Reliability and Redundancy

[murpia]

I read with interest that the 2 leading cars at FSUK, Delft & GFR, failed endurance through electrical issues.

I understand GFR had a cam sensor issue, and Delft a starter relay issue. It surprises me that these things can be car-stopping faults.

Firstly, once an ECU is synced to the correct engine cycle, a cam sensor is no longer required. So why should it's loss stop the engine? In fact I know of a number of race engines that don't have cam sensors, and instead use other methods to sync engine cycle. Maybe this is a question that can be directed at typical FSAE ECU suppliers, to improve the hardware on offer?

Secondly, it's common practice in race cars to parallel up critical systems to add redundancy. Fuel pumps & relays, ignition power relays, etc. I'm sure Carroll Smith has something to say about it. With a separately fused power feed a chafed wire that kills one circuit / fuse can be replaced by another. Clearly Delft saved some weight by not paralleling up the starter relay... What level of redundancy do you build into your cars in the quest for reliability? How do you trade it off against complexity and weight?

Thirdly, I'm sure there are a many more clever ideas for reliability & redundancy out there as well as those I've mentioned. Anyone care to share?

Regards, Ian

[Pennyman]

I know a few teams that swear by a mechanical backup to their pneumatic shifting mechanism.

Personally, I've machined snap-ring grooves into bearing bores that were a press-fit size, just in case the bearing starts to walk out under axial load...

We've also used locknuts to retain our spindles, and then put a cotter in through the spindle anyway.

[Neil_Roberts]

With rare exceptions, none of the race cars that I helped design had any functional redundancy. On the other hand, all of the airplanes I've designed, including unmanned ones, had a backup for every critical function that could be backed up. In every case I can think of, the value of redundancy in terms of both weight and cost made those choices the right ones.

Failure tolerant design is always lighter and cheaper than redundancy. For example, a few lines of code can produce automatic switching for continued running after either the cam or crank position sensor fails.

If you are going to design in redundancy, make it dissimilar redundancy. For example, use a mechanical backup for electrical or pressure actuators.

[Demon Of Speed]

We run very little to no redundancy systems because as stated above: cost and weight. Some times it is better to spend a little more on the first/only thing then to add a backup. Weight is a huge issue, when you take the time to gun drill a bolt to 0.030" wall and chamfer the hex (going from 0.18 lb to 0.06 lb) it doesn't make scene (to us) to then add in an extra 0.1 lb starer relay and extra wiring that you will hopefully not need. If these systems were life threatening or there was a min. weight to the car then this would be a different story.

[MegaDeath]

I would rather have a 500lb car that would finish endurance 90% of the time than a 400lb car that finishes endurance 70% of the time.

That's just my opinion. But a fast car is going to be fast weather it's 450lbs or 475lbs. So in my opinion I would MUCH rather add 25lbs to the car and know that it will be reliable than slim every aspect of it and cross your fingers that it will keep going.

[murpia]

I'm not advocating adding 10kg for redundancy, but 2->3kg should be imperceptable. I bet your drivers vary in weight by that much anyway...

And software is zero weight.

Regards, Ian

[EHog]

I guess this year we needed a redundant powertrain...

Anyway, I believe redundancy in things such as the electrical system are tough to pass up. Anything that could be a critical failure but can become redundant for such a small cost has to be a good idea...

On that note, there is nothing that is redundant on our car...

[JasperC]

To be honest, we never even considered adding a second starter relay for redundancy, because the starter relay is a part we didn't expect to fail in the first place. It usually lasts a lot longer than the 1200km we drove before our failure at FSUK and it is a part that we have never seen fail before (and we're using it since 2004).

At Michigan 2008, we had a DNF due to a camshaft sensor wiring failure, just like GFR had now. After that, we have made the camshaft sensor redundant after the engine has run for a few seconds. I agree that is something you should always do because wiring failures happen relatively often when you're running a single cylinder (although I must say the build quality of the wiring harness can prevent A LOT).

But we didn't leave out a starter relay just to save weight. It's more a matter of not willing to spend any time on something that is already very reliable (and our electronics guy does spend a lot of time on the production of the wiring harness!).

