Resolved [RESOLVED] How does AWD work? [Merged 9-7] torque split division slip drive

[turboawd1992]

Just always wondered, does anyone know how much power a 1g AWD sends to the front/ rear wheels?

 

[Alex7747]

haha this thread kept me entertained and taught me a lil at the same time. Better than school!

 

[4g63dreams]

I now understand awd in our vehicle because of this discussion.

I think what h4 was getting at by saying that the front see more torque in the reg distribution is that our fronts have more forces applied to it. So they technically n theoretically require a higher distribution. The 50/50 physically never changes because those are technically stating that each of the gear set responsible for the front and back is the same size. Stating that this O is the same size as this O.

During acceleration distribution is (60/40 are numbers I use for illustrative purposes) 60/40. During cornering when accel is applied it is still 60/40. During cornering when brakes are applied it is 40/60. This torque I speak about is front and back split. In additional to this front and back spilt, it is also spilt officially between each wheel. Below mentions front/back distribution, as well as left/right distribution.

Front/back , left/right

During straight accel. 60/40 , 50/50
During left cornering accel (exiting a left corner) 60/40 , 40/60 (passenger front having greatest torque
During right cornering accel (exiting a right corner) 60/40 , 60/40 (driver front having greatest torque.

Assuming torque is 100% @ all instances...

During your exit at a right corner torque should be 30L/20R (50 front) , 30L/20R (50 rear). Just apply to the correct situation above and it gives you an estimate based on my numbers.

During left cornering brakes applied (entering corner) 20L/30R (50 front) , 20L/30R (50 rear).

The most torque is sent to the opposing side of the direction of weight transfer because the weight transfer increase grip in the direction it travels.

Also keep in mind the power loss through mechanical transfer between front and back + sides.

I'm sorry to beat a dead horse, this is how I understood it. I'm not saying anyone is wrong, and I'm not calling anyone out. Just wanting to see if I can explain the concept per my understand. And I hope it helps somehow. If im wrong someone please correct me.

 

[steve]

The way I understand it, unless a wheel(s) are slipping there can't be any transfer of torque.

Now when your not traveling in a straight line the wheels are slipping relative to each other which is what the differential is for and there will be some power transfer but it's limited to the difference between the relative speeds of the wheels.

When a wheel is spinning the momentary transfer can approach 100% but as soon as the VC locks the transfer goes back to 50%. Since there isn't a slip limiter up front the available power can be shifted to one side or the other but not more than half because of the center VC.

 

[TunaTalon]

Some of the posts in this thread say that the Viscous Coupling (VC) in the DSM proportions torque to its outputs in relation to the difference in speeds of the outputs, and some say that it is simply open until it locks up. Both can't be right. Let's see what the inventors of the VC say.


Searching the US patent office data base at US Patent Full-Text Database Boolean Search with search terms “viscous coupling” in the title yields 61 hits. Adding the term “lock up” yields only one hit ( number 5,086,863 from 1979) where the term “lock up” is used as: “...when using a spring damper lock-up clutch.” So no support in the USPTO data base for a viscous coupling using “lock up”

Looking only at patents prior to the introduction of the DSM finds many of the patents assigned to Harry Ferguson Limited. For example patent number 3,760,922 from 1973.
From that abstract:

A device for controlling or limiting difference in rotational speed or acceleration between relatively rotatable shafts, for example the front and rear propeller shafts in a four-wheel drive motor vehicle transmission, employs a viscous liquid such as a silicone fluid in contact with relatively rotatable sets of mutually interleaved and mutually spaced annular plates the working surfaces of which are provided with through openings in the form of holes and/or radial slots from which are derived in combination with the inclusion of a pre-determined quantity of air in the viscous liquid advantages including inter alia improved compactness in relation to torque capacity and in relation to dissipation of heat generated during shearing of the viscous fluid. A transmission unit incorporating an interaxle differential gear and such a device is described.

