Nugget

If you've got anything to add that hasn't already been explained in the other thread I care. If not, then not. Hope Cindy is OK.

simsgw

I won't know for sure until she wakes. I think it was just soft tissue bruising. Fifty years ago, we were gymnasts and it looked like a clean hit-and-roll to absorb the energy, but it isn't a good plan to try those things with old bones like ours. She went from full standing to flat on the floor. Scared hell out of me. Sorry if I came across as hostile.

Now back to brakes. I just spent an hour or so discussing this with a fellow engineer off list. It sounds like we've reached the finger pointing stage where people who experience this know it happens and people who haven't -- and especially people who think they know the ABS system -- are convinced it's a driver error of some sort. Either too much braking in a bad situation, 'upsetting' the car, or possibly aftermarket mods or at least R compound tires.

You can make the problem occur more often with R compound tires, but that isn't causing it. Cindy's been asleep for over an hour now and I expect her to wake soon, but I'll paste in the longish engineer-to-engineer description I just sent that friend. Then I'll start editing out the most technical bits and replacing them with simple English. When she wakes, I'll have to stop, so you folks will get a half-edited, half-geekish explanation. Sorry. That's all I'm up to today.

Here's what I had time for:

I think that poor designs are more easily confused than good ones. The quick version between us engineers goes like this. [Between engineers, that means "I could write a book but we'll settle for a chapter."] Simple ABS systems from the sixties looked for lock-up, and eventually by the seventies they watched for incipient lock-up, and they responded by dropping or completely releasing line pressure to the service brake on the affected axle. They weren't even wheel specific at first.

A system like that is in fact *not* as good as drivers trained to threshold braking and that combined with the usual "John Henry" effect to make drivers feel they could improve on ABS if they worked at it. A modern ABS has different design goals than simply avoiding a loss of control. Now they are designed as we did aircraft systems: wring the most performance from a tire that it is capable of delivering under the extant circumstances.

This particular problem is a very subtle effect of tire performance. Basically, a wheel into full ABS mode is being controlled by a feedback circuit that modulates the brake pressure. As always in feedback, the goal is a dynamic stability under changing circumstances. It turns out that two modes of stability exist. It is a bi-stable situation that we usually tolerate. But in extreme cases, we cannot. Or if we must, we have to be aware it can bite back.

You know that tire thrust increases up to a certain amount of slip and then decreases. A 'forgiving' compound is one that has a significant plateau effect. Like an engine we call 'torquey', the compound has a fairly flat area near the peak of performance. Once on that plateau, moderate changes in conditions of traction or in braking force (or cornering force which is where the discussion usually centers) cause little change in thrust produced. Even on racing compounds this is important because no two tires will have the same traction and local tread heating at any time during braking. Close when conditions are ideal, but not even close sometimes. So to get all four tires on their individual plateau at once, we need their response curve to be flattish on top for a workable width.

Now we get to the tricky part. That *is* a curve, not a line at any point. That means it has two solutions for the intersection with a line. At least two solutions in general, but since we don't expect to deal with subtle discontinuities in behavior, let's assume it is a simple curve like an hyperbola, so we have only two solutions. But now we design our ABS to look for a particular level of slip, right? What happens if the tire has shifted onto the far side of the peak before we detect the level of slip we want to stabilize? Well, it was thought that in principle that wouldn't matter. [Quick insert: this is where R compound tires make this happen more often I must suppose. I've only driven full-race and street tires, but I assume "r compound" means a tire that is closer to full race compound. If so, then they are 'peaky' tires and the two points of stability are closer together. It's easier for the system to settle on the wrong one.] [The two bi-stable points don't have the same level of slip. They just look like they do to a simple-minded sensor. As my friend knows, we don't measure slip directly. It is an inferred measurement, so it can be fooled.]

You have a thrust less than the peak whichever side of the curve we settle. We accept that in any feedback situation. We may need to design for getting closer to the peak in some situations, and obviously a Porsche would while a Kia would not, but once we set a design goal that says "this close to peak is satisfactory" then we go forward. The futile pursuit of perfection is the enemy of merely excellent results.

So we designed the systems to recognize acceptable levels of slip producing acceptable levels of thrust. Don't blow the tires on every landing, but don't run off the end of the runway either. On cars, that also meant 'acceptable' levels of cost as the component prices came down.

