MPGuino for Audi

[yetimus]

rm-

You may recall that I replied to your post on 2/15 in the release one workspace thread expressing my interest in tapping the ECU fuel consumption signal on a VW TDI. Well, I finally received and assembled my Fundamental Logic v2 kit and tapped into my VSS and fuel consumption signals. I installed the MPGuino and went for a test drive with no changes to the pre-installed code. I did, however, play around with the preset parameters and changed "Injector Delay uS" from its default setting to 100. IIRC, the default setting was 500 and I somewhat randomly chose 100 uS based on the oscilloscope graphs you posted showing much shorter pulses from the ECU. On test drive #1, I noticed the same behavior in GPH reading as you originally did (i.e. GPH read higher when coasting and lower when accelerating).

So, my next step was to change the code as you did:
attachInterrupt(0, processInjOpen, RISING);
attachInterrupt(1, processInjClosed, FALLING);

The next test drive looked a lot more promising. My calibration values were obviously way off, but not only did the GPH readings display as expected (rising when accelerating/falling when coasting) I also didn't notice any strange behavior with the GPH value rising when coasting. It went to zero and stayed there - which is correct for the TDI because it cuts off all flow to the injectors when coasting in gear.

So I spent a little more time calibrating the VSS and uS/gal settings and everything seems to be running correctly. I still have a little tweaking to do with my uS/gal setting, but it appears that this setup is going to work.

I hope to get around to playing with the sound card oscilloscope program you've been using so we can see if our ECU signals are similar. But I think they are. I recently installed an instrument cluster in my TDI that originally came in the VR6 Passat. The reason I did this is that the VR6 cluster has a built-in MPG gauge that displays average MPG (not instantaneous). The TDI, though it was never originally equipped with the factory MPG guage, obviously uses a nearly identical signal from the ECU as the gasoline powered car. Read more about it here.

So I wish I could provide a bit more insight as to why my particular setup is working so far. Have you tried messing with the Injector Delay uSec preset at all?

Thanks for all the excellent info you've posted on this topic!

[rmccomiskie]

Hi yetimus,

...I did, however, play around with the preset parameters and changed "Injector Delay uS" from its default setting to 100. IIRC, the default setting was 500 and I somewhat randomly chose 100 uS based on the oscilloscope graphs you posted showing much shorter pulses from the ECU...

I'm not sure that the Injector Delay uS is needed with the ECU signal. It's possible that the ECU factors out the injector settling time and presents a pure fuel flow signal. I'll be experimenting with that soon. Hopefully, I can gather a lot of OScope data and graph the relationship of the injector and ECU signals. Perhaps the resulting curves can be extrapolated to show that the settling time is or is not included in the ECU signal.

...I also didn't notice any strange behavior with the GPH value rising when coasting. It went to zero and stayed there - which is correct for the TDI because it cuts off all flow to the injectors when coasting in gear...

I'm surprised at that. Congratulations.

...I recently installed an instrument cluster in my TDI that originally came in the VR6 Passat. The reason I did this is that the VR6 cluster has a built-in MPG gauge that displays average MPG (not instantaneous). The TDI, though it was never originally equipped with the factory MPG guage, obviously uses a nearly identical signal from the ECU as the gasoline powered car. Read more about it here...

Wow! That's an amazing read. It took a long time to figure all of that out. Similarly, my North American 97 Audi A6 did not come with a trip computer. The rest of the world got it but not us. And to add insult, the NA wiring harness doesn't include the connector for the wiper stalk switches. I'm sure that the instrument cluster itself has all of the TC functionality but the lack of any wiring makes a retrofit a big deal.

So, that's why I'm pursuing the MPGuino. My ultimate plan is to install a wiper stalk with switches from a 96-97 A4 (lots around) and fabricate a small LCD display that fits into the instrument cluster in the trip computer location. There's not much room in there so I'll have to use a scavenged cellphone display interfaced externally to the MPGuino. Time will tell whether this all works. In the mean time, it's fun!

I'm glad to see someone else is tackling Audi issues. This thread is a good place to share experiences so other Audi-heads can find us.

Cheers.

Bob.

[rmccomiskie]

I've assumed all along that the ECU fuel flow signal is an exact representation of the injector signal but narrower so all 6 injector pulses can be included within the same time period. Given that it looks like the ECU pulses are 0.123 times the injector pulse width, I expect the settling time to be similarly scaled.

For an injector settling time of 500uS, the ECU settling time would be 0.123 * 500uS = 61.5uS; Injector Delay uS = ~62uS.

I'm going to use 62uS unless experimentation disproves my theory. Keep in mind that the 500uS injector delay is simply an assumed value based on experience across a lot of different injectors. It may or may not be correct for an Audi injector.

These posts give some background information on the injector settling time issue:
http://ecomodder.com/forum/showthrea...html#post31602
http://ecomodder.com/forum/showthrea...html#post34576
http://ecomodder.com/forum/showthrea...html#post34738
http://ecomodder.com/forum/showthrea...html#post34902

[rmccomiskie]

I found injector specs for my 97 A6. Using Yoshi's formula, I calculated the injector settling time to be 615uS. Scaling that down to the ECU pulsewidth, I get 72uS. That's what I'm running now.

