Another Radial - this time 18 Cylinders

[mayhugh1]

Over the past few days I've revised the design of the rocker arm supports. I really wasn't fond of my original 'duck bill' design, but I kept it as a placeholder so I could get on with the remainder of the head design and test its machining feasibility on some trial parts.
I finally decided to go with more conventional rocker boxes especially since I've already made the cutters and developed the CAM know-how to add head-matching filleted fins to their peripheries. I decided to keep the tops open even though I think they would look better fully enclosed. I need to be able to make valve lash adjustments on the assembled engine which means the locking nuts need to be above the top surface of the box so I can easily get a wrench on them while adjusting the valves in the cramped space under the engine. Dealing with the cover plates under the engine would also be a maintenance hassle.
The valve guide design still needs to be finalized, but I was careful to design the heads to accept the maximum o.d. of the guides I developed for my H-9. The valve towers of these heads also have the same deck height as the stock H-9 heads. This lets me use the same valves, springs, keepers, and locks that I developed for my H-9. I had to lengthen the rocker arms, though, for additional pushrod clearance to my larger diameter head. I re-positioned the rocker arm shaft, though, to maintain the stock H-9 rocker arm ratio. The skinny H-9 pushrods may look out of place with these larger heads, and so with the extra clearance I added I also have the option to increase their diameter. Unfortunately, I partially completed a set of H-9 diameter pushrods several months ago. In SolidWorks I was able to overlay my new head/cylinder/rocker arm assembly with my stock H-9 model to prove one is a drop-in replacement for the other.
My next step is to machine the valve towers on my two test heads and to make two pairs of rocker boxes. I plan to make both a front and rear row head assembly in order to verify the designs and their machining feasibilities. The major concern I have about my current design are the potentially fragile ends that are left to wrap around the rear of the rocker arm boxes after the tower machining. These ends narrow down to thin, pointed cross-sections; and I'm concerned they may chatter and become damaged while the tower decks are being faced. I plan to reinforce their exteriors with my favorite Devcon gel before machining, though. If this turns out to be a non-issue I'll begin the milling operations on my 25 heads to bring them up to the current state of completion of my two test heads while the weather is still cool down here. -Terry

 

[mayhugh1]

I enjoy this machining hobby because I like coming up with solutions to the various mechanical issues I run into during the construction of the engines I've been building. When it comes to finalizing esthetics, though, I can't be trusted. I wonder off too easily into the high weeds of minutia and get lost. I've spent countless hours on the cosmetic details of my heads, because I can never seem to be completely satisfied with what I've done. Last week after I thought I had settled on a final design on the rocker boxes I looked back on some of my original designs and decided I didn't like my current design any more. So, I spent another week on the rocker box design massaging minor details and ended up with yet another final version. I decided the only way I was going to be able to get past my OCD and put a period at the end of their design was to machine my test heads to accept what I have now and never look back.
And so the CAM fun began. Since the heads were designed symmetrically about a central axis, new local coordinate systems have to be generated for my CAM in order to handle the off-axis machining of the valve tower tops. This was not as big of a problem for my spark plug cavity machining which was similarly off-axis because the cavity shape was simple. I was able to create a dummy job assignment for a dummy part in a dummy workpiece for my CAM, and it never had to know anything about my actual head.
The valve tower machining is not simple, though. The CAM needs to know about the actual head and, even worse, it needs to know about the exact workpiece that I will be using. Both the part and workpiece are complicated structures with some 10,000 features between them, and the combination of the two slowed my CAM computer down to a crawl. So, in my CAD I simplified the machining models by cutting away most of their lower portions that were not involved with the tower top machining. This solved my computer performance problem, and I was able to create new local coordinate systems for my tower machining operations to match the orientations of my workpieces while being held in my modified H-9 fixture. It really wasn't as complicated as it sounds.
This approach resulted in four separate cutting programs: one each for the exhaust and intake valve towers on both the front and rear row heads. I immediately realized this was going to be a possible source of confusion for me later and probably end up creating some scrapped parts. Each operation requires its own orientation in the milling fixture, and each head requires two different operations. I will eventually need to match the proper program to 100 different part orientations.
When I was ready to verify the programs on my test heads, it was as though I had launched a stand-up comedy routine for a room full of machinists. For my opening act I didn't have my touch-off gage block fully seated in the H-9 head I was using as a reference fixture to establish the machine's center of my local coordinate system. This created a significant error in the first result. During the second operation, I ran into a disastrous CAM bug that violated the part and machined away material that should not have been touched. This didn't show up in the simulation, but if I had carefully studied the CAM generated tool paths instead of relying on the simulation I would have spotted it. I then began the process of mindlessly changing random CAM parameters to nudge the CAM software away from generating the gouging code. Setting the bottom machining depth to .0001" instead of zero happen to solve it in this case, for some reason; but this was another bug in the CAM software that I had never seen before.
While in the middle of my third program on my second head I noticed the result wasn't looking right, and so I stopped the machine. It turned out that I had already confused the four different operations. The confusion had begun early as I had mis-named the two rear row operations when I created them, and I was now machining an exhaust tower when I should have been working on an intake tower. Adding insult to injury, the fourth program that did run as I intended left the part with some unacceptable visible steps due to a poor choice, on my part, of cutting sequences. When the massacre was over, neither of my test heads were usable for verifying the rocker box fits. However, they had served their purpose in verifying the drilling and reaming operations for the valve cage. I had been concerned about the drill and reamer tip clearances to the top of my threaded mandrel as I only had .030" clearance to its top surface while in my fixture.
After a few more days of work I was ready to retest new versions of my programs, but I didn't want to risk spoiling any of my 25 threaded and grooved parts waiting to be finished. I turned a simple blank to match the OD of my head and to fit into my fixture, and then I ran each of the four programs on its own quadrant on this blank to, at least partially, verify them. I won't be able declare success until they've been run on an actual finished head, but I'd say this definitely locks in my rocker box design.
At this point I'm anxious to continue the milling operations on my run of 25 parts while the weather is still cool. The design of the valve cages is finished, and I've made a few trial parts, checked them in my test heads and have iterated the design. I plan to start a production run of 52 cages and alternate machining runs between them and the heads during the next week or so. -Terry