Attention was focused -- at first -- on subtle modulation of brake pressure to deal with all the variables. The brake is heating, the tread is heating, the surface is changing, and we don't want a bang-bang solution. [Like cheap building air systems. Bang, the heat comes on. Bang, it goes off. No modulation.] We want to adjust the pressure, not release it completely or restore it to full, when we detect a tire's performance is sub-optimum. We want modulation based on a feedback algorithm involving multiple sensor data streams in really high-end systems like the ones that aircraft require. Especially at the very high landing speeds we had to deal with in fighters when we didn't know how to build a transonic wing that also provided good low speed lift in the landing configuration. The F-104 landed at some ungodly rate like 150 knots for example. [All this means is you look at more information than just a speed sensor on the wheel. Maybe just add a speed and temperature sensitive response curve to the feedback circuit, maybe some really complex stuff. Lots of control theory my correspondent and I didn't have to make explicit. And I won't bore the rest of you.]

We did all that. I say 'we' but actually I was working on satellites by the time this next set of design issues arose. I was still a pilot though, so I followed the recurring reports of aircraft braking systems. Systems that some pilots were beginning to call widowmakers. Sometimes the system wouldn't stop the aircraft, but it damn sure got in the way of the pilot doing it himself. The discussions were polarizing in ways that old timers (like I am now) said reminded them of the World War II fiasco of torpedoes not exploding when hitting enemy ships. ("Didn't happen. It can't. You missed." Finally, some sub skippers got close enough to take movies and hung around to get film of the bounce back when they should have been leaving the scene.)

ABS engineers insisted it was some handling defect of the aircraft or some mistake of pilots. "If you can't make it happen reliably, it's not my brake, it's your flying." Notice that both arguments have reasonable cause to doubt witnesses. Combat conditions in the first case; and in a lot of the second one as well. But with aircraft we also have a significant problem with close-to-the-runway aerodynamics. It was quite plausible that the dynamics of brake and wing and deck angle were producing momentary lift that reduced the contact pressure and thereby reduced brake performance. No one was being unreasonable, but they sure were polarized between engineers and the pilots. And it should be noted that most pilots in those days, certainly the test pilots always, had engineering degrees themselves. Everybody understood the underlying principles at work. It was a question of theory versus field experience.

Parts of that thread on the GT3 forum sound just like those discussions. Especially the people who think the front end gets lifted somehow and then the ABS can't get the brakes working again when the downforce returns. The trouble is we engineers can't rely on anecdotal reports at the best of times, and even when the witnesses are highly skilled, their perception of events is worthless when we insert a system between their senses and the event being reported. Once we put an ABS into the loop, the pilot (or driver) has no direct way to sense what is happening at the tire.

The whole question burst into flame when the shuttle brakes had to be designed to handle this godawful heavy beast coming down at an unlikely landing speed and descent rate. Like a lot of things in the space program, conventional wisdom wasn't good enough, so money got spent on research that went deeper than the superficial description I just gave above. Actually, I call that superficial, but it's really the distilled knowledge of a whole generation of engineers in the sixties. Still, it wasn't good enough to solve the problems of spaceflight.

It turned out not to be the case that we don't care whether the feedback loop settles on the front side solution to that intersect or the back side. We absolutely must find a way to distinguish the two. What happens in detail is beyond my knowledge of tire design, but basically when we reach the level of thrust that is the design goal for this system at this tread temperature under these traction conditions, we are fine. For about ten milliseconds. If we're on the back side of that peak performance, the tire is heating faster than if we're on the front side. Now it starts to get so hot in some areas that rubber begins to vaporize. This phenomenon progresses rapidly until the entire tread surface of that tire is involved. (All during one braking event remember, but braking lasts a long time on a shuttle so they used a time window that let them recognize this. At least I suppose that's how they hit on it. With enough money to seriously inquire, perhaps they were just being thorough and suddenly got lucky. It happens, as we know.)

