[torsen] SAE885140 vs the Spiderman

Andres Kroonmaa andre at online.ee
Tue Mar 6 15:19:44 EST 2001

On 6 Mar 2001, at 8:47, QSHIPQ at aol.com <QSHIPQ at aol.com> wrote:

> >   "It is presupposed that the torque split limits of the differential are 
> >   NOT exceeded." That is a presumption of this paragraph.
> Ok, remember tho, you have chassis O and U *within* the torque split limits 
> of the differential.

 OK, tho you can't have U in turn unless rear looses adhesion.

> > I think that this paragraph is not about spider bite at all. You should 
> > look more into how torsen behaves in extreme cases when wheels loose all 
> > adhesion.
> Remember, when a torsen loses *all* adhesion Trg goes to 0.  3BR x ~0 = 0.  
> Andres, the key is that we are speaking exactly BEFORE wheels loose all 
> traction.  

 Scott, how can we speak about "exactly BEFORE wheels loose all traction"
 while discussing spider bite? Perhaps I misunderstand you. If wheels have
 enough adhesion to keep torsen within BR limits, then you don't have a bite,
 but are cruising with 100% of engine power delivered to the ground.
 This is what this quoted paragraph is talking about.

 We are talking about exceeding torsen limits, including Trg=0, aren't we?

> >   Please keep in mind, that you can build locked diff from torsen when you
> >   change BR from some fixed value to infinity. Its not that different from
> >   locked diff in that respect. Same oscillations happen with locked diff as
> >   well.
> NO.  A locked diff doesn't allow any speed differentiation.  If that is true 
> (thinking of 2 sets of wheels, one with more slip angle than the other), how 
> do we get O in a locked center diff?  For an exercise, unlock a center diff, 
> take the tightest turn you can.  Now lock it.  What happens? 

 I guess you mean vehicle without acceleration. In that case torsen behaves
 like open diff and locked diff would screech. During acceleration I believe
 front wheel would need to slip and would, causing torque to go rear with
 locked diff. no? OK, not 100%, but still more torque than would go to front.

 What I assumed by oscillations you mentioned, is a required slip of rear
 or front wheels due to overload. Without any slip there can't be any
 oscillations. Torsen will maintain steady state all the time, just as open
 diff is doing. With torsen you have O all the way through the turn,
 _unless_ rear slips.

 You get an O with locked center during front inside wheel lift or slip
 in a turn. Nothing terrible I guess.

> >   Key point in understanding why torsen sucks here is the fact that its 
> > torque split is tied not to axle with better traction, but to that one with 
> > _lower_ traction. It works like open diff here, only with an added bonus of
> > giving axle with better traction BR times more torque than to the spinning
> > axle.
> A torsen torque split is only tied to the axle that rotates slower (or is 
> trying to = torque sensing).  That isn't necessarily a traction argument, yet.

 Scott, look at it differently (and don't confuse with speed sensing)
 1) Torque goes to where least resistance is.
 2) Torsen's internal friction resists to speed differences, and axle with
    less traction force pulls along an axle with more traction. All
    this with a given split limit. Internal friction is a result of gears
    being "jammed" between input shaft and output resistance.
 3) axle rotates slower only if it has more resistance than the other.
    (otherwise applied torque would spin it up to speed).
 So, axle cannot rotate slower if it hasn't enough traction. no way.

 And I mean it when I say that torque split is tied to axle with _less_
 traction (or resistance or higher speed, if you like).
 Point is that torque tries to go out a least resisting axle first, and
 only then internal friction resists that by holding onto slower axle. And
 this internal friction is smaller if speeding axle has little resistance.
 (because the "gear jamming" effect is reduced)

 This is the whole reason why torsen looses all torque (Trg=0) when one
 of its axles looses all traction (resistance=0).

