355 Hacking the F1 TCU

Yes, it is still alive. You will be updated soon, if you like.

It is not.

 

Part 12 of Hacking the F1 TCU

When I started this project I was aware that there would be little progress in the beginning. And this turned out to be right. For a long time there was nothing exciting I could present to my client. On the other hand I knew that at some point I will achieve a breakthrough. And it came when I had all parts together to finally understand the shift process.

For a start it roughly works the way you might expect it: Disengage current gear, select shift track of target gear and then engage it (I leave out clutch control here to simplify things).

Image Unavailable, Please Login

However doing it like this, shifting would be dead slow. Even slower than it is now. That's why the engineers at Magneti Marelli implemented some tricks. So this is how our TCU performs a gear shift in reality:


Disengage current gear
The very first action, even before the clutch gets operated at all, is to preload the control arm into the direction of disengagement. The target of this is to get rid of any wear or play from the mechanics (e.g. joints, shift collar etc.) and to have a some amount of pressure ready. Shifting has not started yet here and the current gear is still transmitting torque.
Few 10 ms later the TCU opens the clutch. It may happen that the preload causes the gear to disengage already now. However usually the next step is to perform the actual disengagement. Therefore the TCU drives the solenoids with a decent current. The amount is proportional to the remaining distance to travel. That is at first it sets up a high current (which results in a high force). Then as the gear disengages, the current is gradually reduced until the gear pops out. This way the risk of overshooting to the adjacent gear is mitigated.

Select target gear
Basically the TCU just activates the solenoids to select the right shift track. There is nothing more to it, as all affected solenoids are of On/Off type. That is the TCU cannot control the force that they apply.
However there is something worth mentioning about the point in time the selection process starts:
- If the shift track shall remain the same (e.g. 5-6 shift), selection already starts during the previous disengagement phase. Due to that the control arm is pushed to the desired selection position right from the beginning. This way no time is wasted for selection.
- If the shift track shall change (e.g 4-5 shift), the TCU starts selection when the gear is close to be - but not yet fully is - disengaged. This is possible as both, the hydraulic and the mechanic system, have some inertia to overcome. That is from the moment the TCU triggers selection until the control arm starts moving, it takes some milli seconds. Interestingly a calibration switch in the software disables this feature though (such that the TCU waits for a slow but safe disengagement).

Engage target gear
As soon as the control arm reaches the desired shift track, engagement begins. A particular current profile is applied for that. First the TCU drives the designated solenoid with an initial current. It is then gradually increased up to a maximum current, which is kept for a certain amount of time. During all that, the gear usually engages at some point. If not, the TCU applies a boost current. This is the last resort where it demands everything the hydraulic is capable of! Now the gear should finally engage. Then the shift process ends. (Otherwise the shift is retried or, if the problem persists, aborted.)

Image Unavailable, Please Login


Each of the phases mentioned is implemented in a dedicated state machine. With another state machine acting as a master. Moreover various limp home strategies exist that jump in e.g. when a sensor fails. Due to this it was so difficult and time consuming for me to discover and understand all that in detail. However now I eventually reached the right spot to proceed with my main goal!

By the way, all parameters (current levels, timings, position thresholds etc.) are completely variable. For example the mentioned current level for engagement depend on shift direction, throttle position and engine speed. Additionally there is a special support for upshifts: Depending on the drivers inputs a "shift style" is derived and according to that an "overdrive" is fired. With that the gear is really punched in! Try it out, you will notice a difference according to
the table below.

- Normal -> No overdrive (100%)
- Advanced -> Moderate overdrive (130%)
- Aggressive -> Full overdrive (300%)

Image Unavailable, Please Login

 

My customer and I planned for 24 months. The target was to reduce shift times by 20% or 50 ms. There is an option for further improvements though.

60 ms would be incredible! But unrealistic for the F1 in the F355 to the same degree. This is what the latest supercar AMTs achieve under best conditions! But reaching 60 ms is not necessary. Knowing the stock F1, you would be excited and highly satisfied with much slower times.

Thanks for the link to TVS.

Wolfgang

 

I wish I had this before. Not because of the technical description. But it would've helped me a lot in choosing concise names for the variables in the software. If I recall correctly, I had one variable that I called "actual demanded target clutch position", because it was holding the current set value of the target clutch position.

If I would've had predefined names it would've been a matter of simply doing the mapping.

To my information the design of the rocker was not changed before the 430. Thus the fastest shift times theoretically achievable would be that of the 360 CS. Which is in my opinion very nice compared to the stock F355.

Thanks for the great pictures!

