The Cam

Although a Harley-Davidson is a "45-degree V-twin," the HT leads of the Bosch ZEV magneto don't fire 45-degrees apart. The configuration of this engine is equivalent to a vertical twin (e.g. BSA 650) whose cylinders have been split to be at 45-deg. rather than 0-deg. with respect to each other. As a result, lobes one and two on the points cam need to be 157.5 degrees apart. The reason for this is, since the magneto operates at half engine speed, 157.5-deg. corresponds to 315 degrees of rotation of the crankshaft. Thus, if one cylinder fires at 0-deg., the second fires after the engine turns 315-deg. further, i.e. 45-degrees less than another full revolution (360-315=45). It then takes another 45-deg. for the crankshaft to get back to the first cylinder, which is now on its exhaust stroke, plus another full rotation before that cylinder is ready for another spark, which is 405-deg total (45+360=405). So, the angular separation between the second lobe and the first needs to be half that, i.e. 202.5-deg. Since 157.5 + 202.5 = 360, this all works out.

[Linked Image]

I used my Talyrond roundness tester to check the contour of the cam itself. The resolution of this instrument is overkill for what is needed, since it lets me center and level components to within 10 micro-inches, measure heights to 10 micro-inches, and determine angles to arc-seconds . However, if I didn't have the Talyrond, the 15 arc-sec. resolution of my mill's rotary table, along with a mechanical dial test indicator, would have been fine for these measurements (although not as fast or convenient to align as the Talyrond). Given how bad other aspects of the restoration were, I was afraid I might find the cam to be a poor tolerance attempt to be 45-deg., or even a 42-deg. cam for an Indian. If this were the case, I either would have to quickly find a replacement (that might turn out to be no better), modify it, or machine one from scratch. However, I was pleasantly surprised to find the angular separation between the lobes at the positions where the ramps would open the points by 0.012" to be precisely 45.0 degrees (or, rather, 157.5 and 202.5 degrees).

Unfortunately, the contour of the cam alone does not determine where this magneto will fire, since the OD of the cam ring locates it in the ID of an adjustable housing which in turn is a sliding fit in an annular slot in the magneto. Because of this, there is no way to know without measurement how the various tolerances add up to affect the final timing. So, once I knew that at least I was starting with a good cam profile, I assembled it in the magneto, centered the axis of rotation of the armature precisely on the axis of rotation of the Talyrond, and again measured the cam profile with respect to that axis.

[Linked Image]

I found when the cam assembly is installed on the body of the magneto the tolerances do add up to cause some deviation of the timing from the 157.5/202.5 values. However, most or all of the problem is intrinsic to the design of the magneto itself, which allows a few thou. side to side motion of the cam because of the need for a sliding fit of the assembly for advancing and retarding the ignition. Because of the slope of the ramps on the cam, a sideways displacement of a few thou. has a significant effect on the timing. However, the deadline for getting this magneto finished and shipped back to the guy who is rebuilding the engine is rapidly approaching, so there isn't time to re-engineer the design and fabricate something with tighter tolerance than it had when it left the factory. I will proceed with it as it is but, if I find the timing too far off in my later dynamic test, I will have no choice but to devise a solution.

Concentricity of the Armature Components

To be sure the components of the armature are properly Concentric after I assembled it, I used a precision bench center with four dial test indicators of resolution 0.0001" (ten-thou.). I reconstructed the setup for the next photograph a few days later using a different armature and only three of the four indicators.

[Linked Image]

In fact, this is not how I hold armatures in place for these measurements, since the races determine the axis of rotation, not the ends of the shafts. However, the fixtures I designed to locate armatures with respect to the races in this bench center would completely obscure what is going on, so what I show here is more of a schematic to illustrate the important points. The indicator at the left is shown measuring one of the races. In this setup, if the race were perfectly Concentric with the dimple in the end of the shaft, the indicator would not waver as the armature was rotated. However, what is important are the races, not the dimple, so the fixtures not shown here are designed to use the races as the reference surfaces. Similarly, the middle indicator checks for the run out of the slip ring with respect to the race, the third indicator would check for the run out of the earth ring if this were a more modern BTH or Lucas, and the four indicator (not shown) would check for the run out of the second race. The Bosch uses a different surface for the earth brush, so actually I used an indicator on the fixture located on the second race rather than the armature housing.

Another subtle alignment issue that also needs to be checked is if the shafts at each end are coaxial with each other. As an example of why this could be a problem, imagine if while reattaching the end cap that I caught a metal chip between it and the body. This would cause one shaft to be tilted with respect to the other. While such an alignment problem may not be likely when carefully assembling an armature that left the factory as one piece, it certainly cannot be relied on when dealing with a rebuilt armature that might be a mongrel made up of components from several others. The operational consequences of this would be that it would misalign the bearing races (causing premature wear).

Since the shaft at one end of the armature is stubby, checking alignment accurately presents a measurement problem. Although I used the Talyrond for this measurement, I could have done it more precisely than the way I did, albeit only if I spent more time setting it up. Instead, it was faster to check if the shafts were coaxial by installing the points block in the taper of the armature's shaft and measuring the run out of the back face of the block near the outer edge. When I did this I found the run out was several thou., but it also varied by a few thou. each time I removed and reinstalled the block. In any case, this is small enough not to be a concern, so finding the source(s) of the small run out and reducing it further wasn't necessary.

Although the armature itself was in good alignment, I determined that there was almost 0.01" of run out of the surface of the slip ring, which would cause unwanted motion of the brushes in their holders. So, I mounted the armature in the lathe using a collet to hold the race at one end (again, because the race is the surface that determines the axis of rotation), and a high precision live center at the other end, as laid out schematically in the next photograph using a different armature. The total indicated run out (TIR) of the spindle of my lathe is less than 0.0002", and the TIR of this precision center is 0.0001", so this machining operation results in an armature that is at least as good as it was when it left the factory c1920. I then polished the track to achieve a measured roughness of just under 1.5 um.

A critical observer will notice the lathe collet clamps on the outer surface of the race, not the track the balls run on. However, I've verified with several races on the Talyrond that these two surfaces are Concentric to at least 0.00002" (20 millionths), and that on this particular race they are to at least 0.0001". The lower precision is because I measured it using a mechanical indicator in the assembled configuration on the bench center rather than with the electronic indicator of the Talyrond. Also, it would have been more precise if the live center somehow could have been fixed to the race at that end of the armature rather than in the inner taper of the shaft. However, to do this would have required machining another precision fixture, and any added precision from going to this effort would have been negligible anyway. That race is ~4" from the center of the slip ring, which in turn is only ~0.5" from the race held in the collet. Because of this ~8:1 leverage, the ~1 thou. run out I measured for the race at the tailstock end with respect to the ID of the shaft as held by the precision center added a "wobble" of only a ten thou. to the slip ring when mounted on the lathe. This is comparable to the TIR of the lathe's spindle, and is at least 10x better than it needs to be.

[Linked Image]

Send questions or comments to [email protected]