Another Knucklehead Build

[Ghosty]

Would the slight rise in temperature due to compression have a crank speed factor. More so on small engines. Temperature would be lost through the cylinder walls, piston and head, but that does not explain why a small engine is sometimes harder to start when warm.

One or two thou in a full sized engine between cold and hot is not that much of a problem, in our small engines it is the difference between a runner and not. We can scale down engines but things like thermodynamics don't scale down the same, the same with drag and momentum.
Cheers
Andrew

 

[tornitore45]

Terry I do not entirely disagree, the length of intake duct and air velocity is a valid consideration. My point was that compression as defined by fit and leakage is affected by the speed of the piston. As long as the piston moves fast enough to make leakage negligible, compression is not lost.

 

[mayhugh1]

I didn't do myself any favors by getting so far into this build without having a design for a working starter system. There isn't much space available, and the gear motor I've chosen will now require a rather elaborate oil sealed adapter to get it into the cam box. The shaft's location requires a system of belts or chains in order to drive the crankshaft. Draw-Tech had pretty much same issue which they solved with timing belts and pulleys. Several of the parts shown in the drawings are non-stocked long lead items, and so I decided instead to machine a set of sprockets and use the short piece of 3/16" roller chain left over from my Merlin build.

A gear motor requires a clutch, and for that I used an RCB-06014 one-way bearing that I had on hand. Its 56 kg-cm torque capacity should easily handle the Knucklehead's requirements. There's no space available on the crankshaft for a starter drive and so, as in the Draw-Tech design, the distributor drive shaft will be driven instead. Since the distributor runs at half the speed of the crankshaft, the starter motor's shaft sprocket was sized for an approximate 1:1 effective gear ratio between the starter and the crankshaft.

Machinist Mate 4.2 http://www.wadeproco.com/ was used to design the sprockets. A single sprocket tooth profile, defined in Machinery's Handbook, consist of five curves and two straight line segments and is a tedious and error prone construction. While shoehorning the Knucklehead's chain drives into the available space inside the cam box, I found that being able to quickly create accurate sprockets for trial-and-error placement was invaluable. I experimented with several online tools and compromise construction techniques which I used to generate sprocket models for comparison with a known accurate profile used in my Merlin's cam drive. The Machinist Mate program gave the best results by far. This program accepts the chain pitch, roller diameter, and number of teeth and outputs a single tooth profile as a dxf file. Its files easily imported into SolidWorks 2010 which I used to create the sprockets. The entire process from tooth specification to completed CAD model was less than a minute.

Another helpful design tool was one that calculated the number of chain links required for a particular center-to-center sprocket spacing. I used a pair of free online spreadsheets:

https://www.chiefdelphi.com/forums/attachment.php?attachmentid=11180&d=1323635128

One spreadsheet uses the chain pitch, the numbers of teeth on each sprocket, and a desired center-to-center spacing in order to calculate the number of required links including any partials. A second sheet uses the pitch, tooth counts, and number of desired whole links to calculate an exact center-to-center spacing. Since I didn't have room for tensioners, I was concerned about the margin built into the calculations. I tested the second spreadsheet using a pair of machined sprockets that I created using Machinist Mate and found the fit to be very acceptable. I learned the hard way, though, that because of the connector link design only even numbers of links are usable. This, in addition to my very short and irreplaceable chain remnant, were painful restrictions.

After completing the designs of the chain drives (with three links to spare) CAD models for the remaining components were created. The photos show the finished model of the starting system. For simplicity, only the cam box components directly associated with the starter or its clearances are shown. My ancient version of SolidWorks doesn't have the ability to functionally play with roller chains, and so I didn't expend the effort to just cosmetically add them. Clearances around the chains and connector links were major considerations though. In the photos, paired sprockets are identified using identical colors.

There seem to be differences of opinions about the minimum number (6-10) of teeth recommended on a sprocket. Since I was interested in conserving links as well as making the motor shaft sprocket as small as possible, I ran my own tests. I found that an eight tooth sprocket didn't seem to like my 3/16" pitch connector links as much as a nine tooth sprocket did, and so I selected nine teeth as my usable minimum.

