Semi-active differential idea

[mech5496]

Got this idea for a while now, came to me while watching the transmission/steering system on a battle tank during my military service (it is actually impressive how many things you could learn by watching around those things). It all came back to my mind while reading a diff-related topic and as it is unlikely for my team to build and use one, I thought of posting it here and get some discussion going.

The concept is really simple (and thats the beauty of it): It uses a simple, open diff (maybe from a small car/ATV, the lightest unit possible). On each side of the diff, there is an inboard brake rotor (on the diff output shaft or on each CV housing), each with a 2x2 (4 pot) caliper, like the ones AP Racing produces. Two pistons (out of four) on each caliper are actuated by the brake pedal, as in any normal 2pot caliper, while the other two are actuated by an electrovalve. The operation is such that the control system brakes the inner wheel both in corner entry (aiding turn-in) and each time the inner wheel spins excessively. The diff then sends more torque to the outer wheel (which has more traction). In extreme cases the system could even lock the inner wheel, acting somewhat like "tank-steering". On the plus side, the system sounds really simple and reliable(in case of system failure or malfunction, you just end up with an open diff) and no heavier than any diff setup, while having the ability of pseudo-torque vectoring (actually speed vectoring). On the downsides I can see overheating rear brakes and possibly lots of testing to make it actually work and make you faster. Any thoughts?

[CWA]

I like it. In principle it seems to be exactly how most brake-based stability systems work on road cars? And with the potential for integrated longitudinal traction control/ABS too I imagine you are implying..

On initial thought, from a practical point of view, getting the interaction between brake pedal and the electo-valve right (as they are both using the same caliper) might throw up some problems. Just checking things like what happens when they are both applied at the same time, what happens when one starts to be applied whilst the other is already fully applied would need to be done. Worst case scenario in terms of a failure of this system when sharing a brake system component would be.. ?

Anyway, in theory, it sounds like what would be a mechanical equivalent to Delft's in-hub electric motor torque vectoring. In it's automotive application, as is my best description, a brake-based stability system is used to ensure yaw rate/body slip angle's do not change more than the driver wants them to when departure is seen. Both axles being utilised (brakes on all 4 wheels), if the front axle departs (terminal understeer), yaw rate drops relative to ay, and the rear (inside) brake is used to apply a yaw moment to the body, whilst also reducing forward speed to restore it's attitude/direction. Likewise, if the rear axle departs (terminal oversteer), yaw rate increases relative to ay, and the front (outside) brake is used to apply a restoring yaw moment to the body, whilst again reducing forward speed. Having experienced these systems on a Volkswagen van, they seem to kick-in not just when terminal departure is seen (where peak ay is reached), but quite a bit before, where the vehicle becomes heavily "non-linear" (yaw rate vs steer).

For a motorsport application, you would be using it to apply an additional yaw moment on "turn-in" to accompany the yaw moment the tyres already create with lateral force and their distance from the CG. By this logic, it can only help in transient situations, which in fairness, most Formula Student endurance tracks/events are heavily biased towards. This meaning it wouldn't really help you on the skid-pad. .

And as for developing a control system that can process the inputs of yaw rate, steer angle (and ay?) signals and turn them into useful brake signal outputs.. well as you have implied, I think it would be damn hard, and I'm still in awe of people who manage to do it, and turn it into an advantage on the track.

As I understand, an old Formula One car (I think a 70s/80s McLaren?) had two brake pedals that applied this very theory in practise, where one pedal would apply full brakes, and the other pedal would apply brakes to only one side, the chosen side, depending on the corner direction, was adjusted by the flick of a switch in the cockpit. I think the system was famously uncovered in the media by a photographer sticking his camera in the cockpit of the car and snap-shooting the additional pedal? Anyway, whether it was ruled out by regulations or never really provided gains I don't know, but this technology seems to work with Delft electrically.

[mech5496]

On initial thought, from a practical point of view, getting the interaction between brake pedal and the electo-valve right (as they are both using the same caliper) might throw up some problems. Just checking things like what happens when they are both applied at the same time, what happens when one starts to be applied whilst the other is already fully applied would need to be done. Worst case scenario in terms of a failure of this system when sharing a brake system component would be.. ?


That's why I suggested using those 2X2 AP Racing calipers (http://www.apracing.com/productdetail.aspx?productid=2538), which in essence are two 2-piston calipers in one body (the pistons are independent in pairs and require two feed lines to operate all four of them). So use two pistons for "normal" braking and the other two for vectoring.

On a side note, if I were to build it I would consider a "spur-gear" differential (Take a look here: http://www.theengineer.co.uk/channels/design-engineering/in-depth/schaeffler-differential-holds-promise-for-oems/1009214.article). It is really early technology, has been used in farm tractors (and battle tanks) for a while and IMO is better due to more compact size, possibly smaller weight and better efficiency of the spur gears compared to the bevel gears used in differentials.