 

[mayhugh1]

I'm using valve cages similar to the ones I designed for my H-9. With my new head design it's not practical to get lathe access to machine a step for the cage in the combustion chamber, and so these cages will be simple cylinders.
It seems that half of the model engine builders have their own best way of coming up with seats for their valves, and the other half dreads the long and chaotic process that always seems to be required. I'm somewhere in between, but I'm becoming fond of a manually cut seat in a stand-alone cage that is sized for a Loctited slip fit in the head. With this approach the seat can be pre-tested before it becomes an integral part of the head, and the slip fit will insure there's no distortion due to installation.
When I made the cages for my H-9, I started with 954 aluminum bronze but quickly decided it was just too much of a hassle to work with, and so I switched to 544 phosphor bronze. About that time I chatted with Dwight Giles at the Gears Show in Portland, and he told me he was moving away from aluminum bronze for his valve seats since he had been seeing excessive wear over time on the stainless valves in his Black Widows. His comments reinforced my decision to go with 544. Unfortunately, after finishing the cages and irreversibly installing them in my H-9 heads I discovered that I had mis-labeled my scrap bronze stock and had, in fact, used 660 bearing bronze. I researched model engine plans in my own library as well as online and found very few examples of 660 being specified for valve seats in a model engine. Most plan sets contained a non-specific 'bronze' specification. Bearing bronze has two possible issues that might limit its use in this application. First, it is very soft - softer than either aluminum or brass. Second, it isn't intended for high temperature operation; and I couldn't find a spec for its metallurgical properties at elevated temperatures. The seat is well heat-sunk to the relatively cool head, but I guessed the exhaust valve seat might still see 300F in a model engine combustion chamber. I did yield calculations on the faces of the seats using pressures I expected to see during combustion, and they showed I had good margins. But these calculations were at room temperature, and I have no idea how the 660 properties degrade with temperature. So far, I haven't seen any issues with my 9 cylinder radial. It could be, in fact, that the softer material might even be of some benefit in the valve seating-in process. For piece of mind, though, I decided to actually use 544 for the cages in this engine.
My T-18 valve cages, except for their seats, were completely machined in three manual lathe operations. After the valve tower machining is completed the cages will be Loctited using high temperature (450F) 620 retainer with the upper ends of the cages flush with the valve tower surfaces. There is a good bit of surface area for adhesion, and I added several glue grooves around the body of the cage. These shallow grooves, which were randomly filed during the first turning operation, will collect excess adhesive as the cage is being inserted into the head and help seal the cage to the head. The rocker box mounting holes are located so that after the holes are threaded, the mounting screw threads will press into the cage for additional security. And, the rocker boxes themselves will prevent the cages from moving upward. Most of the cage will be heat sunk to the relatively cool head, and so there should be plenty of margin against any heat deterioration of the adhesive.
I started a run of 70 valve cages while still working on my run of 25 heads in order to break up the monotony I'm experiencing and to even out the workload on my two machines. In any event, I'll need the completed cages in order to complete the final head machining operation. I've added 20 spares since I don't expect 100% yield, and I'd like a number of practice parts for experiments.
I started making the valve cages by turning the OD on the end of a 544 rod to .0005" under the ID of the bore in the head. I used 400 grit paper to de-burr the adhesive grooves and to verify the cages fit into a gage that I made using the same reamer that will bore the heads. All 70 blanks in my little production run are completed to this point before moving to the next step.
In the next step the blanks are re-chucked, and all the important seat-related operations are completed in the same setup. Only tailstock tooling is used. The end is spot-drilled and then drilled-through with a carbide drill for the straightest and most centered hole I can manage. I used a sharp 90 degree 2-flute v-cutter as a spotting tool to make sure the start of the hole ended up precisely on the spindle's axis even if the cutter, itself, wasn't exactly on center. The carbide drill was held in a MT-2 tapered collet tailstock holder. A floating reamer finished the ID for the valve stem, but it just follows the drilled hole whether or not it is on the spindle axis. Next, the end was pre-drilled and plunged with a stub ball mill that was rigidly supported in a second tailstock collet holder. This step formed a smooth fuel flow area behind the valve and created the edge that will eventually become the seat. This radial edge has a high probability of ending up exactly on the spindle axis due to the rigidity of the ball mill and side cutting action of its flutes. If the axes of the valve stem and the radial seat edge are coincident then a piloted seat cutter can create a very effective seal that is relatively immune to the machining of the backside of the valve. If the two axes are not coincident, the seat area will bear against this machined surface in a skewed manner, and the integrity of the seal can be affected.
My tailstock DRO was used to control all cutting depths. A gage block was used between the workpiece and tailstock tools to reference the Z-axis tools against the through-drilled parts. In order to save time, all operations in this second operation including the gage block referencing were performed without stopping the spindle. Again, all the parts in the run were completed to this point before going to the last step.
For the final step the part was reversed in the chuck, and the cavity for the valve spring was cut. After pre-drilling, the cavity was plunged with a cylindrical end mill for a nice flat bottom.
Due to the number of valves in this engine, it's important to cull out cages as early as possible with any potential for creating sealing issues. I especially don't want them finding their way into finished heads. As a first step even before cutting the seat, I visually inspected each cage under a microscope with a pristine H-9 valve inserted. If the 360 degree virgin edge contact didn't look perfect I discarded it. In order to properly do this inspection a slight burr raised by the ball mill had to be polished out. I used a simple fixture to support the cage while I twisted it back and forth a few times over a sheet of 600 grit paper on a surface plate. Even with all the machining care I took, I ended up scrapping five cages. One of the photos shows one of the rejects. I actually performed this inspection on groups of ten parts during step two so I could correct my process if it started to wonder.
The intake and exhaust ports will be cut after the cages are installed in the head during the exhaust flange machining step. The total machining time per cage was 20 minutes. Inspection and preliminary leak-checking (to be described in my next post) will add another 10-15 minutes for a total cage fabrication time of about 35 hours. -Terry

