Limited Slip Differential (LSD) theory

[DolHaus]

@OdeFinn : Here is an old thread that discusses the LSD with 60/60/60, it was in drifting forum and locked, but the first page has wealth of information and another drifter suggested to use clutch in order to test LSD setting accurately.

https://www.gtplanet.net/forum/thre...ence-inside-also-full-lsd-explanation.307085/

Read the OP and till the end of page 1. Before running tests, be sure to understand what is initial torque/preload in LSD ( particularly clutch plate LSD )

In the drifting forum this was always a hot (and dangerous) topic: Was 5/60/5 locked (welded) or was 60/60/60?

I discussed this with @FussyFez at great length and he was definitely correct in his observation that at 60/60/60 the differential was completely locked and had no change in state between full throttle and off throttle/braking whereas there was a definite change in state with the diff set at 5/60/5. With 5/60/5 the diff opens completely when you lift out of the throttle allowing much faster and tighter rotation.

The problem was that people were only thinking about one dimension of the question, they were only thinking about what was happening when on the throttle and to that end they were correct - under acceleration both 5/60/5 and 60/60/60 were locked. As we know the differential isn't just about acceleration, it has states dependent on what type of load you are putting on it

 

[CSLACR]

@CSLACR great, nice to hear that, anyhow I'm trying to focus my interest on information itself, not formation. But thanks for your opinion.

You're not getting it.

Maybe figuring out how big difference is on traction when banging lock from 5% to 20%, or from 60% to 65%, that's the clue over here, other rips tires out from ground and there you go. Sledgehammer again..

This is not "information", the words are wrong, incorrect, not right, not just in the wrong order, but wrong completely.
"other rips tires out from ground"? That's hysterically bad, do you know what that even means?

So I guess You'll get this thread locked, all you and your friends threads get locked, because you all have the same whiny, self indulgent attitude that everybdy is being a meany and you're perfect.
Enjoy your empty thread, and good luck finding people who speak your broken English and tolerate you blaming them for it.
(That's me leaving, since all of that probably doesn't translate to you)

 

[Ridox2JZGTE]

In the drifting forum this was always a hot (and dangerous) topic: Was 5/60/5 locked (welded) or was 60/60/60?

I discussed this with @FussyFez at great length and he was definitely correct in his observation that at 60/60/60 the differential was completely locked and had no change in state between full throttle and off throttle/braking whereas there was a definite change in state with the diff set at 5/60/5. With 5/60/5 the diff opens completely when you lift out of the throttle allowing much faster and tighter rotation.

The problem was that people were only thinking about one dimension of the question, they were only thinking about what was happening when on the throttle and to that end they were correct - under acceleration both 5/60/5 and 60/60/60 were locked. As we know the differential isn't just about acceleration, it has states dependent on what type of load you are putting on it

I think you meant 5/60/60 as welded like some of the drifters thought from GT5 It's a shame the thread got locked and the OP is nowhere to be seen now There were some good ideas being thrown there and even a mod also took part in the discussion.

 

[Ridox2JZGTE]

@DolHaus , @Ridox2JZGTE ,@oppositelock , & @XS , you have always showed interest in my follies, good or bad, so here is my latest venture into the GT6 world of physics.

I wanted to know how Initial value in the MLSD worked and found this idea from a member called bread82 from an NGU thread (I know bad bad me).

Procedure: I went to Silverstone Int, on the long straightaway and stopped just at the beginning of the bridge placing the 2 left tires on the grass and the 2 right tires on the track. Used AT, just floored it and after reaching 100 kmh got back on track and saved it as a best lap. I then used the data logger with the headings, Speed, RPM, Left Rear Wheel Speed, Right Rear Wheel Speed, and Gear. recorded the data from 0 to 99 using every 3 kmh as a marker. The calculation I used for wheel speed Difference was Left (wheel on grass) - Right (wheel on track). I did every test twice and averaged them out.

I’ve included all the data, and some graphs. The Initial value settings I used were 0%, 20%, 40%, 60%, 80%, & 100% of the MLSD value scale, this giving 5/5/x, 16/5/x, 27/5/x, 38/5/x, 49/5/x, and 60/5/x. I had planned on doing every one of these with the same change in the accel value, this project is on hold for now due to other avenues I want to explore. Have fun and I do hope it doesn't give you nightmares .

