During the last few months I have been reassembling the machine. Adjustable/retractable guide wheels have now been fitted, (7 off), and these should be reasonably easy to connect together and, perhaps, automate.
More angled magnets are now fitted around the circumference of the frame, and held securely with body filler. The stability does not appear to have improved, using slow hand rotation.
I have made a tachometer, using 3 x 1" LED displays. This has been hand built to enable me to interface it to the machine for control. It has its own seperate micro. This will display meters per second rotation.
The new 160mm long drive magnets have proved problematical to time the motor. A magnet length of 120mm (3 magnets) may have been a better choice, as the effective spacing of the Hall Effect sensors is 116mm. There is a short period when 2 mosfets are activated on the one side. That is we may have A High and B Low activated, and then C High will turn on before A High turns off. This was occurring during my original tests, and is not harmful but not beneficial either.
This has led me to make a change to the way the timing is done. Instead of 3 HE sensors, I've made a setup with 6 sensors in 2 sets of 3. The originals are the Primaries, and the new set are the Secondaries. All are mounted on a plate which is adjustable in relation to the coil positions, and the Secondaries are adjustable as a unit, relative to the Primaries. This can now switch the mosfets with great precision. A small set of 6 LEDs is used for the timing. The longer magnets leave much less gap to magnet ratio. 11% compared to about 25 to 30% normally, and hopefully will make the motor more efficient.
The decoding of the Hall Effect sensors was originally by means of two micros. The programming was not too difficult, but I found what seems to be an easier way to do the job, so have not tested the micro setup. When I ran the motor a few months back it was using the new timing method. Each sensor is connected to a dual OpAmp. There are 2 reference voltages. For one pole the voltage will rise over the reference and switch on and switch off as it falls below the ref voltage. The other pole is set to do the opposite. It switches on as the voltage falls below its reference. There is a clearly defined and adjustable off period. These OpAmps provide a Direct Drive to the Mosfets which control the motor coils. No computer or programming required.
To facilitate the extra Hall sensors a new board has been built. This uses quad OpAmps in place of the duals. The output from each OpAmp goes to a 2 input AND gate. That is, A High Primary and A High Secondary will input to one AND gate. Therefore the spacing of the sensor sets will determine the timing or duration of the switching. While I have made these modifications to overcome some inherent difficulties in my design, I have also made a system which will enable me to test the Lutec idea. Lutec is a self running motor/ generator. It is based on the principle of reducing the on time of the switching. They claim to be able to get down to about 12% on time. Effectively the BEMF drives the rotor a small distance, and the generator effect then produces a little more current, all with no loss of performance. Theoretically, my motor should do this with no reduction in switch on time.
Even at 100% of the possible on time I have an extra 33% magnet and coil trailing along doing something useful, hopefully.
Inexpensive drivers are available to decode the HE sensors and switch the motor coils. I have avoided them, so that I can have complete control over the process. I doubt that these could be used with the dual sensor setup, but it may be possible using some AND gates and 2 drivers.
My circuits have been designed using DesignSpark PCB freeware, and drawn by hand on a CAD program for glueing onto Veroboard. I have now purchased a copy of VeeCad which uses the DesignSpark output to autoroute onto a Veroboard template. Very easy to use, and not expensive. A free version is also available, but not as useful.