 

[Ethan D]

Hey 'Mayhugh1'
This is a incredible build! I'm looking forward to seeing more updates.
Great work with it all.
Cheers

 

[mayhugh1]

For my last two engines I made manual seat cutters by drilling pilot holes through the ends of 90 degree HSS countersinks. I recently came across a better solution from Brownells. They sell a 45 degree muzzle reamer which is a turn-key solution to the part I've been making. It also has many more flutes than the counter sinks I've been using for smoother cutting. I've found there is some art involved in using these cutters with bronze. Very light pressure of the cutter against the seat seems to give the best result. Held vertically, the weight of the Brownells cutter is almost enough force on its own to cut a nice .005" seat. Pushing the cutter hard into the seat can cause it to grab the bronze and the seat will end up wider than desired. A seat width of about .010" seems to be optimum for model engines.
For my twin-18 I need more assurance, than just a visual inspection, that the seats are going to seal after they're installed in the heads. For my last two engines I used a vacuum leak-down (up?) test to verify my seals. I used a MityVac to pull a vacuum on the cage containing the installed valve after it was installed in the head. The vacuum was pulled through a rubber adapter inserted into the intake or exhaust port while the valve stem was sealed to the valve guide bores with a silicone plug. My subjective 'pass' criteria was a leak-down time from 25 inHg to 15 inHg of 15 seconds or greater.
I want to do a similar test on my T-18 cages but before they are installed in the heads. To create a test fixture, I cut a v-groove along the stem of a pristine H-9 valve that was left over from that build. For my test I insert this valve into the cage-under-test and draw the vacuum at the rear of the cage. The groove allows any leaking air to escape. Without it, the close fit of the valve stem and guide significantly affects the leak-down time and can completely mask a badly leaking valve.
With unlapped manually cut seats my H-9 notes show I was typically achieving a 5 second leak-down. Under a microscope the seats showed perfect concentricity with the valves and layout fluid on the seat showed a uniform 360 degree contact line. Microscopically visible circular grooves left in the seat by the seat cutter, though, limited the integrity of the seal. My stainless valves, on the other hand, had mirror finishes on their seat contact areas due to a final polishing step I did before removing them from the lathe.
While I was making my H-9 valves, I made a number of stainless steel laps, which were nothing more than polished rear valve shapes with handles. Using these with very fine TimeSaver lapping compound I achieved my desired leak down time after about a minute of lapping.
I prefer to do lapping with a separate lap instead of using the actual valve. Lapping damages the fine finish on the rear of the valve because the process starts by transferring the coarse circular grooves in the seat to the valves themselves. Getting a smooth surface finish on both parts by lapping takes a long time, isn't really necessary, and ends up leaving a trough around the valve that's equivalent to a lot of engine running time. There's a temptation to go to a coarser lapping compound to speed things up, but this really only speeds up the damage to the valve. The TimeSaver compound breaks down quickly, helps protect me from myself, and minimizes the damage to the lap. When I hear the squeaky sound of metal-against-metal as I wring the two parts together, it's time for a leak test or, if necessary, a new charge of lapping compound.
When lapping, it's difficult to tell when you're done from just a visual inspection of the smoothness of the seat. A seat can appear to have a mirror finish to the naked eye, but under a microscope it can contain circular grooves that are coarse enough to prevent the valve from sealing. Conversely, under a microscope the seat can appear to still have grooves; but seals just fine.
To illustrate this, I took four microscope photos while experimenting with one of my T-18 cages. The first photo shows a de-burred but uncut seat. The measured leak-down time with my grooved H-9 valve was 5 seconds. For the second photo the Brownells tool was used to cut a .005" wide seat. The circular grooves are very prominent, but the leak-down time measured a passable 20 seconds. By the way, these grooves are nearly identical to the grooves created in my H-9 seats by my shop-made cutters. For the next photo the seat was widened to about .008", and the leak down time worsened to about 9 seconds illustrating that thinner is better. The fourth photo shows the same seat after 2 one-minute lapping sessions with TimeSaver and a well-used H-9 lap. There's only a small visual improvement in the grooves, but the leak-down time improved by almost a factor of two to 35 seconds.
Before I finally installed the H-9 valves in their heads, I lightly lapped the valves to their cages and achieved final leak down times approaching a minute. I've not yet decided if I'll do this same last step with this engine. If I achieve my pass criteria with a sacrificial lap, I'll likely stop and avoid any damage to the valves.
I duplicated the above steps on five more of my new cages, and the results were pretty much the same. My plan now is to wait until the head machining is completed to the point where the cages must be installed. At that time I'll cut the seats, lap them, perform the leak-down tests, and Loctite the cages in place. The heads will be numbered and the results recorded for later comparisons with the final leak down tests using the actual valves. The last step in the head machining is the milling of the surface for the intake/exhaust flange. In this step the intake and exhaust ports are also drilled into the cages.
For perfectly concentric valves and seats it seems that the circular machining grooves ultimately limit the quality of the seal. Thinner seats with fewer grooves seem to produce better seals. One can imagine the additional complications that come into play if the valve and seat aren't concentric. For example, if the valve axis is skewed to the seat axis because the valve guide wasn't drilled straight, the circular grooves can straddle the the seat and act as tiny pipelines to allow air to escape past the seal. The grooves will have to be completely lapped away to solve this, and the final result may a deep, wide seat with other issues. Another problem that can arise with skewed axes involves the actual shape of the contact ring of the valve. If the rear of the valve was dressed with an abrasive in an attempt to polish it and the resulting surface became non-spherically curved, then the 360 degree contact ring can be lost. -Terry