In order to make sure that I have an actual working design before doing the final machining on the cam box, the next step will be to machine the starter components and assemble them into a starter that I can bench test on a mock-up plate. - Terry

 

[kvom]

I was out of the country the past 3 weeks. Your progress during that time made for a delightful early morning read.

 

[mayhugh1]

The motor adapter was machined from aluminum for a snug fit around the ServoCity series of HD planetary gear motors:
https://www.servocity.com/motors-ac...uty-gear-motors/premium-planetary-gear-motors
The motor I'm currently using is the 313 RPM version that I previously tested using my Howell V-twin as a load. Its gear-case makes the motor a little longer than I'd like, and so I'm using the adapter to hide some of it inside the cam box. The adapter was machined for a shaft seal, and a .010" teflon gasket between it and the cam box should prevent oil leaks.

The motor sprocket was machined from 1144 to closely fit the 6 mm D-shaft. There's a hub for a grub screw, and between it and the sprocket is a clearance groove for the chain. One of the issues that I wrestled with during the design was providing enough clearance around the chains for the connector links which are somewhat wider than the chains themselves. Space around the sprockets is so tight that some of those clearances are only .010".

The primary chain connects the motor to an over-running clutch. The clutch is actually an RCB-06014 one-way bearing running on a countershaft located between the motor and distributor drive shaft. Its exact location had to satisfy the 'whole link' requirements of both chains and in doing so influenced the sizes of all four sprockets. The result compromised the 1:1 effective gear ratio that I was trying to maintain between the motor and the crankshaft. I ended up at 1:1.2 which reduced the available torque at the crankshaft by 20%. I had been toying with the idea of replacing the 313 RPM motor with the 437 RPM version, but now seems counterproductive. A 165 RPM version with twice the torque is available if necessary.

Although the countershaft has its own sprocket and spins on its own pair of end bearings, it's also the inner race of the one-way bearing. During starting, the countershaft is locked by the clutch to the starter motor through the primary chain. When the engine starts, the secondary chain linking the distributor drive shaft to the countershaft over-runs the starter and the clutch disengages it.

The sprocketed countershaft was also machined from 1144. Manufacturers recommend hardening the inner shafts used in one-way bearings, but I was concerned about my ability to correct any warpage that might be created by the heat treatment. Instead, I decided to rely upon Stressproof's extremely high tensile and yield strengths to withstand the bearing's sprags without deforming. Although it's not a guarantee that 1144 is up to the task, I've been running this same bearing inside the Merlin on a similar shaft with no issues so far. Inside the Knucklehead it will be running at less than half its rated torque.

A one-way bearing has a thin drawn outer shell that requires a pressed-on backup sleeve. If the shell's o.d. is measured, it may appear to be a couple thousandths oversize and out-of-round. Regardless, it's important that the i.d. of its pressed-on sleeve be exactly that specified by the manufacturer. It's also advisable to use an arbor similar to the one in the photo during the pressing operation in order to support the bearing's internals against damage. It's easy to become confused about the bearing's direction, and over time I've learned to check twice and press once. My particular sleeve including its integral sprocket required the full capacity of a two ton manual press.

The distributor drive shaft is geared to the crankshaft, but it's also linked to the countershaft through the secondary chain. In addition to its integral sprocket, this shaft contains additional features for mounting the crankshaft-to-distributor reduction gear as well as a miter gear for driving the distributor. The reduction gear is bolted to a flange on the shaft, but the miter gear's hub will be inserted into a recess machined into a rear face on the shaft. The miter gear will be Loctited in place after the distributor is completed and the gear's exact location on the shaft can be determined. This recess was machined using a tiny face grooving tool ground from a .042" HSS drill bit that I soft-soldered to the end of a piece of steel.

My confidence in the design grew while machining its components, and so I decided to scrap the idea of a trial assembly on a mock-up plate. Instead, I finished the cam box machining and assembled the starter inside it.

