diy mpg gauge

[jim-frank]

I've recently gotten this to work just fine. Here's my take on it.

1)Any silicon small signal diode will do. Radio Shack or scavenge. (Actually, a germanium diode will work, but you'll get a maximum signal of about 0.3 volts instead of 0.6 volts.)

2) Yes, use the injector wire from just one cylinder, whichever wire is easiest to get to is fine.

3) If your Palm Pilot has a stereo line level sound card input, you can do it. Otherwise, no deal. You'd also have to get the Java app ported to Palm OS. (I don't know how that would be done).

There are only four variables that need to be calibrated to your vehicle. These are called fuelFudge, distanceFudge, injThreshold, and vssThreshold.

They are found in the mpg.properties file that the program creates the first time you run it. You can open this file with any text editor, (with the program NOT running) and then save your changes before running mpg.java again.

I found some details that are scantily covered in the information above, and these are relevant.

Once you get the interface electronics wired up and connected to a soundcard, it's a good idea to run a recording of the signal. I use CoolEdit or Adobe Audition, but anything that will let you record and then zoom in on the waveform will do.

A peek into the java code reveals that the waveform is decoded for the program as numbers from -127 to +127. This number is the number of samples above or below zero that the waveform goes. This number is also what is referered to as the threshold variables in the program.

In my case, the injector waveform makes a pretty clean vertical line at the opening/closing of the injector, (like a square wave), so a threshold of near zero works OK. I adjusted mine to -8, which prevents false injector readings when the engine is off (noise signals will vary around the zero line by some amount.)

The vehicle speed sensor waveform in my car is a steep sided quasi-sinewave. A look at the waveform shows that it varies from -65 to plus 65 or so, and a threshold of about 10 works for me.

After you have set the threshold factors, you can set the fuelFudge and distanceFudge to correspond to your car.

Set the distanceFudge first by running down the road and having a helper (for safety's sake) monitor the speedometer versus the program's calculated speed. The program counts the number of speeds sensor pulses in a one second interval, then divides this by the distanceFudge to get the speed.

So, if the program shows that you are going faster than the speedometer, INCREASE the fudge factor. If it shows a calculated speed lower than the actual, DECREASE the fudge factor.

You can get pretty close with a little math. If the distance Fudge is 3200, (the beginning default value), you're going 60, and the program shows 70, then you're off by (70/60)*3200, or about 373. So try a fudge factor of 3573. When you get within a few percent, it's easier to just tweak the fudge factor by two to five units at a time, say from 3573 to 3575, then 3580, etc.

Over a long trip the TANK total miles (bottom box in the program) will be more accurate to tweak the fudge factor. So if you have 400 miles on the tank, and the program has calculated 375, you will have to decrease the fudge factor a bit.

Now that you have the speed calibrated, it's time to get the fuel factor set up. It works the same way, the program divides the total number of fuel injector pulses by the fuelFudge to get the actual fuel usage. You can get pretty close by topping off the tank, running the program while driving a quarter to half a tank off, then comparing the total gallons calculated with what it actually takes to fill the tank back up.

Same math will get you close, so if the program says it took 10 gallons, and it actually takes 8, you have to increase the fuelFudge by about 1.25 times. (The default is 8 million, so you'd want to try 10 million) It takes a bit of tweaking to get this one set up, but just a tank or so of fuel should get you pretty close indeed.

I hope this helps you all out.

Jim