Actually the only redundant parts we have that I can think of right now, are some buttons for shifting. Despite that, in 1600km of driving we only had the failure at FSUK and we had to replace our sprocket adapter because it had fatigue cracks. Apart from that, the car just kept driving and driving and driving. We have never had such a reliable car before.

Murpia, just out of sheer interest, what team are you from?

Kind regards,

Jasper Coosemans
Chief Drivetrain 2009-2010
DUT Racing Team (Delft)

[Demon Of Speed]

I would like to add that a "redundant" "starting system" is implemented for the endurance event to make sure the car starts after the drivers change. What it is, is confidential however.

[MH]

Originally posted by Demon Of Speed:
I would like to add that a "redundant" "starting system" is implemented for the endurance event to make sure the car starts after the drivers change. What it is, is confidential however.

Well if you want to bypass the electronics and make the system independent, then it doesn't take a genious to figure out you want to implement a manual "kickstarter" which the driver can operate http://fsae.com/groupee_common/emoticons/icon_wink.gif

[TMichaels]

It seems a bit odd that nearly everyone is talking about redundancy in electronic systems. If done properly, (burn ins, iron birds etc.) electronics are extremely reliable. We never failed endurance because of an electronics failure and we had and have a lot of this stuff on board. Nothing saves you from a production failure or wrong data sheets, but that does not really occur often and can usually be overcome by testing.

Regards,

Tobias

[murpia]

Originally posted by TMichaels:
It seems a bit odd that nearly everyone is talking about redundancy in electronic systems. If done properly, (burn ins, iron birds etc.) electronics are extremely reliable.
I'm not. I'm interested in opinions on electrical and software redundancy. I agree that electronics are normally reliable. It's wiring looms, connectors, relays, motors & sensors that are not (in my experience).

Jasper, I'm not from a team, I'm too old for that... I have been a Design Judge for some years at FSUK and have also presented at IMechE's Learn to Win on Testing & Development.

Regards, Ian

[Thrainer]

Originally posted by murpia:
... Clearly Delft saved some weight by not paralleling up the starter relay... What level of redundancy do you build into your cars in the quest for reliability? ...

As far as I know, bikes don't have two parallel starter relays and I don't know of any FSAE team that uses two. Clearly, most racing series have a minimum weight, so it's much easier to justify "unnecessary" weight.

We had a lot of redundancy in our electrical drivetrain this year, because we chose a symmetrical layout with two battery boxes, two controllers and two motors. In testing, one motor broke and we could continue with only one left, at a slower pace. We can also run with one battery box connected while keeping the pace, as you could see in Germany. Our vehicle software has a modular architecture and the driver can switch off individual functions like traction control and torque vectoring.
In the future, we'll try to give the driver even more information and control over the car, because it's difficult to handle all possible failure possibilities in the software, but the driver can react creatively.

Concerning reliability, I'd prefer finishing three endurances in the top 5 and failing one, rather than finishing all four endurance events and not reach top 10 in any of them.

Regards,
Thomas - AMZ

[Demon Of Speed]

Originally posted by MH:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Demon Of Speed:
I would like to add that a "redundant" "starting system" is implemented for the endurance event to make sure the car starts after the drivers change. What it is, is confidential however.

Well if you want to bypass the electronics and make the system independent, then it doesn't take a genious to figure out you want to implement a manual "kickstarter" which the driver can operate http://fsae.com/groupee_common/emoticons/icon_wink.gif </div></BLOCKQUOTE>

The "kickstarter" idea had been thrown out there, however never used.

[exFSAE]

In my experience, all of our part failures or electrical issues or whatever... really came down to inexperience in design and fabrication. Solution would have been better knowledge transfer, design reviews, etc.

When your front hubs shear in half the first time you use the brakes... not much you can do about redundancy there. Just need better design (and in this case, FEA element sizes NOT larger than the smallest feature dimensions).

Electrical gremlins were anything from voltage regulator getting no airflow and overheating, to poorly built harnesses and electrical connections.