At any rate, what happens acquired the name "everted rubber hydroplaning", a term which is terrible English and not all we could wish technically, so I have to suppose they invented it to have a simple -- well, relatively simple -- phrase to use in presentations of their research results. Not having research papers at hand from the sixties, we'll settle for a layman's explanation. Enough rubber vaporizes in a short enough interval to provide a buffer between the surface and the tread. Not completely. Road surfaces are not polished, they have a texture. But enough to fill some of the gaps in the surface texture that normally would provide a 'tooth', a geared contact, that is what permits rubber and runway to produce as much thrust as the downforce in optimum circumstances. For context, remember this was a time when "common knowledge" among engineers held that you could not get more drag/friction/thrust from two materials pressed together than the force holding them together. Basic physics of friction, right? Well, too basic. In tires, the contact with the runway/road is more complex than the basic model contemplates. And in space the contact can be too good for the basic model. (Hence vacuum 'welding' and 'stiction' as we called it.) The result is not a loss of braking thrust, but a stable amount of thrust that responds to modulation by increasing or decreasing, but without ever reaching the performance level predicted for the tire. Pilots said the same thing some of those GT3 drivers do: "hard brake, pushing it as hard as my leg would permit, and I got nothing." Well, it isn't nothing, it just isn't nearly as much as it should be for the situation, especially with the end of the runway approaching. The feedback loop knows the tire is not in lock-up and it is delivering retarding thrust if you have a system that measures that as well. It just isn't delivering enough. You can see why those earlier engineers insisted it had to be aerodynamic lift killing the braking potential, not their system. But it was. The system had reached a secondary stability point.

Getting back to practical evidence, I have watched this happen a couple of times in a Formula Ford. No ABS, but a great deal of importance placed on threshold braking. If I'm pumping my brakes, I look like an idiot as the backmarkers pass me. You absolutely must learn how if you weigh forty pounds more than the average driver in a class without a starting line weight target. So a couple of times I got exactly that effect people describe, but I could watch it happen because the tires are exposed.

At Big Willow on a testing day, I was coming out of the carousel, the giant 800-ft radius turn two that lets you explore the high-speed g-limit of your tires. It's like a giant skid-pad, so the outside tires are fully up to temperature by the finish of that turn. The inside tires are cooler, but definitely not that much difference in a race car. Maybe fifty degrees worth compared to 400 F at the tread of the outside tires. At the track-out point, we are somewhere around 110 mph I suppose in a Formula Ford of those days and the upcoming turn complex is an off-camber left of about 100 degrees as I remember. But it's up a hill, so you want to carry as much momentum as you can into that corner while not pitchpoling into the desert as I had watched a couple of guys do at the most recent race. Good time for threshold braking done right. Lock up and you just spin into the desert flat. Don't use enough and you can put on a real show that would be on YouTube these days. Come out of turn two slower and... well, you lose, don't you? This was test day, so I was out there to find limits. I did.

As I rolled on threshold braking, the car began slowing dramatically of course and the tread surface changed color at the same time. On Formula cars, the tire being visible is an important benefit. I'm getting good braking when the right front, that inside tire relative to the corner, slows. No lockup, and I can't get off the brakes unless I want to get up close and personal with a Mojave Tortoise, but it damn sure isn't going the speed it should be. Rotating, but not being effective. I kept braking, extending the zone somewhat, and the other three tires brought me down to a workable speed for the corner. Finished the lap and pitted.

The tire was out of balance to a sensible degree, but not flat-spotted. We had it ... term escapes me, it's been years... but we had the tire guys grind it back into round and the tire was fine. More compound gone than that lap should have used, but tires are consumables.

The point is my braking force at the pedal had not been reduced. I was at threshold in the left front and given the set-up rear braking force would have been only slightly below the threshold. Yet the right front slowed. It did *not* stop. If the brake pressure is too high for the situation, the tire stops rotating. If it's too low, you don't get all the retarding thrust you want, but the tire damn sure doesn't stop rotating. A stable condition of turning without creating the drag it should. This was almost uncanny, because at the time I'd not connected shuttles and everted rubber and all that research with my own racing. But there it was. I figured it out that night.

Now why was it so rare in race cars and now it's so common that it's a thread on Rennlist? Because modern ABS on high performance cars takes the tire much closer to that peak of thrust, and the systems have four independent channels. They explicitly target (as we do on aircraft) the goal of each tire giving its best. A practical best. Not perfect, but as close to the peak as we can design the system to be stable.

How did we solve it? Well, in the heat, you have do just what people describe. Back off the pressure, get out of ABS mode briefly and then resume braking. Almost always the feedback loop will settle on the correct point of stability the second time around. Especially if the pilot/driver is more smooth in rolling on the pressure. As for fixing the systems, you must either be less subtle by half or a lot *more* subtle. It being space work we had to go up, never down, so they worked at it. [This being Porsche, their situation is similar. I can't see them going back to a less high performance ABS, a less subtle circuit that can't be fooled but because it doesn't try to get so close to maximum performance.] I quit being interested in design details once they did solve it, but the presentations I got spoke of multiple-sensor feedback. I inferred that they built a multiple-variable feedback equation. As a computer geek at heart, I'd say they found a way to discriminate between the desirable slip condition and the one that leads to rapid outgassing of the tread compound.