 

[camm-1]

Love your build! Outstanding!
Do you have a link were to by this valveseat cutter?
Ove

 

[mayhugh1]

Here you go:

http://www.brownells.com/gunsmith-t...ing-cutters/45-chamfer-cutters-prod41716.aspx


Terry

 

[mayhugh1]

The milling operations on my run of 25 heads was continued while I worked on the valve cages. The cages need to be finished and installed before the final milling operation, and so I alternated between the machining of these two parts to help with the tedium. Except for tool changes, the mill can run unattended; but the valve cage work is manual.
The first of the five head operations was the most grueling as it required five tool changes and some 50 minutes each of run time. This on-axis operation removed excess material from between the valve towers and created the finished fins in this area. The holding fixture for this step was a 5C collet chuck bolted vertically to my milling table. My threaded expandable mandrel which gripped the head was held in the collet. The stacked TIR's of the workpiece and mandrel combined to move the effective center of my fixture around by as much as .004", and so I re-referenced the machine for each part. An error this size would have caused visible flaws in the fin profiles.
Due to my own carelessness, I trashed a head early on in this operation when an un tightened roughing tool pulled down slightly in its collet leaving me with 24 parts. This first operation consumed about 25 hours spread over several days. The carbide tooling still looked like new when this step was completed, but I did consume two HSS roughers.
The second operation created the spark plug cavity and drilled the hole for a CM6 spark plug. After the cavity was completed, I manually tapped the hole using a spindle tap starter while the spindle was still in position over the drilled hole. This operation was 25 degrees off the head's center axis, and the workpiece was gripped by my threaded mandrel which was, in turn, held in my modified H-9 head fixture. This fixture was calibrated only once at the start of the run with the help of a spare H-9 head and some some trigonometry. The part cycle time for this operation was about 30 minutes for a total of 12 hours.
The first parts machined by the next four operations were pretty stressful. These four operations created the 25 degree surfaces on the valve towers for the rocker boxes and blended the tower fins to the rear of the rocker boxes. I had to create different operations for the intake and exhaust towers on both the front and rear row heads. After this step was completed there were two flavors of heads - one for the front row and one for the rear row. In addition to my own confusion with four similar but different operations, I had a lot of trouble getting them to compile as I wanted. The issue was the tower fins that were being cut and terminated at an angle. The tool paths that my CAM wanted to generate very visible machining steps in the result and I had to over-complicate the tool paths to get a clean result.
I screwed up both of my test heads in the process of getting these operations to run properly and really never saw any of the four operations run successfully on an actual head blank. Each operation took 20 minutes to run and there were 48 of them for a total of 16 hours. I decided to plow through the whole lot in an exhausting day and a half.
At this point, only the operation to cut the opening for the intake/exhaust flange remains. (I also need to hand-tap 192 2-56 holes in the tower tops.) So far I've accumulated about four hours of machining time on each head, for a grand total of about 100 hours. The valve cages, with their cut and pre-tested seats, will be installed before this step is started. The intake and exhaust ports will also drilled through the head and into the installed cages at this time.
I'm anxious to get on with the rocker boxes since it looks like their machining will also be pretty involved. This run of 50 (or so) parts is going to be another big deal that I want to finish up while the weather is still cool, and so I'm turning my full attention to them. I've begun the programming, and it looks risky enough that I'll also need to do a small trial run. I've also ground another custom round-over tool for my CAM to work with since I made yet another change to my final, final rocker box design. I'm now planning to radius the topside of the rocker box bottom. There's also two significant work holding fixtures that need be designed and built before starting. - Terry

 

[creast]

OMG!!!
I think I am going to die looking at such work!!
AMAZING!!

 

[mayhugh1]