As I understand, an old Formula One car (I think a 70s/80s McLaren?) had two brake pedals that applied this very theory in practise, where one pedal would apply full brakes, and the other pedal would apply brakes to only one side, the chosen side, depending on the corner direction, was adjusted by the flick of a switch in the cockpit. I think the system was famously uncovered in the media by a photographer sticking his camera in the cockpit of the car and snap-shooting the additional pedal?


Actually it was late 90's (98 or 99) and it was Mika Hakkinens' car that had been photographed.. ;)

[bob.paasch]

Harry, we ran exactly this setup in BajaSAE in 2006 and 2007. We used a Polaris caliper with two independent circuits, and dual linear actuators acting on small master cylinders. The last instantiation used super-caps to provide quick response. The controller input consisted of rear wheel speed sensors, brake line pressures and a steering position sensor. Straight ahead, the system acted as traction control, braking a low traction spinning wheel. At high steering angles, the inside rear wheel received braking to enhance turn in. The controller had a simple map based on steering angles and predicted wheel speeds, and compared that to the actual wheel speed to adjust brake pressure. The system was effective but expensive for BajaSAE at around $1000. In BajaSAE, that's about 10 points, and we couldn't justify the expense.

[CWA]

Harry, we ran exactly this setup in BajaSAE in 2006 and 2007. We used a Polaris caliper with two independent circuits, and dual linear actuators acting on small master cylinders. The last instantiation used super-caps to provide quick response. The controller input consisted of rear wheel speed sensors, brake line pressures and a steering position sensor. Straight ahead, the system acted as traction control, braking a low traction spinning wheel. At high steering angles, the inside rear wheel received braking to enhance turn in. The controller had a simple map based on steering angles and predicted wheel speeds, and compared that to the actual wheel speed to adjust brake pressure. The system was effective but expensive for BajaSAE at around $1000. In BajaSAE, that's about 10 points, and we couldn't justify the expense.

Hi Bob,

Just wanted to clarify something about the operation of your system. Did it apply brake torque to aid "turn-in" based on high steering wheel displacement, or high steering wheel speed? I would have thought the latter more appropriate to what you are trying to achieve with it. If the former, as you described, the system could be braking the inside rear wheel during steady-state, for example, on a skid-pad (I'm thinking if this were on a hypothetical FSAE car, I'm not sure if Formula Baja has equivalent to a "skid-pad" event), where really an unbalanced yaw moment is not needed? In this case, under those circumstances, it might have even slowed the car down? Perhaps my understanding is a little off.

Did you guys manage to quantity any improvement made by this system? It sounds very interesting any way!

Look forward to your reply.

Chris

[MCoach]

If you check here you can actually find out quite a bit of the nitty gritty details you're looking for:

http://ir.library.oregonstate.edu/xmlui/bitstream/handle/1957/2850/Brake%20based%20wheel%20speed%20control.pdf?sequen ce=1


It helps to know what you're looking for. I was actually referencing this paper sometime last year for an unrelated project.

Thanks, Bob.

[bob.paasch]

Hi Bob,

Just wanted to clarify something about the operation of your system. Did it apply brake torque to aid "turn-in" based on high steering wheel displacement, or high steering wheel speed? I would have thought the latter more appropriate to what you are trying to achieve with it. If the former, as you described, the system could be braking the inside rear wheel during steady-state, for example, on a skid-pad (I'm thinking if this were on a hypothetical FSAE car, I'm not sure if Formula Baja has equivalent to a "skid-pad" event), where really an unbalanced yaw moment is not needed? In this case, under those circumstances, it might have even slowed the car down? Perhaps my understanding is a little off.

Did you guys manage to quantity any improvement made by this system? It sounds very interesting any way!

Look forward to your reply.

Chris

As I recall, the system used steering wheel displacement. For turn-in, it didn't brake the inside rear wheel until almost full lock. There's no skidpad-like event in BajaSAE. There is "maneuverability" which is like an off-road autocross except tighter, and the system was useful there. BajaSAE also has an endurance event (100 cars wheel-to-wheel, 3-5 km track, 4 hours long, extremely entertaining :) ). The traction control part of the system could have been useful in endurance in muddy conditions, but the corners are generally high enough radius that the yaw moment part didn't kick in.

[mech5496]

Hi Bob,

good to hear that someone else has considered it useful, and even better to hear that it actually works! Have you ever tested it to FSAE conditions? IMO it should give IC cars some of the torque vectoring advantage E-cars enjoy (that's why I thought of it in the first place, and that's why we will never gonna develop it; if you want torque vectoring on an e-car, just go all the way and use multiple motors).


