So, what I understand about Torsen diff: -It is kind of opposite of open diff -Transfers power to wheel that has more grip, open diff transfers power to wheel that has less grip -If there is two driven wheels and one is lifted up, Torsen does nothing differently from open diff, it needs some resistance from wheel that has less grip, with brake you can help Torsen a lot in such situation. I can of course make LSD type and have low preload, high locking under power, but that is not really doing it quite right, still wrong wheel has more power and having it welded under power is not really right either. So I think Torsen is something that would be nice addition, ETK, I mean Audi has used such a lot for example.
That was nice video! However they kinda forgot to mention the situation where you loose completely traction to one wheel, then Torsen can't apply any torque to another either, it kinda multiplies the weak side torque to side that has more grip, for example 5 times multiplication, but if there is zero it is not possible to multiply, but use handbrake and you get brake torque multiplied by that said 5 times, which gives you more torque to grippy side than what brake torque is. That is why handbrake trick works well with a Torsen diff. Then there was a British manufacturer which name I have forgotten, makes gears for rally transmissions and also kind of variation of torsen that had differently shaped gears, helical/hexagon or something like that, sorry limits of my English and memory. Anyway that British made diff could transfer power even if one wheel was off the ground and still had same features as Torsen. Oh, now I found it https://shop.quaife.co.uk/differentials those some have claimed to be the best. Both Torsen and Quaife are really smooth and does not initiate power slide as easy as other types. Can I has both, please?
An lsd with low/no preload is pretty much exactly how a torsen works. Because there is no preload, when a wheel is lifted off the ground, the torque across and into the diff drops to 0 (other than the torque to accelerate the wheel freely which is short lived) which means no leverage to enact the locking behavior. A regular clutch lsd with no preload would do the same. Once you add preload to a clutch diff, that gives it a baseline torque that can allow the locking function to act on (the baseline resistance forces the pin up the ramp, which increases the locking, and the effect runs away to lock up the diff) Furthermore torque ratio is wholly determined by the wheel traction in combination with the differential. If one wheel is off the ground with an lsd with preload, the torque ratio is basically infinite, because the wheel off the ground has no reaction, and the wheel on the ground can react the preload torque. A torsen is just capable of more torque ratio than a clutch lsd without the downsides of clutch lsds that can chatter and act unpredictably with steep locking ramp angles. For a basic model there isn't a fundamental difference.
Thanks, I think I might understand a bit. In wiki there is information that going higher than 0.5 cross torque could cause problems, but with 1:3 or 1:5 multiplication ratio my understanding is that I have to set lockingCoef to 0.67 (1:3) or 0.8 (1:5). I did test 0.67 and I think it might be how thing should work, I could not see any problems either, so if I understand this correctly it might be something to add wiki too. This is really useful stuff I believe
I've been working at this, but I can't seem to get it to unlock very much when turning. This was a good turn where it worked well, but the next turn ended like this:
At a lockCoef of 1, the differential can produce as much cross torque as input torque. That might not be the typical definition for multiplication ratio or something used in literature, it might need some conversion. With 0.5 lockCoef I remember seeing 3 or 4:1 torque bias ratio during cornering on throttle with some rwd cars (by printing to console from the differential lua), but torque ratio is an end effect, a dynamic, measured value, not an actual mechanic of how things work. Example: with 1000 input torque, and locking coef 0.5 you could achieve 1000, 0 torque L/R. With lockingCoef 1, you could achieve 1500, -500 L/R.
Hmm, I have this kind of diff, no preload and I guess 0.5 is what I should have. ["differential", "differential_R", "driveshaft_R", 1, {"diffType":"lsd", "gearRatio": 3.266, "lsdPreload":0, "lsdLockCoef":0.5, "lsdRevLockCoef":0.5, "friction":8}] It seems not to be issue that one wheel is up in to air: There is a mystery, I must say this is very confusing, but based on what you have told me, I think that diff should be correctly set, but perhaps it is effect of friction, that might need to be lower and for helical type higher, if I have understood correctly (that is a rare event though!). So 75% of torque to slower wheel (when setup for 3:1 TB ratio), but if faster wheel is airborne there can't be torque to slower wheel as it will just multiply torque of faster wheel, that I think I'm seeing, or maybe even bit more, but unless I'm understanding something wrong, I have here more torque being applied to slower side: With open diff, it seems to be same amount to both sides, with ridiculous amount of wheel spin: Only issue is that wheel up in the air, but lowering lockCoef from 0.5 to 0.4 solves that, however with 0.4 there is ridiculous amount of wheelspin, with 0.45 wheels seem to be better locked up while still having correct behavior when one wheel is up. Of course one wheel on ice, there should be only amount of torque ice side wheel has multiplied by torque bias ratio to other side, rest of the torque should become wheelspin, so maybe 0.4 is near upper limit, going above that might be too far. This is of course not very scientific way of setting up diff, but maybe works too, at least until all this sinks into my brain
As the one on ice tries to spin up, some torque is taken away from it and passed to the other side. That is the restoring force or locking force trying to force the wheels to go the same speed. With a perfect limited slip differential you would end with the wheels at the same speed and each only receiving as much torque as it can react without slipping. If the ratio of tire traction is 4:1 you would hope for 4:1 torque bias. There is a little bit of "softening" to the diff code to improve stability. So you can expect a speed difference of around 1mph before the full amount of locking torque can occur. It needs a smooth ramp up of action to prevent unstable oscillation. About the video with no preload and 0.5 lockCoef still going up the ramp, I think that is because the friction and inertia of the wheel allows some torque to be input, which is enough to cause a runaway locking effect. Essentially, the more locking effect, the more input torque is reacted, the more locking effect, and so on. Its like a positive feedback loop. I think 0.5 lockCoef is still too much for a Torsen, as it allows for infinite torque bias, whereas Torsens are advertised at 4 or 5:1 maximum. So yeah, somewhere probably in the 0.4 range is where you would see Torsen like behavior... I think. The slow one wheel problem is sort of a corner case that probably depends on the balances of friction in all the components. I've noticed this sort of treshold problem in my MX5 which has a Torsen as well. In snow or wet mud, if I ease into the throttle it spins one tire easily, but it locks for a few seconds if I drop the clutch to produce a sudden impulse. Similarly, on gravel, it will start spinning one tire while I am trying to be gentle on somebodies gravel driveway, then I hit some larger rock which produces a sudden reaction, and it locks up after that for a few seconds. In these cases the grip and input torque is so low that it is right on the edge of locking or unlocking behavior, and it behaves unpredictably. However on wet pavement, it seems to remain above the treshold.
