While re-thinking the starter design, I realized that I had forgotten about the pair of gears inside the wheel case that interconnect the manual and electrical starter shafts. This beveled gear set is a 2:1 reducer. That is, when the manual shaft is driven, the electrical shaft follows at twice the rpm. These gears divide by two the load seen by the electrical starter compared to that seen by the manual shaft, but from the manual shaft's perspective they are yet another gear set that will contribute to its lock-up. It now seems that any gear reducer used in conjunction with the electrical starter should be isolated with a one-way clutch.
Before further complicating my already complex Nichibo design with the addition of a such a clutch, I decided to take a serious look at the high torque starter motors I've been recently collecting. I have no electrical data on any of them, and I really hate blindly lashing electrical stuff together. But, I realły don't know the starting torque requirements of the Quarter Scale well enough to be sure the Nichibo is actually capable of starting the engine even while running at its ideal operating point. It loks like some experimenting is going to be required after all.
I just received the third and last eBay motor that I had ordered earlier. This one is a component in a starter for a John Deere JS-30 (6.75 hp 90 cc) lawn mower, and it was another suggestion from Naiveambition. This motor looks similar to the Nichibo, and its measured unloaded current draw and rpm are also similar; and so I'm not sure it fits my current definition of 'high torque.' The starter includes a 6:1 gear reducer. The motor has a 12 tooth 30 DP pinion gear pressed onto its shaft, but for some reason its 72 tooth driven gear is a non-matching 32 DP.
I thought my next approach to a starter might be to simply power one of the high torque motors with an electronic PWM controller and completely discard the troublesome mechanical gear reducer. My reasoning was as follows.
A typical gasoline engine requires a cranking speed of 100-200 rpm in order to reliably start. The small engine starters based upon the permanent magnet motors that I'm currently looking at typically use a 6:1 gear reducer. This means that these motors are capable of supplying the torque needed to start their engines while spinning between 600 and 1200 rpm. Of course, this assumes there is no additional gear reduction between the starter and the engine.
The Quarter Scale should have similar rpm starting requirements. Although its 350cc displacement is considerably larger than the engines these starters were designed for, the Quarter Scale's displacement is distributed over a number of small cylinders with small compression bumps rather than one or two large ones. Inside the Quarter Scale wheel case there is a built-in 10:1 gear reduction between the electrical starter shaft and the crankshaft. Therefore, the Quarter Scale's electrical starter shaft will spin need to spin at least between 1000 and 2000 rpm while cranking the engine at its starting rpm.
Hopefully, at these rpms, one of the fit-for-purpose starter motors will be capable of supplying at least enough torque to start the Quarter Scale. The downside of using a motor with the potential of supplying even more torque than required is that the engine's starting system must be stout enough to handle the starting power including the occasional abuse of an inadvertent fault such as a temporarily blocked prop. With some modification, a PWM controller should be capable of limiting the available torque to just what is required and at the same time provide a soft start to reduce shock to the gear train.
A new housing was designed. With no gear reducer, the need for an offset went away, and the motor was centered on the axis of the starter shaft. The diameter of the new housing was selected to fill the available space and included a notch to clear the pesky coolant pump. I designed the complex upper portion of the housing to attach to the wheel case independently of the particular motor used. The bottom plate of the housing, which is also the mount for the motor, was designed around the John Deere motor with the intention of modifying it for one of the other motors should testing show it to be inadequate.
An issue arose with the motor's pressed-on pinion gear. I didn't want to pull the gear from the shaft for fear of damaging the motor should it need to be replaced later. The steel Oldham shaft had to be blindly splined so it could be driven by the pinion. I didn't need to perfectly match the splines to the gear teeth, but they had to be stout enough to handle the torque that the resulting socket would be required to handle.
The only way I could think of cutting the splines in my shop was to mill them. I laid out the tooth profile of the pinion gear and then approximated a socket around it using a circular pattern of twelve .075" diameter drilled holes arranged around a plunge-milled center hole. Standard small diameter end mills typicalły don't have the half-inch flute length needed to remove the material left among the holes. But, in my collection of eBay carbide circuit board cutters, I found one that did. A .075" diameter cutter running at 0.5 ipm took a while to do the job, but the result turned out great. When completed, the pinion was a perfect slip fit inside the socket.
It was satisfying to finally start machining the components of the housing since I'd been drawing and re-drawing them for a couple weeks now. The housing and its bottom plate were machined from aluminum, and the splined Oldham shaft was machined from 12L14. The Oldham coupler was turned from 1144 Stressproof.
A CCM9NW 40 amp PWM controller, available everywhere including Amazon:
https://www.amazon.com/dp/B00RFDFL54/?tag=skimlinks_replacement-20
was used to control the motor from a 12 volt UPS battery. I selected this particular controller since its PWM frequency is switch selectable for 240 Hz, 2.2 kHz, or 22 kHz, and without some testing I didn't know the best frequency to use. An external pot is used to control the duty cycle of the waveform, and therefore the power, applied to the motor.
The first test involved spinning the crankshaft at 150 rpm without the spark plugs installed. I began with a PWM frequency of 22 kHz. The measured voltage at the terminals of the motor was 2.1 volts indicating a PWM duty cycle of about 18%. The average current draw from the battery measured 5 amps which, under an 18% duty cycle, indicated a peak current of 28 amps.
I then gripped the prop shaft as tightly as I could with a gloved hand to add some additional load to the starter. The motor easily overcame my grip, and at 150 rpm the voltage rose to 2.8 volts and the average current to 13 amps. At the new 23% duty cycle the peak current had risen to 56 amps which was actually beyond the recommended 40 peak amp maximum of the PWM controller. Assuming a motor efficiency at this operating point of, say 40%, I estimated (.4 x 2.8 x 13 x 5252 / 746 / 150) the torque delivered to the crankshaft (after the wheel case 10:1 gear reduction) to be about 0.7 ft-lbs which is considerably less than my earlier guesstimate of 10 ft-lbs at this point needed to eventually start the engine.
I performed similar tests using the other two available PWM frequencies, but 22 kHz worked best by far. The available torque decreased with decreasing frequency, and the duty cycle had to be considerably increased to obtain the same rpm. Switching noise emanating from the motor was very audible at 2.2 kHz, and at 240 Hz the available torque was so low that I could easily stall the prop shaft with my hand.
The next step was to repeat the tests with the spark plugs installed to provide a more realistic load on the starter. I re-ran the cranking test as I installed the plugs one by one. I could detect trouble as soon as the first plug was installed. The motor baulked at turning over as soon as that cylinder hit its compression stroke. Running the PWM wide open wasn't sufficient to get the starter to turn the engine over with just two plugs installed. Fortunately, my battery powered drill, driving the manual input, could still easily turn the engine over under the same conditions.
After a bit of actual hands-on experience, I'm gaining some appreciation for the torque levels I'll need to be dealing with on this starter. My concern about breaking parts inside the wheel case has also gone up a few degrees. I now better understand why John Ramm starts his engine by hand slapping the prop. - Terry