All the cosmetic detail that I added to the rocker boxes turned them into fairly complicated parts to make. They're irregularly shaped and need to be machined on three sides, but fortunately most of the complex milling can done from the top. My first attempt at the topside CAM went pretty smoothly as I was able to re-use some of the fin profiling tricks and custom cutters that I created for the heads.
My plan was to make about 50 parts in cookie sheet fashion but to do them in small batches of six pieces or so. My simulator predicted about 3/4 hour of topside milling per part, but the actual cycle time with its ridiculous 12 tool changes was closer to an hour per part. I prefer working with small batches especially on complex parts because it can sometimes take several runs to get all the process kinks worked out on a part this complex.
It didn't occur to me during the design, but the finished thickness of these parts ended up at a miserable something over .500", and so I wasn't able to make use of the pile of 1/2" 6061 scrap that I've collected over the years. I was able, though, to scrounge up some pieces of 5/8" thick material that had lots of deep gouges and had evidently spent much of its life out in the weather.
For my test batch of four parts I surfaced both sides of a 5" square piece of this material and then glued it down to a piece of MDF. The test run uncovered a number of minor issues with my code; but, more seriously, I ended up crashing my new shop-ground 1/2" corner round-over cutter. This caused the workpiece to shift in my set-up and the machine to probably lose steps. I wasn't able to re-reference the workpiece, and so the run was scrapped before the parts were completed. Sometimes you get into trouble when you try to outsmart the CAM by lying to it about the exact size of the workpiece. In this case, a retract trajectory that this huge tool tried to use was in the path of workpiece material that it didn't know anything about. I did get enough information from this run, though, to generate an improved version of the code. And, I managed to salvage a few partially finished parts to test out the two work holding fixtures that I made to machine the other two sides.
Some of my scrap was long enough to yield two batches of parts, and so I planned to run multiple batches on a single workpiece rather than cut up the workpieces. I faced both sides of all my plates to remove as many of the dings as I could, and I planned the layouts to avoid the ones I couldn't remove. Since I ended up getting very close to my finished thickness, I tried to keep the long surfaced workpieces flat to better than .001" mostly, though, as a practice exercise. The final skim pass left me with less than .010" excess stock, and it was done after the blank was glued down to the MDF. I also added 8-32 hold-down screws in the four corners for extra support during the heavy roughing that I was planning. Although the 16 inch long pieces of MDF measured flat to better than .001" over their entire lengths, I was getting some significant 'drumming' in the center due to the lifting forces of the surfacing cutter. I was able to hold the MDF down with my hand to maintain a smooth finish during surfacing, but I made some quick-and-dirty hold-down clamps from oak for the actual machining later.
I ran another three pairs of parts during my second run to verify my new code. An .087" drill that I was pecking to full depth in the half inch workpiece broke due to its built-up edge and spoiled one of the second run parts. I replaced it and began seeing the same issue in the next run. I ordered a Guhring deep drill from MSC and had no more trouble.
The roughing step that removed most of the excess stock from the workpiece was done with a three flute 3/8" rougher running at 20 ipm and at a half-diameter depth. Just after the operation completed, I measured the temperature of the residual workpiece at 155F which is uncomfortably close to the maximum service temperature of the Devcon adhesive. So, I decided to add hold-down screws to the centers of the parts for extra support during the tool change pause of the final contouring operation when they are cut free. As it turned out, the adhesive in about one out of every five parts released from the heat of the roughing pass.
I don't run flood coolant; but, instead, I use Trico's Micro-Drop system. This coolant system is very marginal for the deep aluminum roughing that I was doing. During my fourth run, I evidently didn't have the nozzles positioned optimally, and the flutes filled with hot sticky aluminum. The cutter broke, and the workpiece shifted slightly. I was able to re-position the workpiece, replace the cutter, and re-reference the machine so I could continue. The finished parts in this batch all turned out perfect.
The total topside machining time for all the parts was about 50 hours, and I was able to run about 10 hours per day. It wasn't until my fifth batch of parts, though, that I really had a stable and reliable process with which I was comfortable.
The next step is to complete the simpler machining on the other two sides. The bottom must be faced to final thickness before its inner and outer perimeters are rounded over. A tapped 2-56 hole is also added so the rocker shaft can be secured with a set-screw. And, of course, the rocker shaft hole must be drilled and reamed. I've made two fixtures to perform these operations. -Terry

 

[mayhugh1]

There was just some simple machining left to do on the rocker boxes, but the huge number of parts involved took some of the fun out of it. I didn't expect to end up with so many extras because I didn't expect my yield to be as high as it was. At this point, though, it seems a shame to not finish them all out.
The first step was to surface the bottoms, and bring each part to its finished height. It's important to get them all at the same height so a common z-axis reference can be used for the corner rounding operations in the next step. I made a very close fitting fixture to hold the parts one at a time in a vise with their bottoms up. This was set up under a small surfacing cutter in the spindle of my manual mill, and then I continually fed all the parts into the fixture in one large batch. Over half of the two minute cycle time was taken up cleaning the chips out of the fixture between parts.
For the second step I moved the fixture to the Tormach. It was referenced once at the beginning of the run with the bottom of a surfaced scrap part facing up. I used my shop-ground corner rounding cutter and some simple contouring code to round over the edges of the inner and outer perimeters of the bottom. I want the bottom surfaces of the rocker boxes finished because they will be visible in the engine's lower cylinders. A few scrap parts were used to fine tune the code for the best edge blending before I started the actual run. The line that a corner rounding end mill can leave on the top surface drives me crazy.
In the third step I drilled the hole for the rocker arm shaft. I created a second fixture to do this operation under the spindle of my manual mill. The hole has to be spotted, drilled, and reamed; and by design it's centered between two fins. I designed the fixture to be a close fit to the parts so I could swap them in quickly and easily hold them in place with a simple pinch clamp. My procedure was to do all the parts in three steps. The first step spotted all the parts, the second step did all the drilling; and in the third step all the parts were reamed. Performing each individual operation on all the parts in complete batches was a lot more efficient than dealing with 150 tool changes, but it required a repeatable fixture.
The final machining operation drilled the hole for a setscrew to secure the rocker arm shaft. This, again, was done on the manual mill using my first fixture. The 2-56 holes were hand-tapped, and then a few hours were spent de-burring the shaft holes and manually re-reaming them. A close-fitting dummy rocker arm was used to verify the shaft holes are perpendicular to the rocker arm clearance slots. I spot checked about a dozen parts and found no rejects.
After all the work during the past several days, I couldn't wait to see how the rocker boxes looked atop the heads. The photos I took are are staged in that the mounting holes in the heads aren't yet tapped, and so the rocker boxes are just sitting on the valve towers. Keep in mind that the intake/exhaust flanges have not yet been cut out of the rear of the heads. I also set one of my new assemblies in one of the crankcase sections, along side one of my H-9 spares for a sanity check on the pushrod angles and clearances. I wasn't able to create a SolidWorks model that I trusted for a clearance study, but it appears I have plenty of clearance. The second piece of good news is that the half-completed batch of pushrods I made several months ago have enough excess length to be useable. I still may increase their diameters by gluing light-weight slip-fit aluminum tubes over them for improved appearance.
I have some de-burring to do on the rocker boxes and the heads. It will probably take several days because it is one of my least favorite things to do. I'll likely start on the rocker arms next so the rocker box assemblies can be completely finished. I still have to install the valve cages in the heads and run the last milling operation on them. - Terry

 

[aonemarine]

looking great! Does the rocker box get a cover on top of it?

