Another Knucklehead Build

Home Model Engine Machinist Forum

Help Support Home Model Engine Machinist Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
More Photos ...
341.JPG
342.JPG
343.JPG
344.JPG
 
Just beautiful Terry ! Like all of your builds, I am really enjoying this .
Many thanks for the time you take to document and photograph all of your setups and work. I am sure it adds considerable time to the build, I for one really appreciate it.

Scott
 
Spray bars inside the rocker boxes lubricate the rocker shafts with pressurized oil from the pump. Some of this oil will end up inside the valve boxes where it mustn't be allowed to accumulate and flood the valve guides - also a common source of leaks in full-size engines with aftermarket oil pumps. Four 1/8" copper lines, for which scaled compression fittings were machined earlier, drain these boxes and return their oil to the crankcase and cam box.

The two lines that drain the inner valve boxes have to be routed through the space between the heads. In the original design they were intended to be tee'd into a fitting in the center of the roof of the crankcase. The intake manifold, however, takes up a big chunk of this space leaving little room for connector'd lines. The Draw-Tech CAD renderings aren't real clear about the routing of these lines nor the design of the particular fittings associated with them. None show the manifold and oil lines together in a common view.

After installing the intake manifold, it was obvious there wasn't room for the fittings that I'd made. For these two lines I had to machine a pair of banjo fittings. One of the photos shows their component parts. Since the space inside the valve boxes is also pretty limited, I machined a notch in the end of each banjo bolt to provide a freer flowing oil entry. In order to insure the notches face upward where they will do the most good, each bolt was custom machined for its particular box so the thread orientations could be accounted for.

Earlier, I'd moved the return line for the outside front valve box from the crankcase to the the cam box because of a possible clearance issue with the front cylinder. This freed up its crankcase fitting which I was able to use for the front inner return line and therefore do away with the tee.

After finishing the inside lines, I formed and installed the return lines to the outside valve boxes as well. This time I used a piece of 1/8" diameter plumbers' solder for the trial-and-error bending/fitting which was much easier to work with than the aluminum rod stock I had been using.

After assembly, each return line was tested by filling its valve box with oil in order to make sure it was capable of draining the box in a timely manner and without leaks. The boxes drained their room temperature 40w oil in just under a minute. The oil that accumulated in the crankcase was nicely contained by the o-ring seals between the crankcase halves and between the crankcase and cam box.

The oil pump's external line was the final and most complicated leg of the engine's oil line plumbing. A copper line from a fitting installed on the top rear of the cam box feeds oil to a tee located on the engine's flywheel top side. The output lines from this tee supply oil to the spray bars through compression fittings mounted on the rear of the rocker boxes.

The tee itself was made up from 3/16" silver-soldered copper tubing since its i.d. nicely matches the o.d. of the 1/8" tubing. It took a couple tries to get an acceptable part, though, because of the 'quick and dirty' setup used to hold the parts together during soldering. The 1/8" copper lines were soft soldered to the tee with the entire assembly in place on the engine. The three joints were fluxed with activated rosin and wrapped with ringlets of low-temp solder. Although a tiny butane torch was used, the tee's small size made it difficult to simultaneously heat all three joints without one or more of the rings melting and falling away due to the radiated heat. I don't like hand feeding solder because of the extra cleanup that's usually required, but in this case I didn't have a choice.

Again, the intake manifold proved to be an obstacle, and this time it forced the routing to be unsatisfying asymmetrical. After studying photos of the routing of the lines in the full-size engines I noticed they weren't all that tidy either, though. - Terry


345.JPG
346.JPG
347.JPG
348.JPG
349.JPG
350.JPG
351.JPG
352.JPG
353.JPG
354.JPG
 
You are a master of engines .
All the details of the engine are great !
 
I'm blown away by your attention to the smallest details & engineering you're putting into this Terry. Very well executed & I just love that bead blasted finish. Just awesome!

John
 
While out of town during the past week, I managed to do some groundwork for the Knucklehead's starter. The Draw-Tech starting system was designed around a Mega AR390031 (15 turn) 540 size brushed dc motor. This is a popular and widely available motor once used in a number of cordless tools and in some RC cars and trucks. Draw-Tech designed a system of pulleys and belts to connect it to the engine's crankshaft with an effective gear ratio of about one. The BOM also calls out a 4.8 volt Li-Ion battery pack to power both the starter and ignition.