[PeterK]

Firstly, once an ECU is synced to the correct engine cycle, a cam sensor is no longer required. So why should it's loss stop the engine? In fact I know of a number of race engines that don't have cam sensors, and instead use other methods to sync engine cycle.

I find it interesting that in a post about reliability and redundancy you would remove one of the key ways for an ECU to check for reliable control. If you are running full sequential injection/spark and don't have the ability after first start to correct for a bit of noise on your crank trigger than you are going to be SOL pretty quick. All it would take is one voltage spike and your timing would be off a whole tooth... on an F4i with a stock crank wheel we are talking about 30 degrees of extra advance!

I have to agree with Tobias, if you do a good job with your wiring design and build then you shouldn't have any major issues. When I joined my team the wiring on the car was a mess, mostly because they didn't have someone with enough electrical experience on the team.

Sometimes things fail, but I don't think relays fail often enough to double up on them for our team.

[wagemd]

Originally posted by MegaDeath:
I would rather have a 500lb car that would finish endurance 90% of the time than a 400lb car that finishes endurance 70% of the time.

That's just my opinion. But a fast car is going to be fast weather it's 450lbs or 475lbs. So in my opinion I would MUCH rather add 25lbs to the car and know that it will be reliable than slim every aspect of it and cross your fingers that it will keep going.

I would say it's about half educated optimization and half gamble. You said you would rather the 500lb car in this case. Would you also rather a 600lb car that finishes 94% of the time? or 700lb car that finishes 96% of the time? Its about choosing that cutoff point. I think some redundancy (possibly replaced by proper/proven design) is wise in some cases.

It also depends on your design philosophy. Just finishing endurance might ensure you get at least x far into the standings, but not necessarily first. For example, you could build a extremely fast, minimally redundant car that is theoretical designed well enough and IS "fast" enough to win the competition outright but has a 50% chance of complete failure and doing very poorly that year. On the other hand, you could could build a 100% redundant car that weights 800lbs and is almost guaranteed to finish, but I am going to go out on a limb in saying the design judges wont be impressed with the lack of risk taking (that's what racing IS). Not that either of those extremes are the best way to go, but are examples of very different design philosophies.

In the end, its completely up to the team leadership to decide what they are willing to risk.


Originally posted by murpia:
I'm not advocating adding 10kg for redundancy, but 2->3kg should be imperceptable. I bet your drivers vary in weight by that much anyway...

And software is zero weight.

Regards, Ian

This might be true, but its really easy to say "just a couple kg (or even g)" here and there and by the time the car is finished, you are 20 kilos overweight and scratching your head.

Driver weight certainly isn't something to ignore. We baby them enough as it is, put them on a diet and make them earn it http://fsae.com/groupee_common/emoticons/icon_wink.gif

Even if Bobby is your fastest driver and 30 lbs overweight, he will be even faster fueled by carrot sticks.

Everything IMHO...

[Brian Barnhill]

Originally posted by wagemd:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by MegaDeath:
I would rather have a 500lb car that would finish endurance 90% of the time than a 400lb car that finishes endurance 70% of the time.

That's just my opinion. But a fast car is going to be fast weather it's 450lbs or 475lbs. So in my opinion I would MUCH rather add 25lbs to the car and know that it will be reliable than slim every aspect of it and cross your fingers that it will keep going.

I would say it's about half educated optimization and half gamble. You said you would rather the 500lb car in this case. Would you also rather a 600lb car that finishes 94% of the time? or 700lb car that finishes 96% of the time? Its about choosing that cutoff point. I think some redundancy (possibly replaced by proper/proven design) is wise in some cases.

It also depends on your design philosophy. Just finishing endurance might ensure you get at least x far into the standings, but not necessarily first. For example, you could build a extremely fast, minimally redundant car that is theoretical designed well enough and IS "fast" enough to win the competition outright but has a 50% chance of complete failure and doing very poorly that year. On the other hand, you could could build a 100% redundant car that weights 800lbs and is almost guaranteed to finish, but I am going to go out on a limb in saying the design judges wont be impressed with the lack of risk taking (that's what racing IS). Not that either of those extremes are the best way to go, but are examples of very different design philosophies.