 

[mayhugh1]

Hi,
No, I'm leaving the tops open so the valve lash adjustments in the lower cylinders won't be so difficult.

Terry

 

[mayhugh1]

I took some time off from my build to visit my ill mother-in-law and to attend the Cabin Fever Expo which happened to be within a few hours drive. I was able to purchase the last three 'old style' Exciter ignition coils from the Jerry Howell booth, and so I now have TIM-6 back-up options to my CDI ignitions. Lee Hodgson was showing his impressive slide valve radial. I was tempted by the massive two volume plan set he was selling at the show, but I think for now I'll let the design simmer for a while in the hands of some other ambitious souls.
I also made a pilgrimage to Edison's lab in West Orange, N.J.. It was another 2-1/2 hour drive from the York show, but it was well worth the effort. Edison was my childhood hero, and his biography had a major influence on my career. His R&D machine shop is huge and was one of the highlights of our trip. It was pretty much a self-guided visit, and we spent the better part of a day wondering around the site which is now a National Park.
Getting back to the build, my next task is to finish up the rocker boxes which are lacking only their rocker arms. Since I'm not planning to make covers for the boxes, I want the arms to completely 'fill' them, and that means their profiles will be a bit complicated. Like the rocker boxes, they will be machined on three sides, and a special fixture will support them during the final side machining. I designed ball socket ends to mate with the hemispherical pushrods. The sockets also have the clearances needed for the skewed pushrod angles. The two CAD drawings show sections taken across the sockets that I used to verify their clearances at the extreme positions of the arms. Because of the complex shape and the number of pieces involved, I decided to make the arms out of 7071 aluminum instead of steel. For durability, I inserted phosphor bronze slugs into the arms for both the pushrod sockets and rocker shaft bearings. I'm using oval-ended socket head set screws as bearing surfaces against the valve stem ends.
The first step in the arm construction was to saw some 3/8" thick stock into several lengths sufficiently long to make the arms in batches of ten. An array of holes was first drilled in each blank for the lash adjusters and the bronze inserts for the ball sockets. The lash adjuster holes were free-hand tapped for the 4-40 set screws. Even though I've tapped almost 400 holes on this engine without breaking a single tap, I managed to break two 4-40 taps on the lash adjuster holes. I was using the tap in an electric drill to speed things up, and so maybe I got what I deserved. The inserts were lathe-turned for a Loctited slip-fit, and the 680 adhesive was allowed to cure overnight in an oven.
The ball sockets were then milled from the bottom side of the blanks. I used my Tormach Speeder for most of the arm milling operations. It increased my spindle speed to 15k rpm and saved several hours of machining time with the small cutters I was using.
The blanks were then flipped over and the outer profiles of the arms were cut to their full depth. This avoided any 'parting' line in the side of the arm and left about .040" excess stock at the bottom of each arm which still connected them to the blank. The tops of the arms were then filleted with the same round-over cutter used during the head and rocker box machining. I then used my favorite Devcon epoxy gel to re-bridge the arms to the topside of the blanks for support while the bottom side machining was done. Unfortunately, I didn't get the blanks cleaned properly before I applied the epoxy. Normally, for smaller batches of parts I use acetone as a final surface cleaning step, but I skipped that step this time and just cleaned the vegetable-based cutting lubricant from the blanks in the kitchen sink using dish soap. Normally this would have been fine, but I evidently didn't do a good job of scrubbing the cutting oil out the troughs cut around the parts because I ran into an adhesion problem during machining. After looking closely I could see the still-oily surfaces under the epoxy that had broken free. I added a glued-in MDF bridge as well as a lot of additional adhesive to stabilize the arms during the bottom-side profiling, but I still ended up trashing some half dozen parts. The bottoms were machined in two steps using a cylindrical cutter for the flat areas and a ball cutter for the contour around the shaft bushing. Normally I use a heat gun to loosen the epoxy from the finished parts. Since I was dealing with nearly sixty parts I tried, instead, baking the blanks in an oven at 190F for half an hour. This worked beautifully, and all the parts broke free with very little effort with absolutely no clean-up required.
After the bottom-side machining was completed the rocker shaft bearings were installed. A fixture was machined to hold the arms one at a time under the spindle of my manual mill while the through-hole for a phosphor bronze slug was drilled. A carbide vee-drill was used to get a precise hole with a single drilling operation. The slug is .010" longer than the width of the arm, and so when it was Loctited in place I was left with a .005" spacer bearing on either side of the arm. After the adhesive cured, the parts were returned to the drilling fixture where the hole for the rocker shaft was drilled. Again, a carbide vee-drill was used to save spotting and reaming steps. Machining this hole after the bearing was inserted into the arm speeded up the bearing fabrication, and gave me the best possible perpendicularity to the arm axis and provided some serendipity. After fitting the first completed arm into position in a rocker box I realized I had mis-machined my rocker box shaft drilling fixture; and, as a result, the rocker arm was .015" too far forward in the box. With the minimal clearances I had designed between the arms and the rocker boxes I now had a bind that wouldn't allow the arms to reach the full limits of their swing. I solved the problem by moving the shaft holes in the rest of the arm bearings to compensate for my error. There's plenty of bearing material around the offset shaft holes so there should be no operational problem and, cosmetically, my mistake will be hidden inside the rocker box; but I hate it when these things happen. I'm really glad these jinxed parts are finally finished. -Terry

 

[sbdtasos]

just incredible.....

