noise."Rationality and Circuit Checks · 2:05am Dec 19th, 2017
Failure Management
We've got a two-parter here! Mostly because it occurs to me that there's some groundwork that y'all need for the Intrepid to make any sense.
Unfortunately, this one's gonna be heavy on theory, which means that I don't have too many topical pictures. Luckily, Present Perfect and I are having a little, uh, disagreement at the moment, and so I happen to have lots of batpony pictures cued up.
EDIT: link to said disagreement
Back in the early days of onboard computers, they were really simple, and they didn't do too much. According to the service information, the computer in my 1988 s-10 can monitor up to sixteen different sensors, and it can control fuel, spark, and a few other things.
Note the "up to."
As far as I know, it monitors manifold pressure, engine temperature, engine RPM, vehicle speed, knock, and the single oxygen sensor, and it makes its calculations from there. It would also monitor power steering pressure and whether or not the AC is on, if the truck had either.
And they aren't so much calculations, really. The computer has a set of lookup tables, and it decides what to do by looking at the table. Say the engine is at operating temperature and I'm mid-throttle at cruising speed with light engine load, it's got a chart for that, and it decides timing and fuel delivery from that. The knock sensor tells it if it got the ignition timing wrong, and the oxygen sensor tells it how it did on fuel delivery.
One thing that the computer can do that the old mechanical systems couldn't is adapt itself. So if anything wasn't as good as the computer wanted it to be, it could make corrections. Say that the oxygen sensor says the exhaust is a little rich . . . it'll modify its strategy accordingly, and next time it won't give quite as much fuel.
It is capable of learning over time, and for fuel it has two parameters—long term, which is the constant correction, and short term, which is the instantaneous correction. Eventually, short term becomes long terms in that particular load cell.
If things get too far out of whack, it turns on the check engine light.
Modern cars do the same thing, essentially, although they monitor a lot more parameters, and there's a lot more that they can adjust.
Now, assuming that the computer was programmed correctly, it's plain to see how this allows the vehicle to adapt to the owner's driving habits, and how it can adjust for wear, or unexpected load. The car will know if it's summer or winter, and it also knows roughly what altitude it's at, and can make adjustments accordingly.
Of course, things don't always go like they ought to go. Sensors wear out, and things go wrong with the wires leading to the sensors.
Back in the old days, when the computer monitored up to sixteen different inputs, they were really bad at figuring out that something was wrong.
For starters, the computer didn't have feedback when it did something. My s-10 had an EGR valve that was computer controlled (instead of a thermal switch). When it was appropriate, the computer activated a vacuum solenoid, and the EGR valve opened, probably.
If it didn't, the computer had no idea. As far as it was concerned, it sent the command, therefore something must have happened, right?
Likewise, it didn't have rationality checks. For those of you who wonder what that is, that's the question of whether a particular sensor's operation is logical.
Old oxygen sensors switched between roughly 0V and 1V (we'll get into more detail on that later). The ideal mixture was around 450-550mV; that was the computer's target.
When the sensor in my s-10 went bad, it stuck at about 450mV, and the computer was happy as a clam. No matter what happened, as far as it knew, the exhaust mixture was perfect. I bet it was patting itself on the back and thinking about what a great job it was doing, when in reality it ought to have been wondering how it was possible that every single load cell in its system had zero long term and zero short term correction. That doesn't even happen when the car rolls off the assembly line.
But the PCM had no way of knowing. It had no way to check itself.
Finally, a common failure on older GM vehicles was the coolant temperature sensor. It would get grounded, which gave a temperature reading of -40F (or -40C, for you metric people). Unless it was really that cold, the vehicle wouldn't start; it would flood itself out. But it never knew that it was doing something wrong. It had no choice but to trust that sensor.
It was fairly obvious to everyone that this was not a satisfactory situation. A computer needed to know if it was doing the right thing, and it needed to know if its sensors were lying to it. Plus, for reliability, it needed to have a way to manage if something went wrong.
[Ironically, that's one of the reasons why computer controlled cars got a reputation for unreliability—the computer would do its best to keep the car running well until things went so far out of whack it couldn't do it any more, and then it would fail 'instantly.' (Never mind that the Check Engine light had been on for six months; obviously if the car still started that light was a mistake.)]