It was a great relief to find that the shafts lined up properly between the cam box and its cover. The cover is doweled to the cam box, all the bearings are snug press fits, and the shafts are close fits inside their bearings; and so there was little room for error. Once the alignments were verified, I fine-tuned the lengths of the shafts for a .003" thrust clearance while keeping the paired sprockets at the same height. The chains appeared to have the proper fits, and they ran interference-free with the cover in place. I was disappointed, though, when I realized I wasn't going to be able to watch them running. The sprocketed shafts require the support of their outer bearings, and the starter can only be safely run with the cover in place.

Additional good news was that I somehow managed to properly install the clutch. The flywheel spins in the correct direction when the motor is energized and freewheels when it isn't. Since the engine isn't far enough along to build compression, the only load I could put on the starter was my uncalibrated hand on the flywheel. The starter sounds good though and has a whine that's reminiscent of a full-size engine. - Terry

 

[kvom]

How did you cut the teflon gaskets? I've been using an xacto knife. I like the drill fixture. I've been using a manual punch that is likely less accurate.

I must have missed what the purpose of the miter gear is.

 

[mayhugh1]

Kvom,
I cut the gaskets using a drag knife on my Tormach. I should have taken a picture of that step, but I forgot. It was just a matter of taping the teflon down to a sacrificial surface under the spindle and cutting a pair of circles. The 2-56 holes were too small for the knife, and so I stacked a few gaskets and drilled through them using the fixture shown.

The miter gear is for the distributor that still has to be made. It will sit on the roof of the cam box and protrude down inside it. Its shaft will have a matching miter gear on it so it can be driven by what I've been calling the distributor drive shaft. The distributor drive shaft is geared down from the crankshaft by 2:1 and the 1:1 miter gear set will ultimately spin the distributor at half the speed of the crankshaft. - Terry

 

[mayhugh1]

A final addition to the starter was a cosmetic cover for the motor. It was turned and polished from a chunk of 1-3/4" diameter aluminum and then machined with cutouts for the battery terminals. I covered up the air vents over the brushes in order to keep exhaust grunge out of the motor. Hopefully the motor will never have to be run long enough to overheat. Clearance slots for the adapter's mounting bolts allow the cover to slip over the motor and completely enclose it.

Large permanent magnet dc motors often have an external spring clip over the outside of their housing to shunt any residual flux from the stator magnets. The low permeability path provided by this steel sleeve reduces nuisance external magnetic fields and provides a small torque improvement by increasing the flux density in the motor's air gap. My interest was in sleeve's inherent springiness which I used to keep the cover in place.

I didn't end up with the final gear ratio that I had been assuming during the design of the starter, and so I ordered the 165 rpm version of the motor as a backup. It spins the crankshaft at 200 rpm and provides plenty of torque. Unfortunately, it has an even longer gear box and requires its own cover. I wanted to move on to the distributor and be done with the starter, and so I also machined a cover for this motor while the fixtures and the setups were still fresh in my mind. With this motor spinning the engine, it's nearly impossible to hold onto the flywheel. - Terry

 

[mayhugh1]

The Knucklehead's distributor is something of a misnomer because it really doesn't distribute anything. It has a rotor that opens and closes a set of points or, in my case, triggers a Hall device; but the plug voltages aren't brought inside the distributor. Not having a high voltage section greatly simplifies the design, and the Hall sensor catches a big break. As in the full-size engine, each plug will be connected directly to its own coil, and the trigger will simultaneously fire them both.

This 'wasted spark' or 'dual-fire' system is used today in modern distributor-less automotive ignitions. In even-firing engines there's no adverse effects because the cylinders can be paired so the wasted spark always occurs on an exhaust stroke. This isn't quite the case with uneven-firing Harleys, however.