 

[mayhugh1]

I thought I would spend the next week or so making some of what I call the 'nuisance' parts of the engine. I'm going to start with the gaskets. There are three sets of gaskets in this engine. The first are the soft aluminum gaskets that go between the threaded cylinders and heads to seal the connection between the two. A common way of making these is to bolt a stack of dead soft aluminum sheets of the proper thickness together and then turn and bore the sandwich into a set of completed gaskets. When I built my H-9, I finish turned and bored a piece of 6061 tubing that I had on hand, and then I just parted off the gaskets. After annealing, they were quite a bit softer even though they were still 6061. Since this approach seemed to work OK for my H-9, and since I still have some of the same tubing, I'm doing the same for this engine.
The parting operation leaves a thin but easily removed lip on the headstock-side of the gasket. After pulling it off with a pair of side-cutters, I polished both sides of the gasket using a wet sheet of 600 grit paper on a lapping plate and a simple shop-made tool to hold the gasket down flat. I left the corners of the inside diameter sharp. After polishing, the gaskets were checked for uniform thickness at four points on their circumference. I then annealed them in my heat treat oven for three hours at 775F and then cooled them at a rate of 50F per hour to 500F after which they were allowed them to cool to room temperature inside the oven.
The next set of gaskets are the thin paper gaskets that seal the cylinder flanges to the crankcase. For these I used some sheets of linen or 'rag' paper that I removed from a 30 year old university dissertation. I cut these gaskets on my Tormach using a 60 degree vinyl cutting tool and the same program that I had previously generated for my H-9 gaskets.
The last set of gaskets are the ones between the heads and the intake/exhaust flanges. For these I used a 1/64" thick black rubberized FelPro automotive gasket material. This material is not designed for automotive exhaust applications but should be fine in this application. My testing showed that it begins deteriorating somewhere above 450F. I cut the gaskets for a .002" peripheral clearance around the opening in the head for the intake/exhaust flange. After a slight burnishing they become a snug fit. -Terry

 

[mayhugh1]

My final set of 'nuisance' parts involves the hardware used to retain the valve springs. I'm using the same two piece retaining scheme that I've used on my previous two IC engines. It consists simply of a stainless steel retaining cap held in place by a U-clip. These particular parts have all the features that I personally don't like dealing with including trivial design, tiny size, and large quantity.
In use, the cap slips over the end of the valve stem and is held in place against the valve spring by the U-clip that rests in a recess and is retained in a groove on the valve stem. The spring caps are simple to make; but since many are required, the challenge is to figure a way to make them using as few Individual operations as possible.
I started by drilling a one inch deep hole in the center of a length of a 5/16" diameter 303 S.S. round. I used a sharp carbide vee-drill of the finished valve stem diameter in a tailstock collet to avoid the center-drilling and reaming operations. I then bored the shallow U-clip recess using a 1/4" cylindrical end mill also held in a tailstock collet. I began the manual parting operation using a .035" thin bade insert, but before it was completed, I chamfered both sides of the groove using a small triangular file. This simultaneously chamfered the bottom corner of the current part and the top corner of the next part. I repeated the boring, chamfering, and parting operations for seven or eight caps until I ran out center-drilled rod. I then re-positioned the the stock in the chuck, drilled another one inch deep hole, and repeated the process for another seven or eight parts. I couldn't figure a way to speed things up with CNC, and so the parts were made manually. The final step was to polish off about .0005" from each side of the cap using wet 600 grit paper on a lapping plate.
The U-clips were started by again drilling a 1" deep hole in each end of a 1/4" diameter stainless rod, but this time slots were milled along the edges to form the U-grooves. Working with both ends of the workpiece doubled the number of clips I was able to produce per grooving set-up. The clips were then simply parted off in my CNC lathe in an interrupted cut operation using a .020" wide carbide insert. The U-clips themselves are only .017" thick and are difficult for arthritic fingers to handle, but they were also lapped on both sides.
My yield on both parts turned out to be rather poor due mostly to my clumsiness in handling the finished parts. My fingers just don't seem to handle tiny parts as well as they used to. It was important to make a number of spares of these particular parts since during final assembly there will be plenty of them flying across the shop under spring force.
Although I don't consider the valve springs to be nuisance parts, this seemed like a good time to make them as well. The stock H-9 springs are very pricey commercial items, but they are easily shop wound. The specs I used were 8.5 turns per inch of .020" diameter stainless steel music wire with a finished o.d. of .250" and an uncompressed height of .450". Before I got my CNC lathe I used the threading capability of my manual lathe to wind springs using a simple shop-made tool-post wire feeder. With my CNC lathe I have the ability on most days to include a pair of close wound coils on either end for a more 'commercial' look. I say 'on most days' because I need to run the lathe spindle considerably less than its minimum spec'd rpm in order to handle the feedrate required for the close-wound coils. Some days the VFD just doesn't want to cooperate. The required mandrel diameter in my set-up was determined by trial and error to be .140 inches. Some experimenting with the feed-rate is also required to arrive at the proper turns per inch.
After cutting the wound mandrel coil into individual springs, I ground the ends flat using a simple holding fixture over a sheet of 200 grit dry paper. I then stress relieved them at 500F in my heat treating oven for one hour before allowing them to slowly cool.
I've been rounding up the metal to machine the intake/exhaust flanges, and this looks to be the next step in this project. -Terry

 

[mu38&Bg#]

Very nice work!