Go to the post and download the attached file to see the test result

 

[DolHaus]

I think you meant 5/60/60 as welded like some of the drifters thought from GT5 It's a shame the thread got locked and the OP is nowhere to be seen now There were some good ideas being thrown there and even a mod also took part in the discussion.

Between Gt5 and 6 the belief changed from 5/60/60 to 5/60/5 somehow and not many people noticed
Yeah I miss Fussy, he was a good dude, the thread had to be closed for exactly the same reason every camber thread gets closed, people refuse to believe something else and it turns nasty

 

[CSLACR]

Between Gt5 and 6 the belief changed from 5/60/60 to 5/60/5 somehow and not many people noticed
Yeah I miss Fussy, he was a good dude, the thread had to be closed for exactly the same reason every camber thread gets closed, people refuse to believe something else and it turns nasty

This implies people can't discuss and disagree without getting nasty. I've discussed tuning with a significant amount of GT players, and the only group where it continually gets nasty roams around the GTPlanet tuning forum.
In the WRS, it doesn't get nasty. In racing series, it never got nasty, and on another site, it never got nasty.

But every single time on GTP.

The difference is on some of these, people just say "ok I disagree", but on GTP, they post a picture with captions trying their damnedest to make an ass out of anyone that doesn't agree with them.
Must be in the water.

 

[DolHaus]

This implies people can't discuss and disagree without getting nasty. I've discussed tuning with a significant amount of GT players, and the only group where it continually gets nasty roams around the GTPlanet tuning forum.
In the WRS, it doesn't get nasty. In racing series, it never got nasty, and on another site, it never got nasty.

But every single time on GTP.

The difference is on some of these, people just say "ok I disagree", but on GTP, they post a picture with captions trying their damnedest to make an ass out of anyone that doesn't agree with them.
Must be in the water.

I know what you mean, its a shame people can't have a discussion without turning it into another flame war. Not claiming to not have been involved with a lot of the arguments and slinging my fair share of 🤬 in the past but I'm done with it now, no point getting involved any more because no one ever wins. We've already lost too many talented tuners to the petty squabbling, I just want to build fast cars and share my insight with those who are willing to listen

 

[CSLACR]

I know what you mean, its a shame people can't have a discussion without turning it into another flame war. Not claiming to not have been involved with a lot of the arguments and slinging my fair share of 🤬 in the past but I'm done with it now, no point getting involved any more because no one ever wins. We've already lost too many talented tuners to the petty squabbling, I just want to build fast cars and share my insight with those who are willing to listen

I can say pretty much the same story, if I could like this more times I would. 👍

 

[xande1959]

friends I'm a little confused many answers, but none convinced me. LSD on how to put on a physical GT6 car one LSD so complicated. where we can only change the numbers of the binary, acceleration, and braking sensitivity. the numbers never correspond with reality. example: (starting torque 5 / minimum and 60 maximum). someone can tell me the exact value of LSD a street car 500 HP.

 

[Grippy]

@DaBomm4 my understanding is if you have a locked/welded differential IRL both tyres spin at the same time, usually overloading the outer tyre cuasing that to slide first as it is already having more weight while cornering.

whereas having an open LSD the inside wheel spins first due to having not much grip whilst cornering.

Also if you mount two wheels on an axis and turn them around a corner the outer wheel has to rotate more to get to the same distance as the inside wheel. Meaning the outer wheel will always travel more distance than inside.

 

[OdeFinn]

@Grippy there were test done with LSD values 100/0/0 and it locks 100% all time, 60/0/0 is 60% all time. 0/100/0 is 100% only during acceleration, open differential on coast and brake.
A) 60/40/40 is 100% during accel&brake, 60% on coasting.
B) 40/60/60 is 100% during accel&brake, 40% on coasting.
C) 60/60/60 is 100% during accel&brake, 60% on coasting.

Difference between A and B probably is locking effect speed.
100% is max and A has locking curve from 60% to 100%, and it is ready to go there in one time unit, but example C is having locking curve from 60% to 120% and it's also ready to go there in same time unit, this means at it will reach 100% before example A.

 

[DolHaus]

@Grippy there were test done with LSD values 100/0/0 and it locks 100% all time, 60/0/0 is 60% all time. 0/100/0 is 100% only during acceleration, open differential on coast and brake.
A) 60/40/40 is 100% during accel&brake, 60% on coasting.
B) 40/60/60 is 100% during accel&brake, 40% on coasting.
C) 60/60/60 is 100% during accel&brake, 60% on coasting.