Over the weekend, I calibrated my uS/gal setting on a 250 mile trip. I filled the tank and reset everything. When I next filled up, my MPGuino said I had used 13.28 gallons. That was with a uS/gal setting of 123000000. The fillup was 13.53 gallons. So, I decreased the uS/gal setting a couple of times until the gallons read 13.53. I'll continue to check it over the next couple of tanks but I think it's real close now.

[rmccomiskie]

Two tankfulls later...I've recalibrated uSec/Gal to 128800000. My MPGuino now indicates fuel used within 0.7% of the fillup amount. For example, MPGuino showed 16.71 tank gal used and the pump was 16.6. After changing to 128800000, MPGuino said 16.60. Still calibrating but that's close enough for my needs.

[dcb]

That is sounding pretty useful Might need to confirm with a city tank or two but you are way better than in the ballpark it sounds like

edit: I have to wonder if 128800000 is the value that yetimus used? or that dremd could use, or anyone with a newer vw/audi? And where did you leave injdelayus?

[yetimus]

Originally Posted by dcb

I have to wonder if 128800000 is the value that yetimus used? or that dremd could use, or anyone with a newer vw/audi? And where did you leave injdelayus?

I'm still dialing in my calibration numbers - I recently (last week) replaced my injector nozzles on the TDI with some larger orifice nozzles which obviously is going to effect my uSec/gal. However, I'm currently using 135100000 uSec/gal and I'm within 2% of actual MPG.

As far as injdelayus, I had it set at 100 for a few weeks and then tried it at 0. I had forgotten about Bob's research suggesting 62 uSec. Either way, the mpguino provides fairly accurate and very usable feedback for increasing FE. On that note, just achieved my best-ever tank MPG at 47.7 on the old Passat (with the help of the Mpguino and new nozzles!)

[rmccomiskie]

Originally Posted by dcb

...I have to wonder if 128800000 is the value that yetimus used? or that dremd could use, or anyone with a newer vw/audi? And where did you leave injdelayus?

Recall that I'm using the ECU fuel consumption signal. I scaled injDelayuS down to 72uS. I've run through 2 tanks highway and 2 tanks around town. Keeping track of all adjustments to see if the highway versus city usage is different. Haven't figured out how to tell if injDelayuS needs to change.

[dcb]

Originally Posted by rmccomiskie

Recall that I'm using the ECU fuel consumption.

I am 100% aware of that and seeing that it may cross over directly into TDI/vw land too I am especially glad you stuck with it. I'm sure there are a lot of tdi folks who would be glad to have consumption feedback.

[rmccomiskie]

I've resolved the high MPG reading when coasting. It turns out that the ECU shuts off the injectors when coasting. The injector signal from the ECU is normally negative going. However, when the injectors are shut off, there's still a very small positive spike for each injector pulse. MPGuino inadvertently interprets the small spikes as verrry long injector on pulses. The behavior is observed as an instantaneous jump to a very high MPG value after a couple of seconds of coasting.

I've made the following code mod to fix this behavior.

//This code resolves very high MPG values when coasting (1997 Audi A6)
//Based on MPGuino v0.75
//Injector signal taken from ECU, negative going signal, injectors shut off when coasting
//Adds pulse width limit parameter = 4000
//Assume any pulse width greater than the limit is invalid and sets it to zero
//Results: Out of range pulsewidths are never added to total injector open time

//insert new parameter, in this case "PW Lim", 4000ul for Audi A6

unsigned long  parms[]={69ul,13406ul,128900000ul,6ul,420000000ul,19800ul,72ul,4097ul,0ul,2ul,4000ul};//A6 defaults w/ECU signal
char *  parmLabels[]={"Contra","VSSPPM", "uS/gal","P/2rev","Timout","Tk Gal","iDeluS","WgtLbs","Scrach","VSSdel","PW Lim"};

//define new parameter

#define pulsewidthLimit 10  //insert new parameter

//modify processInjClosed() routine

void processInjClosed(void){      
  unsigned long t = microSeconds();
  unsigned long s = parms[injectorSettleTimeIdx];
  unsigned long pwLimit = parms[pulsewidthLimit];
  unsigned long x = elapsedMicroseconds(injHiStart, t);
  
  if (x >= s && x < pwLimit) {  //if elapsed time > settling time and < upper threshold
    x -= s;                     //then normal pulse, subtract settling time
  }else{
    x = 0;                      //otherwise pulse width out of range and assumed to be zero
  }
  
  tmpTrip.injHius += x;       
  tmpTrip.injPulses++;      

  if (tmpInstInjStart != nil) {
    tmpInstInjTot += x;     
    tmpInstInjCount++;
  } else {
    tmpInstInjStart = t;
  }
  tmpInstInjEnd = t;
}

I determined the 4000 uSec pulsewidth limit for my car by creating a display screen that showed the currently measured pulsewidth. I observed that the maximum normal pulsewidth was below 3000 uSec and that it would jump immediately to about 10000 uSec when coasting. So, I set my limit to 4000 uSec. Anything greater than that is assumed to be invalid and the pulse is ignored, i.e. set to zero.