Measuring car RPM is giving weird answers

Ok so I have a bit of an update:

I still couldn't figure out why the numbers were randomly jumping around so I decided to rewire my board and go at it a different way.

I decided to use D6/D7 for the analog comparator. Rather than trying to pick up an edge between 2 thresholds like 1 and 4 volts, I figured I'd make life easy for myself and pick 1 single threshold value on D7 and to make things even easier, I constructed a voltage divider out of a simple pot grounded from the 5V pin with the wiper going to D7 so that I could rotate the knob on the pot, sweep the voltage from 0 to 5 and watch my display to see what the numbers do. Here's what happened:

I started at maximum voltage and the number was stuck on 0 RPM. I rotated it almost all the way to the left when numbers finally started to change. Just blipping at first and as I rotated more they went full on random and updating constantly just like before. I could not find a spot where I would get the correct values, only a vary narrow range at the bottom end where I'd get random ones. At 0V the Arduino would reset itself over and over again, suggesting some sort of floating voltage was making its way into the reset pin? Anyway I went back to a higher voltage on the pot and slowly turned it down until I started to see the numbers just barely blipping. I got a multimeter and measured the wiper voltage at this point --> 0.5V.

So there you have it. The whole square wave is 0.5V Pk-Pk with not so negligible noise levels. No wonder I can't pick up anything off of it. I guess the PCM can't put out anything larger.

So my question now is: can I use the internal 1.1V band gap reference to increase the resolution of my ADC so I can effectively grow the wave to use the entire 10 bits of ADC precision? I know it can be done but I'm also measuring 4 other 5V sensors on the same MUX so I don't want to affect them. Can the band gap ref be used only on 1 channel or do I have to switch it every time I poll that channel and then switch it back? What about clock cycle delays if this has to be done? Can I use the band gap ref on the comparator without affecting the ADC? Or is none of this feasible and I just have to amplify the square wave, filter and try again?