Difference between A and B probably is locking effect speed.
100% is max and A has locking curve from 60% to 100%, and it is ready to go there in one time unit, but example C is having locking curve from 60% to 120% and it's also ready to go there in same time unit, this means at it will reach 100% before example A.

The other difference between A and B is that A will force open the diff initially as you get on the throttle, thats why I never recommend initial being higher than either accel or decel as it tends to destabilise the driven wheels

 

[OdeFinn]

The other difference between A and B is that A will force open the diff initially as you get on the throttle, thats why I never recommend initial being higher than either accel or decel as it tends to destabilise the driven wheels

What you mean? Initial doesn't override anything, it's just base locking value and other values are added to that.

When playing with close to open differentials as many here are then it might be harder to feel how lock actually works, most locks done here are just really mild locks sitting on initial locking rate and parachute braking with higher brake side during corner entry.

 

[eran0004]

@OdeFinn I've been wondering what those LSD values represent. Thanks for providing an answer

 

[DolHaus]

What you mean? Initial doesn't override anything, it's just base locking value and other values are added to that.

When playing with close to open differentials as many here are then it might be harder to feel how lock actually works, most locks done here are just really mild locks sitting on initial locking rate and parachute braking with higher brake side during corner entry.

The number represents the most open the diff can be in relation to torque/power.
If we imagine we are approaching corner apex and coasting at low rpm the diff is as open as the Initial will allow (lets say 20). We then begin to accelerate and the diff switches to Accel, if the accel is set lower than the Initial (lets say 10) then the diff opens first before tightening as the torque/power increases with rpm.

It doesn't make sense in terms of a mechanical differential but thats how it appears to work in game. The Initial is described as a pre-load setting in game, if that was the case then it would be impossible for the diff to open further than the Initial is set as it permanently has X amount of load on it.

 

[OdeFinn]

The number represents the most open the diff can be in relation to torque/power.
If we imagine we are approaching corner apex and coasting at low rpm the diff is as open as the Initial will allow (lets say 20). We then begin to accelerate and the diff switches to Accel, if the accel is set lower than the Initial (lets say 10) then the diff opens first before tightening as the torque/power increases with rpm.

It doesn't make sense in terms of a mechanical differential but thats how it appears to work in game. The Initial is described as a pre-load setting in game, if that was the case then it would be impossible for the diff to open further than the Initial is set as it permanently has X amount of load on it.

But not, game never opens lock under initial value, neither in real-life.
Reaction speed may give you that impression.
Testing between different variations of 30% lock as on your example you'll find differences on locking speed and what's good for your current car setup (suspension, gearbox, compound etc).
Example 30% locks: 5/25/# : 10/20/# : 15/15/# : etc. Car setup is key what works best there, and how locking effect reacts on currently available grip on powered wheels.

Forget, if I remember right you're also prone to do your LSD configurations as "parachute"way, are you sure at your test isn't coming from overrun side to accel?

 

[DolHaus]

But not, game never opens lock under initial value, neither in real-life.

It shouldn't, but it does. Take a rear wheel drive car with plenty of grip, find a long corner and then keep raising the Initial until understeer stops you following line, set the Accel lower than the initial. When you reach the point during coasting where the understeer is forcing you off line start accelerating, you might be surprised

 

[OdeFinn]

It shouldn't, but it does. Take a rear wheel drive car with plenty of grip, find a long corner and then keep raising the Initial until understeer stops you following line, set the Accel lower than the initial. When you reach the point during coasting where the understeer is forcing you off line start accelerating, you might be surprised

If I test this like:
X/5/5 and changing initial until no understeer, then doing what?
You probably have some "usual" starting point for your accel side?
During cornering switching from accel to cost and to brake side is really easy thing to happen, specially if running Triple-Plate clutch and Carbon Propeller Shaft, if on standard clutch and standard propeller lock won't jump so quickly from side to side, works this way on real life too.

 

[DolHaus]

If I test this like:
X/5/5 and changing initial until no understeer, then doing what?
You probably have some "usual" starting point for your accel side?
During cornering switching from accel to cost and to brake side is really easy thing to happen, specially if running Triple-Plate clutch and Carbon Propeller Shaft, if on standard clutch and standard propeller lock won't jump so quickly from side to side, works this way on real life too.