In the end, its completely up to the team leadership to decide what they are willing to risk.


Originally posted by murpia:
I'm not advocating adding 10kg for redundancy, but 2->3kg should be imperceptable. I bet your drivers vary in weight by that much anyway...

And software is zero weight.

Regards, Ian

This might be true, but its really easy to say "just a couple kg (or even g)" here and there and by the time the car is finished, you are 20 kilos overweight and scratching your head.

Driver weight certainly isn't something to ignore. We baby them enough as it is, put them on a diet and make them earn it http://fsae.com/groupee_common/emoticons/icon_wink.gif

Even if Bobby is your fastest driver and 30 lbs overweight, he will be even faster fueled by carrot sticks.

Everything IMHO... </div></BLOCKQUOTE>

So true - it's something that is very crucial (and sometime difficult) to find the balance of.

As for driver weight - I am down 30lbs since the middle of June, so I have NO problem being strict on the rest of the drivers as well http://fsae.com/groupee_common/emoticons/icon_smile.gifI was already on the lighter end of the drive spectrum as well....

We always debated the redundancy stuff as well, and were never really able to get much actually implemented. It was hard enough to get people to finish the critical systems sometimes.. I'd be curious to hear if anyone has done it and actually had it make the difference between finish and DNF

[MalcolmG]

Originally posted by Brian Barnhill:
As for driver weight - I am down 30lbs since the middle of June, so I have NO problem being strict on the rest of the drivers as well http://fsae.com/groupee_common/emoticons/icon_smile.gifI was already on the lighter end of the drive spectrum as well....

You lost 30lbs in 2.5 months? What did you do, have a leg amputated?

[Brian Barnhill]

Originally posted by MalcolmG:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Brian Barnhill:
As for driver weight - I am down 30lbs since the middle of June, so I have NO problem being strict on the rest of the drivers as well http://fsae.com/groupee_common/emoticons/icon_smile.gifI was already on the lighter end of the drive spectrum as well....

You lost 30lbs in 2.5 months? What did you do, have a leg amputated? </div></BLOCKQUOTE>

To be honest... I have no idea how it was that much (when I think about it it was closer to the beginning of June when I started... but I've also been back down to this weight since the beginning of August) I just basically cut all the sugar, refined wheat and startches out of my diet, and upped my running to 3-5 miles 3 to 4 times a week (that's about my usual distance, just stopped "skipping" days of running.)

Probably very a-typical results (went from 165 to 135lbs) the weight I am again now is my normal weight, and I had almost always been a skinny guy up until recently. I put on the weight fairly quickly, and was lucky I guess and able to drop it just as quick.

[wagemd]

Originally posted by MalcolmG:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Brian Barnhill:
As for driver weight - I am down 30lbs since the middle of June, so I have NO problem being strict on the rest of the drivers as well http://fsae.com/groupee_common/emoticons/icon_smile.gifI was already on the lighter end of the drive spectrum as well....

You lost 30lbs in 2.5 months? What did you do, have a leg amputated? </div></BLOCKQUOTE>

Well, getting rid of a leg would be really quick and wouldn't be an issue at all as long as you built a 2 pedal car...

[Bemo]

Getting out off the car within 5s would be more difficult. I don't say it's impossible, but more difficult...

[wagemd]

Originally posted by Bemo:
Getting out off the car within 5s would be more difficult. I don't say it's impossible, but more difficult...

Touche... FSAE ejection seat?

[BrendonD]

You'd probably need to blow out the front roll bar/bulkhead structure as well to avoid the breaking of the legs.

[wagemd]

Originally posted by BrendonD:
You'd probably need to blow out the front roll bar/bulkhead structure as well to avoid the breaking of the legs.

*Leg

[murpia]

Originally posted by PeterK:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Firstly, once an ECU is synced to the correct engine cycle, a cam sensor is no longer required. So why should it's loss stop the engine? In fact I know of a number of race engines that don't have cam sensors, and instead use other methods to sync engine cycle.

