Shift points: EPA testing of manual transmissions (why beating MT ratings is easier)

[MetroMPG]

Admin note --

this discussion originally came from here: http://ecomodder.com/forum/showthrea...age-27384.html

---

If you've eco-driven / hypermiled a range cars and have compared the ability of manuals vs. automatics to beat their EPA ratings, then you have probably seen that manuals typically trounce the EPA city ratings by a much bigger margin than automatics, even without using advanced hypermiling techniques like pulse & glide or engine-off coasting.

Why is that?

That's what this thread is about.

(Note: we're talking about city/combined driving here, not highway driving. Don't think because you can trounce a manual's EPA rating in the city that you can necessarily beat the equivalent automatic car in real world driving (cruising) on the open highway. On the highway, it all comes down to gearing, and if the automatic is geared significantly "taller" than the manual, it might win.)

Originally Posted by cbaber

It would be interesting to see the EPA tests and how they are done. Obviously the M/T is not being driven to it's full potential.

I've done a bit of searching on this...

Not surprisingly, test drivers performing an EPA certified fuel economy/emissions on a dynamometer must accelerate & decelerate the vehicle at precisely defined rates.

They essentially stare at a compter screen hanging in front of the windshield and follow a trace, trying not to "colour outside the lines". It's a big video game. (Deviate from the prescribed speed by over 2 mph, and the test is thrown out.)

This is easy with an automatic: Put it in "D" and work the pedals.

When testing a manual, though, the driver is instructed on the computer screen exacty when to upshift or downshift.

Source: The Truth About EPA City / Highway MPG Estimates
Car & Driver, August 2009, by DAVE VANDERWERP

So there's our answer in a nut shell. And it explains why there's a much bigger opportunity to trounce the EPA in sub/urban or combined driving with a manual than with an automatic.

But what I didn't find (and I'm not going to keep searching right now), is just how the shift points are determined.

---

UPDATE, Nov 13 -- Found the answer. In current EPA testing, shift points are provided by the manufacturer. See post #17 for description & link to source.

----

I've found references that at various times in the past, the EPA used:

(A) Pre-1976 ... exclusive use of universal shift points for all vehicles based on speed thresholds: 1st to 2nd @ 15 mph; 2nd to 3rd @ 25 mph; 3rd to 4th at 40 mph (source)

(B) Vehicle-specific shift points as provided by the manufacturer (which, predictably, led to MPG boosting shenanigans like skip-shifting and/or lugging the engine);

(C) Vehicle-specific shift points that were based on a percentage of max engine RPM or bracketed the engine's torque peak (maybe... that one was part of an old comparison study on the subject);

(D) Shift points based on a vehicle's dashboard shift light/indicator;

(E) Shift points based on research of how actual drivers shifted;

(F) Some combination of the above.

But I didn't find anything definitive for current testing.



Source: EPA-AA-SDSB 81-8
A Summary and Analysis of Comments Received in Response to the EPA/NHTSA Information Request Regarding the Effects of Test Procedure Changes on Fuel Economy
By: James Hourihane, Glenn D. Thompson and Edward LeBaron
November 1980

[Daox]

I wonder how much variation there is from test to test, and how many runs they do.

[MetroMPG]

The C&D article said there's little variation, even with hybrids: "the EPA claims the results are repeatable within one to two percent."

Don't know how many runs they have to do, though.

[niky]

From what I know, it's a set shift point (like you said... we don't know if it's set by the EPA, calculated per vehicle or set by the manufacturer). If the vehicle has a shift indicator light, the tester follows that. If it has skip shift, then the tester is forced to skip shift, so the tester follows that.

[UltArc]

I would think shift points should be matched to the automatic. Of course, the CVT could cause some difficulty in precision, and if there is a difference in gears, it could also make an issue (4 speed auto vs 5 speed standard).

I am uneasy on the statement "the EPA claims the results are repeatable within one to two percent." because of the posts on the sticker.

It had the estimates as the average between the high and low results for the sections. So if they ran the EPA tests, wouldn't it be more like 17-19 City, showing 18, rather than 14-22 showing 18? This is under the impression there is one test, not 10 tests with different variables, and the results all mixed up and used to calculate it.

[PaleMelanesian]

"expected range for most drivers = 14 to 22 mpg"

That's the standard "Your mileage may vary" disclaimer, not the test results.

Edit: I pulled out my calculator and both the city and highway ranges are +-20% from the epa number.

[MetroMPG]

Think I'll split the EPA/shifting discussion into a new thread. It's interesting in its own right, and I can see people searching for this topic...

[UltArc]

Originally Posted by PaleMelanesian

"expected range for most drivers = 14 to 22 mpg"

That's the standard "Your mileage may vary" disclaimer, not the test results.

So they should all be uniform, I would guess.

Stickers
2011 Toyota RAV4- 22 (18-26)/28 (23-33)
2011 Subaru Outback- 18 (14-22)/25 (20-30)
2012 Subaru Outback- 22 (18-26)/29 (24-34)
2012 Ford Mustang- 19 (15-23)/31 (25-37)
2012 Chevrolet Cruze Eco- 28 (23-33)/ 42 (34-50)

At first it seemed uniform, but it seems like there are other variables than just giving a +/-4 range for city, +/-5 for highway.

[PaleMelanesian]

I edited my post above, but it's +- 20%. A couple of your examples are off by a rounding error, so I'm guessing they start with the unrounded number.

[UltArc]

Originally Posted by PaleMelanesian

I edited my post above, but it's +- 20%. A couple of your examples are off by a rounding error, so I'm guessing they start with the unrounded number.

Ah, that would explain the deviation. I see, that is rather disappointing- I always assumed there was some basis other than just a percentile, it seems pointless this way.