BMW MSS65 Torque Management and SMG control

Hey guys, thought I'd address a couple of things here:

So Randy is testing the SMG mods for me and he's on round 1 of the changes. To go in greater detail, there are numerous 2D maps associated with the SMG shift characteristics in the DME as well as a *ton* of single byte variables and constants. Much what is done is also handled by PID control and some of the parameters that feed that are handled in the DME. So, by modifying these, we can control some of the characteristics of how the SMG shifts.

I think what many people don't realize about the mss65 and all modern BMW DMEs is they operate on a "torque model". What I mean by that is the DME actually knows how much torque the engine is putting out and what loads to the engine are consuming that torque. This is necessary to provide the level of integration we see on these cars and even more so in their current day offerings with hybrid drivetrains having more than one torque production means.

In order for the DSC/ABS to interact with the DME and provide their functions, they actually request torque increases(!) and decreases via the PT-CAN bus directly from the DME. The DME takes into account the driver's requested torque (from here on out, consider it a torque pedal and not a gas pedal), the torque requested by the cruise control, the torque requested by the DSC (positive or negative), the torque needed to drive the alternator and AC compressor (yes, the actually talk to the DME and communicate this), and finally, the most obvious, the torque applied to the SMG. All you 6MT guys can tune out here (pun intended) for the SMG part, but everything else is the same.

For the longest time, we all understood the to properly shift, the SMG had to match RPM on the layshaft side and flywheel side of the clutch. We hear this on downshifts, we know the DME/SMG do it, but there is so much more to it than that. If you think carefully what a MT driver must do to execute a smooth, lurch free shift, we can start to understand it a little better.

In order to shift smoothly, we need to not only match RPM on either side of the clutch a re-engagement, but we have to take into account how rapidly the vehicle is accelerating or decelerating as well. Think about that, for the smoothest acceleration to happen, ideally we want the drivetrain to put out the exact same torque to the ground on the incoming gear that the outgoing gear was. The difficulty in that is the change in torque multiplication (or division in 7th gear) that happens when changing gear ratio. This means the SMG needs to know it's acceleration rate. This is where the longitudinal sensor comes in. We've all seen it and heard of it when running the SMG adaptation process, but only vaguely understand why it's there. Quite simply, the SMG monitors longitudinal axis acceleration and does some fancy math taking into account vehicle weight (ever wonder why the M5 and M6 have different SMG flashes?), current gear ratio, etc and can determine *exactly how much torque* is being applied to the driveshaft. It doesn't care about knowing the final drive ratio, because it never has to calculate anything changing for ratio. So, in order to maintain an exact amount of longitudinal acceleration, the SMG looks at how much torque it's putting out, then looks at the gear the driver just requested, and calculates the exact amount of torque the engine should produce to maintain driveshaft torque constant when the clutch is released. Yes, the SMG actually requests a specific value in newton-meters from the DME.

Similarly, when wheel spin is detected, the DSC requests a specific newton-meter reduction in torque from the DME to reduce the slip to zero (DSC in normal) or to some given allowable slip rate (M-dynamic). Additionally, the DSC can request an increase in torque in a sliding situation and then clamp a single brake caliper on the rear to force the diff to accelerate a single wheel. Yep, all that happens via the torque model in the DME.

So, understanding the torque model allows for smoothing the SMG a few different ways. First off, how does the DME know how much torque the engine is producing? Does it have strain gages measuring torque output? A handy idea that has made its way into cycling for the last 15 years, but not to our cars (yet). No, the DME just has a plain dumb map of torque stored in it. When BMW went through the long initial calibration process for the mss65, they recorded the torque output of the engine in a 2D table. This map is then referenced by the torque model and when a specific N-m output is called for, it looks up the correct throttle plate opening for the current engine RPM.


What happens when we start tuning the engine and increasing power output? Does the DME know it's more powerful? Nope. Not unless the torque maps were altered to reflect what the new engine calibration was capable of. For the most part, the effects will be minimal, but in some cases when the output is significantly increased (e.g. the 6.0 strokers), things can get a little tricky. There is also a self-check routine that if the DME believes that torque output is significantly higher than the map, something must be wrong and a safety feature kicks in and can put the DME in a limp mode.

It should be clear at this point that if anyone is still of the belief that performance tuners are just "reverse-engineering" a bunch of hex code and changing maps randomly that would be physically impossible to ever find anything useful.

All of that comes back around to the SMG modifications and a good, solid fundamental understanding of how the SMG works is essential to get the best performance out of it. Remember, BMW had to somewhat tone the car down for sale to the masses. We know it can be full on mental (S6 no-lift shifts), but a lot of improvement can be made knowing the target driver audience better.

Remember back in the day BMW claimed the SMG3 GS7S46BG could execute a gear change in 30 milliseconds? Anyone timed a no lift shift in S6? It isn't 30 msec. I was excited to see Randy's datalogs showing the ignition timing cuts there are made during a shift at WOT. This is the clearest indicator, looking at the duration of that cut. I still think there is room for improvement, but with the obvious analysis and consideration of durability and longevity.

The change Randy is testing, I'm very satisfied with, but waiting for him to accumulate some miles and incorporate his input before I make it available to everyone. Stay tuned, more to follow.