9. Removing and redesign ...

After many not successful tries to fix the problem with tachometer i have decided to remove all the electronics from the bike , and redesign the circuit. I must to test and fix all the problems starting from the design.

I must fix some problems:
- the size of the metallic box (which is an ATX power supply from an old computer), i have to reduce the size.
- reduce the number of connections between the two PCB (you can see in the picture above, the wire mess between controller and power source), This can be done by using only one PCB.
- The main problem it is the wrong reading of the RPM!


I decided first to visualize Variable Reluctance sensor output signal. Because i don't have an oscilloscope i use an external USB sound card (i could have used the sound card of my laptop but it is safer with an external card).The generic sound card can capture signals up to 20KHz(aprox.). The engine it is designed to go up to 12000 RPM, equivalent with 200Hz => I can use the sound card!
I start the engine and i pick up the signal from sensors throw an 3M ohms resistor. The signal can be visualized with zelscope or virtins. I used the trial version and you can see the result in following image:

The above signal it is the typical signal for Variable Reluctance sensors. You can find here some info about this sensor and how to calculate the RPM (i tried all that methods to eliminate the reading error ).
Searching on the net for the best circuit to transform the sinusoidal input signal into rectangular or pulse signal. I found two IC that can do this work: LM2907 and LM1815 .

I choose first to test the LM2907. The test schematics and response looks like this:

The above tests was made with engine running at idle RPM, and all the components used was just like that in the schematics. I choose to use the second schematics connected with the controller (all the electronics still working ) and the results was pretty good. The RPM displayed was stable and according to engine speed.

8. First test with engine running .

Hello.
I made first test with the engine running. The overall results was just good , not very good.

Configurations:
- the carbs are in place => the test was made without injectors and throttle body.
- the ignition system it is the original one.
- the power supply and the controller are mounted and connected to 12v battery.
- the air filter it is home made

Good:
- the controller works well; I've made 200km, and not even one hang or restart!
- Sensor temperature (LM35) returns the correct values
- Pressure sensor works.
- Battery voltage is monitored properly.
- Injectors works. The test was made without pressurized fuel (so this is not quite good)
- The command to stepper motor it is working -> the stepper responds properly.
- The serial port connections works. I can update the software on the controller without removing it from the PCB.


Bad:
- the three buttons are affected of rain
- the air filter it is not very good. The air-fuel ratio are dependent of the air filter.
- the big problem that i have it is "bad readings of the engine rpm". Some times the readings are good ,
but suddenly from, for example, 2400rmp jump to 8000 rpm. I recheck and change the software and the problem
become to occur rarely.

7. installing circuits...

Picture from this post was removed due to a ugly bug from picasa!

6. Work in progress...

5. Solving the problems

First of all i must reduced the voltage variations on the power supply output.
To stabilize the voltage i used a zener diode of 12V (5W) connected on the base of a 2N3055 transistor,
followed by a LM2940T-5.0 regulator.
Another problem that a i have to solve it is to eliminate the electromagnetic influences. This can be done beginning with the design of the circuit/pcb . It is important to:
a) place the controller in to a metallic box
b) all the conductors (power and signal) must be shielded.
c) the pcb must designed with ground plane and components must be SMD if it is possible.
d) avoid ground loop.

I change the controller 16F876 with 18F2550 which has same layout (28pins). The 16F876 has some difficulties at memory access because the memory are split in banks. The 18F2550 memory can be allocated withowt problems. Another advantage of 18F2550 is 32K program memory.