Basically, once the computer got good at self-diagnosis, it also had failure management modes. Those are the subject of the next blog post; first, we're gonna talk about how the computer knows it's got a problem.
The computer needs to know if a signal can't be trusted. First, it needs to know that the wires that go from the computer to the sensor are good, and not shorted to power or ground.
The way that that's usually handled is by having the sensor not be able to travel its full range. Most sensors are on a 5 volt reference, and if you make their sensible outputs be from .5V to 4.5V, that solves the problem of shorts to ground or voltage: if the throttle position sensor (for example) is reading 0V or 5V, the computer knows that it should be ignored. It knows that the throttle isn't more than 100% closed or more than 100% open.
Amusingly, I once worked on a Blazer where the ambient temperature sensor (that measures how warm it is outside) could read -40F as its low point . . . and +235F as its high point. Honestly, if the outside air is +235F, you've got more serious problems than an inefficient A/C system.
Oxygen sensors won't read all the way down to 0V, or up to 1V—if they do, they can't be trusted. In fact, some automakers (Chrysler in particular) bias the ground voltage, so the sensor reads from 2.5V to 3.5V, which helps eliminate sloppy grounds causing engine performance problems.
Secondly, computers now get feedback from their sensors to see if what they asked to happen actually did. Rather than have a fairly static fuel mixture, the computer varies it up and down a little bit, and it watches the oxygen sensors, to make sure that they saw it. When it opens the EGR valve, it expects to see the oxygen sensors change—if they don't, than it knows that the valve didn't actually open.
Some circuits can be tested by sending a small amount of electricity down them and the computer will watch to see if that comes back. If not, there's a problem with that circuit, and it can't be trusted (this is normally only done with safety-critical systems, like the airbags or the brakes).
On startup, it also monitors a bunch of stuff. These days, cars tend to have intake air temperature sensors and coolant temperature sensors (and transmission fluid temperature sensors, and Chryslers have battery temperature sensors, sometimes, too). If the car's been sitting for a while, it's obvious that all of those should be close. Maybe not exactly the same, but the computer knows that if the IAT sensor and the TFT sensor both say it's 60F, and the CTS says that the coolant's -40, the coolant sensor has gone rogue.
While the vehicle is running, there are additional rationality checks. If the crank sensor says that the engine is turning, and the cam sensor says it isn't, the cam sensor is probably broken . . . or else the engine has just suffered a critical failure. Other sensors can confirm or deny the critical failure, if need be.
Vehicles are also good at remembering things now. They didn't used to be; for the computer in my s-10, every day was its first day, as far as it knew.
But now the car is smarter than that. It knows that the outside temperature can't swing too far in one day, so it's going to be mighty suspicious if it was 100F outside when it was parked and it's -40F the next day.
It also knows that when you shut off the engine and a few minutes later it's got more fuel than it did before that you probably just put some in, and it'll run an evap gross leak test right then, and if it fails, it'll turn on the Check Fuel Cap light (not all vehicles have this).
For that matter, it knows how much fuel it should be burning, and if fuel usage changes a lot, it's going to be very suspicious of the fuel sending unit.
Truth is, we've barely scratched the surface here. There are test your car carries out while it's running—some of them, all the time (misfire monitors, for example); other tests only when certain conditions are met (5 seconds of closed-throttle decel, for example).
Heck, there are some tests that the car only runs when you've left it alone for a while. Chrysler is famous for checking its oxygen sensor heaters like that—after the car has sat for a few hours, the PCM wakes up and turns on the heaters and watches to see if the oxygen sensor readings change as expected, then it takes note of the results and falls back asleep.
If it fails the test, the next time you start the car, you're going to be greeted with a check engine light.
Before I wrap this up, I should also say that there are different types of codes. Some of them set as soon as a failure is detected, while other codes might take more than one trip to set a code. Generally, codes will go away on their own if the vehicle has a certain number of failure-free trips; depending on what the fault is, that could be a few, or it could be a lot.
