MPGuino for Audi

[rmccomiskie]

Yes. I really do want to use existing signals. They're easy to get to. It's just a matter of calibrating them.

PW at idle seems to be about 400uS.

After some experimenting, VSS = 8400 is close. Haven't gotten to uS/Gal yet.

I'm wrestling with what looks like noise. When the car is stopped with engine off, speed registers about 14-15 mph and mpg keeps ticking slowly. I'm reworking my connections to the car. Alligator clips now.

My MPGuino is on a breadboard and seems to be working correctly.

[dcb]

what is INJECTOR pulse width at idle? So that you know the conversion factor for that signal you are looking at.

[rmccomiskie]

Ah! I'll have to dig for that. Not this week. Soon though.

[dcb]

Ok, I'd suggest measuring the injector signal along side that plug signal under a couple rpms and maybe it would make more sense.

[rmccomiskie]

I replaced my breadboard device with the MPGuino v1 from Fundamental Logic. Hooked up to my Audi, I adjusted VSS pulses per mile to 13418. That's within 0.10 miles distance over an 80 mile trip. So it's very close.

I adjusted fuel flow uSec per gallon to 050000000 which seems to be in the ballpark. I noticed that the GPH and MPG were responding in reverse. Up a hill, on the accelerator, GPH decreased and MPG increased. That's reverse of what it should be.

I remembered that the fuel flow signal was negative going. Assuming that the code is set up to measure a positive going FF signal, I changed the code like this:

attachInterrupt(0, processInjOpen, FALLING); changed to RISING
attachInterrupt(1, processInjClosed, RISING); changed to FALLING

The idea being to start measuring the pulse on the negative going leading edge. Now the GPH and MPG are responding in the proper direction.

Now I'm trying to zero in on the uSec per gallon number and I noticed that when I lift off the accelerator going down a hill, the GPH quickly goes to zero but then jumps up to something like 70 GPH. I'm guessing that the engine computer cuts off the fuel flow entirely when coasting and the signal is being interpreted as a very high flow rate.

Have you seen this before? Is my code change correct? Or is there something else in the code that should be changed instead?

Thanks.

[dcb]

Sorry, but we are flying blind here. The mpguino was designed to hook up to an injector, not a computer. We do not know what the computer is doing to the signal without first accurately comparing it to an injector signal and running several scenarios.

The computer could be adding offsets, it could be porportional, it could be non-linear, it could be some really weird function. If you want to avoid using an injector lead and still be accurate, you have a lot of work to sort out with this proposed signal, and it all starts with a known point of reference (i.e. a calibrated mpguino) which requires you to tap an injector line anyway.

It would be pointless for me to look at code without knowing anything about the signal it is supposed to process, and even then I'm not sure of the value of it if it means a code branch for an unknown number audis that could just tap an injector lead. If I didn't have other priorities I would be more inclined I suppose, but I can't even look at this at the moment. You need audi ECU specific help or reverse engineer what you have there if you are bent on this path.

[rmccomiskie]

OK. Thanks for the quick reply. I'll go ahead and validate that the fuel flow signal is identical to the injector signal. If it's not, I'll dig into the code to see how it can be interpreted.

As I pursue that approach, am I correct in assuming that the MPGuino code is looking for a positive going injector signal only? And that a simple method of changing from positive to negative signal is not provided.

[dcb]

Originally Posted by rmccomiskie

And that a simple method of changing from positive to negative signal is not provided.

It is probably simpler overall to change the code than to present an option for the few cases that might need positive switched, but you change these two lines exactly like you mentioned:

Change:
attachInterrupt(0, processInjOpen, FALLING);
attachInterrupt(1, processInjClosed, RISING);

To:
attachInterrupt(0, processInjOpen, RISING);
attachInterrupt(1, processInjClosed, FALLING);

[rmccomiskie]

The speedometer face in my 1997 A6 has "K=6703" printed on it. You can see it here at the bottom.



That's the number of VSS pulses per mile expected by the speedo. For the MPGuino, simply double that number and enter it as the VSS pulses parameter in setup. So, mine is 13406 and the distance is dead on. The number must be doubled because the MPGuino counts both the rising edge and the falling edge of each pulse.

I've seen other Audis with a similar, but not identical number, printed on the speedo. You might have to dissemble the instrument cluster to see it. Remember, assembly is the reverse of disassembly! (8'D

[rmccomiskie]

Here's a comparison of the injector signal versus the ECU fuel flow signal. I frankensteined both the injector and ECU fuel flow signals into a stereo audio plug and captured them with my SoundCard Scope. The green signal is from the injector and the red signal is from the ECU.

This is about 30 mph steady on level ground:

You can that there are 6 ECU pulses for every injector pulse as expected.

...and this is about 30 mph uphill accelerating hard.

The ECU pulses appear to be scaled down from the injector pulse width but I expect them to be proportional.

Then I started coasting downhill at about 30 mph...

When coasting, the signals start out looking like the engine is at idle.

...after a couple of seconds, the signal looked like this...

The ECU signal goes completely flat while the injector signal shrinks to a very short pulse. The flat ECU signal may present a problem because the external interrupt routines will not execute.

I compared the injector pulse width versus the ECU pulse width across a range of speeds and throttle settings to get an average ratio between the two. Here's the graph of the relationship:



The Raw data is for the full injector pulse width and the -500uS data is for the inj pulse minus the settling time. So, the ECU pulse width seems to be about 1/8th the width of the injector pulse and it's consistent across different speeds and acceleration.

If I assume that the ratio is 0.123 and I know that the injector uS/gal is 200500000 for my car, when I switch to the ECU signal input, the uS/gal should be 148000000. Calculated as 0.123 * 200500000 * 6 inj pulses. That gives me a starting point to calibrate the ECU signal.

Of course, it still remains to figure out how to handle the ECU flatline signal when coasting.