I find it interesting that in a post about reliability and redundancy you would remove one of the key ways for an ECU to check for reliable control. If you are running full sequential injection/spark and don't have the ability after first start to correct for a bit of noise on your crank trigger than you are going to be SOL pretty quick. All it would take is one voltage spike and your timing would be off a whole tooth... on an F4i with a stock crank wheel we are talking about 30 degrees of extra advance!

I have to agree with Tobias, if you do a good job with your wiring design and build then you shouldn't have any major issues. When I joined my team the wiring on the car was a mess, mostly because they didn't have someone with enough electrical experience on the team.

Sometimes things fail, but I don't think relays fail often enough to double up on them for our team. </div></BLOCKQUOTE>
OK. Firstly, I know of no ECU that would use a cam trigger for something as critical as ignition timing. If they did, I would avoid that brand. Cam triggers are there to tell you engine phase, so the tolerance of the trigger edge can be +/-30deg or more without issue. Try ignition timing off that...

Secondly, if your crank trigger gets a bit of noise and sees an 'early' tooth, the ECU should reject that as it represents an implausible acceleration of the crankshaft. If it misses a tooth it should 'unsync' and wait for at least one full revolution of the crank to re-sync to the missing tooth and one full engine cycle to get back to sequential injection. Yes that will cause a stumble, no it shouldn't put you out of the race. Yes, if that happens you'll be needing your cam sensor again, but only until the engine phase is detected.

Any ECU brand that fires ignition off an early crank tooth needs avoiding.

The main issue is that you'll find it hard to get the info on all this behaviour from an ECU supplier. At least, not until you are already a customer.

Note also that I would regard 2 cranks sensors as a sensible redundancy measure, but once again your ECU needs to support it properly.

Regards, Ian

[wagemd]

Originally posted by murpia:
OK. Firstly, I know of no ECU that would use a cam trigger for something as critical as ignition timing. If they did, I would avoid that brand. Cam triggers are there to tell you engine phase, so the tolerance of the trigger edge can be +/-30deg or more without issue. Try ignition timing off that...

Secondly, if your crank trigger gets a bit of noise and sees an 'early' tooth, the ECU should reject that as it represents an implausible acceleration of the crankshaft. If it misses a tooth it should 'unsync' and wait for at least one full revolution of the crank to re-sync to the missing tooth and one full engine cycle to get back to sequential injection. Yes that will cause a stumble, no it shouldn't put you out of the race. Yes, if that happens you'll be needing your cam sensor again, but only until the engine phase is detected.

Any ECU brand that fires ignition off an early crank tooth needs avoiding.

The main issue is that you'll find it hard to get the info on all this behavior from an ECU supplier. At least, not until you are already a customer.

Note also that I would regard 2 cranks sensors as a sensible redundancy measure, but once again your ECU needs to support it properly.

Regards, Ian

I don't know of any that use ONLY a cam trigger either, you really need both: its a built in redundancy that has become standard among ecus.

Cam sensor "issue" could be anything from complete loss of signal to very noisy signal. How do you program in noise detection/sensitivity at the crank - i.e what do you consider "implausible" acceleration. These motors are capable of revving extremely quickly, how do you tell the difference between noise and a lead foot throttle blip? The ecu can guess (and it does, at least in our case) but unless it knows the rotational inertia of the internals and the expected torque at the current load point, its just a guess and not something to rely on (for more than one cycle).

We run the Motec M4 it essentially does what your talking about. If you watch the diag page you can see see it throw ref signal errors on very fast revs. It recovers after 0-2 revolutions once it sees the sync trigger again.

This system seems to work great unless there are issues with both signals...

Then do you add an extra crank AND an extra cam sensor? How does the ecu decide which set is correct if there is a problem with both? I have yet to see an ecu that will take 2 ref and 2 sync signals. Maybe then a redundant ecu with offset triggers and a different crip would take care of that? Then then you are back to deciding if its worth the extra 5-6 lbs and possibly thousands of dollars to double up a system that doesn't seem to fail that often (compared to other fsae parts) and could be very cheaply protected electrically and mechanically with EM shielding and crafty cable routing.

