Tech Analysis: BMW E46 M3 Shifting

BMW's Sequential Manual Gearbox is a major technical achievement. It was the only significant change to the M product lineup for 2002, but it is newsworthy. european car provided an overview of the system in the April 2002 issue, though it was several months before BMW actually had an SMG M3 for us to drive. We took it to the dragstrip along with an '02 manual transmission M3 and a radar gun, and compared launches, speeds at shifts and time taken to shift. Our goal was not to decide which was best but to characterize the system objectively. When is SMG an advantage? How much of an advantage is it? What difference do the various settings make?
We learned more than we expected to.

Driving Impressions
The manual M3's six-speed is the same transmission used by the M5, but for some reason is not as buttery-smooth in the smaller car. Around town, the M3 can have a slightly notchy throw and jerky clutch engagement. In fact, the first time I ever felt smooth and liked the car was a redline upshift on a freeway onramp, when it dropped into gear effortlessly and resumed acceleration absolutely seamlessly. With some seat time, it is easy to match revs on downshifts, but the manual car is definitely best when doing what an M car was built for--going fast.

The SMG M3 uses exactly the same transmission and clutch, but they are actuated by an electro-hydraulic control system, which also has authority over the throttle. There is no clutch pedal in the car, and the conventional shift lever is replaced by a complex switch, supplemented by paddles mounted to the back of the steering wheel. Shifting can be accomplished using either control. Sequential mode responds to driver input. In automated mode, the computer does its own thing, as if emulating an automatic transmission. There are five settings for shift aggressiveness (speed) in each mode, plus a "maximum-attack" sixth level available only in sequential mode with DSC turned off. BMW states that SMG, in its fastest settings, is faster and more precise than any human could be.

Around town, the SMG car moves away from a stop smoothly, but one must understand that "automated" mode is not the same as an "automatic" transmission. Up- and downshifts are annoying if the car is driven like an automatic. On an upshift, a lurch is felt as the clutch disengages and acceleration ceases, then a pause while the transmission shifts. Finally, acceleration resumes just as abruptly. On downshifts, the engine is often 500 to 1000 rpm short of having matched revs when the clutch engages, causing another lurch. These patterns are felt in sequential mode as well, but they are less annoying because they are responses to driver inputs.

To drive an SMG car smoothly in everyday conditions, you must realize that even though you are not moving a shift lever, you still have to drive it. Lift the throttle lightly for upshifts, rev it a little extra for downshifts. Then it is smooth. Still, it seems as if BMW missed the mark on this part of the calibration and didn't quite deliver all the technology's promise. The pause while shifting in the less aggressive settings is often longer than even gentle shifts with a manual transmission. One solution that seems natural is, instead of lengthening the shift actuation so much, use the extra shift time to breathe off and back on the throttle more gradually.

On a mountain road or racetrack, however, the situation is entirely different. SMG feels as good as advertised. Put it on setting 6, or 5 if you want DSC on, and you quickly feel you couldn't match its speed or precision, whether shifting down or up. Certainly, the art of driving, the complexity and beauty of the dance we love, is diminished, but letting the car do the shifting allows the driver to focus more on his/her line and controlling the chassis. Given equal drivers, an SMG M3 will be faster than the six-speed.

Only two issues arose when driving fast with SMG. First, it ignores downshift requests rather than storing them. If the downshift paddle is pulled too early in braking to be completed without over-revving the engine, the request is ignored. When getting back on the throttle, the driver finds he/she is still in the higher gear--which quickly leads to the other issue: column- versus wheel-mounted shift paddles. In a very tight, autocross-type turn, the wheel may be upside down, switching which side to pull to shift up or down. So when intending to shift from third to second, it becomes easy to inadvertently shift to fourth instead. With column-mounted paddles, they would be exactly where they always are.