In a Harley engine the rear cylinder is ahead of the front cylinder by 45 degrees which is the angle of the engine's V. Initially, assume zero spark advance. When the rear cylinder fires at TDC, the front cylinder will be in its exhaust cycle and 45 degrees before its TDC. This cylinder's wasted spark causes no problems so long as the burn from the previous cycle was complete. However, when the front cylinder fires, the rear cylinder will be in its intake stroke and 45 degrees past its TDC. If conditions are right (long duration cam and/or overly-rich carburetor) there could already be a combustible mixture in the rear cylinder, and its wasted spark could create a backfire through the carburetor. Fortunately, timing advance helps to mitigate this. With 30 degrees of advance, the rear cylinder will still be in its intake stroke but only 15 degrees after TDC. It's still possible, however, for exhaust gas to suck in enough fuel to be combustible due to cam overlap. In any event, Harley switched their ignitions from dual-fire (wasted spark) to single fire in 1999 even though the aftermarket had already been offering them for several years. As an aside, that rear cylinder takes even more punishment by not receiving the same amount cooling that the front cylinder enjoys. Anyway, back to the Knucklehead's distributor ...

I reduced the diameter of the distributor from one inch to 0.8 inch because of clearance issues with that pesky oil fitting at the front corner of the front cylinder. I also reduced its main body height to clear the front cylinder's intake fitting so the distributor could be installed or removed without major engine disassembly.

Although distributors are typically negligible loads on the gear trains driving them, the 48 pitch 18 tooth brass miter set specified in the drawings seemed a bit lightweight to handle the original distributor's sleeve bearing'd rotor shaft. Just in case, I changed the design so the rotor shaft runs in ball bearings.

Because the camshaft will be redesigned, the rotor was modified to accommodate the reversal in the crankshaft's direction as well as the conversion to the 315/405 degree firing angles that the engine is famously known for. Looking down on the top of the distributor, the rotor will now spin clockwise, and its second trigger will be located 315/2 = 157.5 CCW degrees behind the first.

I decided to use an aperture disk to trigger the Hall device rather than multiple magnets. I've used both methods in the past but felt like I got better cylinder-to-cylinder consistency with the single magnet approach. Before finally deciding, I tried to compare the relative trigger distances among a number of supposedly identical magnets that I had on hand using an admittedly crude test setup. I could easily see up to .010" differences in the turn-on distances of my test Hall device which would translate to 5 (crankshaft) degree timing errors in my little distributor. The differences in the distances needed to turn the device back off (this is the trigger edge I actually use) was smaller, though.

I decided upon an Infineon TLE4905 Hall device (obsolete, but still available from digikey.com) and a stacked pair of 1/16" diameter magnets as a flux source. Bench experiments and some trial disk machining were done to derive the specs for the disk. I eventually settled on a .020" thick 12L14 disk with .125" diameter holes rotating over the center of the Infineon device. The total gap between the magnet and the sensor's face will be .035".

For an ignition, I plan to use the twin CDI unit from cncengines.com. I emailed Roy and was assured that this new style potted unit contains two separate coils and a 2X storage capacitor. This means each plug will receive the same energy that would have been available from a pair of his conventional single coil modules.

I've included photos of the CAD models showing the components of the redesigned distributor. A strain relief for the Hall sensor cable that will pass my 'tug' test still needs a little work before I start making chips. - Terry

 

[mayhugh1]

The distributor components were machined according to the dimensions behind the previous CAD models, and assembly went pretty much as expected. If I were to do it all over again, though, I'd use 303 stainless instead of aluminum for the distributor body. The fit of the distributor inside its bore in the aluminum cam box is snug, and even with a bit of oil, I think about galling. An extra .015" material was left on the lower end of the distributor body, the rotor shaft, and the miter gear hub. During assembly these were trimmed as needed to fine tune the mesh between the distributor's driven gear and its driving gear inside the cam box.

I tried using Loctite 680 bearing retainer to secure the magnet pair inside their recess, but even after three attempts it refused to set up. It's possible the chemistry behind the curing process is affected by magnetism. Although I don't know if they're necessary, I learned later that Loctite has adhesives specifically formulated for use around magnets. A drop of super glue seemed to work, but I top-coated the around and up to the magnet with a thin layer of JB Weld for good measure.