Another patent assigned to Ferguson number 4,022,084 from 1977 refines the earlier invention and calls the invention a “control coupling” in the field of the invention:\

This invention relates to devices for controlling or limiting differences in angular velocity or angular acceleration between relatively rotatable members. Hereinafter, and in the appended claims, such a device is referred to as "a control coupling". Control couplings find use, inter alia, in controlling the relative rotation between parts of a differential gear, particularly an inter-axle differential gear in a four-wheel drive vehicle. The invention also relates to a four-wheel drive vehicle having an inter-axle differential gear associated with a control coupling.

Also in 1977 patent 4040271 was assigned to Ferguson Limited. From the patent abstract:

The quantity of viscous liquid, in the range 85% to 100% of the minimum volume at ambient temperature, and the spring strength are chosen to give a predetermined pressure build-up within the enclosure with rise in temperature during shearing of the viscous liquid to tailor the torque-transmitting characteristics of the coupling to particular needs.

The DSM VC will lock up when stressed beyond its design limits but the normal operation is to distribute torque so that more torque goes to the slower moving output.

 

[steve]

Some of the posts in this thread say that the Viscous Coupling (VC) in the DSM proportions torque to its outputs in relation to the difference in speeds of the outputs, and some say that it is simply open until it locks up. Both can't be right.

If you thought I was claiming the latter you introduced something in the reading that wasn't written.

Whenever there is a difference in rotational speed between the plates drag caused by the shear of the silicone oil will pull on the slower one to speed it up and pull on the faster one to slow it down attempting to make the difference 0.

The DSM VC will lock up when stressed beyond its design limits but the normal operation is to distribute torque so that more torque goes to the slower moving output.

I don't read any of your citations as stating locking up is outside the design limits.

The transfer function is not linear so at low differential speeds the torque transfer is minimal. It works well enough that off the ground in neutral both rear wheels will turn in the same direction but if you hold the pinion you'll see normal differential action causing the wheels to turn in opposite directions.

The 1G FSM in it's typical Jnglish implies that the locking of plates in the Hump region is how the VC provides for keeping the car from being stuck in the worst case. It talks about Normal operating range and Hump region but don't say that hump is outside the design limits but a part of them.

 

[TunaTalon]

If you thought I was claiming the latter you introduced something in the reading that wasn't written.

Whenever there is a difference in rotational speed between the plates drag caused by the shear of the silicone oil will pull on the slower one to speed it up and pull on the faster one to slow it down attempting to make the difference 0.

I did not keep track of which posters took which position, only that both proportional torque distribution and lock up mode were both described in this thread. No finger pointing by me, just an attempt to go to the most reliable documents on the issue.

I don't read any of your citations as stating locking up is outside the design limits.

The transfer function is not linear so at low differential speeds the torque transfer is minimal. It works well enough that off the ground in neutral both rear wheels will turn in the same direction but if you hold the pinion you'll see normal differential action causing the wheels to turn in opposite directions.

The 1G FSM in it's typical Jnglish implies that the locking of plates in the Hump region is how the VC provides for keeping the car from being stuck in the worst case. It talks about Normal operating range and Hump region but don't say that hump is outside the design limits but a part of them.

True the Ferguson patents do not mention the hump mode. Like most inventors he talks about how the device is supposed to work and leaves the bad side effects unsaid. Avoiding Wikipedia, a dictionary of automotive terms at Automotive Dictionary - "Hu" describes the Hump Mode as:

An operating condition where the transmitted torque in a viscous coupling rises to a value several times higher than the value produced in the so-called Viscous mode, due to internal clamping, i.e., metal friction of the coupling discs

The different posters positions are reasonable. The hump mode is real and mentioned often in the literature. Lock up differentials are real and were the cat's meow in my youth. Torque distribution as a function of differential speed is necessary for a AWD car.

For the intended and normal operation of the DSM LSD I will stick with the words of the inventor.