One solution (besides the obvious one of not downshifting early) is to make a habit of using the lever on the transmission tunnel. Unfortunately, that would eliminate many of SMG's advantages. In WRC cars, which must be shifted with the steering wheel doing crazy things, the shifter is usually either a dash-mounted lever or a push-pull ring behind the steering wheel. Overall, I find SMG to be an excellent system from the driver's seat. If I only drove on the street, I might stick with the manual, but for track use against a clock, I'd go for SMG every time. I'd like to see a software revision within a year or two, but it's very good right now.

Track Testing
The test used to learn about SMG's calibration was essentially the same procedure used for our review of Volkswagen transmissions in the September '02 issue. Both M3s were stock, fitted with Michelin Pilot Sport tires. DSC was off. Table 1 contains acceleration figures. Our normal procedure is to round these figures to one decimal place to ensure repeatability, but we do some hair-splitting in this test, so we left them at two decimal places to more accurately represent the differences we found.

Table 2 shows the vehicle speed at which each shift occurred and the time required to execute each shift. These measurements were made for the manual car, for the SMG car in automated mode with settings 1, 3 and 5 and for the SMG car in sequential mode with settings 1, 5 and 6. The shift time was measured from the first significant decrease in acceleration to the return to full acceleration. Because the measurements required visual judgments about curves, there is some uncertainty, estimated to be 0.02 sec. for each measurement. The engineers who calibrated the SMG's shift parameters may disagree with the precise figures. european car's numbers are based on recorded vehicle acceleration, not the actuation of the transmission itself.

The impressive thing about the manual transmission car on the dragstrip was the ease of driving it. Having barely driven an E46 M3 before the test, our driver's fourth run was just about perfect, with a 2.47-sec. 60-ft time. He then backed it up with two more basically identical runs. The M3 was fast and easy to get off the line, despite not being intended for this. Our driver was also fast, with a total time spent shifting of just 0.87 sec. I was worried the test would end with a clutch explosion, but no sign of trouble was detected before the car went home.

In automated mode with the SMG car, we were reminded that setting 1 is the "winter" mode. The car started in second gear and shifted at 7000 rpm, giving it drastically slower times across the board. The total shifting time was only 1.05 sec., but that was for two shifts instead of three. Runs one and two had strange oscillations in the acceleration rate before and after the 2-3 shift, as if the car were still trying to figure out what to do with the throttle. Runs three and four were consistent and smooth, as well as faster. However, the "flutter" in the acceleration curve was present in all runs before and after the 3-4 shift, suggesting a possible feedback control issue with the throttle. In any case, it wasn't noticed from the driver's seat and isn't a big deal in "winter" mode. Automated setting 3 was more than a second faster in the quarter, with a decent launch and 7500-rpm shifts. We found that no special tricks or "launch control" algorithm was required to make the SMG do a burnout. Just stomp on the throttle. If "stomp speed" is quick enough, it will light up the rears. It will also shift into second gear at 15 mph, making that the slow way. Interestingly, this 15-mph shift took only 0.20 sec., the fastest shift we saw all day.

A lot of the credit for the ease and consistency with which the M3 can be launched goes to the forgiving nature of the Michelin tires. On these and subsequent runs on setting 5, acceleration with the tires spinning and hooked up was almost indistinguishable. Neither was quite as good as a good launch with the manual, however. Shifts took approximately twice as long as our driver achieved with the manual, and the total time spent shifting was a second greater.

On automated setting 5, it was even easier to light up the rears, with 0.20-sec. shifts to second gear at 15 mph again making it slow. The computer was less consistent, or slower to dial in to this pavement surface, than our human driver. The best launch with the manual was better than the best with SMG, and the average manual launches were the same as the very best with SMG. Shifts were now executed at the S54 engine's 8000-rpm redline, taking between 0.30 and 0.35 sec., not quite as fast as our skilled driver. On the other hand, our driver does this all the time. While testing the SMG in automated mode, he just kept his right foot planted and steered in a straight line. The SMG car was within three-tenths of the manual car in the quarter mile.