> >   Driving with Torsen presupposes that it always can get at least some grip.
> >   As long as it can, it works better than open diff or locked diff. But when
> >   it looses grip as during you racing, it is not "much" better than open 
> diff,
> Not with you here.  The problem with a torsen is that it's always locked on 
> torque, so it has some bad properties asscociated with that, the biggest of 
> which is that allocates torque based on slip angle differences.  Add to that, 
> it *can* allow speed differentiation while maintaining torque split applied.  
> However, once that differentiation becomes significant (spin up) it will 
> shift torque again, this time based on a traction variable.  Looking at the 

 No. there is no such dual shifting in torsen. it's oneway device. I agree
 that it applies more torque to rear in turns, but not because of speed diff
 per se, but only because it is "easier now" to propel front.
 You say that torsen is always locked on torque. If so, I can't see how it
 behaves differently to locked diff during that time. And if its not locked,
 I can't see why we can't compare it to "tight" but still open diff.

 And again, we can't speak of O-U-O changes if we don't allow rear spin.

> paper, we are looking at very small slip angle differences causing torque 
> shift.  What does that mean when we speak of spin up, isn't that ideed a very 
> small amount.   I've seen torsen car wheels spinning, but rarely is that a 
> bite phenomenon, IME.

 I agree that when rear spins, we have O-U with torsen. I also agree that
 torsen may be the culprit in spider bite. But I can't agree that torsen
 can cause this _without_ rear spin. _With_ rear spin, we have to compare
 torsen to locked diff, and we see that there are similarities. Now I
 propose that with locked diff regaining of rear grip is less noticable
 than with torsen. Thats ok with me. From here I try to deduce how the
 bite happens.

> >   and looses to locked diff badly. Thats my thought on why they don't race
> >   with torsen centers.
> More specifics please.  Why does it lose to a locked diff badly exactly?

 When rear axle spins, torsen can provide to front only as much torque, as
 it still can get grip from rears, multiplied by TBR. No more. Zero grip
 from rear = zero torque to the front, no matter how good traction fronts
 may have. Between any extremes, there is infinite "between".
 Locked diff can provide 100% of torque to the fronts in the same case.

 Now as per your other post on 205T vs torsen we realise that rear spin
 is actually wanted for racing. Torsen with spinning axle is loosing
 engine power compared to locked diff. Torsen without spinning axle
 probably has better overall traction, but on average, locked diff has
 a little edge in racing. IMHO.

> >   Torsen is sacrificing total traction effort "in hope" that it can reach
> >   back to better grip conditions. This is good for fat old man getting his
> >   beers home intact, but not so good for speed racing. ;)
> Torsen is an absolute traction device that needs to turn.  A VC is an 
> absolute traction device that can.  What and where are the differences Andres?

 Come on, what are we arguing here? VC is purely speedsensing device with
 some problems torsen doesn't have. They are similar, yet quite different.

> >   You push the torsen car too far, its torque split far exceeds BR limits,
> >   and rear axle starts spinning. Torsen tries to "grip to" front axle, but
> >   still looses engine power towards rear as it can only be as good as rear
> >   grip times BR, so it has not enough torque to get you out with fronts.
> Andres, I dispute this scenario, because I presuppose that the axles *aren't* 
> spinning, in fact, most of my "bites" happen without spinning tires.  Again, 

 Hmmm. Now I'm puzzled. If you don't spin/slip/slide the car, wheres the bite?
 So you're running with 100% engine power to the ground, and you have a bite?
 Would you care to describe one of the bites once more, in detail, please?

> you need to think in terms of a torque sensing diff, not a speed sensing 
 I really do.
> diff, and in terms of O and U in a given chassis.  If *within* the BR I can 
> get a car to O and U, I only need the conditions that cause O and U.  Turning 
> without wheel spin allocates torque rear, before you get spin (torque 
> sensing) torque is allocated forward.  

 You can't have torsen providing torque forward in a turn without rears
 loosing grip. Whats the difference between fast spin and slow spin?
 Rear wheel on the ground can only loose traction force if it spins. right?
 So with torsen, for a car to have U in turn, its rear has to spin. ok?
 At least virtually. If torsen stops the spin as it starts, then so be it,
 but this doesn't change the fact that spin was about to evolve.
 If so then we are back to my proposal of bite, built on spin precondition.

 I feel like you have a confusion about how torsen "allocates" torque.
 It doesn't do anything fancy. It just grips onto an axle with better
 traction. So does locked diff. no difference.