Lets take the long right hander at Big Willow (turn 2), start at 10/10/10 (as long as this doesn't cause the inside wheel spin up under acceleration) and then increase the Initial until you can no longer hold your line during the neutral phase (no throttle/brakes). Ideally you want to be at low RPM well below the torque/power peaks to show the maximum effect but what you should notice is that when you blip the throttle the rear frees up and rotation increases. I can't test if this corner is ideal for the test as my PS3 has died but as long as you're not at the top of a gear then you should be ok.

 

[OdeFinn]

Lets take the long right hander at Big Willow (turn 2), start at 10/10/10 (as long as this doesn't cause the inside wheel spin up under acceleration) and then increase the Initial until you can no longer hold your line during the neutral phase (no throttle/brakes). Ideally you want to be at low RPM well below the torque/power peaks to show the maximum effect but what you should notice is that when you blip the throttle the rear frees up and rotation increases. I can't test if this corner is ideal for the test as my PS3 has died but as long as you're not at the top of a gear then you should be ok.

Will get to this later, but do you realise at you're hanging on brake side lock during that "neutral" phase?
Brake side=initial+ deceleration

 

[eran0004]

Lets take the long right hander at Big Willow (turn 2), start at 10/10/10 (as long as this doesn't cause the inside wheel spin up under acceleration) and then increase the Initial until you can no longer hold your line during the neutral phase (no throttle/brakes). Ideally you want to be at low RPM well below the torque/power peaks to show the maximum effect but what you should notice is that when you blip the throttle the rear frees up and rotation increases. I can't test if this corner is ideal for the test as my PS3 has died but as long as you're not at the top of a gear then you should be ok.

But when you blip the throttle the LSD tightens, which causes the car to lose grip in the rear and makes it rotate.

Test this with an FF car:

1. Set the LSD to 5/5/5
2. Make "donuts" by turning hard to one direction at full throttle and holding the e-brake.

What you should see is that the inner front tyre goes red (as well as the rear tyres, because of the e-brake) as most torque is directed to that wheel. The outer front tyre is white.

3. Now change the Initial to 60, but leave the others at 5.
4. Do the donut test again.

What you should see now is that both front tyres go red, as the torque is split between both front wheels.

If the acceleration slip is only determined by the accel value, then the results of both the tests should be the same, as the accel value was 5 in both cases. But at Initial = 5 most torque was sent to the inner front wheel, and at Initial = 60 the torque was split between both wheels.

I think the reason why they write that a low Initial value might result in unexpected handling changes when the LSD kicks in is because it is assumed that if you set the Inital low, then you'd set the Accel high, which means that you go from a very loose diff to a very tight diff when it kicks in. But if you set the Init high and the Accel low, you go from a tight diff to a slightly tighter diff, and the handling change isn't as dramatic.

 

[DolHaus]

Will get to this later, but do you realise at you're hanging on brake side lock during that "neural" phase?
Brake side=initial+ deceleration

The three settings operate like 3 separate entities depending on what you're doing with the controls

Initial = No throttle/brake
Accel = Throttle
Decel = Braking

They are connected vaguely in the sense that there is a transition between modes but they have no direct effect on each other despite how it feels.


It is an extremely simplified system which I would guess this is done for ease of programming across 1200 cars and a million variables in between

 

[Ridox2JZGTE]

But when you blip the throttle the LSD tightens, which causes the car to lose grip in the rear and makes it rotate.

Test this with an FF car:

1. Set the LSD to 5/5/5
2. Make "donuts" by turning hard to one direction at full throttle and holding the e-brake.

What you should see is that the inner front tyre goes red (as well as the rear tyres, because of the e-brake) as most torque is directed to that wheel. The outer front tyre is white.

3. Now change the Initial to 60, but leave the others at 5.
4. Do the donut test again.

What you should see now is that both front tyres go red, as the torque is split between both front wheels.

If the acceleration slip is only determined by the accel value, then the results of both the tests should be the same, as the accel value was 5 in both cases. But at Initial = 5 most torque was sent to the inner front wheel, and at Initial = 60 the torque was split between both wheels.

I think the reason why they write that a low Initial value might result in unexpected handling changes when the LSD kicks in is because it is assumed that if you set the Inital low, then you'd set the Accel high, which means that you go from a very loose diff to a very tight diff when it kicks in. But if you set the Init high and the Accel low, you go from a tight diff to a slightly tighter diff, and the handling change isn't as dramatic.