The SMG M3 was tested in sequential mode on settings 1, 5 and 6. In sequential mode, the calibration is a little more aggressive overall than in automated. On setting 1, the runs were very consistent. Launch was comparable to automated mode on setting 3, and shifts took just slightly less time to execute. The sequential mode's advantage was gained in shifting at redline, according to the driver's input.

At sequential setting 5, the launches were again extremely consistent, with all the 0 to 30 times falling between 2.40 and 2.44 sec. Our driver thought the car was shifting faster than he did, but the radar data shows shift speeds comparable to his, with a total time spent shifting of 0.84 sec. versus 0.87 sec. for the manual. The uncertainty in those measurements overlaps. However, with other aspects of driving to attend to, such as braking and turning, none of us would question that SMG could do its job faster and more consistently.

At the "maximum attack" setting 6, shift times decreased again, bringing the total to 0.80 sec. This is indeed faster than our human driver achieved, and the launch was getting closer. The best quarter mile was 13.61 sec., two-tenths slower than the manual M3. The difference was seen by 60 mph, the SMG car making that sprint in 5.46 sec. versus 5.22 sec. for the manual. On setting 6, the SMG car was astonishingly consistent, laying down three runs that were virtually indistinguishable on the radar traces, with quarter-mile times separated by only 0.02 sec.

Comparing those runs to the best manual run, we discovered where the difference lay. All the SMG runs seemed to fall off a little between 6 and 12 mph, then resume with the same acceleration as the manual. The ground lost by the SMG at that point translates into area under the curve through the whole run, with even faster shifting not making up the difference. There are two possible reasons for the data to look this way. One is that the SMG car leaves hard with a slipping clutch, then bogs due to its tires slipping less than the manual car's tires.

The other is hidden within our test methodology. The radar gun is inconsistent at low speeds, showing speed-versus-time curves that are clearly impossible. To eliminate this problem, we have to delete the first data points from a run, typically below about 4 mph. The software then interpolates a straight line to a theoretical starting point. Thus, the "bog" we detected could simply be the end of a slow, non-tire-spinning launch that we were unable to capture with the equipment available. The effect of that on this test would be to make all the SMG runs appear to have a better launch than they actually did. That would agree with the visual impression shared by everyone present, which was that the SMG car never left the line nearly as hard as the manual car. There is test equipment that would not have this shortcoming, but it is unfortunately well outside our budget. (european car would, of course, be very receptive to any company that wanted to lend it for to us for review)

Conclusion
We learned that, for drag racing, a manual transmission M3 is faster. It is easy to launch in a way that the SMG car just won't match. Once moving, however, SMG in its most aggressive setting can shift about 10% faster than a very skilled driver going in a straight line. It will also shift that fast while the driver is focusing on other aspects of driving, such as braking and cornering, or traffic. Thus, for road races, track lapping days or autocross, SMG will undoubtedly be faster and easier to drive.

As for day-to-day use, it's a matter of personal preference. SMG could be improved dramatically in this area, as a skilled driver shifts smoother and faster, even when relaxed, than SMG when left to itself in its moderate settings. SMG does have the advantage of no clutch pedal, a help in stop-and-go traffic and convenient if one would like to share the experience with a driver who is not adept at shifting manually. Some of us still prefer the physical, emotional and ritualistic connection with the vehicle provided by the conventional manual transmission.

Perhaps the most impressive thing about SMG is the commitment it shows on BMW's part to keep the M3 pure. There are many otherwise fine driver's cars that are often ordered with automatic transmissions instead of manuals and are greatly diminished because of it. The ability to control the chassis' balance precisely in a turn and to put the power down exactly when it is wanted is traded for an easier and more mindless freeway commute, or the ability to drive the car without having learned to use a third pedal. That will never happen with the E46 M3. In the end, the philosophy is as impressive as the technology itself, and may be more important to BMW.