> >   You fly over the track and at some point rear wheels find better grip.
> >   Together with increasing rear grip increases traction effort to rear, AND
> >   traction effort to front ALSO (rear traction effort times BR). Front
> >   traction gets you forward. But because you've lost rear traction for some
> >   time, and rear grip return increases also front traction, BR time more so
> >   than rear traction, you perceive it as sudden and jumpy torque shift to
> >   the front (note the steeper curve from no-traction to max traction in
> >   Chocholek's paper fig.2, compared to open diff). This confuses.
> A reduction in Trg when you 'lose' traction is what keeps the spider bite 
> from happening when you lift a wheel (or spin one, same thing).  A lifted 
> wheel isn't in my scenario at all.  As trg approaches 0 it really doesn't 
> matter what the torque spit is, as Trg approaches 0, the car will U, since 
> baseline chassis dynamics will prevail

 Yes, but I'm talking more about the 2 points on the graph (fig.2) between
 which things happen. On that steep line, between torsen giving 100% engine
 power to the ground, and Trg=0, it behaves weird. If track surface is with
 uneven grip, you'd feel that alot more with torsen than with any other diff.
 More because torsen takes rear torque, "multiplies" it by TBR and provides
 it on front axle. Small variations in rear grip can cause large variations
 of front torque. I think this is what you are talking about, O-U-O-U
 oscillations. I merely try to show where and when they occur and are more

> >   So, in part you are right, imho, this happens. But this happens with other
> >   types of diffs as well, only with some other amounts. For eg. open diff
> >   would give you similar torque jump to forward, but its amount can't be
> >   more than as much as rear torque.
> A open diff can't shift any torque.  All torque in an open diff is split 
> 50/50.  You lift a wheel on a open diff quattro, put the car in gear, it will 
> go nowhere.  

 right, exactly what I'm saying: front "torque can't be more than as much
 as rear torque". You should note the moment rears regain some grip. What
 I describe happens, just with open diff it isn't very noticable. With
 locked diff also less, because it can fluctuate between 100/0 and 50/50
 split. What I want to say is that during a slip, torsen can go like this,
 in terms of engine power to the ground: 4/1, 16/4, 60/15, 80/20, 70/30,
 50/50, 30/70 (if in turn)
 (given BR of 4 times). Notice how front torque goes multiplied upto 80,
 and then back to 50 as rear axle regains grip and axle torques equalise.
 This is O-U-O which is different from other diffs, because in the range
 of grip loss it changes alot faster.

> >Locked diff would ignore rear slip and
> >   would give all torque out to the front, and as soon as rear grips it would
> >   give out all torque to rear. Torsen would give out BR times more torque
> >   to the front, suddenly, but not ignoring rear spin. So, open diff won't
> >   change behaviour of a car, locked diff widely jumps between fwd/rwd, and
> >   torsen jumps from weak fwd to strong fwd. In a straight line this torsen's
> >   peculiarity is totally ok, but in turn you need to learn how to handle
> >   this safely. And you don't like this "style". ;)
> No.  Andres, again you must think in terms of chassis dynamics.  A 50/50 
> split in a quattro is U, in a locked center or an open center.  A torsen 
> quattro can O, the conditions need to be right in the matrix, but it can O.  

 Alright, I follow. I admit I made an error with locked diff, it probably
 can't go 100% rear so easily.

> But the torsen is like a rubber band, it always wants to go back to it's 
> static state of 50/50.  So the O will prevail, *until* another condition 
> (steering, cf, trg, traction) causes it to shift again.  The problem with the 
> diff is that it can't accomodate a turn without allocating torque.  It's too 
> dumb to know it's turning.  The exact problem with torque sensing, it senses 
> all inputs to the device as traction arguments.  Turning isn't one.... yet

 I fully agree that this is a weakness of torsen. It should provide 50/50
 during turns to have better results. But I can't see the bite moment here ;)
 I can imagine it elsewhere, but not here ;) Locked diff also can't imagine
 its turning.
 When going from straight to a turn on power, locked diff split is what?
 Fronts have to slip, right? how does their slip change torque split?
 How predictable it is, based on unknown track surface?
 Riding an open front locked/c and having a wheel lift in turn.
 Is this a bite? I guess not, but here we also have clear O.