 

Part 13 of Hacking the F1 TCU

Finally reversing the shift process was an important success! For my next milestone, I would now leave theory behind me and move towards practical work. But before I would start calibrating and tuning, I had to prepare a few things. The first was to properly record the current situation as reference for my future improvements. Or in other words to find out the time an untuned, "Stage 0" TCU takes for a gear shift

What's the definition of shift time anyway? Usually it is defined as the time during which no acceleration is possible. Having a look at the WSM, Ferrari mentions three phases in this context: Torque reduction (I), Actuation (II) and Torque return (III)

Image Unavailable, Please Login


Torque reduction and return mainly refer to actuation of the clutch. However, they also include Torque adaption by the Motronic as shown at the top of the above figure. At the very beginning as well as at the very end of a shift, our TCU asks the Motronic to adapt torque in the way to support the shift. Both greatly increases comfort. Unfortunately, torque adaption is comparatively slow until it becomes effective. Thus, depending on the car maker's philosophy, it is sometimes skipped at heavy load (WOT). If so, the clutch operates (almost) under full torque. You certainly noticed in your F355 that shifts at high load and high rpm are rough (making clear, which philosophy Ferrari shares).

By the way, there is a trick when it comes to advertised shift times: Of course actuation of the clutch also takes some time. Therefore, some car makers take the above definition of shift time literally. As long as there is any positive torque delivered by the drivetrain (and thus accleration possible), shifting is not considered as ongoing. That is even a half open clutch does not contribute to the shift time, as long as overall torque stays positive. This way the advertised shift time improves by a few milliseconds

I'm not interested in advertisement. My definition of shift time exclusively depends on whether the clutch left its rest position (shift) or not (no shift).
Image Unavailable, Please Login




To get a feeling of the stock shift times, I arranged a meeting with my client. When we met I installed my measurement equipment in his F355. Then we had a test drive. Below you find the result of a 2-3 shift under low load condition (throttle around 10%)

Image Unavailable, Please Login


The red graph shows the clutch position, green is engagement position and blue selection position of the control arm. At first the clutch is closed. It then opens quickly (marker a, top left). After that the TCU disengages the current gear until neutral is reached and the control arm is freed. Then it moves the arm until the shift track for the target gear is selected (marker b). Now engagement starts. The arm moves the collar into the direction of the target gear until the sync disc touches. It stays there until the speed of input and output shaft match (c). Then the gear is fully engaged (d). Finally the clutch is closed. At first this happens quickly (e), but then rather slowly (f), until it is fully closed at the end (g).
Have a look at the cursors (white bars), which measure shift time. As you can see, the entire process took around 1.5s. Obviously the TCU was not in a hurry. It clearly preferred comfort over shift speed. Especially the phase of closing of the clutch consumes a lot of time.


Side note: You may have noticed something odd in the measurement. The selection position jumps though gear engagement has already finished (x1). As explained in a previous post, all solenoids are switched off as soon as shifting finishes. At this point the control arm automatically moves into its rest position. Usually this is hardly noticeable, however in the above measurement the selection position changes a lot. (The engagement position also changes (x2), but only a little.) Thus, the shift linkage definitely is misaligned in terms of gear selection!
The selection position is set by turning the gear linkage shaft. Back then I supposed that this shaft moved a bit after all the years. I explained it to my client. He then asked his mechanic to check, who confirmed and finally corrected the misalignment.


All right. With this setup I was now able to indepandantly measure gear shifts in reality. However, before starting with calibration and tuning, I had two more things to prepare. Of course it's my pleasure to tell you about them in my next posts.

 

You're right here. The most relaxed shift strategy though is for down shifts due to low speed, e.g. when approaching a red traffic light.


I took a lot more samples back then. However, when I came back to my office and analyzed them, I noticed something interesting: When we did our test drive, we had a warm-up, some specific driving scenarios I asked him for (including harder accelerations) and in between a lot of normal driving. The sample you see above is a typical example for the latter one. It is an ordinary shift, I had plenty of these in my measurement.

Technically it is perfectly fine. The increased shift time doesn't matter when you accelerate moderately, e.g. like you do in normal traffic. But emotionally it is a disappointment. The car feels lazy and not like a Ferrari.

I know the F355 as imposing, impulsing, appealing, fun! With these shifts, which you experience most of the time, the F1 cannot keep up. In my interpretation that's the true reason why most people dislike the F1. Technologically it is superior to the manual. Unfortunately the F355 is not about technology, but about passion and fun.

That's why I chose this sample. It will really help the 355F1 to speed up these shifts! Along with the benchmark shift at full throttle, of course.

 

Please keep in mind that I'm doing business when reversing the TCU. I'm getting payed for everything you find in this blog.

Although I had no look at the damper ECU yet, I assume that it is not overly complex. So you might be better off to find an enthusiast with the right skills and then convince him of your idea. But please keep in mind that hacking will only be one part of the story. You'll also need an expert for chassis setup, an application engineer and test driver for re-calibration, access to a test track etc. Otherwise your result will not get any better than what's currently inside the ECU.