Certain things are checked every time the vehicle is started, and that's when the computer makes a go/no go decision. Wheel speed sensors for ABS are often like this—the computer can't decide if they're good or bad until the vehicle starts moving. It'll remember which one failed, but it won't disable the ABS on a new ignition cycle until it sees that failure again.
Next week, we're gonna apply what we've learned on a Dodge Intrepid and a Ford Mustang. Yay case studies!
I can't tell you how many times I've had to replace O2 sensors on US made cars. It got to the point that I was seriously tempted to wire up a "Everything's perfect" widget for them.
My Honda, not so much. Just hums like a sewing machine.
By the way, you forgot the computer on the VW Diesels (and some other cars) that have a "Oops, I'm on a dyno, so I better change my performance specs to absolute squeaky clean even if it makes my acceleration suck." Personally, I thought that was a perfectly valid mode. The EPA set the rules, VW followed them exactly. To the letter. Every. Letter.
I like batpones. Batpones are cute.
You need to link to that disagreement, sir. :V
I have nothing more to add because I am listening to my friend recover in the background after puking his guts out. Fun!
4752632 I concur and am curious to know the fight. Also, good thoughts for your friend, PP.
16 sensors. 16 maps. Each one is 16 by 16 of 256 points? Say 4K, Rom. and for ajustable? same again. 4k Ram. you can go for unified space, 64K, or Seperate pages. You can use Vectored Interupts, which is where the 16 comes from.
Strange, Because if you use the 1976 available Z80 that was liked for miniframes, home PCs and spacecraft, you could have 256 interupt vectors with direct table lookup on the data bus?
Maybe I just simplified too much aso am missing something really basic and important.
Personally, Id just get a mechanical diesel, put a cyclonic filter on the back end, and a oxygen concentrator on the front, and use it to drive a generator with a triple cell battery pack and propulsive electric motor. Then run it on sunflower oil or renewable crude or rendered zombies, whatevers available.
Some nice ideas there. Even better if sensor wires acted as data ransmission, so when the CPU pin goes high, it measures the current draw, and then uses pseudo random pulse density modulation. I mean, if its good enough for natures nerve cells and its so simple and cheap to do?
4752613
They got a bad rap on early domestic models (and to be fair, they did go bad a lot). Now that there's better management by the computer, they live a lot longer.
The everything's perfect widget is a good idea when everything is perfect, but when it's not . . . well, sometimes shaking bits of melted catalytic converter out of an exhaust system is actually the least of your worries.
Well, I could write a whole 'nother blog post about sneaky computer programming, or faults that the computer should have seen that it didn't, or faults that didn't exist that the computer found, starting with the notorious GM ABS code caused by the blower motor spinning down--electric motors, when spun, make voltage. You know that, I know that, GM forgot about that.
4752615
Totally agreed. Somebody, though, doesn't like batponies, and I'm trying to bring him over
to the dark side.4752632
Done!
Not cool. Hope he's okay.
4752642
It is now linked in the blog post.
4752651
I'm not sure how many maps/lookup tables each one has. It might vary by model.
Yes, although consider production cycle (engines have to be designed around what the computer can do or modified to work with it) and durability. You've got a computer in a sealed box that's supposed to work reliably for twenty years or 200,000 miles with no servicing whatsoever, in an environment that not only vibrates and has variable voltage, but also ranges from something like -40F to 300F. It has to be able to get wet, and be impervious to all automotive fluids and cleaners which might come in contact with it. Strap your computer under the hood of your car and see how long it lasts for.
It's a really challenging environment for a computer to live in.
That's an option. You trade some performance for that kind of simplicity, and that's not what most people want. Some people do, though, and if you're willing to live with the trade-offs, I'm not going to be the one to say you can't do that.
A lot of stuff has to be brute designed, because connections get iffy due to vibration and corrosion, grounds stop working, some damn fool jump starts the car with the cables backwards, etc. Heck, I've seen a PCM that still mostly worked even though it was full of hot coolant (like, we literally poured it out).
4752666
4752632
H-how long have you two been doing that...
I'm starting to feel like you should write a sequel to Silver Glow's Journal featuring Silver Glow's daughter, the first Equestrian car mechanic.
4752700
Two weeks, at least.