Not to say that a lot of teams don't end up with electrical gremlins, but I would say a majority of them are not as much the fault of the actual hardware. We are all students and we make mistakes. Finding someone willing to learn the proper methods goes a long way here. I would rather spend the extra time to carefully crimp every pin on the car and use new, high quality wire than to double up 2 shoddily built systems. Since we are an ME dominated group, wiring seems to be an afterthought many times...

Again, it all comes down to weighing the probability of each system failing. Improve or add redundancy to the systems most likely to cause issues, and you have to gamble on the others, or accept having a heavy car. If you don't have anyone that can put together a solid electrical system, that might mean redundant ecus (or sensors, switches, relays, etc). If you've had shifting issues in the past but never an engine stumble, special attention to improving that design might be more cost/time effective.

Also, not having new members design or manufacture critical parts goes a long way.

Sorry if I seem rambly, its raining up here in Seattle (go figure) and I have nothing else to do http://fsae.com/groupee_common/emoticons/icon_wink.gif

[bob.paasch]

Originally posted by murpia:
I read with interest that the 2 leading cars at FSUK, Delft & GFR, failed endurance through electrical issues.

I understand GFR had a cam sensor issue, and Delft a starter relay issue. It surprises me that these things can be car-stopping faults.
Regards, Ian

First, the failure was in the cam sensor wire, not the sensor itself. The heat shield for the engine were fabricated incorrectly and had a sharp edge. The wire was routed incorrectly and was resting on that sharp edge. Finally, the wire was wrapped in wiring harness tape, which both provided little protection for the wire, and hid the chaffing.

I would classify GFR's failure at FSUK as a management issue, not a technical one.

1) The heat shield was manufactured in Germany, based on a design developed in the shop on the US car. The design was not modeled in Catia, and there was a breakdown in communication between the two parts of GFR as to the construction of the shield.

2) The wire routing was a mistake by a rookie team members, and was not caught by more experienced members as they were very busy in the two weeks before Silverstone preparing other areas of the blue car for competition. One of the lessons of this first year of GFR is that many times two "identical" cars are in fact twice the work of one. Such was the case with shakedown testing of the cars. As of the end of June, the orange car was much more race ready than the blue car.

3) There was a discussion Fall 2009 about the use of tape verses heat shrink tubing for the wiring harness. Heat shrink is tougher and more reliable. Tape allows more flexibility and easier manufacturing of the harness, and is less expensive. The team decided to use the tape. Per this discussion, perhaps our design priorities should have more strongly emphasized reliability.

This has been a very challenging year for GFR, with a lot of new lessons learned about management of a global enterprise. http://fsae.com/groupee_common/emoticons/icon_smile.gif

[murpia]

Originally posted by wagemd:
I don't know of any that use ONLY a cam trigger either, you really need both: its a built in redundancy that has become standard among ecus.
My original point was that GFR failed because they lost a cam sensor, when a bit of thought indicates this is an unecessary failure mode for any ECU.

If it has 'become standard' then you need to be more selective as to which standards you follow.

For no added weight, and just a bit of care with the software, an ECU can run with 2 crank sensors instead and if one loses sync for any reason the other is there to take it's place.


From your descriptions it sounds like you run very few crank teeth and a very precise cam sensor setup. Care to share?

Regards, Ian

[wagemd]

Originally posted by murpia:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by wagemd:
I don't know of any that use ONLY a cam trigger either, you really need both: its a built in redundancy that has become standard among ecus.
My original point was that GFR failed because they lost a cam sensor, when a bit of thought indicates this is an unecessary failure mode for any ECU.

If it has 'become standard' then you need to be more selective as to which standards you follow.

For no added weight, and just a bit of care with the software, an ECU can run with 2 crank sensors instead and if one loses sync for any reason the other is there to take it's place.


From your descriptions it sounds like you run very few crank teeth and a very precise cam sensor setup. Care to share?