Within the same case size, manufacturers fine tune a motor's performance by varying the gage and number of turns of the wire on its armature. The AR390031 was tuned for torque. At 4.8 volts, its no-load speed is roughly 20 krpm at 1.9 amps, and its stall torque is .8 kg-cm at 38 amps.

85E07A0C-5DB7-4375-897B-E05E99258753.png


Jameco sells the Nichibo KP5FN-3255 which is representative of what 12 volts can deliver in the same size case. Its no-load speed is 5 krpm at .2 amps, and its stall torque is 1.6 kg-cm at only 7.4 amps.

These motors can be operated anywhere on their speed-torque curve so long as attention is paid to duty cycle. Maximum output power is achieved when the motor is run with both its torque and speed at half their maximum values. The motor's recommended or nominal operating point is somewhat below this. If maximum torque is the primary goal, such as in a starter application, the motor can be run above this point and sometimes well above it for short periods of time.

The first step in selecting a candidate starter motor is to determine the engine's peak torque and cranking speed requirements. Earlier, using a torque wrench, I found that 15 inch-lbs (17 kg-cm) was required to turn the crankshaft of my Howell V-twin through its compression bumps. With nearly identical pistons, c.r., and flywheel, the Knucklehead should react similarly.

Most full-size engines require cranking speeds on the order of 200-300 rpm, and my experiences with using drill starters on model engines have been similar. Cranking speed has to primarily satisfy the needs of the engine's carburetion. A carburetor that's too large for the engine and has difficulty creating enough manifold vacuum to draw fuel at low rpms may require an excessively high cranking speed. Poorly sealing valves and/or rings in an engine with a marginally low compression ratio may also add to the requirement.

There's more to engine starting than might first be apparent, and successfully covering it with a starter spec will invariably include testing. A starter's first few revolutions are its toughest. After overcoming the flywheel's inertia and the engine's friction, the starter must then deliver its maximum torque for brief periods of time just before TDC of each cylinder's power stroke. After spin-up, stored flywheel energy may help the starter through the engine's compression bumps.

Cranking speed typically isn't constant, and this will be especially noticeable in an engine with only two cylinders. Even with flywheel assistance, an engine's cranking speed will rise and fall as its cylinder pressures change. In a twin, these peak loads will be present for less than 25% of the engine's cranking time. For once, Murphy's law doesn't apply, and the rpm rises between compression bumps will happen at just the right times to aid fuel/air flow.

Based upon the above, I've chosen for my target a conservative 25 kg-cm at 300 rpm at the Knucklehead's crankshaft. This works out to be about 75 watts of peak mechanical power. The AR390031 is far from being able to deliver such power without help from a gearbox. Gearing has the effect of changing the slope of the motor's speed/torque curve. Its no-load output rpm will be reduced, and its stall torque increased by the gear factor. For example, if the the 4.8 volt motor's output shaft is geared down by 40, its stall torque will increase to 32 kg-cm, and its no-load speed will drop to 500 rpm.

87245949-2361-4F43-ABA9-E9A9A4A0A164.png


With gearing, this motor would meet my requirements but with a pretty high current draw due to its low operating voltage. As a result, I also decided to run the entire engine from a 12 volt battery. A gear ratio of 40 will require a large number of gears for the space available, though, as well as a one-way clutch to prevent the engine from trying to reverse drive the gear set after it starts.

Gear motors are available from a number of sources. They efficiently pack gear boxes inside an extension to the motor's head without increasing the diameter of its case. Robotics suppliers are good sources for these because they include torque numbers in their specifications which is something RC suppliers don't do.

An example of a gear motor that should meet the Knucklehead's requirements is:

https://www.servocity.com/313-rpm-hd-premium-planetary-gear-motor

Its basic 12 volt motor has a no-load speed of 8.5 krpm at .52 amps and a stall torque of 1.1 kg-cm at 20 amps. With its 27:1 internal gearing, its no load rpm drops to 313 rpm and its stall torque jumps to a whopping 30 kg-cm at 20 amps.

47B9B8BC-629A-4A1D-8EED-FA1E9FAC707B.png


There are many 'similar' gear motors available on eBay and Amazon at much lower cost, but in many cases their torques aren't specified. Several of these are sold as 'high torque' but don't contain enough copper for this application.

