you think a GPS or a barometric altimeter will be accurate enough for getting a gradient of a road? I may also check with the city for topo's of the area. Just wondering how I can get accurate data on the road without any surveying background.

any other ideas? are you ever in Columbia? ;-)

Robert Jordan said...

Rich: I wouldn't worry too much about the absolute elevation data. GPS units (unless they're survey grade) have very poor vertical resolution - usually about 3 times as much as horizonal error. And barometric altimeters vary widely in accuracy. Take the measurements and get as close as you can and use those numbers. Also, contact the Columbia GIS department. I'll bet they'd pick a couple of elevations (high and low points) on a Chung Course for you from thier digital elevation model of the city. If it's in an area where there's gravity sewer, you might also identify a couple of manhole rims and ask them for those elevations. MH rims are often asbuilt surveyed after they're constructed.

It's very helpful, but not critical, to have exact elevation data to getting good Chung results. Mainly it allows you to better separate out Crr from CdA.

If you can find a course that has:
1 - as much elevation change as possible in a relatively short course (maybe 1 mile max).
2 - has the best side-road tree protection possible (the biggest threat to good results seems to be a 2mph breeze that you don't know is there).
3 - allows you to stay in the bars without ever hitting the breaks.
4 - has little or no vehicular traffic (even one car passing you will cause that lap to be bad data.

Then you can do the loop over and over. When analyzing the data, just assume Crr=0.0053 and solve for the CdA that best lines up the tops and bottoms of all your virtual profile runs. It's not critical that you know the absolute elevation change - only that the virtual profile have a very consistent amplitude.

Email me if you want, then I can contact you by phone and we can discuss it in more detail.