And I've already got ones picked out for Christmas, so we've pretty much gotta keep going.
4752702
Well, I do have Poppy in two of my stories. Not quite the same, but. . . .
I never knew that. THE MORE YOU KNOW.
Not entirely related to all this, but I thought I'd share. 03 VW Jetta, belonged to a long time customer who recently sold it to us. The trans computer fried a driver, replaced the computer. Now the radio doesn't work. No worries, this happened before, when we rebuilt the transmission. VW in their infinite wisdom, decided they needed a fuse on the back of the radio. It was blown, like last time. Pop the radio back in, turn it on, everything is ok, right? Wrong. Turn off the ignition, the radio decides it wants to stay on, and that it should illuminate itself. Now, VW, in their infinite wisdom, thought that instead of having hot-always and accessory positive connections, they would instead have only a hot-always, and wake-up the radio via data bus.
So the radio knows to wake up and also illumination is controlled by data, and it works just fine while the ignition is on. So data works, but the radio isn't getting the go-to-sleep data. As the radio has no accessory wire, I'm unsure what to look for. I just poke all the fuses, and find a 5A burned. Turns out, it's the keep-alive power for the instrument panel... and for the transmission computer. Dammit Fahrvergnügen!
4752651
I'm not entirely sure what you're suggesting, but there are certain sensors on certain cars that minimize wiring by using only 2 wires for data, and taking power right off the data lines. Kinda like certain things like phones and security cameras can work on only an ethernet connection. A lot of airbag crash sensors are like that, 2 wires, data sends a wake-up signal, and they continually talk back and forth with the airbag computer. It's especially good because the computer knows the exact millisecond it stops talking back.
My poor, innocent eyes! Too much lewd!
I move forward a motion to forbid any other comments to be added to this, so that it would remain untouched as a monument for the future generations.
A monument to what? I am not sure, but a monument still.
A couple of week ago, during a morning meeting at the store, I presented a new product that already had bad review from customer that did not use it properly. I concluded my exposé by "As you know, customers' intelligence can vary."
I stand by my claim.
Well, I finally had my last exam for this session, I was NOT ready for more last minute studies, I may fail that ford mustang test next week.
4752700
save me D:
4752752
Actual picture of me when you ask for my help
i.pinimg.com/originals/9c/f5/bf/9cf5bf229bb897cc28fdb881fbdd7b56.png
4752663
I remember that one! We had a... '93(?) Cavalier that always had ABS issues... and after 3 years AC issues (every summer queue a $300 repair bill on it)
I'll stick with mechanical injection diesels, manual gearboxes, nothing made after 1994 and nothing that isn't badged "Ford" for personal use. Most gassers, autos and other manufacturers have failed me spectacularly.
4752780
REEEEEEEEEEEEEEEEE
So wait, why does PP hate batponies? I never stuck around long enough for him to stop alluding to his hatred and start explaining it.
4753031
I dunno. Maybe one of them beat him up, or maybe they owe him money. Maybe he doesn't like tufted ears or leathery wings.
4753045
I mean, I know he thinks we shouldn't fuck them, but I don't know anything beyond that.
4753068
Like, just batponies, or ponies in general?
4753070
Obviously just batponies...
4753070
Just the batponies. 'Never fuck the batpony', I believe he says.
4753074>>4753077
Well, that's just silly. Batponies need love, too.
Wow, this is the first time I knew everything already, that's a first.
5V is really annoying, most modern microcontrollers are 3.3V and some are lower. My last mechatronics project had a bunch of 5V stuff. Had to add a bunch of dedicated ICs to do level shifting. Made the board bigger (and pricier) and more difficult to route. Not to mention that it cost an extra $10 or so.
Echo-the-Bat-Pony-6
She will always be Batsy Fluffentuft the Magnificent to me. Best batpony.
4752719
Wait until I talk about the Intrepid. The MIL was on and flashing for over two years.
Ford traditionally did their small leak test overnight (I think Chrysler does, too). That was a problem for taxis in places that checked emissions, since the car was never shut off long enough to run the monitor, and it of course would fail an inspection for a "test not run" flag.
Chrysler also did that on the 95 Jeep Wrangler, and possibly others. I've replaced one of those fuses before.