[I very much suspect the latest Porsche system has additional measures to recognize whether brake response at a tire is the best that can be expected. But it isn't an obvious distinction and takes very sophisticated design measures. Looking at that tire in my Formula Ford, I knew something was wrong, but even human judgment couldn't be sure it wasn't something extraneous causing the loss of brake power at that tire. Designing a circuit to make that distinction would be fun. Which is another engineer's term: fun=hard work=expensive. But so is the 911. Expensive, I mean. We can't afford space shuttle levels of design effort, but don't expect bargain brakes either.]

Cindy just woke. Gotta go.

Gary

ADias

Fascinating!

simsgw

I have time for a couple of more comments. Instead of skimming, I re-read as much of that thread on the other forum as I could handle. I got bored after eleven pages or so when people became seriously frustrated and began accusing Porsche of malfeasance and each other of everything from insensitivity to posting apologias for deadly design flaws.

First observation. Lots of different fixes were proposed. Many of them are things you bloody well ought to be doing anyway if you take a car this fast to a race track. Just ordinary care and maintenance items.

Second, a lot of those things (and a few others like setting the modified shocks too stiff) are things that can make it more likely the system settles into that secondary stable position. It's a tire phenomenon and anything using rubber tires can encounter this problem, but you can make it more likely when you start changing things from the factory configuration.

Other comments just got desperate.

Hitting the brakes too fast? Well, jeez. The system's primary role is to help 'civilians' who have a panic stop situation. An ABS engineer was on the thread and I almost picture him grinding his teeth. Hit the brakes as damned fast as you like. They won't fail and neither will the ABS. You won't be as fast around a track as you would like, but that's between you and the timer. Cars respond to smoothness, but the brakes will do their job.

However... if you also have done some of those other things, like running R compound tires, aftermarket shocks set up hard, and I don't remember what else, then you've created a fast transition for the tires from very low slip to very high. If you then hammer the brakes instead of rolling onto them smoothly, you shorten the time interval for the tires to climb that slip vs thrust curve. Road conditions they describe can also shorten the time drastically. As we all know, a tire near the limit that goes over something, or if the car unloads from a "vertical curve" as Taruffi calls it, then it loses some of the force holding it to the road. When we speak of g-loading in terms like the 1.1g a road tire produces, we really mean the thrust available is 110% of the force holding the tire to the road. If we are braking at even 0.8 g when the car unloads half its weight, then we're suddenly attempting 1.6g of braking. That ain't gonna happen. The tire locks up. Or it wants to. With a decent ABS, it won't reach lock up, but it will be approaching that very fast. Much faster than your own reflexes in moving from throttle to brake pedal and then depressing the brake pedal. All that takes perhaps 250 milliseconds with a fast driver using right-foot braking. The car (or just one tire hitting some irregularity) can unload weight in 50 milliseconds.

Okay, why does that matter? Because the system needs a finite length of time to respond to changes at the tire surface. When the tire moves too fast up that slip vs thrust curve, it is more likely it will overshoot before the basically hydraulic system can react. It can reach that position past the peak braking point the system is designed to approach. It reaches the far side, the dark side if you will.

Now the point I didn't hammer in that previous note, because my correspondent is a fellow research engineer who knows it quite well, is that a system that arrives at a secondary stable condition is by definition stable there. It hasn't failed. Its sensors are reporting inadequate data by our human terms, we don't want to run off that runway so our broader view has merit. But a system in this condition is stable by definition, not failing, not falling apart or likely to do anything you won't appreciate.

Let me put it another way. We don't put supercomputers that win chess games in automobile environments. Circuits rugged enough for that role don't have that much bandwidth at this state of the art. So this problem is like dealing with a good natured child. Within the scope of its understanding, it won't do anything nasty. For lack of understanding, it may create situations you must manage with your greater experience and judgment.

I can live with that and I think we all can once we understand the issues.

In this case, the problem isn't new and isn't unusual. Could I build a system to detect it and handle it without my attention? Probably. Almost certainly in fact, but I ain't cheap. And I don't mean my pay. I mean that I was used to working on multi-billion dollar design projects. I definitely wouldn't promise I could build a braking system at road car prices that could identify this phenomenon and keep it from reaching the level that drivers could notice.