Regards, Ian </div></BLOCKQUOTE>

It would be great if the sensor wasn't required. However, I'm not going to rely on a (or two) mag or hall sensor to pick up a tooth 3033 times per second without ever missing a beat. Having a cam senor check in with the ecu 116 times per second makes me a lot more comfortable. Maybe you could run two cranks, but again, how does the ecu know if your rpms increased very quickly on sensor 1 or you missed a tooth on sensor 2. (look at a bad downshift, you could go from 6k to 12k rpm in very short order indeed)

Maybe I shouldn't have really called it redundancy per-say. Its more of a ecu-system level redundancy that allows it to recover from an internal error that occurs relatively frequently. Its not as much a global-car-level redundancy. You could try implementing that yourself if you insist.

You wouldn't be able to run two crank sensors INSTEAD of a crank and a cam. At the very least, the ecu would need to know the engine cycle at start up.

Also, at least with our Motec, you don't have to lose that whole power cycle. The timing might be +- 26 degrees, but that's not the end of the world for one cycle on our restricted NA motors.

We run the stock trigger for one reason. It works just fine for us. We currently use an F4i motor and it uses a 13 tooth wheel (with one missing tooth). Cam sensor isn't particularly sensitive. We could try running more teeth on the crank, but a). we don't have issues with 13, b). at some point the stock sensor can no longer pick up discrete trigger teeth well enough (being, when the ECU can detect AND process them in time). A better sensor might pick up the teeth then, but if 13 works, cutting a new hole in the crankcase cover and playing around with sensor alignment adds a new (and worse imo) failure mode.

Maybe you would be happier working with two stroke karts?? http://fsae.com/groupee_common/emoticons/icon_wink.gif

[wagemd]

Originally posted by bob.paasch:
I would classify GFR's failure at FSUK as a management issue, not a technical one.

...

1) The heat shield was manufactured in Germany, based on a design developed in the shop on the US car. The design was not modeled in Catia, and there was a breakdown in communication between the two parts of GFR as to the construction of the shield.

2) The wire routing was a mistake by a rookie team members, and was not caught by more experienced members as they were very busy in the two weeks before Silverstone preparing other areas of the blue car for competition. One of the lessons of this first year of GFR is that many times two "identical" cars are in fact twice the work of one. Such was the case with shakedown testing of the cars. As of the end of June, the orange car was much more race ready than the blue car.

3) There was a discussion Fall 2009 about the use of tape verses heat shrink tubing for the wiring harness. More experienced members of the team wanted to use heat shrink (as had been done on the 09 OSU car), as it is tougher and more reliable. An EE advisor with somewhat limited involvement with the team insisted on the tape, as it allowed more flexibility and easier manufacturing of the harness.


You can never trust those EE guys without a little common sense help from ME... (see below) http://fsae.com/groupee_common/emoticons/icon_wink.gif

Daniel Wageman
Mechanical, Electrical Engineering
University of Washington FSAE
Team 19, 20, 21, 22 - Technical Director

[Wetmelon]

We ourselves almost lost our car to a CAM sensor issue, but we managed to repair it in time to finish endurance with no penalties.

99% of our electrical failures have stemmed from lack of attention to detail when wiring. I know, because I wired the entire car. Spending a little more time to, as Mike Holmes says, "Do it right the first time" would result in much high reliability and much less need for redundancy.

[PeterK]

OK. Firstly, I know of no ECU that would use a cam trigger for something as critical as ignition timing. If they did, I would avoid that brand. Cam triggers are there to tell you engine phase, so the tolerance of the trigger edge can be +/-30deg or more without issue. Try ignition timing off that...

My post wasn't about using the Cam sensor for timing, I agree with you completely that it is for phasing, this is what I was talking about. Also, you don't necessarily need a missing tooth on your crank if you have a Cam sensor... but if you really want redundancy why not just switch away from sequential and switch to wasted spark... and there's always 2-stroke as Daniel suggests. http://fsae.com/groupee_common/emoticons/icon_smile.gif

Like others in this thread I can't imagine any reasonable way for the ECU to correct for all the possible types of fault/noise which could occur here. I completed disagree with your thoughts on implausible acceleration, last I checked my engine revs PDQ with no load.