I've got a Chrysler that was doing a similar thing, due to a bad driver's side door latch. Shut it off and take the keys out, radio keeps playing until the Accessory Delay Relay finally shuts it off.
One of the local cop cars would not go into power saver mode until the battery was too weak to restart the car. Luckily, the police equipment installer was the one who had to fix that (it was some bit of cop equipment that caused it).
And on the plus side, your VW wasn't new enough to have the
laserfiber optic ring network on the infotainment system. It's a good idea, except the way they wired it, one component goes down, they all do.I think he's thinking like the data lines already work (2 wire bus, high/low pulsed signals). But maybe not. . . .
4752745
Best not click on this, then.
derpicdn.net/img/view/2016/9/27/1259362__safe_artist-colon-rhorse_artist-colon-somewritefag_oc_oc-colon-corona_oc+only_bridle_collar_colored_earth+pony_female_harness_mare_pony_pony+p.png
4752746
We can't forbid all comments; PP and I are still going at it. We've got to make it to Christmas, at least, 'cause I've got some Christmas batpony images set aside for that purpose.
That is very true. On more than one occasion at the dealership, I had to 'fix' a vehicle that the customer was just using improperly.
I'm not gonna make y'all guess what went wrong, but y'all got to know how the computer thinks and reacts in order to fully understand why it did what it did.
4752817
Cavaliers were pretty much disposable cars, anyway.
I don't know what year it started (I think it was later than '93, though), but there was a recall to install a new relay and some wiring to the starter to avoid "thermal events."
4752821
Here's the two vehicles with the most miles we've seen in our shop:
A 1990 Chevy 1500, with over 1.3 million miles on the ODO. Fuel injected 350ci.
A 96 Dodge Ram with 550k. Cummins inline 6 (which is a great motor).
The one crappy thing about the mechanical injected diesels is that if the injector pump goes, that's a big expense all at once. Individual injectors on newer common-rail diesels are much cheaper.
Manual transmissions . . . I'd stick with the older ones with cast iron cases, if possible. The Saginaw 3-speed side-load trans I had in my '78 Chevy was nearly indestructible.
If you live somewhere that there's lots of road salt, Ford had issues with suspension pieces rusting off the frame (spring shackles and mounting brackets, radius arms and brackets). Otherwise, they're not bad trucks. I'm a GM guy, but I wouldn't be opposed to owning one of those.
4753333
I'm assuming that's because that's a common output voltage for lithium batteries?
Yeah, I don't know why they went with 5V, except that it was probably the early 70s that they came up with that as the standard, and they've just used it since.
There was talk of going to a 48V system on cars, but so far that hasn't been done. I think some of the standards for it have probably been written, though.
4753693
Batsy Fluffentuft is super-adorable.
4753732
Oh, that's not so bad. No bit.
4753742
You make valid points. However, most injection pumps that I've worked on that have failed have either 1) stripped a gear. 2) suffered shaft failure. 3) suffered major seal failure or 4) the fuel meter valve stuck shut. Rarely ever do I have to replace an entire pump. On average, it costs a customer two to three times in parts alone for an injection system failure on a common rail engine. Stars forbid that the HPOP fail on a Cat
As for iron case transmissions, I'm part-and-parcel with Rockwell 9-speed, Eaton-Fuller 10-speed, 13-speed and 18-speed transmissions. They hold up the best out of the bunch. But, I've got to give an aluminum case its due; the ZF5-42. It's pretty tuff for what it is. Other than that, the T-18 and T-19 transmissions are my go to top-loaders.
You are right that Fords can have rust issues in high salt use areas. I don't know about rusting off, as I've never dealt with that. Then again, I tend to buy a truck, then build what I need. Everything gets reworked to my specs.
Can't buy? Build it! That's my logic.
4753738
My mom's lasted from 93' to around 2010. The hood rust was pretty bad by that point, if you've worked on it then you know the spot. I think every single one of them for about 5 years had it in the same spot on the hood.
4753735
This is what we call "An exception that confirm the rule".
