Exhaust brakes flaw [Edit] : True cause : a clutch flaw

[michbret]

Hello,

I like a lot the exhaust brake simulation and I think that it is a very nice feature (thanks SCS).

However, I think there is a flaw in the way it is implemented in the game.

When you activate the exhaust brake, your truck engine immediately looses something like 50 or 100 rpm.

When you release the exhaust brake, your truck engine immediately regains something like 50 or 100 rpm.

Basically, this is not correct since there is a direct relationship (gear and transmission ratio) between the speed of the truck and its engine rpm as long as the clutch stay engaged. And the effect of the exhaust brake activation/ de-activation is not instantaneous !!!! (I don't think such big truck are using variator gears !)

This would be a minor or cosmetic/visual problem but there is a side effect (that I suspect strongly).

If you drive in sequential gear mode, there is some sort of automatic clutch management : When you engine rpm become too small, the clutch is disengaged so you will not stall (like a beginner with manual gearing).

I have the strong feeling that this happens sooner when you are using the exhaust brake as a side effect of the 50-100 rpm lost.

I am on the way to instrument some logging using the telemetry interface in order to sort things out but I choose to share this before going to far in code writing.

All version impacted since exhaust brake implementation.

ETS2 1.8.2.5s with all DLC. No mods

Best regards and thanks for this feature.

Michel

[Cadde]

This is not a flaw of the exhaust brake per say, more so the fact that the "clutch" has a very low viscosity. (200 by default)
I have set my clutch viscosity to 800 (makes it really difficult to get off the line though) and it behaves just like it should in game. (Perhaps a bit too good with 800 though)

The one problem i have with the game in it's standard configuration is the lack of a setting to adjust the clutch viscosity value in game. Even more so, automatically adjusting it when you switch from manual (H-Shifter) gearbox to automatic/sequential.

And also some way (unless i totally missed it) to adjust where the clutch starts (not offsetting the axis, in the game sense) and how much travel there is between no grip and 100% lock.

[michbret]

That's very interesting Cadde. Thanks for the analysis.

So this clutch viscosity factor should also have an effect when you down-shift (e.g. gear 10 to gear 9).

Another point is that, in manual gear, with a H-shifter, you can easily down-shift "brutally" (mismatching between buttons and so on) thus engaging e.g. gear 2 at 80 kph (a big virtual mechanical breakdown). I never tried this in game but will certainly do soon. This is probably impossible on modern truck and on old one that quite sure that the gear would never engage if the differential speed is too high at the pinion level.

[ohaha]

normally, the synchros (don't know technical term) in the gearbox will never allow you to do that, unless they're really bad... sometimes it's impossible to shift down even by two gears, not to mention six...

[michbret]

You are clever, Cadde.

Definitively, the clutch simulation doesn't look correct (at least in sequential mode).

Even with an heavy load (truck and trailer ~ 40 t), when you switch from idle to full acceleration you immediately gain rpm without immediately gaining speed. In the other way, release the accelerator from full and you will loose rpm even if you are still gaining speed (going down a slope) !!!

Based on all what I know about mechanics and physics, that's not correct.

Should I rename this thread with something like "wrong clutch simulation" ? This may be more accurate for the developers, if they ever have a look here ?

[Max]

devs know that for a long time (its one of oldest issues).
and they are watching as well ...

[ohaha]

I'm watching, too...
I made a few enemies with this

[michbret]

Wow, everybody is watching
[ external image ]

If the developers are aware (Thanks Max), I think that the best thing to do is to wait and see. Who knows, may be in the announced 1.9 and its new physics.

Meantime, I will play a little with def/vehicle/physics.sii following Cadde example.

Thanks for watching !

[michbret]

Hello again,

I continued to make some investigations around this clutch bug/flaw.

The game "propulsion" engine is apparently based on torque and viscosity effect. That's OK for me since it provides something close to the real physics without going into intensive computations (never forget that the game must run on a wide range of computer).

