Digital Air/Fuel Ratio Gauge Under Development

[Flatland2D]

I've created this thread to document the status of my AFR gauge project. I will update when advances in design are made.

Tonight I completed the first step in developing a microprocessor-based AFR gauge. Use of a programmable chip allows for easy manipulation of the signal and analysis. It currently reads a signal from 0V to 5V and outputs it on a 10 segment LED bar graph. This is analogous to the 0V to 1V signal produced by the oxygen sensor - all I have to do is change some numbers around in the code.

The next version will incorporate a tuning function that will allow the driver to scale the O2 sensor up or down, effectively making the engine run lean or rich. Running lean should allow for large mpg gains. I'm hoping this will really come into play during constant speed highway driving when acceleration is not needed.

I was also thinking about adding a wideband simulator feature that would basically allow your cheap narrowband sensor to output as if it were a $150 wideband sensor. This would also benefit fuel economy.

I plan to test the inital version (just the gauge) this weekend if it isn't too stormy from hurricane Rita.

The final product will either be released strictly as plans or possibly for sale as completed units through this site with the approval of Matt.

[SVOboy]

How would it work to simulate a wideband o2 sensor? I thought the difference was the data that was returned from the sensors and therefore it could not be manipulated to make a narrowband act like a wideband 02 sensor.

[Flatland2D]

It would simulate a wideband sensor by implementing a moving average filter over the O2 signal. So, let me explain that some more. A narrowband O2 sensor is constantly fluctuating from rich to lean, up to 10 cycles per second. The advantage of a wideband is that it does not overshoot each cycle - it outputs a constant reading and is supposed to be more fuel efficient. A moving average filter samples the input signal (reads the value at some time t) and calculates an average based off past values. The output is (f[t] + f[t-1] + f[t-2] + ... + f[t-n]) / (n+1) where n+1 is the number of terms in the average, which is also called the bandwidth of the filter. This filter dampens out the peaks into a constant average if the sample rate and number of samples are properly chosen. This is how I would use a narrowband sensor to act as if it were a wideband through some very basic digital signal processing.

Now here's the catch. The sample rate times the number of samples must be greater than one period of the input signal or your result will be off. Imagine implementing this filter on a sine wave and the bandwidth of the filter is only pi/2. Your average consists of only the positive portion of the sine wave and doesn't have any of the negative values to bring the average down to zero where it should be.

I wrote a program on my calculator today to simulate this concept. I could get very good results by using only 5 samples but sampled at a little over 3 periods of the input. If the frequency was about that of an O2 sensor at 10Hz, the bandwidth would be about 1.5 seconds, meaning that if you suddenly went WOT, there would be a 1.5 second delay for the pervious values in the average to get kicked out and the output of the system to be reading the right value. This would be unacceptable for normal driving, but possiblty useful in highway driving. I am going to look into other methods of processing the signal that have a much smaller bandwidth, preferably nothing more the the period of the O2 sensor at 0.1 seconds. I'll have to look into the limits of the PIC, but if I could sample the input fast enough I could get by with only 1 period. Analog to digital conversion is a time consuming process relative to other commands.

[SVOboy]

So really what the wideband sensor does it take a running average to give a steadier output rather than a spiky annoying thing? That's a good concept, I think, and if you can get it working I think it'd be worth it even if just for highway. You could have it acting as narrowband be default and then poke this buttom to turn on wideband mode or some such thing like that.

[Flatland2D]

Well I don't think a real wideband sensor is actually taking an average (not entirely sure about that though) but I believe it just doesn't overshoot like a narrowband does. It's like a spring that is fixed at one end is compressed then let go. It constantly overshoots the value it's trying to hold.

[SVOboy]

Okay, that makes sense to me. I'll be interested to see how it plays out.

[Flatland2D]

I finally found a nice place to mount this device that isn't going to draw attention. I'm thinking about mounting it in a plastic box (Radio Shack and electronics stores sell them, called project boxes) that will sit on the steering wheel column, but not on the part that rotates with the wheel. It won't cover up any of the other gauges and shouldn't be very noticable. It might be a little difficult to stick your hand through the wheel to make adjustments or turn things on/off, but I can live with that. I was hoping to make some progress this weekend but I think I will have to hold off until next weekend.

[SVOboy]

I can't recall where I saw this but someone made two pods for on top of the steering column that didn't get in the way of things and looked really nice. If you could rig up controls through the column and mount them somewhere on the wheel to poke at easily without lots of jostling that would be good.

[Flatland2D]

I just looked them up and I see what you're talking about, but they only make them for a very limited number of cars. They are also a bit expensive. It'd be nice if I could mount a normal dash pod on there, but I think it'd get in the way of the tach and speedo.

[SVOboy]

They make them comercially? I didn't know that. I think this person just got some normal pods from wherev cheaply and then made them into his steering column doing whatev that would require. I don't see a real problem with the box though.