Another source for useable motors and gear boxes are battery-powered drills. A real advantage to using them is that their 'good' parts can be cannibalized after testing them in a still convenient form factor. An example is the Black & Decker BDCDD12C 12 volt 550 rpm cordless drill which is widely available for about $30. Although removing the chuck can be a bit tricky, these tools are easily to disassemble, and a number of Youtube videos are available to help. Their dual planetary gear sets can save a lot of work even if a suitable housing has to be machined.

While still on the road, I placed an online order for one of the ServoCity gear motors. I plan to run some tests on it using my Howell V-twin as a mule before trying to design it into the Knucklehead. - Terry
 

Attachments

  • 9576A381-4A44-49BB-B8E4-77FEFE55ED4D.png
    9576A381-4A44-49BB-B8E4-77FEFE55ED4D.png
    43.5 KB · Views: 230
Last edited:
This is good stuff, Terry. I was not aware of the servo-city type geared motor availability. The gear ratio is quite high (27:1) compared to my RC experience of planetary drive inrunners, usually in the 5-7 : 1. Is the servocity a parallel double or triple reduction type train do you happen to know? Anyway, that looks well suited to the purpose.

I just assumed (but maybe I'm wrong) the majority of variable speed drill motors are brushless. If so, there may be other considerations.
1) The nominal voltage may be different, anywhere from 9.x volts to 20 volts depending based on the cell type & chemistry used in the tool. So you would either have to replicate that by also harvesting the pack or just ensure your intended power supply is suitably matched at discharge load. Your graphs illustrate this perfectly.
2) if truly brushless (3 wire vs 2 wire for likely identification), then I assume you would also have to extract the ESC (electronic speed control) module because it is delivering commutative voltage. This could also be an advantage if you desired variable speed, but I'm not actually sure how that would be implemented.
3) The current levels you are talking about probably wont factor into excess demands on the battery. But another factor to consider for motor matching particularly using smaller cells or portable packs is C-rating of the battery itself. Just an arbitrary example, if you have a 12 volt pack made up of 10C rated cells of say 2000maH capacity, then max current = 10 * 2.0 = 20 amps. Its not a hard number but basically the voltage will sag or harder on cells. Motorcycle type batteries refer to cold cranking amps, but I believe different standard again. Current while maintaining some suppressed voltage, like 7.x volts on a 12v nominal. Anyway, at your low anticipated current should be no worries.
 
Last edited:
Peter,
I don't know for sure what the gear reducer is in the ServoCity gear motor, but my guess is that it's a dual planetary gear set like the one in the Black&Decker drill.

I believe you're right that cordless tools are going to brushless motors, but the Black&Decker mentioned above is definitely a brush motor and maybe one of the few remaining with brushes. Most of those tools are also going away from 12 volts and are already at 18V and higher.

My plan for the battery is to use the 12V 7AH sealed lead acid battery I used on my Merlin. It delivered some 80 amps to the starter on that engine.

Thanks for the interest and comments. - Terry
 
Awesome build. Thank you for posting. It is inspiring me to develop my skills further.
 
After completing the Quarter Scale Merlin, I took several months off to work on a number of projects that had piled up around the home and shop. High on my to-do list was assembling a number of backup XP computers while the parts were still readily available. I've had an ongoing concern that my ten year old homemade shop computers as well as those running my wife's embroidery machines have been living on borrowed time. I could, if necessary, convert my Tormach to Linux-based PathPilot, but the hardware associated with my Wabeco lathe is still tied to Mach3. I also built up a couple Windows 7 machines so I could have at least one foot inside the modern world. I tried migrating to Windows 7 entirely, but I wasn't able to get some of my ancient CAD/CAM software nor my wife's embroidery software running on their 64-bit operating systems even in their so-called compatibility mode. Replacing all that software was pretty much off the table for me.

Committing to a new long term engine project involved a lot of procrastination and eventually came down to a decision between Ron Colonna's 270 Offy and Draw-Tech's Knucklehead. In order to shake the bugs out of the new shop computers, I modeled the Offy's crankcase as well as the Knucklehead's cylinder head assemblies in SolidWorks. I felt the Offy would probably be of wider interest to others since I'm not aware of any detailed published builds for it. In the end though I felt like I needed more time to consider some alternate approaches to the Offy's one-piece crankcase, and so for now I chose the Knucklehead.