 

[jtmcinder]

You have a little bit backwards, Steve. When a VC locks solid, then any and all torque can be moved from one end to the other. When the outputs from a diff are turning at different speeds (due to, e.g., tire-slip), then the action of any LSD, such as a VC, is to shift torque towards the slower-turning output. It does this by slower the faster output and speeding up the slower output. That shifting in speed - or, maybe better said as: that bringing the two speeds closer together - is what moves the torque. When a VC is working but not managing to lock solid, then less than the maximum amount of torque is being shifted, but some is still being moved across.

Note that "hump mode" corresponds first to something that happens inside a VC: the plates come into direct contact. The effect of hump mode is to suddenly raise the odds that the VC will be able to lock solid.

 

[4g63dreams]

Cinder, Was I off?

In regards to the transfer when cornering and weight?

 

[jtmcinder]

Not really, but it wasn't awful (and you were cautious, too, which is nice). For one, it's the front that usually slip first on a hard launch, so transfer is from front to rear, so you're much more likely to end up at 40/60 than 60/40. Most of all, as Steve said, no transfer occurs until there is slip, so you don't get all the dynamic changes that you talked about in everyday driving.

Now, if we switch and start talking about predictive or pro-active systems, like the active centers in an STi or Evo, then we'll have the torque distribution going all over the place without any wheel-slip. And a helical can start locking before noticeable wheel-slip occurs, too. But a VC doesn't do anything until the two outputs are turning at different speeds.

 

[4g63dreams]

Ahhh ok. I see now.

Comparing active centers in sti and evo to our vc static-until-slippage, what system is easier to drive? And if actives are better, can we have those installed on our vehicles?

 

[jtmcinder]

Anything is easier to drive than a VC. VC's are non-linear and depend on previous conditions (e.g., the heat already in the fluid). A well-programmed active, such as that in my Evo, is very predictable and not too intrusive. Better than the STi, but not by much, as the STi's is darned good, too. Never drove a car with a helical center, but that would probably be way better than a VC. Again, the only thing worse than a VC is an open center. A VC is a "don't get stuck in mud or on ice" gizmo; it isn't a performance LSD.

You can do anything with enough money, but it would outrageously expensive to add an active center to a DSM. They need a lot of sensors ... wheel speeds, yaw, acceleration both directions, etc.

I'd look at the Cusco 35/65 center before anything else. You don't need a LSD device is the native split matches the grip.

 

[steve]

You have a little bit backwards, Steve. When a VC locks solid, then any and all torque can be moved from one end to the other.

I'm not sure I follow where I got it backwards. When the VC is locked the most it can to is insure a 50/50 split since the differential is involved. Up to that point of locking I can see the torque transferring because of the differential but the VC works to limit that transfer.

If you don't have a slip limiter, then 100% of the torque can move to the slipping wheel because of the differential but as you point out with a limiter some of the torque is shifted to the slower moving wheel until both are moving the the same speed.

If I'm not getting it, help me figure out why.

 

[jtmcinder]

A locked "diff" is not a diff anymore. The two outputs must turn the same speed, regardless of how much torque is being used by either end of the car. For example, a car with it's front wheels on oiled glass with a locked center will have a torque distribution of 0/100 (at first).

The key, in my opinion, is to keep torque split - which is set by the gearing of the diff and never changes - completely separate from torque distribution - which also involves any LSDs and can change from moment to moment.

Then add in the fact that a locked "diff" is no longer a diff. The native torque split is completely irrelevant when the diff is locked. All that matters when the diff is locked is how much grip each end of the car has.

 

[jtmcinder]

I just reread what you wrote. Maybe the issue is mixing torque and shaft speed. If you have no limited slip device and one end of the car has very little grip, then, yes, the wheels at that end will be spinning like mad, but they still are being sent 50% of the torque. They just aren't using that torque for anything useful. Likewise, the end with grip isn't wasting any of it making tire-smoke, but it only can receive 50% of the torque, as that's what the diff sends it.

 

[TunaTalon]

Anything is easier to drive than a VC. VC's are non-linear and depend on previous conditions (e.g., the heat already in the fluid).

