Hi fellow game devs.

I'm trying to implement a center differential in my racing game. However, there are a few things I don't understand about its operation. I'm going to implement it with 2 basic parameters: locking percentage and torque split percentage. Locking percentage is basically how strong the locking torque is before 2 sides break away and rotate at different speed. Torque split is how much torque is going to either side, in case of a asymmetrical differential.

1 - Let's say it's a STi with 40 Front/60 Rear torque split, so 40% of the torque goes to the front diff, and 60% to the rear. Assuming equal inertia at both ends, and the diff is open, the front and the rear differential should be rotating at different speed. The confusing thing is how locking percentage works with torque split. The purpose of locking percentage is to keep the 2 sides rotating at same speed. Then with the asymmetrical split like above, then would this split defeats the purpose of locking percentage? Would it just be better to have symmetrical 50/50 split and a desired locking percentage? Also with the asymmetrical split, would the diff be under heavy load all the time, trying to make 2 sides rotate at same speed, thus wearing it out?

2 - Also there is one thing that bothers me about viscous diff. Let's assume a situation where the differential torque is larger than the drive torque (from engine) to the current wheel, resulting in a negative total torque output, causing the wheel to rotate backwards. This led me to think about how to actually integrate the diff torque. Should I be integrating it as a braking/friction torque instead of the driving torque like the engine? Which means the torque is applied only if the integrated speed is ≥ 0, otherwise the speed will be 0)

I'm sorry for a wall of text. Thank you everyone for bothering with this boring question ?

Great topic ? It bothers me too, I've implemented mine, but not completely happy with it.

The torque split was a riddle for me too, and I read somewhere that in real life the center diff axles always rotate with the same speed, and the torque split is achieved with different diff ratios at front and rear. I'd be glad if someone could second that or is it nonsense?

This could solve the initial problem of center-locking, but as the car moves the feedback torques will generate different speeds at front and back, and that would somehow be reflected back to the center-diff.

The “best” approach I could come up with was, using 1:1 gearing and the same ratio at front and back, but in the diff I just split the output torque according to the setting. The torques are only accelerating the wheels, and all rotations are derived back from wheel rotations. The rear wheel torques are indirectly transferred to the front via the chassis so rotation speeds should be the same after all.

I've read around and thought I understand until I stumbled upon the word “power distribution” and “Torque Apportion” ?
For power distribution, I really can't grasp the idea as well compared to “torque distribution”. Many sites use torque and power interchangeably, making things more confusing.

For locking center diff, I'm making it just like a 2WD locking diff: Same diff ratio front and rear, 50/50 torque split, acting like an open diff, and apply an amount of locking torque to bring both sides to an average velocity.

I use the same approach implementing the torque splitter in some on-demand AWD system. The system will be 2WD most of the time, then only when the primary drive axle is slipping (spinning faster than the other unpowered axle), then an amount of locking torque is applied to the 2 axles with different sign to keep them locked at an average velocity. The maximum amount of locking torque will be the percentage of engine torque that can be routed to the other axle, stated in the torque apportion of the diff manufacture. For example: up to 50% of the engine torque can be apported, the the maximum locking torque will be 50% of the current engine torque, higher than that and the 2 axles slip. Sort of like a proper limited slip center diff with a certain locking percentage. However, I don't know if this is how the system actually works in real life, as some systems uses viscous clutch coupling instead, so there will always be slipping between axles.

For a center diff with asymmetrical torque split, the only thing I can think of is a geared diff (torsen), where the gear inside creates asymmetrical torque split and remains open even if the velocity at 2 sides are different. Then only when the difference is above a certain point then the gearing starts biasing torque around to keep the 2 sides spinning at the desired average velocity. And the hard thing in implementing this is to find out that average speed, when the gearing creates asymmetrical torque split. I can't think of anything for an asymmetrical torque split clutch-based diff, eventhough Subaru has been doing this all the time.

I don't know what @edy would say about this. I love the differential implementation in his sim on Unity Engine.

http://forum.projectcarsgame.com/showthread.php?52783-Magic-inside!-Need-Help-With-Differentials

I just read on Project Cars 2 forum and this is what I think (focus on the Edit on post #2):

a rear wheel drive car that then transfers a bit of torque to the front based on front and rear differential gearing differences and slip differences between the front and rear. The rear power balance value can be used to adjust for the different front and rear final drives (so that some torque is always sent to the front), while the locking settings will affect the dynamic behavior.

I think what this means is the torque split is handled by the different gear ratio at front and rear diff. The “rear power balance” setting is used to compensate for the different speed resulting from different ratio at front and rear diff by sending the correct amount of torque to each diff to prevent one axle spinning faster than the other under full grip. Which means, if “rear power balance” setting sends 50/50 torque to front and rear diff with different gear ratio, then speed difference will occur. So basically it is similar to what you theorized, at the center diff there will be no speed different under full grip.

To further explain the "front and rear differential" bit: If you make the front differential's ratio longer than the rear differential's ratio, when the tyres are rotating at the same speed the driveshaft to the front differential wants to rotate slower than the rear differential. If there's for example a viscous locking connection between the two, the faster spinning rear will cause the viscous LSD to transmit some torque to the front. The bigger the difference in ratios (or the higher the locking in the viscous LSD), the more torque will be sent to the front in all conditions.

Based on the above, with the mentioned setup, if the torque split can be changed on the fly, like Subaru DCCD, then I'm guessing what's changed is the center diff locking percentage, so that an amount of torque will always be transferred to the other axle, resulting in a different torque split ratio.

Also another helpful guide on 4WD and AWD:

https://sites.google.com/site/awdand4wd/