If I were building a racecar in one of the classes that allowed custom ABS (which has to be very few since braking skill is so important on the race track), then [boring detail deleted after thinking about it. This isn't design class.] With all those simplifications, I think you could get it down to a five digit cost. That's a stationery budget, so we're talking really cheap as high performance hardware goes. Or you could lift slightly, come out of ABS and then roll the pressure back on. People go off corners... well, never mind the commentary. Doing that pressure cycle won't put you off the road. Watch the fastest cars in the world some time. It costs time, but not enough pavement to cause a crash.

I believe that in practical road cars, with dynamic behavior we tolerate in road cars, you're not going to see the second stable point arise -- on the road. On the track, you might push it that far. So lift pressure briefly and soldier on. It is us driving the car after all. Tires do that. You wanted to be a pilote, right? This is part of the job.

Oops. Gotta run. No time read this in preview. Hope it makes sense.
Gary

ADias

I wonder how much computing horsepower (MIPS) is available in current Bosch ABS systems. Probably a fraction of a decent i7 processor. With a better processor and decent 3D g-sensors it should be able to analyze a 2-state solution ABS feedback system. As Gary says it's only money but I guess its time will come. Traditionally, automotive pricessors lag the desktop by a very long time.

Thanks Gary for a thoughtful essay on a subject you know so well!

simsgw

Always welcome, Tony. Writing isn't work when people care to hear.

Gary

utkinpol

in all honesty i think Porsche went way too far with those computer assists, it went to the level where it becomes a distraction. at last 2 DEs my PSM very actively interferes where it should not have been, it starts pretty chaotically engaging front brakes right at 1st apex in lime rock`s big bend at merely 75mph which should be fine by all means and does same in west bend. as well as sometimes generates 'psm error' in big bend right in the middle. i turn it off now but boy is it annoying when your car starts braking and makes you slide right at the apex if you forgot to press this dreadful button.

same i bet is the issue with ice mode - it is just too much brains for its own good. i understand it makes no sense from engineering standpoint to blame speed of brake pedal application, but most people i spoke to and same as it was described in the GT3 forum thread - all folks say the same - you lift your front, then quickly slam on brakes and your ABS goes bananas. go figure why. what is really depressing is the fact that on same cars using same rubber - on one car it may never ever happen and other car may show this stuff on any track, any conditions. go figure why.

ADias

Gary explains above that ABS behavior is due to the wrong solution point being picked up by the ABS processor. A more sophisticated algorithm and a better processor would fix that. Re PSM... the PSM in the 997.2 gen is virtually perfect and non-intrusive; some say that if it intrudes the driver is at fault. There again PSMII is better due to better algorithm/processor.

Yomi

Hard to follow up on Gary's comments, but just wanted to add that this isn't new to GT3s, Porsche's, etc. Go to a Subaru forum and you can read about the STi's "ice mode" issues (also using big Brembos which give excellent performance 99+% of the time) as well as plenty of other cars.

I had it happen one or two times, on a car with adjusted suspension and R compounds, while braking hard on a bumpy surface. I remember one in particular because a fellow competitor chose that moment to explain to a new person how good the brakes on my car were. Just in time for me to sail right past the braking point. This is 1-2 times out of literally thousands of corners going in hard with ABS.

Afton Thomas

Sounds like you are having an early braking intervention of the ABS System,am i correct ?

Afton Thomas

I have had a number of issues relating to this black ice effect and normally these causes are the fact that the compound of the Tyres are aged,cracked,hardened or expired.I am interested in what you have tried to solve this particular issue ? Looking forward to your views on the Black Ice effect !

MICHAELWWW

I enjoyed the part about torpedoes and U-boats . Maybe someday I will understand

MagnusB

As far as I'm concerned this is not new to the 997's either. My 996 had the same "issue".
The braking is more effective just before ABS kicks in.
Have seen it so many times while doing AX, braking after the finish line.

equ

It happened to me once in my 06 Cayman S... Braking hard on a very bumpy off-ramp there was a second or so where the car oddly seemed to glide over bumps but not slow down. Definitely felt like the ABS system was being confused the by the bumps.

Yes, I was hooning on that particular off-ramp and I was very alone. I've experienced it only once in over 40k combined porsche miles (33k on that cayman).

The only thing similar that I've felt is the normally excellently tuned bmw DTC system (akin to porsche's PSM) getting confused on a bumpy, 90-degree turn once. It actually swung the car slightly the wrong way and required a save on my part. First generation run-flat tires + bumps = bad.

alexb76

+1.

He either has a problem with his ABS, or has been driving very poorly to engage it in such a fashion.