The mount for the Hall sensor was machined from PVC so the JB Weld used to bond the two would have some bite. A less rigid adhesive like GOOP might have been a safer choice for use over the device's sealed leads. I liked the idea of using JB Weld, though, since its metallic filler might provide some electrostatic protection.

Some careful soldering was required to attach the three 30 AWG insulated wires used to make up the sensor cable. The joints were carefully covered with JB Weld to insulate and keep them separated from one another. A piece of heat shrink tubing was embedded in the epoxy for strain relief, and the entire junction will eventually be hidden in a machined pocket in the distributor cover. The far end of the cable was terminated in a Futaba male servo connector. Since there was some technique involved with mounting and cabling the sensor, I made up some spares while the process was still fresh in my mind. I ended up modifying my design for the distributor cover shown in the earlier CAD drawings in order to provide a nicer exit for the sensor cable.

I received an email this week asking how a Hall device could possibly detect the presence of a magnet through a hole in a steel disk when the hole is only a bit larger than the magnet itself. It was a good question because one would expect the disk to capture nearly all the flux leaving the north pole at the rear of the magnet. On its way back to the south pole at the front of the magnet, the flux would just be channeled through the disk and around the hole. The disk, even with its hole, provides a low reluctance path compared with the surrounding air, and so one would expect very little flux in the air above the disk.

Modern Hall sensors, however, contain internal flux concentrators in order to improve their sensitivities. Although its design is often proprietary, the concentrator is essentially a low reluctance path integrated into the packaging and used to optimally channel flux through the sweet spot of the sensor. In doing so, it offers an alternate path to the magnet's flux while the sensor is over the hole. With the right size hole and sensor spacing it can offer an even lower reluctance path than the disk and draw some of the flux out of the hole. - Terry

 

[mayhugh1]

More Photos ...

 

[Scott_M]

Hi Terry
Beautiful work as always ! I really liked the nice job you did with the JB Weld. The connection looks very well terminated. Did you mold it or shape it while it was still soft ? It looks really good.
Did you use the regular JB Weld or the new fast curing stuff ?
And thanks for the match sticks, that tells all.


Scott

 

[mayhugh1]

Scott,
Thanks. I just used regular JB Weld and dabbed in on with a toothpick under a magnifying glass. Gravity took care of the rest. - Terry

 

[Buchanan]

Terry.
The way you make the distrubter blend into the overall design is perfect.

 

[XD351]

That is brilliant !

 

[mayhugh1]

Although it was probably a little premature, my next step in the build was to fabricate a set of drag pipes for the engine's exhaust. These were formed from half inch diameter 316 stainless tubing using a Rigid three-roller tubing bender. Although I received mine several years ago as a Christmas gift, I noticed this bender (along with most everything else nowadays) is available through Amazon:

https://www.amazon.com/dp/B008ULY74G/?tag=skimlinks_replacement-20

The rear pipe was simple and easily formed, but it took three tries to get a matching front pipe.

The exhaust ports on the heads are terminated in floating o-ring'd flanges that require matching header flanges on the exhaust pipes. These were machined as separate items and then silver-soldered onto the ends of the formed tubes. For joint strength, my approach was to insert the tube into a close-fitting recess in the flange and then coax molten solder to wick up between them. Since the clearances between the tubes and the heads of the flange screws will be minimal, it was important to not end up with an interfering solder fillet. In what turned out to be a lame and totally unnecessary attempt to prevent this, I blackened the flange surfaces with a Sharpie pen as I often do when low temperature soldering. I'm certain the coating vaporized long before the melting point of the solder was reached.

For solder, I used .030" Silvaloy 355 wire (available from Brownells) and white fluoride flux. This wire has the stiffness of spring wire, and a short piece was used to form a tight fitting ring inside a recess partially bored through each flange. The diameter of the recess was .005" over that of the tube and provided the necessary capillary space between the two. The ring as well as the tube's o.d. were coated with flux and the assembly setup on a thin steel plate so the flange could be indirectly heated from below with a torch. This prevented the flux from becoming scorched before the solder reached its melting point. At just the right time, the torch was moved above the plate and splayed around the tube in order to draw the solder upward. After cooling, the baked-on flux was removed by pickling the assembly in sulphuric acid (drain cleaner from a local builder's supply). Neutralization in a baking soda/water solution followed by a thorough water rinse completed the job.