4753745
5V is a standard logic voltage (along with 3.3V and 1.1V) so it's not out of no where. But 3.3V is the standard for newer microcontrollers (i believe) and you can find 1.1V ones. Lower voltages consume less power and generate less heat (especially at higher frequencies), but you need tighter tolerances.
probably not for the microcontrollers though. That'd be a really power hungry.
4753724
Oh, so the damn car didn't know you opened the door for it to shut off the radio. I wonder if you could just pull the relay, or is it required to use the accessories at any time?
Never seen that in a VW. But have worked on that in a BMW. Luckily, the scanner revealed communication codes for the 'entertainment bus' or some such nonsense. At least tracing faults is linear, just follow the line of working stuff until you get to the thing that doesn't work, the amplifier had taken a dump. Interestingly, they have a fiber optic loop in the glove box... for whatever option the peasants didn't get, I suppose. You can order those on ebay for like $5 or $10, and plug it in place of the broken computer, to just bypass it on the data line. But to test it, I very carefully removed the 'pins' from the fiber optic plug, and held them together end to end, and watched as the radio started working.
Luckily, the amplifier on that bmw only controlled the back speakers, so the customer didn't care that it wasn't there.
No machine should ever be intelligent enough to guess wrong and confuse itself...
That just puts them up against the stupidest users, and they have far more experience at it.
4753824
I suppose it only gets kinky when the bit gets added.
derpicdn.net/img/view/2017/10/23/1568058__safe_artist-colon-aerthmanolo_starlight+glimmer_sunset+shimmer_equestria+girls_spoiler-colon-eqg+specials_beanie_bit_bridle_crossed+legs_digit.png
4753855
I've done one, on a 6BT. Whole pump. Not sure if it was rebuildable, since that's something that we don't do.
I've heard horror stories of them failing, but to be honest we don't get enough diesels at the shop for me to know if that's just hyperbole or if they really do fail somewhat often.
It really depends on what fails, and how, at least in my experience. But that's also something we haven't done too much of. We had to do all eight injectors on a 6.0L Ford, and those sold at about $250 each, so yeah, it was a fair chunk of change for that beastie. IIRC, it also got some other internal engine repairs (which were partially the cause of the injector failures).
GM apparently had some issues with the ones in the Duramax truck.
I've got a Muncie SM465 out of I don't know what, and that thing's built like a brick s
house. I don't think you could break it, to be honest. I know my much lighter duty Saginaw side-load was still driveable after ingesting the front main bearing and fracturing the nose shaft. It was kinda loud, though. 
I think it's less of an issue on the heavier duty trucks (250s and 350s). They might have used better steel, or rust-proofed it better. But I can't tell you how many shackles and mounts I've done on F150s, Explorers, and Rangers.
As long as the salesperson charges right, it's not a bad job, actually.
It's not a bad idea if you've got the time, skills, and tools. I'm actually toying with the idea of putting a 4bt into a S-10.
4753876
Around here, most of them failed almost a decade ago. They were common as dirt when I started as a mechanic, and then around 2006 there weren't any more of them. I think that they rusted up from the bottom, mostly.
It's actually kind of weird, the cars that we used to work on that were really commonplace and now there aren't any of them to be found.
Ford Contour, Mercury Mystique, Plymouth Breeze, Chevy Cavalier, Corsica, and Beretta, Honda Civics with the reverse-mounted engine. . . .
4753922
That's true.
Aside from a few other commenters the first few days, it's just been him and I. And it looks like he's not going to surrender anytime soon, which means that I'll get to send him the Christmas pics.
4753956
I could see tolerance being an issue; on a car, you've always got slightly corrupted signals. And I'll be honest, I don't know what voltage the microcontroller runs at; it could be something else. It probably isn't, because I bet it would cost more for a computer that had to run lots of different voltages (they already have to do 5V and 12V [sort of; I don't know if it counts as 12V when you're only grounding the circuit]).
It's also possible that some other computers on the car use different voltages for internal processing, depending on what they have to do. Engine sensors are generally 5V; HVAC equipment might not be (for example).
No, just for the main power supply.
IIRC, (and don't quote me on this) the FCIM module on Ford diesels is 48V. I can't remember if that's because the glow plugs need that, or if it's for the injectors.