The time delay is true for Harry Ferguson's original vision. But that was over three decades ago. The delay problem has been solved by filling the VC with a Silicon fluid with a viscosity that increases with shear rate. The faster the slip the higher the viscosity and the greater the coupling.

Google failed me when trying to find if Mitsubishi used such a fluid in the DSM. Wikipedia doesn't count so I don't look there for reliable documents. I did find one patent (Number 5,526,912) referencing such a fluid.

28. A coupling according to claim 3, wherein the viscous fluid is a dilatant medium whose viscosity increases as a function of the shear speed.

Note that this is claim number 23 so the inventor is not claiming invention of such a use, just this application.

The time delay problem may be solved but the non-linearity remains. Not as non linear as the good ole lock up differential which is either on or off but not the slick computer controlled gee whiz technology of 2012.

But a VC doesn't do anything until the two outputs are turning at different speeds.

Are you trying to tell me that the DSM limited slip differentials only function is to limit slip and if there is no slip to limit then the limited slip differential doesn't do anything? Well OK then. I thought you had slipped up there on first reading.

 

[jtmcinder]

The silicone fluid is still temp-dependent, which is why I say that the behavior of a VC is dependent on previous conditions, which is one of the reasons that I don't like them. In autocrossing, for example, it's nice to be able to shrug off a mistake and do the rest of the run as best as you can and as planned. But with a VC, one tire-smoking error can trigger another, because the approach you take to the next corner - which worked great last time - now doesn't work because the center VC is acting up.

I know that the last paragraph was a joke, but it was a rear change for me switching to a car with pro-active (instead of reactive) limited slip. Having a center that locks before tire-slip happens is uber-cool.

 

[TunaTalon]

I have only one data point from an anecdotal source about the speed of the DSM VC response.

Back in 1991 when I bought my first new AWD talon I read about the need for temperature change to change the behavior of the center and rear differentials. Hmmmm I thought that sounds like a bad design. The next winter I was first off from a traffic light and before me was the perfect test road for LSD response. The highway curved gently to the left and snow had melted and ran down from the right shoulder and refrozen in 3 foot to five foot stripes across the pavement. Ice, pavement, ice, pavement, ice, pavement for hundreds of feet.

At the time I was young and foolish, still in my early fifties, so in the spirit of science I went to WOT. The differentials, suspension, and tires performed perfectly. There was no tire chirps on changing from ice to pavement and no loss of directional control at any time. The engine management system on the other hand was thoroughly discombobulated. The RPMs were going all over the place with no apparent relation to the load.

OK, on re-reading the above it probably says more about the training of the driver than the drive train. But the event still has an effect on my confirmation bias.

 

[jtmcinder]

All I can say is that my AWD 2G didn't act like that. If I was on the gas and one tire hit ice, that tire would immediately spin, even if it was a rear tire (which connects to the engine via 2 VCs). I doubt that my car had significantly more power than yours, so that's not likely the reason for the difference.

But I can say this with confidence: Only hump mode can lock the center hard enough to reroute half of the available torque at WOT, even if you have only OE levels of power. And hump mode requires some pre-heating to occur.

 

[TunaTalon]

The ice stripes went all the way across my lane so they were encountered two tires at a time.
The tires probably did spin on the ice, just not enough to lose control or chirp when transitioning to pavement.

When your say "Only hump mode can lock the center hard enough to reroute half of the available torque at WOT..." I read the "re" in reroute imply that 50% is normally routed to each output and with 50% rerouted that would be 100% to one output. True, but not required to explain the behavior of my 91 Talon or excuse the behavior of the driver.

I still think of hump mode as a bad thing that causes the VC to fail soon like driving with mismatched tires or towing with the front tires off the road. I don't trust Wikipedia or forums so I will hit the USPTO site again soon.

Or I could just buy a new Infinity with intelligent torque split system. Nah.

