hillpc

About 6 or 7 times today, idling in a parking lot, and then on the highway in a slow moving traffic jam. Every time it cranked and started right up immediately after it died. The only codes I got were

16711 - Knock sensor 1 (G61) P0327 - 35- - 00 - signal too low, and
16535 - Oxygen (Lambda) Sensor B2 S1 P0151 - 35 - 10 - Signal too low - intermittent

As far as I know neither of these devices (or their connections) failing would kill the engine. Is that right?

If the engine speed sensor were going intermittent, I'd get a code, wouldn't I? I've had those wires melted before.

Would the fuel pump dying just suddenly kill the car but then start right up again?

jfrahm

A dying fuel pump can stick and unstick a few times in it's final days. Mine did. A trick is to hotwire it backwards to unstick it, then reconnect it normally and get home.
I have not heard of codes from the crank sensor, if that signal goes away the ECU just thinks the engine is stopped. I suppose there could be codes for implausible or shorted though.

Kneale

Typical symptoms of failing fuel pump

d2noob

Since we are on the topic of the fuel pump, is there any diy tips for doing the fuel pump job?

hillpc

Somewhere on this forum I've seen warnings not to buy some cheapo generic replacements, which don't even work (or fit?). There's a recommendation for what to order.

hillpc

This suggests that the failure mechanism is some kind of mechanical wear, doesn't it? Cool.

hillpc

Now I'm fairly sure it's not the fuel pump stalling the engine, for 2 reasons - I recently replaced the fuel pump with a new, correct VDO pump, and when I pulled the fuel pump fuse at idle, it took a few seconds of stumbling before the engine died. My stalling problem occurs suddenly and rapidly, with no warning. I had another thread going on about putting a fuel pressure transducer in the fuel rail, https://www.audiworld.com/forums/a8-...ng-id-2982580/ but I think that's a moot point now.

I did recently put a 4 channel oscilloscope (Picoscope) on it, and caught the stall in the act, but I wasn't monitoring whatever it was that made it stall. See the scope trace below. I monitored the crankshaft sensor G28 (blue squiggly trace), #5 coil primary voltage, #5 fuel injector voltage, and fuel pump voltage at the fuse. I was hoping to see what dropped out first. Had the car idling collecting these signals continuously for 10 - 15 minutes until the car stalled, after which the rest of the data collection was triggered by the disappearance of the fuel pump voltage when the ECU shut down the fuel pump relay. The triggering had been a real problem until I switched the type of software trigger I used. The details of the triggering to catch an unattended stall are at the Picoscope forum https://www.picoauto.com/support/vie...100900#p100900. Here's part of that thread (in italics) -
The ECU will kill the fuel pump after a stall. Caught an intermittent stall in the act, twice. From the trace, below, zoomed in to see the action, it looks like I can rule out the crankshaft sensor signal going open or shorting, which had been my main suspect, because it seemed to behave as expected the whole time. It also looks as if the fuel injector is lengthening the fuel spray time as the engine slows down on the way to the stall. Maybe in an attempt to keep the engine running? One can also see the fuel pump voltage dropping slightly as the alternator slows on the way down to the stall.

I think the next step is to find out what other sensors or actuators could kill the engine immediately if they failed while running.

The fact that the vehicle has 8 fuel injectors and coils makes it more difficult because of the limited amount of channels available. There are 2 “power output stages” (some people call them “ignition amplifiers”) that each drive 4 coils when commanded by the ECU. If the ground connection to either of these is interrupted, I think it'll kill 4 cylinders. Not sure if another tool (such as the exhaust pressure/intake vacuum pulse sensor (Picoscope FirstLook) could help; I have one.



Both the #5 fuel injector and #5 coil primary were stopped within a half second of the engine starting to slow down. The fuel pump was killed 1 second after the engine came to a stop. Whatever it was that started the event, I didn't have it monitored. You can't see it on the 5 second long trace, but on zooming in it can be seen that the fuel injector is squirting longer for the last two squirts before it is shut down as the engine continues to slow.



zoomed in to see engine starting to stall

So now I have to look for other stuff to monitor. Maybe I'll repeat this whole exercise with each of the other 7 cylinders in turn. Maybe injector current instead of voltage; I don't know. But it'd be nice if it were just one sensor dropping out. What sensors or actuators would kill the engine if they dropped out momentarily? Certainly the crankshaft sensor G28 would, but that behaved as expected the whole time in this stall. Mass air flow? Throttle position? Camshaft sensors, according to the Bentley, would not shut down the engine or prevent starting, since the ECU would, to compensate for a failed cam sensor, squirt fuel every revolution of the engine, instead of every other rev. Power output stages ground connection? O2 sensors would throw a code and still keep the engine running. The only code I found was P0441, evap purge incorrect airflow. Somehow I doubt that's it.

What would kill the engine so suddenly? jfrahm had suggested looking at fuel pump current; I may do that to see if there's anything to see there. Sudden loss of fuel pressure, even with the voltage on? What keeps the backpressure on in the fuel rail to keep the injectors supplied? Can something fail open and release the pressure back to the fuel tank, and then close again before I restart the car? Or is the piping itself what keeps the backpressure on in the fuel rail?

I hope it's not the whole ECU dropping out for a fraction of a second.

jfrahm

That scope setup is really cool.

A failing fuel pressure regulator I suppose could fail sometimes and drop the fuel rail pressure, if that happened I think you'd see the ECU trying to correct for a lean condition and perhaps giving up at some point. But there should be a lean code? That said I do not think I ever got codes from my dying fuel pump even though I had some long cranks and it died a few times.

Maybe fuel pump relay or some sort of wiring issue? A current tap should show if the pump was disappearing or turning into a little toaster.

jfrahm

P.S. I'd probably replace the fuel pump relay on spec. I understand that at idle it should take a bit of time for the engine to die but I'm low on ideas. Maybe a fuel filter issue or blobs of water in the gas. The latter should be a problem that gets less and less frequent over time. A used fuel pump relay from a pick-n-pull vulture should be pretty cheap.

hillpc

I will put a current clamp on the fuel pump wire at the fuse; good idea. That would also show if a connection anywhere in the fuel pump circuit is intermittently opening. Voltage wouldn't show that unless it was being monitored downstream of any break, say at the pump itself.

I don't think it's the relay, since the relay sends voltage to the fuse, which then feeds the pump. I would have seen the voltage at the fuse box drop out.

Your comment about the ECU trying to correct for low fuel pressure made me go back and look at the lengthening of the injector open time (the time the injector voltage is grounded) in detail. There's support for that in the traces, but only if the O2 sensors can respond quickly enough for the ECU to notice it and try to do something about it. I don't know the response speed of the O2 sensors, but since it's a diffusion-based gas analysis (of sorts), at this point I'm somewhat doubtful the O2 sensor could sense it and transmit it in time for the ECU to respond. That said, the 3rd to last injector length was about 3.3 milliseconds, and the last two were over 15 msec long. 2 revs of the engine take about 155 msec, so maybe the O2 sensors are that fast. I'll have to study that; I know they do switch up and down at some rate. It may not throw a code for mixture unless it stays that way for a while

Now that we've seen that (and if we believe it), it makes me doubt that ignition is an issue (like spark dropping out), because the engine would run rich, and the ECU would shorten the injector squirt instead of lengthening it.


injector squirt lengthened just before stall