After finishing the pipes, I began having second thoughts about using (even hard) Viton o-rings in between the exhaust flanges. I wasn't concerned about the exhaust heat so much as I was about the weight of the pipes. The floating flanges greatly simplified the drilling of the screw holes around the header flanges, but the friction provided by the compressed o-rings is all that prevents the pipes from rotating on the heads' exhaust ports. This friction is holding the pipes in place now, but the engine's vibration might create problems later. I considered adding a clamp to hold the ends of the pipes together, but I wasn't sure I liked the added clutter.

Instead, I made up a Plan-B set of copper o-rings that I turned from a piece of a discarded electrode from a spot welder. In order to improve their sealing ability, I turned a .010" deep face groove around the center of both faces of each o-ring. This divided their flat faces into two .020" wide rings which will increase the effective pressure available from the flange screws. After completing their machining, the o-rings were annealed to improve their conformity to the surfaces of the flanges.

Finally, the pipes were polished using 400, 600 and 1000 grit papers followed by buffing with red jewelers rouge. The complex reflective surfaces nicely hide imperfections left behind by the bender that weren't removed by sanding. - Terry

 

[petertha]

Terry, now that I see your complete exhaust fit-up, I am seriously thinking about re-engineering the intake/exhaust system on my radial. The head design inlets are tapped M10 for a coupler nut which clamps the 8mm tube to an inner counterbored face via a raised flare on tube end. There are other bits & pieces I wont go into detail, but collectively seems a bit fiddly & iffy to me seal wise & strength wise compared to a permanently attached header boss stub glued & set screwed into the head like you appear to be doing. And then the pipe gets attached via bolted flanges + gasket. Coincidentally this is the second instance I have seen a similar technique, so perhaps divine intervention.

Anyway I wanted to draw your attention to link below starting ~ post #245 where Mike discusses his choice of adhesive (JB weld vs Loctite type retainer) on similar header bosses & also his torch & wrench bench test. I know you have used HT retainer on valve cages etc. with good results, just wanted to mention for reference. In my case I don't have as much insert surface area working for me. And from what I could determine from JB 'specs', the high temp capability did look pretty decent all things being equal. Did you ever do elevated temp bench tests yourself?

Mikes W165 Grand Prix engine http://www.modelenginemaker.com/index.php/topic,5142.240.html

 

[mayhugh1]

Peter,
Thanks for the great link. I just don't seem to get over to the other forum as often as I should.

I've never done any direct comparative tests between the temperature performances of JB Weld and high temp Loctite, but I did do some testing on JB Weld during my 18 cylinder radial build:
https://www.homemodelenginemachinist.com/threads/another-radial-this-time-18-cylinders.21601/page-12

See post #237. The JB Weld seemed to work to 500F, and from other testing I've done, I'm sure the Loctite would have given up before that. The prop wash over the heads of your radial will keep the head and exhaust temperatures well below the critical temperatures of either of these, and so either will probably work well for you.
Terry

 

[petertha]

Thanks for reminding me exactly where you discussed all that adhesive stuff. Excellent resource.

Another question. It seems to me you used a similar Ridged style tube bender in the past for smaller diameter pipe. I get the impression they come nominally sized like 1/4, 3/8, 1/2 etc. You don't replace the forming head & rollers, you buy the tool for the tubing size, correct? What was the wall thickness of your SS pipes & did you have to do any pre/interim heating or core filling on those particular bends? They tuned out beautiful btw.

 

[mayhugh1]

Peter,
They do come in standard sizes for tubing, and the rollers aren't replaceable. I fixed the Amazon link I originally posted and so you should be able to search around for the others. I have a couple of these benders that I've used to do all my tubing bends since I like using stainless. No heat or core filling is required. The three rollers do about as good of a job as one could hope for without going to a mandrel bender. The wall thickness on the half inch tubing I used is .050". - Terry