 

[jtmcinder]

Be careful to not take seriously the claims made in ads for cars like a G-series Infiniti. The ads are almost always written by people who have no idea how the car works. The G-series "AWD" is really an updated version of the old 4-Matic-type system (used by Mercedes-Benz and VW). The engine is directly connected to the rear wheels (i.e., via gears, shafts, etc), while the fronts are linked in through a LSD-like system. In the old 4-Matics, the fronts were connected via a VC. So, normally, no torque went to the front, since the fronts and rears were turning at the same speed. When the rears start turning fast (e.g., they are spinning, because the car is stuck), then the VC starts to send torque to the fronts. Well, a G-series' "AWD" is the same thing, but replaces the VC with a clutch almost exactly like the clutch on an active center in an STi or Evo. Now it's a computer that decides when to send torque to the front, instead of a VC. But note - and here's where the ads are nonsense - because the clutch acts to lock the front and rear together, it doesn't switch the car from 0/100 to 50/50; it switches the car from RWD to 4x4. And a 4x4 (i.e., a vehicle with a locked "center") doesn't have a meaningful torque split; it only has a moment to moment torque distribution. In other words, those ads are making the same false claim that so many people make: that a locked center produces a 50/50 split.

In general, all of those 4-Matic-type systems, whether they use a VC or a computer-controlled clutch are really only good at getting unstuck. Do not base a performance AWD car on it.

 

[TunaTalon]

Don't take ads seriously you say? Now there's something we can agree on. I don't even take the experts seriously, not since the mid seventies. But I do like my son's G35 coupe. And now it's available in AWD.

I just checked the first five patents related to viscous couplings and the hump mode and got one inventor trying to get to hump mode smoothly one who can go either way and three inventors trying to avoid hump mode.

The first hit in Google for {patent “viscous coupling” hump} was patent number 5,133,438 from 1990 that mentions preventing the hump mode torque in the abstract and in claim 1. I don't quote full sentences because lawyers conserve periods. Only one period per claim is allowed.

“...whereby the so-called hump torque is prevented from being produced...”

Patent number 5,562,191 from 1996 claims a way to get the torque transfer of the hump mode without overheating or physical contact between the plates. Again the old way of getting hump mode is considered a negative to be avoided.

“The invention takes advantage of the fact that, in principle, the hump mode also occurs independently of the coupling temperature provided the conditions of a completely filled operating chamber and an increased internal pressure are met. As a result of the progressive increase in the internal pressure as a function of the differential speed, the increase in torque is also progressive. The operating mechanism is independent of the stationary coupling temperature. Therefore, it is only a function of the temperature gradient in the operating chamber and thus entirely a function of the speed differential. “

Patent number 5404978 from 1993 describes the hump mode as a problem in prior viscous couplings because it causes wear and eventual failure.

“The friction locking effect, especially during the hump mode, causes individual plate particles to be abraded. The separated metal particles are absorbed by the viscous fluid, but have a damaging effect on the service life and on the effect of the viscous fluid. It has been found that, in use, the viscous fluid gels (silicone oil with a high viscosity of 5,000 to 300,000 cSt). This, in turn, limits the service life of the coupling. “

Patent number 5080211from 1992 describes a technique to assemble viscous couplings to achieve hump mode or not. A mixed bag in this search.

“...thereby forming a kind of labyrinth to facilitate the transfer of the coupling into the so-called hump mode. However, with different applications it is necessary to achieve different characteristics of the torque/speed curve. For instance, a 100% torque transmission is to be avoided; only part of the input torque is to be transmitted. “

Patent number 4,940,123 from 1990 takes a contrarian approach and claims a way to get to the hump mode smoothly.

“Thus, it is possible in this embodiment of a viscous coupling to be capable of preventing decrease in torque due to the shear force, and thereby enabling escape from skidding of one of the rear wheels smoothly by hump phenomena, so that the
running stability of the vehicle can be improved.”

It looks like if we look far enough any viewpoint could be supported by some “reliable source” I'm still going with the three out of five results that direct metallic contact between rotating plates is a bad thing and to be avoided. Anyway a glass of wine is calling my name so....