Not sure that would work at low revs. The lighting coil voltage will be even lower than the ignition coil voltage.
I think I've come up with a circuit to by-pass a blown bulb automatically. Bit more work to do on it.
The brake light bulb is 15 W, 6 V, so designed for 2.5 A. That will cause a drop of 17 V if the resistor alone is in circuit. That probably won't matter. However, the power dissipation will be 42.5 W, so the resistor may not last very long.
It's not that simple, bearing in mind the coil inductance and the points dwell time.
Running the sim with the reg disconnected I get ~15W in a 15W (2.4 Ohm) bulb at 2800rpm. With the 2.4 Ohm replaced by 6.8 Ohm I get 10.5W in the resistor at the same rpm and points dwell angle.
It's not that simple, bearing in mind the coil inductance and the points dwell time.
Running the sim with the reg disconnected I get ~15W in a 15W (2.4 Ohm) bulb at 2800rpm. With the 2.4 Ohm replaced by 6.8 Ohm I get 10.5W in the resistor at the same rpm and points dwell angle.
OK, I haven't run the simulation, but the points are closed most of the time. In one half-cycle, and in about half of the other cycle, the current is limited by the inductance of the coil and the resistance. The ignition is supposed to work with a short to ground or with the 2.4 Ω of the bulb in the way. On your calculation, the current drops to 1.24 A with an 8.45 V load (the 6.8 Ω resistor). You say that with 6 V load, the bulb, the current is 2.5 A. With a bit of trial and error, I think that works out at about a 9 V emf from the coil, with around 2.5 Ω of inductive impedance.
However, if the engine speed is higher, the emf will go up, as will the impedance. If the speed is 8400 rpm, then there will be 27 V of emf and 7.5 Ω, and the power will increase from 28 W in a 2.4 Ω resistor to 48 W in a 6.8 Ω resistor.
The resistor is a far from ideal solution. As well as risking burning out at high revs, it will significantly reduce the current at low revs, and take quite a bit of current away from an intact bulb.
My local electronics store showed NTE2991 as a substitute transistor for the one Alec called for (AUIRF3205) in his design. I'm not sure how close the match is, but I ordered NTE2991 and it should be here next week.
1. Will the regulator still work with this component?
2. Should I change other components to match this transistor's specs?
Data sheets for both transistors are attached.
Regarding the brake light resister, I put it in because some vintage mopeds had such a design. Once the regulator get's built, the resistor is coming out!
Not sure about the NTE2991 FET. The datasheet gives Rds(on) as "8.0" but confusingly gives the units as "-<" . If that's 8mΩ then fine; but if it's 8Ω then that FET unfortunately won't do, even with other components modified. When it arrives we can design a test to tell us. BTW, any MOSFET is static-sensitive, so take due precautions.
Hence the effort to design both a regulator and an efficient failed-bulb bypass circuit. I think I'm getting there with the latter; a bit more simming to do.
Not sure about the NTE2991 FET. The datasheet gives Rds(on) as "8.0" but confusingly gives the units as "-<" . If that's 8mΩ then fine; but if it's 8Ω then that FET unfortunately won't do, even with other components modified. When it arrives we can design a test to tell us. BTW, any MOSFET is static-sensitive, so take due precautions.
I see, how disappointing that the specs are incomplete. Rather than test the mosfet, it might be easier to insert in the circuit and test the device performance overall.
The test circuit could be just:
1 off PP3 9V battery,
1 off 100Ω resistor,
1 off 3k3 resistor,
1 off FET.
Here's the bulb bypass circuit sim:
Operation: When the bulb is good, the voltage dropped across D1/D2 is enough to turn on Q1/Q2/Q3 thus keeping sensitive-gate SCR1 and SCR2 off. When the bulb fails open-circuit, even if the magneto coil voltage is as low as ~2V (as in the instance shown), Q1/Q2/Q3 are off and current through R3/R4 turns on SCR1/SCR2 thus providing a near short across the bulb.
I don't quite see why you need a failed bulb bypass circuit, if a shunt regulator is fitted. Surely keeping the peak voltage to 6 V or so would keep the ignition running?
Hopefully yes. However, if at low revs you inadvertently applied the rear brake the magneto voltage would not be enough to turn on the FETs in the regulator and the engine would stop if the bulb had failed. That could be annoying in traffic.
Here's a simple test circuit for the FET. A Vds measurement > 50mV would indicate the wrong FET type.
The same circuit can be used to test an SCR as used in the bulb bypass. R1 is optional for the FET test, but not for the SCR test.
Don't keep the battery connected for long, or it will die .
That would certainly be simpler, but won't work well. To get an impedance ~2.4 Ohms at low/modest revs would require a cap value > 2000uF. Fine with a blown bulb, but the current in a good bulb at low revs would be reduced beyond the useful point. At high revs the good bulb would get an even smaller share of the current.
Diodes: any general purpose silicon, e.g. 1N4148, 1N400x, ...
Resistors: 1/4 Watt
FETs: as shown, or other automotive type, low Rds(on), each able to handle at least 15W for as long as the brake is 'on', package easily mountable to heatsink, must have body diode which can handle >5A.
Note: Without electrical isolation the FETs would need separate heatsinks and the heatsink(s) could not be bolted directly to the bike chassis.
I thought the FETs had insulated cases, so I could mount them directly onto the heatsink (the heavy die-cast body of the enclosure). Do I need to add an insulator between the FETs and the sink?
... and some have a metal tab but it is isolated from the FET silicon. Unfortunately, the NTE2991 datasheet doesn't specify an isolated tab and we have to assume the tab is connected to the FET drain, so in the regulator the FET that doesn't have its drain connected to ground will need amounting kit to insulate the tab from the heatsink. It won't then matter if the heatsink contacts, or is fixed to, the bike chassis.
I'm hopeful the ambiguous NTE2991 Rds(on) spec is ~8mΩ, since the datasheet lists one feature as "ultra-low on-resistance".