I think I got this working some what well. ["differential", "differential_R", "driveshaft", 1, {"diffType":"lsd", "gearRatio":3.42, "lsdPreload":0, "lsdLockCoef":0.45, "lsdRevLockCoef":0.0, "friction":5, "uiName":"Rear Differential","defaultVirtualInertia":0.5}] I notice the effect of the wheel inertia on the spinning wheel. It's just enough force to kick the wheel on the ground into moving. If you can stop it mid air and get on the gas, you'll never move. I put the Torsen truck on the pads on the grid map and the standard LSD can use the engine torque at idle to roll over it whereas the Torsen will spin helplessly in the air. In extreme situations, the Torsen and the standard LSD act similarly and I assume this is because the preload of the S-LSD isn't enough to move the wheel with traction. All pictures are Torsen.
That sounds about right I just tried an experiment, with one tire locked down with node grabber and the other one ice (to make sure there is some reaction torque to get things going), and printing out a torque bias value. 1.0 lockCoef = huge torque bias, > 100 in the test, the clutch failed first, no wheelspin 0.5 = more or less the same, slightly softer behavior, no wheelspin 0.4 = ~5:1 bias before wheelspin occurs 0.2 = 2.5:1 bias before wheelspin occurs 0.1 = 1.4:1 bias before wheelspin occurs 0.05 = ~1.2:1 bias before wheelspin occurs
This is really cool data, thank you! Now what I have gathered is that worm gear type has usually 3:1 bias while helical type has more resistance and can have up to 5:1 bias, with this data it is possible to make different types quite easily. Perhaps because of higher internal resistance, helical type is said to be better in situations where other wheel is on really slippery ground, but also I have read with higher bias it can be bit rough, noisy and juddering. I did little graph, I did not input datapoint for 0.5 as if it was 100 it would just go straight up after 0.4 and 0.4 being pretty much realistic maximum this should be sufficient for all Torsen needs, just picking desired bias ratio from Y-axis and read needed lockingCoef from X-axis, for 3:1 ratio I would estimate 0.24 lockingCoef based on this graph: @CTJacob You seem to have revLockingCoef 0, should that be same as lockingCoef as I think Torsen locks to both ways?
I was under the impression that the locking of the differential happened under power, my mistake. The Detroit Truetrac is a Torsen type diff that has a 3.5 bias. This calculates to .28 lockCoef. It works pretty well. I got wedged like this pretty bad but, a simple tap of the brakes was all it took to launch me up the rocks. Edit- Does this mean the stock Limited-Slips in the game are have a higher lockCoef than they should?
yeah basically its already been mentioned. Torsens have no acceleration/braking lock but they can lock in case that the difference in speeds its too great. IIRC the point in which the diff lock depends on the angle of construction of the worm gears inside the torsen.
One thing to keep in mind is that Torsens can act a little silly. As mentioned earlier, if one of the drive wheels has 0 traction (say it's in the air or something) and the other has full traction... it will actually let you spin the 0 traction wheel just like a normal differential IF there is only a very small amount of torque applied to the differential. Torsen differentials work in differences torque and when it's a small difference, it will spin just like a normal differential. Put your food down a little and increase that torque and the Torsen will come to life and begin to lock the wheels together. A pretty good example can be seen here... Start watching at 2:00 minutes and you will see exactly what I mean. The Hummer H1 used dual Torsen differentials... granted AM General decided to use it with Torque Trac 4 to make the system more robust. In the Hummer it looks like it locks up almost instantaneous with just the smallest amount of throttle... which it does... but also remember that the Hummer uses gear reduced portaled axles and automatic break modulation. That means that those torsion differentials are spinning at almost twice the speed and with half the torque as they would normally... so just the smallest increase torque is enough to enable them to lock up a bit and power off of the obstacle. In low traction environments, a good trick to use with Torsens (to increase that difference in torque) is to apply light pressure to the brakes... this will in turn feed some torque back into the system and allow for it to lock up that much easier. This trick works even better if your vehicle is equipped with ABS as it almost acts like a rattle gun on the other tire freeing it from its situation with ease.
I think that all this good information should be put into wiki, or link from wiki's differential page to this thread so that people can find information how to make Torsen type diffs.
Got it backwards! Torsens only lock if there is acceleration or braking torque going through the worm gears. It is not strongly sensitive to speed difference. And if one tire is in the air, the system often cannot develop enough torque (because there is not enough resistance from the airborne wheel) to enable the locking, so it acts like an open diff. Press the brake, now the wheel is providing some resistance which means a torque can develop in the diff, and it locks up.