The ATTS equipped Honda Prelude Type S '96 with 35/-/- stock LSD or the Prelude Sir '96 with 50/30/0 stock LSD, that should show it ( inner wheel spin ) I would also call initial/preload as breakaway torque, the point where the torque overcome the resistance and the diff engage ( starting locking action ), when the locking action is weak, the inside wheel may spin. This can be seen on posi traction / eaton LSD with worn clutch pack and high preload spring.

 

[eran0004]

The three settings operate like 3 separate entities depending on what you're doing with the controls

Initial = No throttle/brake
Accel = Throttle
Decel = Braking

They are connected vaguely in the sense that there is a transition between modes but they have no direct effect on each other despite how it feels.

But the initial value does have an impact on acceleration slip. Also, in the description it says that it's the torque exerted when the LSD is not engaged, i.e. when travelling in a straight line. It's not for no throttle/brake.

It is an extremely simplified system which I would guess this is done for ease of programming across 1200 cars and a million variables in between

That's not an issue with object-oriented programming. Rather than coding each car from the ground up, they're most likely just instances of one and the same object class. A change that you make to the class affects every instance of that class, so rather than working with a million variables you'd actually just be working with a fraction of that amount. An LSD script wouldn't have to be able to deal with a million different variables, instead it would collect the same couple of variables from all the cars and the only difference would be that the variables have different values.

Consider this simplified script that calculates a C based on inputs A and B:

C = A + B

This script doesn't care about what values A and B have, or what object is passing those values to the script. As long as there is an A and a B, the script will happily calculate C.

If you think about it the same thing applies to all the components of a car: engine, gearbox, suspension, aerodynamics, etc.

 

[DolHaus]

But the initial value does have an impact on acceleration slip. Also, in the description it says that it's the torque exerted when the LSD is not engaged, i.e. when travelling in a straight line. It's not for no throttle/brake.



That's not an issue with object-oriented programming. Rather than coding each car from the ground up, they're most likely just instances of one and the same object class. A change that you make to the class affects every instance of that class, so rather than working with a million variables you'd actually just be working with a fraction of that amount. An LSD script wouldn't have to be able to deal with a million different variables, instead it would collect the same couple of variables from all the cars and the only difference would be that the variables have different values.

Consider this simplified script that calculates a C based on inputs A and B:

C = A + B

This script doesn't care about what values A and B have, or what object is passing those values to the script. As long as there is an A and a B, the script will happily calculate C.

If you think about it the same thing applies to all the components of a car: engine, gearbox, suspension, aerodynamics, etc.

How can you prove the initial value has an impact on acceleration slip?

 

[OdeFinn]

Whistling break here, before continuing this conversation please watch few videos, start with Engineering Explained video "limited slip differential - explained", then their torque transfer video. Then continue thru clutch pack lsd's.
LSD collection: http://www.youtube.com/playlist?list=PL5YaJzsNBJoU-Pni3Sw1KUkBvoSs0365Y

 

[DolHaus]

Its not a real differential, it doesn't have to follow the way a real differential works, it only has to emulate a similar effect

 

[OdeFinn]

Its not a real differential, it doesn't have to follow the way a real differential works, it only has to emulate a similar effect

Do you really believe at PD/Kaz would put out something extraordinary on one easiest calculated part of physics engine?

 

[DolHaus]

Do you really believe at PD/Kaz would put out something extraordinary on one easiest calculated part of physics engine?

Ride height...
Rake...
Camber...
Toe...
Roll physics...
Drag....
Aerodynamics...
Gearing....
Dampers....
ABS....
Etc.
Etc.

Doesn't take much to add LSDs to the list of things that don't really work like they should.

 

[eran0004]

How can you prove the initial value has an impact on acceleration slip?

Here:

Test this with an FF car:

1. Set the LSD to 5/5/5
2. Make "donuts" by turning hard to one direction at full throttle and holding the e-brake.

What you should see is that the inner front tyre goes red (as well as the rear tyres, because of the e-brake) as most torque is directed to that wheel. The outer front tyre is white.

3. Now change the Initial to 60, but leave the others at 5.
4. Do the donut test again.

What you should see now is that both front tyres go red, as the torque is split between both front wheels.

If the acceleration slip is only determined by the accel value, then the results of both the tests should be the same, as the accel value was 5 in both cases. But at Initial = 5 most torque was sent to the inner front wheel, and at Initial = 60 the torque was split between both wheels.