Note also that I would regard 2 crank sensors as a sensible redundancy measure, but once again your ECU needs to support it properly.
Is there a commercially available option for this? Has any team run them? How would you be able to tell which one is right? How would you even implement them (positioning of the sensors is critical)

I have to agree with Wetmelon, it isn't very often that VR sensors fail... and the first thing I would check if I was having problems with one would be the wiring and then possibly the detection circuit (especially if there are biasing pots) ala Megasquirt. Chances are if the actual sensor has failed it's because something mechanical has hit it (which means you likely have bigger problems). Reliability is about looking at your most likely points of failure and mitigating the potential of failure. I think wiring is very high on that list, VR/Hall sensors? Not so much.

[murpia]

Originally posted by PeterK:
Like others in this thread I can't imagine any reasonable way for the ECU to correct for all the possible types of fault/noise which could occur here. I completed disagree with your thoughts on implausible acceleration, last I checked my engine revs PDQ with no load.
Generally the algorithm works as follows: Store the last 2 periods between crank teeth. Take the difference between them to get an acceleration delta. If you see a new crank tooth sooner than 0.X times the expected next tooth (last period minus delta) it's a glitch. Choose X according to your fastest possible engine accel or decel (X need not be the same for both).

A late tooth (sensor dropout) is detected the same way, although there is a complication associated with the true missing tooth.

If you do the maths from your own data you'll see it works unless you go really low on crank teeth (like 6 or something). 12 should be fine and anything above 20 is better especially if you like your ignition timing to be accurate.


<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">
Note also that I would regard 2 crank sensors as a sensible redundancy measure, but once again your ECU needs to support it properly.
Is there a commercially available option for this? Has any team run them? How would you be able to tell which one is right? How would you even implement them (positioning of the sensors is critical) </div></BLOCKQUOTE>
Not yet that I know of, I think there should be. Maybe, if they wrote their own code. Simple boolean OR on sync state. With two independent crank algorithms as above, feeding a single internal angle clock. Positioning is no more critical than with one crank sensor.

I find it fascinating that people here are really ready to defend the status quo, and dismiss new ideas. Yes, of course you need to do a proper wiring job on your car if you want reliability, but why stop there when smart thinking can get you so much further?

Regards, Ian

[PeterK]

Generally the algorithm works as follows: Store the last 2 periods between crank teeth. Take the difference between them to get an acceleration delta. If you see a new crank tooth sooner than 0.X times the expected next tooth (last period minus delta) it's a glitch. Choose X according to your fastest possible engine accel or decel (X need not be the same for both).

What you are suggesting here is windowing the expected value and I agree with you that it is a good way to detect the error. I would hope that most modern ECUs already do this (even the Freescale supplied functions for crank and cam on their eTPU platform do something very similar). But this is not a way to correct for the problem, it is error detection, not correction. It's what you would do when you experience an error that matters. Also, how would you know if your engine is in accel or decel to have different 'X' parameters (choosing the bounds of your window)?

I would think that you could toss away not getting one tooth or an extra tooth, it wouldn't destroy the engine anyway... but what if you had a problem on 2 successive teeth? or right around the missing tooth? At some point you should have a re-sync being triggered... which needs the cam sensor.

All this extra work to add redundancy for a sensor which doesn't fail all that often. I can think of at least 10 other critical components which would stop the engine that have a much higher probability of failure than a VR sensor. Has anyone run redundant fuel pumps?

I'm not dismissing the ideas, I'm just saying that there are way better areas to look at first than the actual sensors.

[timWHS]

i worked for 20 yrs in formula one as a wiring technician and we had very little redundancy built into the electrical system. In fact the only example i can think where there was any parallel system was gearbox drum position pots. This is because they were know to fail due to their harsh enviroment location, and a failure would literally destroy a gearbox. The other thing we would do is to identify which connectors carried critical wiring (critical being car stop if failed) and wire locked any mating halves. When you hear of an electrical failure, 9 times out of 10 its not a component failure, but a human installation error.