> YMMV, mine did.  Andres, reread that post carefully.  And remember that I'm 
> speaking of spiders biting before slip.  With slip (as I've already pointed 
> out many times to Dave Eaton) Trg isn't significant and really isn't an O 
> moment at all, it's a U condition.

 I can read you. Somehow I feel that you read my wording "slip" as a full
 loss of any traction (Trg=0). I'm sorry if I cause a confusion here, but
 I really mean slow slip, while not all traction is lost, but it is reduced
 alot. In this case Trg is an important factor, as is traction force on
 slipping axle.

> >   How to handle the bite situation is another matter, and not within bounds
> >   of my experience, but I tend to believe that this could be accounted for
> >   and perhaps even used for good.
> When you figure it out, let the boys at audisport know.  They never did.  I 
> really think you should consider that if torsen is *so* good, why is audi 
> standing alone with it's commitment to it in a center axle.  Many cars use it 
> at the front or rear of the car as standard equipment and in racing.  No one 
> uses it in the center diff but audi, and even audi *rarely* used it in their 
> own racing efforts.  Stig, hated it, WR was rumored to dislike it strongly as 

 Hey, where did I say torsen rulez in racing? I think on contrary. But also,
 for street driving, it is "sticking" better to the ground than open diff
 or locked diff would, so it is a good "dumb" device in the center, mostly.

> well.  I read all these *good* things you post here, Andres, and I truly feel 
> someone has missed something really basic here.  I'd love it to be me, but I 
> can say that I know exactly what Stig spoke to me about wrt to the chassis 

 Its no problem with me that torsen has this issue. Just I don't buy its
 current explanation.

> device itself and explain why.  What are we missing?  You need to be more 
> specific.  If I have O at some point within the BR of the device, I don't 
> need to "lose" traction to have O and U in a turn.  Hint:  Forget wheel lift.

 You have to have *less* traction in rear than in front. During a tight
 turn when torsen already "sees" less traction in front, this means quite
 a bit of lost rear traction to get U. I'm at loss how this can happen
 without a loss of traction in rear...

 I don't know. imho we don't miss anything in relation to how torsen works,
 maybe we need to know more on how specific Audi's torsen is constructed?

> >   I agree with you. And I agree with Dave E also. is that possible? ;P
> Anything is possible.  885140 is pretty clear that O and U can happen before 
> wheel spin.  If I can set up a turn like what was setup in the section of 
> 885140 I quoted, and I can get rear torque bias and O. What happens when I 
> reach the limit of adhesion at the rear, and "the drive torque is divered to 
> the other axle, independent of the forced slip resulting from the vehicles 
> circular path?"  I claim U.  Then, if you've done nothing  else, the rears 
> have regained traction, you have a forced slip from turning, which results in 
> O, which overloads the rears and "the drive torque..."  Ad infinitum.  

 I perfectly agree with this, except with "Ad infinitum" part.
 If that is the case, then you have varying adhesion on track surface.
 Then we should talk about how this is noticed differently vs other diffs.

 Take VC, rwd vehicle. Same turn. O most of the time, after a rear spin
 you have U, rears regain traction, you get O again. Ad infinitum? Not
 that different, only that with torsen fast spin is not needed in rear.
 But we are searching for something inherently specific to torsen here.
 So that isn't quite what I hoped.

 And if you insist that torsen is unable to achieve a stable state, then
 we need to think more on why this can at all happen.

> Don't at all think of "reaching the limit of adhesion" as wheel lift Andres.  
> The torque shift has already happened by then.  

 ok, its just simpler to say "spin" than "reaching the limit of adhesion".
 Torque can't go to the front unless rear is providing less "support" to
 torsen's rear output. This is "spin", be it even as slow as front axle
 rotation speed. Isn't it?

 Andres Kroonmaa <andre at online.ee>
 CTO, Delfi Online
 Tel: 6501 731, Fax: 6501 708
 Pärnu mnt. 158, Tallinn,
 11317 Estonia

More information about the Torsen mailing list