I really liked the looks of Draw-Tech's CAD rendered Knucklehead but wasn't even aware of its existence until I came across Steve (Driller1432)'s HMEM thread:

http://www.homemodelenginemachinist.com/showthread.php?t=24705
http://www.homemodelenginemachinist.com/showthread.php?p=301687#post301687

His successful build validated the plan set and proved the model could be made to run using the original Harley timing. So I decided to do a thread on its build and, along the way, fill in some of the machining steps that Steve left out to perhaps encourage others to build one of their own. There was so much effort put into that engine's drawings that it seems a shame to allow them to languish on the forum's download site.

Even though it has only two cylinders, this engine isn't a beginner's project, though. It's considerably more complex than a Hoglet or even Jerry Howell's V-twin, but the finished result will be more reminiscent of an actual full-size engine.

I decided to begin the build by machining the exterior components of the head assemblies which I had already modeled. This included the heads, cam brackets, valve boxes, and rocker arm boxes. At first glance, the head assemblies appear to be the most complex parts of the engine, and their individual parts must fit precisely together.

My first step was to get hard copies of the pertinent downloaded pdfs since I've never been comfortable with working directly from drawings on a computer screen.

http://www.homemodelenginemachinist.com/downloads/draw-tech-297.html

Because some of the key drawings were intended for E-size sheets, I dropped a flash drive off at our local copier store so they could print them out for me on their huge cut sheet printer while I ran some errands. When I returned, though, I was informed that the store's policy was to not copy or print out copyrighted material. They pointed out the title blocks in the lower right hand corner of the drawings that contained words to the effect that the drawings were not to be reproduced without written consent from the original owner. No amount of common sense reasoning could get me past the clerk I was dealing with. Instead of coming back later when someone a little less literal might be on shift, I printed the large size drawings out in poster board mode on my home printer and then carefully taped them together to create the large sheets.

It'll feel good to be making chips again, but with only two cylinders to deal with this time there won't be as many of them. -Terry
 
I don't have a CNC mill so I am doing a lot of hand work. I have a question and maybe you showed it on the gear box, the length crank shaft and the a picture of the gears arrangement I made patterns and cast the gear box and cover
 
Rodue,
Have you looked through the Draw-Tech .zip download named gear-system? It contains all the gear information, but you will have to use the dimensions in the cam box drawings to locate their shafts. I've not gotten that far with mine, and so the only photos I've posted are of the drive gear for the oil pump. I'll soon be working on the starter gearing, but I'll probably end up doing something different from the original plans that may not apply to your build. - Terry
 
The ServoCity gear motor arrived, but I evidently entered the wrong part number and received a 437 rpm motor instead of the 313 rpm motor that I wanted. The two motors are actually identical except for a small difference in gearing that adds another 8 kg-cm of stall torque to the slower motor. I machined a simple adapter for use between its 6 mm D-shaft and a 3/8" drive socket so I could test it by spinning the crankshaft in my Howell V-twin while holding the motor in my hand.

For my first test, I used an analog ammeter to monitor the motor's average current draw and an optical tach to measure the cranking speed. The ammeter's needle momentarily kicked up to about 9 or 10 amps at the start of each test before settling down and oscillating between 3 and 4 amps during cranking. The tach indicated a steady cranking speed of some 400 rpm.

In order to see the detail in the current waveform, I borrowed an oscilloscope with an accessory current probe. I've included a photo of a typical current waveform acquired over a two second cranking period. At the start of cranking, before the motor has overcome the rotational inertia of the flywheel, there is a 14 amp spike. This inrush is the motor's stall current as measured in my setup. My wiring resistance was on the order of a quarter ohm, otherwise this spike would have reached 20 amps which is the motor's advertised stall current.

During the next 125 ms, while overcoming the flywheel's angular momentum but before full spin-up, the motor ran into its first compression bump. Sometime after spin-up the current dropped to about 4 amps. The current continued to briefly spike up to some 9 amps at the peak of each succeeding compression bump. Two crankshaft revolutions are required to complete the engine's 4-stroke cycle, and the recording shows that the time to accomplish this was 300 ms which works out to be 400 rpm. Between compression bumps the current fell to 2 to 3 amps which corresponds to the torque supplied by the motor to overcome the engine's frictional losses and to replenish the flywheel's energy.