Hi Bob,

Just wanted to clarify something about the operation of your system. Did it apply brake torque to aid "turn-in" based on high steering wheel displacement, or high steering wheel speed? I would have thought the latter more appropriate to what you are trying to achieve with it. If the former, as you described, the system could be braking the inside rear wheel during steady-state, for example, on a skid-pad (I'm thinking if this were on a hypothetical FSAE car, I'm not sure if Formula Baja has equivalent to a "skid-pad" event), where really an unbalanced yaw moment is not needed? In this case, under those circumstances, it might have even slowed the car down?

Chris

Chris, as far as I thought of my system, it would be really simple in its' first iteration; just use steering angle and wheel speeds to sense if the inner wheel slips and brake line pressure to get feedback on what the system does. When the inner wheel slips above a certain percentage (compared to the predicted wheel speed), the system would brake it, transferring more torque to the outer wheel. That way I think it should not brake a wheel in skidpad, but even if it does you should be able to de-activate the system (from a cockpit mounted switch maybe?) for that event only.

I think it would be fun if someone tried it out!

[Rex]

Actually it was late 90's (98 or 99) and it was Mika Hakkinens' car that had been photographed.. ;)

McLaren actually uses a similar concept (although done by the computers rather than a separate pedal and switch) in the MP4-12C. It uses a lightweight open diff, and the car actively brakes the inside rear on corner entry under certain conditions. I do not know the exact details (i.e. what sensors are used and what conditions are required to activate the system), but the end result is that you can enter a corner at a speed that should be quite a bit too fast, and the car still turns in and makes the corner when it should have otherwise pushed straight off the track. It's a very strange feeling when it happens!

[bob.paasch]

Hi Bob,

good to hear that someone else has considered it useful, and even better to hear that it actually works! Have you ever tested it to FSAE conditions? IMO it should give IC cars some of the torque vectoring advantage E-cars enjoy (that's why I thought of it in the first place, and that's why we will never gonna develop it; if you want torque vectoring on an e-car, just go all the way and use multiple motors).


We didn't pursue the idea for Formula. In that timeframe, we were pretty busy developing the whole GFR organizational concept. And given our GFR design philosophy, we would have to be very convinced there is a significant competition point gain to be had to invest the human resources on development.

[mech5496]

McLaren actually uses a similar concept....the end result is that you can enter a corner at a speed that should be quite a bit too fast, and the car still turns in and makes the corner when it should have otherwise pushed straight off the track. It's a very strange feeling when it happens!

MP4-12C magic is most on the e-diff and the Kinetic interconnected suspension system; and everyone that has ever driven the car claims that might feel strange but it is actually faster than it should be! I'm quite sure McLaren engineers are much better than any of us in developing the control system though...

[Rex]

MP4-12C magic is most on the e-diff and the Kinetic interconnected suspension system; and everyone that has ever driven the car claims that might feel strange but it is actually faster than it should be! I'm quite sure McLaren engineers are much better than any of us in developing the control system though...

Well sure - lots of the 12C systems are magic, especially the two you mentioned (although some people dislike the ediff stuff). I'm just speaking specifically to the brake steer because it's a real-world example of the concept the old F1 car had, and it's very tangible. When it happens you KNOW it happened. I didn't fully appreciate it until entering a corner much too quickly and feeling it kick in right as I expected the car to understeer off track. So you're absolutely right that it's faster than it should be. So much faster that it can be a bit scary at first. If I could do FSAE over again I would probably want to investigate these concepts on an FSAE platform if the team had the time/resources to spare - would be a really cool learning experience in any case!

[mech5496]

I was referring to brake steering too when I said e-diff, as it is an integral system, at least at my proposed version.


I'm just speaking specifically to the brake steer because it's a real-world example of the concept the old F1 car had, and it's very tangible.

Fun fact: The "old F1 car" code name was MP4/12... ;)

[DannytheRadomski]

That isn't coincidental. The 12C was named after the F1 car because of the "brake steer" feature.

[onemaniac]

As I understand, an old Formula One car (I think a 70s/80s McLaren?) had two brake pedals that applied this very theory in practise, where one pedal would apply full brakes, and the other pedal would apply brakes to only one side, the chosen side, depending on the corner direction, was adjusted by the flick of a switch in the cockpit. I think the system was famously uncovered in the media by a photographer sticking his camera in the cockpit of the car and snap-shooting the additional pedal? Anyway, whether it was ruled out by regulations or never really provided gains I don't know, but this technology seems to work with Delft electrically.

McLaren called it 'Brake Steering' where the 4th pedal on the left side would bias the driving (or braking) torque in conjunction with steering wheel angle.