I go on with the reverse engineering (so things are not certain ...).
- When the clutch is engaged, the viscosity is applied in 9 steps at the rhythm of the physical update. This one is 60 Hz and is not (loosely) correlated to the rendering rate (the FPS we get). So the full viscosity is applied in 160 ms (approx) which is a very short time so the humans feel this as something instantaneous.

- When the clutch is dis-engaged, this is done the same way, in 160 ms (approx)

- The clutch is automatically engaged (except when using a pedal and manual gear mode / H-shifter) when the engine RPM reach 1000 (approx)
- The clutch is automatically dis-engaged when the RPM fall below 600 (approx).

So we have an hysteresis and that's nice (good SCS).
But the clutch is engaged too brutally : 160 ms is short in regard with the engine inertia .... That the first problem

If you have a clutch pedal, it is used by the game as an on/off switch, the on/off point being around the middle of the area where your pedal is operating (limited by deadzone and range). That's the second problem

The default viscosity is 200 (thanks Cadde for putting me on the right way).
Since the clutch engage too quickly this value is to high for engine with low torque (small one) when you start the truck of flat. This is worse if you start on climb with an heavy load.
In other words, you cannot make the clutch spin properly as we (those who drive) learned to do. That's the third problem

But the default viscosity is also too low for a proper simulation of a truck with clutch engaged : In theory this value should be infinite or very high in this situation.
I have made some tries and even a value of 2000 is not enough. But with 2000 applied in 160 ms, starting the truck is a real challenge.

The side effects of this too low value are :
- RPM non realistic change with exhaust brake activation/de-activation
- RPM non realistic change with fast change applied to the accelerator (especially in low gear with powerful engine and heavy load ==> big inertia)
- Big loss in propulsive power/torque especially when climbing with heavy load : This create an artificial limitation in the performance of the truck. The problems that arise with a load of 75t or more take their root here (In the real life, I have seen some overweight - exceptional - heavier than 80t moved with a single standard truck).
That's the fourth problem


What is needed.

- For the manual case (pedal clutch available) that's simple : change the switch behaviour in something that exploit the clutch pedal position : But please not a basic linear interpolation. Something more complex is needed like quick passage from 0 to 50 (viscosity value), the a long progressive between 50 and 1500 and a quick raise to the infinite value (10000 ?)
- For the other mode which are "automatic", this is more complex since a feedback loop with the engine rpm (you never stop the engine in automatic) and the automatic gear box (non sequential) is needed. Truck speed may also be a useful information.


I think this would be a nice improvement for ETS2 but also for ATS (heavier trucks in the America) or, let's dream, some "Extreme Trucking" new game ....

How would I do this if SCS doesn't improve it.

That not easy to do this outside the game. The overall design would be
- Writing some external program that would run in background but may be only a plugin dll can do this.
- Get necessary informations from the telemetry SDK interfaces
- For the manual clutch position, I may need to register to Direct Input or use Logitech SDK
- Compute the required viscosity using the algorithm mock-up mentioned above.
- Patch the required viscosity into the game memory as the ACC program is doing for the cruise control

When I see all that has been done by talented and clever gamer (ACC, head movement, camera move, ...) I have the feeling that all the basic program needed (except the automated clutch algorithm) are around in pieces.

There is a big drawback : Doing such thing at home would remind me a lot, the avionics systems I design during my work hours

[Cadde]

My idea on how to "fix" the issue (outside of SCS intervention) michbret describes is to just use a virtual joystick (as i am already messing with, but not enough time) to move the clutch in and out of the "bite" area to simulate an average viscosity. It would be jerky getting off the line and may not feel entirely proper but it would still be an improvement over the physics model in existence right now.
It would rapidly engage and disengage the clutch depending on the real device input.

Another option would be what michbret already described, feed the game new viscosity values based on physical device axis position BUT it's going to be difficult finding the location of this resource in memory because it doesn't change and i am not capable or willing to find it in a disassembler.

Final option would be to find the location of the method that controls how the clutch is applied and re-write it in assembly so it doesn't do what has been (accurately as i see it) described. This is also not something i would be capable (maybe i am underestimating my abilities) and/or willing to do.

So best case is SCS provides a method to adjust how the clutch operates in game so an (advanced?) user can alter it to his/her liking.