[torsen] RE: Haldex differential

Dave.Eaton Dave.Eaton at clear.net.nz
Fri Nov 22 03:51:14 EST 2002

no, with a locked differential the tbr is infinite (the result of dividing
by 0) as there is no way of controlling the torque shift (the traction
available does that for you).  with the torsen, the tbr is a *limit*.
torque will be shifted around up to that amount - whereupon speed
differentiation occurs.  in other words, with a torsen the tbr is the limit
up to which the diff will remain locked.  with the haldex there is no tbr in
the normal case, as there is no clutch engagement.

btw in these discussions, torque is by definition that which is *supported*
by the tyres on the ground (traction in other words).

there is also no 5% rear as you suggest with the tt haldex lsc - as there is
no method of initiating the shift (pump) until slip occurs.  technically,
once slip has occurred, you could initiate rearwards torque shift and hold
it there i suppose, but i'd be extremely surprised to find that audi do that
just for the tt.  there is certainly no mention of this as a feature of the
haldex lsc - i think you are getting confused with a classic vc operating in
a "hang-on" mode (as opposed to one integrated with a true differential).
in this sort of application (vw synchro in other words), it is common to
have a nominal torque bias due purely to the type of silicone used and it's
shear properties - it is advantageous to reduce the engagement time.

fwiw, the porsche 959 system also used a hang-on clutch driven by an axial
pump powered by the front/rear speed difference - the difference in that
application was that the difference in tyre size front/rear created the
pumping pressure to allow clayton's active control.

lets look at it another way:

open diff:   normal tbr 1:1,  limit tbr 1:1
locked diff: normal tbr 1:1*, limit tbr: 1:infinity
torsen diff: normal tbr 1:1*, limit tbr: 1:3 (depending upon applciation)
haldex lsc:  normal tbr 0;    limit tbr: 1:infinity

*actually not really 50:50% - the static case is affected by vehicle chassis

once again, with the clutch fully locked, the haldex lsc has no ability to
control torque shift, and remember that, once engaged, it cannot effectively
increase lockup until there is further slip.  what it can do though is to
monitor wheel speeds and reduce clutch engagement if the rear is sliding.

the modification to an active clutch would be a good 1st step - but this
requires a completely different operating principle (why continue to use
axial pumps?), new software and a powerful external pump.

however, regardless of the nature of the lsc, i think the biggest
improvement would be in the incorporation of a differential, then you would
have a normal static torque split, and use the clutch to supplement control.
the haldex lsc is too reactionary to my mind.

this is where zexel have gone with the t3 - although this is aimed at a
centre application, it is by definition (it uses an epicyclic gear set)
aimed at providing more rearwards torque static bias than forwards.

a true active differential (a la wrc) is simply a clutch with a
differential.  i hope that this is what haldex are developing now.

'95 rs2
'90 ur-q
 -----Original Message-----
From: QSHIPQ at aol.com
Date: Thu, 21 Nov 2002 08:35:56 EST
Subject: Re: [torsen] RE: Haldex differential

In a message dated 11/20/02 9:46:02 PM Central Standard Time,
deaton at tranzrail.co.nz writes:

>2) shift and capacity are one and the same.  if the diff is supporting a
>particular bias ratio, then that is where the torque is going.  in a open
>diff, the bias ratio is 1:1 so that input torque is differentiated evenly,
>with the input torque determined by the traction of the least tractive
>in a locked diff, the bias ratio is 1:infinity, so that there is no limit
>the ability of the most tractive axle to support torque (100% in other
>words) - to the limit of adhesion.  with the haldex you have the ability,
>courtesy of the throttle valve stepper motor to have a variable bias ratio
>determined by the ecu, based on input from the abs sensors, the throttle
>the "g" sensor etc - provided there is slip in the 1st place.  what you
>*don't* have is a static torque split (bias ratio).

Dave, please math that statement out in terms of T1/T2.  TBR is speaking of
shift IMO, cuz you can't "solve" for a locked differential.  Remember, the
TBR has specific formulas put forth in Chocholek's paper.  You can't
multiply/divide by zero.  In a torsen for example, in it's 78/22 state, it
also "locked" by definition, regardless of Trg.  That doesn't at all mean
TBR is infinity:1

>this is the reason that your "45% torque shift rearwards" scenario makes no
>sense in a haldex.  at various times the ecu will be doing exactly that of
>course, in reaction to road conditions, but at other times, it will be 80%
>rearwards, other times (the "normal case") 0%.  the beauty of the haldex is
>that, provided there is pressure in the system enough to create a desired
>bais ratio, you can infinitely vary the bias ratio *below* this using the
>throttle valve, *regardless* of the slip conditions, or even increase the
>bias ratio if slip is continuing to happen.

That's not changing The Bias Ratio based on the definition of it.  The
capacity of a locked diff creates the "infinite", it's not a function of the
Bias Ratio.

>as i see it, there are 2 weaknesses with a haldex: there is no torque bias
>in the normal case (i.e. 2wd until slip occurs).

Completely depends on how the electronics or the clutches are preloaded
 Even now, that's not a truism, the TT haldex is 95f/5r.  The haldex is
capable of shifting up to 45% of input shaft torque to the output shaft.
Once in that locked state, either axle can support 100% Trg.

> to solve this you need a
>differential (keith you are right about the visco lok - good and bad - i
>have the sae paper).; and secondly that once slip occurs the haldex is
>working to minimise the slip so your pump has increasingly less ability to
>build pressure.  whether or not this is a real issue, i don't know - it
>certainly appears that way on the surface.

It can retain pressure OR change how it "builds" pressure, either concept
work, right now it can hold "peak" indefintely thru the throttle valve.
Right now the "designed" slip before pump pressure is 15degrees of wheel
rotation differences.  That can be modified several ways, either by an
external pressure resevoir (bomb) an internal modification to the pump
valves/clearances, or by adding more preload pressure, etc.  Gen II haldex
sure appears to take a more concentrated look at center diff applications vs
hang-on.  Regardless, a great first step in active differential technology,
the electronic component creates a massive jump in how the same coupling
concept can be changed significantly with the addition of electronics.  What
can happen in  the Jeep  Halex type is massively different that what can
happen in the TT, same device with ecu.

>the haldex gives the chassis engineer some interesting tools to play with.
>i wasn't the only person wondering if the tail happy nature of the early tt
>had something to do with the ecu opening the clutch on throttle-off
>conditions where a more prudent approach would be to ensure that there is
>least a measure of rearwards torque bias, particularly at high speed.
>the weakness of the haldex in these situations and a nice summary of the
>above 2 points is that if slip hasn't occurred in a high speed corner, in
>the throttle-lift scenario where you would want to apply rearwards torque
>bias to limit oversteer, you can't do a thing.  with an active clutch you
>could do what you liked.

This is summary is *not* the weakness of the haldex, it's a weakness of the
programming of the haldex.  The device is capable of reducing rear torque
without release, it just needs to be programmed to do so.  I don't agree
the problem with the TT is the haldex necessarily, as the FWD version of
vehicle were known to have the same problem.  Thinking in terms of the fixes
involved (no programming changes to haldex, adding rear spoiler and already
added rear weights off the back bumper) this is a chassis dynamics problem
inherent to the chassis design, not the driveline components within it (IMO)

More information about the Torsen mailing list