 

[mayhugh1]

I've had major concerns about two aspects of this build from its beginning. The first was my ability to get a built-up crankshaft aligned and spinning freely to my satisfaction in a four piece crankcase. Somehow, that hurdle is long behind me. My second concern involves the intake system. This engine uses a two-into-one intake pipe with silver soldered flanges on the two legs of the 'Y' that mate the pipe assembly to the heads. Each fabricated 'Y' pipe is shared between a front row and a back row head. This means that a fairly complex and rigid pipe assembly will have to be accurately fabricated and sealed without leaks to three different surfaces in the engine. And, it will have to be repeated nine times. Three steps in order of increasing difficulty are required to build the assemblies: 1) accurately forming the stainless tubing bends and silver soldering the flanges, 2) neatly welding the 'Y's, and 3) actually installing the completed assemblies. I think the first step can be done with the proper care, and I hope the second one can be accomplished with enough practice. The third one, though, may not be possible with much of the design that I've already completed and grown fond of.
I had a chance to examine Hodgson's 18 cylinder twin at the Cabin Fever show. His engine uses similar 'Y' pipes, but with his head design the pipe assemblies slide easily into place from above the engine since the pipe flanges mount to easily accessible exterior vertical flats on the heads. Both he and the Chaos guys also used soft brass tubing that can be man-handled some to coax the flanges into position.
In my engine the mounting flanges are sunk into pockets in the heads (mostly for cosmetic reasons), and I'm using stainless tubing which is more difficult to deal with. And, if all this wasn't enough additional challenge, I'm starting out with a known interference problem between the bulbous rear row heads and the front row intake pipes. My modeling shows that after notching one side of each of the rear row heads I'll have the space needed for the pipe assemblies. What I haven't been to tell from my modeling, though, is whether I have the needed 'wiggle room' to actually maneuver the completed pipe assemblies into final position.
So far, I've only milled the recesses for the intake/exhaust flanges in my two test heads. My next step is to machine a few test flanges and do some experiments with actual parts in my hands to see if my current design can actually be assembled.
My only stainless scrap of the dimensions I need for the flanges is of an unknown alloy, and so I was prepared for a frustrating experience. While at Cabin Fever, though, I purchased two $10 unused 1/4" end mills from a vendor who told me they were specifically designed for cutting 304 stainless. They're center-cutting with a sharp corner, have five wavy flutes and a blue-violet coating, and they are razor sharp. Whatever my scrap alloy was, these cutters chewed through it with no effort or noise and produced a beautiful surface finish. I wish I knew more about them, but they have no markings of any sort.
My piece of scrap was long enough to make five flanges. I spotted, drilled, and reamed all the holes before contouring their peripheries. The workpiece was .080" thicker than my flange and so after contouring I simply cut the parts free using my bandsaw and milled them to finished thickness on my manual mill. My bandsaw whispered to me that the alloy was probably 304.
I partially assembled the crankcase and then set the two test heads in place atop cylinders - one in the front row and one in the adjacent position in the rear row. After several tries, I eventually had two pieces of tubing bent by eye that approximated a crude 'Y' to which I temporarily added the flanges. After playing with the pieces for a while it quickly became obvious that it was actually was impossible to install and remove the intake assemblies in my current design.
For one thing my flanges are thick and their fit in the deep recesses is fairly close with only .003" clearance. I designed them this way for a good looking and leak-free fit. There is also too little 'slop' in the fit of the intake tube in the crankcase plenum for 'wiggle room' before the o-ring compression nut is tightened. This was also intentionally done to reduce the chance of a leak at this connection. The intake tube assembly can slide vertically, but there is very little forward-aft clearance to rock the flanged ends into position in their recesses. The wide widths of the flanges combined with their close fits in the head recesses mean they can only be inserted when they are closely concentric to the recesses. The front/rear row cylinder-head pairs cannot be simply raised to accomplish this because there is a 20 degree angle between them. When they are lifted high enough to slide the flanges in place the lateral distance between them has increased to nearly 1/4" and prevents it. The close fit of the cylinder skirts in the crankcase as well as, eventually, the cylinder mounting studs prevent the heads from being tilted to gain the needed clearance.
There seems to me to be three possible solutions. The most radical and the one I least like is to re-design the flange mounting surfaces on the heads so they are vertically accessible like those on the H-9 heads and the pipe assemblies can be dropped into place from above the engine.
The second solution is one that I've gleaned studying the Chaos Industries photos. They are using the same deep recesses for their mounting flanges that I am using. However, their flanges are considerably thinner than mine even though they are mounted at the bottom of the recesses in their heads. In my opinion, the deep mounted thin flanges have a poor appearance because they expose the machined flat ends of the radius'd fins. But they were probably have enough clearance around them to rock the intake assemblies into position with the head pair slightly raised. Burying the flanges deep inside the recesses will tend to hide the these clearances.
Before the third solution occurred to me, I thinned a pair of my flanges and then machined two spacers to partially fill the flange recesses. This created a finished appearance identical to my original design but moved the now-thin flanges out toward the front of the heads where some misalignment during insertion could be tolerated. As far as I could tell without building up an actual welded assembly, I may have been able to coax the assembly into place. I might be more difficult to remove it, though, if the need should arise.
The third solution was one of those things that finally smacks you upside the head after you've spent too much time thinking inside the same box. This solution involves simply cutting the long intake pipe into two pieces in order to separate the troublesome connection between the front and rear row cylinders during assembly. This allows the two sub-assemblies to be easily and individually installed. A sleeve is then slid over the junction to seal the connection. In my design the sleeve will be mostly hidden from view within the notch of the adjacent rear row cylinder.
I'd like to come up with low profile stainless sleeves since some of them will be visible in the lower cylinders, but so far I haven't been able to come up with a suitable seal.
I've started a fuel compatibility test with a piece of clear Tygon tubing That I had on hand, but I plan to do more thinking about a suitable seal in the next few days. A sleeve of innocuous clear heat shrink tubing failed my fuel test during the first few minutes. -Terry

 

[Sparky_SC]

Amazing work ! I find this thread a "graduate course" in workholding techniques which is often a huge challenge.

I am curious how the tubing bends were done? Some of the tubes are pretty short and most have multiple precise bends. Once again, amazing work !