The instantaneous torque supplied by the motor at any time can be found using the current curve on its speed-torque diagram shown in the next photo. The 9 amp peak currents at the compression bumps correspond to about 9 kg-cm of peak torque. My earlier torque wrench measurements predicted this torque would be nearly twice this value at 17 kg-cm. The discrepancy is due to the flywheel dumping its energy into the crankshaft at the compression bumps rather than slowing down. Its energy is replenished in between the bumps.

The speed-torque diagram shows the average cranking speed is dominated by the relatively long periods of time during which the motor is in between compression bumps. If more cylinders were added to the engine, the average cranking speed would fall because the motor would spend more of its time dealing with compression.

Although I seem to have lucked out, after seeing the current waveform I was probably too focused on the compression bumps when coming up with the starter motor's torque spec. The flywheel's momentum can be just as important during spin-up, especially in a two cylinder model engine where the flywheel tends to be relatively massive. The current waveform shows this particular motor was within a couple amps of its stall current when it ran into its first compression bump only 50 ms into spin-up. The 313 rpm motor will have some additional torque available during this initial start-up time. I plan to use this motor while continuing on with the starting system's mechanical design, but I'll likely re-order the 313 rpm motor. - Terry

358.JPG
359.JPG
360.JPG
 
This is very enlightening sizing information. So if this approach was applied to your radial engines, do you think its fair to say negligible flywheel effect from the prop & focus on the compression bump + whatever desired headroom?

I've always wondered if starter rpm in itself makes for better or worse fire-up conditions. I have nothing to base this on other than RC engines. With fading starter motor (low rpms even though its managing to turn the crank) engine just seems to take its time to pop over faster turning starters all things equal. Maybe that's a function of timing? Example 10% of a 4000 rpm motor = 400 rpm, but 400 rpm on a 20,000 rpm motor is only 2%. Sorry for the tangent. The things we rarely think about :/
 
"I've always wondered if starter rpm in itself makes for better or worse fire-up conditions"
I would say not. In my the model airplanes times, we used to start even worn engines with a flick of the finger. At old engine shows I have seen starting hit and miss engines by hand turning the flywheel at speed much lower than normal operating speed.
As long as the piston velocity is sufficient to make the leakage effect on compression negligible the engine should start. The flywheel helps but the explosion/compression energy is much greater than the flywheel stored energy as witnessed by the fact that an hit and miss engine will get the flywheel spinning up with the firs firing.
 
To say this is impressive is short changing this build. Perfection would be closer.
 
Sorry to be late answering, but my wife and I are dealing with incredible sinus infections, and my mind is still like a dusty dark attic. I was afraid that we might have kicked off the flu season for central Texas, but the doctors say no, our flu shots are working.

Peter,
The huge (18"-24") props on those radials (and the Merlin) have quite a bit of angular momemtum of their own. I've never done any calculations, but I just went over into the shop to spin a spare similar-size prop in my hand, and it felt very much like the Knucklehead's flywheel. I'll probably include the effects of the flywheel next time.

As far as cranking speed is concerned, Mauro, I think we'll have to agree to disagree on this one. The old 12v drill I initially used on my 18-cylinder radial had barely enough torque to turn that engine over, and wouldn't start it. After a Dah... moment (actually much longer than a single Dah), I realized the drill had an alternate torque setting. Changing it helped a little, and the engine eventually started. I replaced the drill with an 18v version and after final tuning it still took a 2-3 seconds of cranking at 300 rpm to start a cold engine.

That said, though, as I mentioned earlier, it has to do a lot with the engine's carburetion. In both the radials and the Merlin the fuel reaches the cylinders through a long contorted path through the diffusers and intake runners. The starter needs to spin fast enough to create enough vacuum and for long enough to get the fuel in there.
I've had to crank every non-fuel-injected full-size car I've owned for a few seconds to start, and even longer on cold mornings. My fuel-injected puck-up with its built-in help getting fuel to the cylinders still requires a second or so.
I can see why simple one cylinder engines with little or no intake manifold to deal with can start much quicker. I think the guys hand starting those small RC engines also frequently have to prime them.

Thanks all for your thoughts and comments. - Terry
 
Last edited:
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.
 

Latest posts

Back
Top