The last post talked about the stripped spark plug threads. Well, today I got the Time-Sert repair kit in the mail so it was time to see whether the repair would work.

The Time-Sert repair starts with a thread tap that has an initial part that is the size of the spark plug thread and then steps up to the larger thread for the inserts. The start thread makes it possible to start the tap in the old threads and thus get it aligned properly. This works great, assuming there’s anything left of the old threads, of course.

After cutting the thread for the insert, you leave the tap in place and slide a seat cutter onto the tap, which then guarantees that the seats for both the insert and the spark plug itself are aligned with the new thread. Quite nifty.

Of course, in my case several of the holes had these crappy repairs that needed to get ripped out, so in several holes the start thread didn’t even engage. The threads looked quite awful, in fact:

One of the old spark plug threads, clearly in distress. I think this was the thread that had the “half insert” chunk shown in the last post.

Another sketchy thread, slightly better. This one had a rusty helicoil that I ripped out.

And the third sketchy thread that had a helicoil. This one was in decent shape but the seat was so bad that I decided to go ahead and rip it out anyway. This hole is one of the “extra” spark plug holes that are machined into the heads for aircraft use. Those holes come in at quite an angle.

The previous spark plug hole after cutting the insert threads and seat. The insert threads clearly have not been completely cut, but the helicoil and the insert have the same pitch, so the insert actually locks in place securely. Note the freshly cut seat for the spark plug washer, which is quite a bit misaligned from the casting surface.

After cutting the insert threads, you screw the insert in place. The bottom of the insert has threads that are not fully formed, and the insert driver is a thread-forming tap that, as the insert bottoms out against its seat, will expand it and lock it in place.

The insert in place. I also rounded off the sharp edges of the hole with a grindstone. The sparkplug threads are just a little bit longer than the insert, so it actually sits pretty shrouded into the hole. Maybe that’s why the engine seems to always run worse on these spark plugs than on the ones in the standard holes?

Nice, new seat and insert in place. From the different depth of the seat cut around the hole it’s apparent the old, cast surface was not very perpendicular to the hole.

So that actually seems to work. I have a few holes where I haven’t decided whether I should rip the helicoils out yet, but when I sent some pictures of the repairs, one of the other Sonex owners here pointed out that it looked like there was a crack between the valve seats.

I did a little grinding to see whether it was just a surface feature. Nope, it looks like a crack. And actually three cylinders seem to have cracks in that area.

This sure appears to be a crack between the valve seats.

Another crack.

They certainly don’t seem to be just surface features, but they also don’t seem to go all the way through to the exhaust and intake ports. I’m not sure what the implications of these cracks are. Cylinder head cracks are bad on standard aircraft engines because the entire head is screwed onto the cylinder and is in tension. This means that a crack in the cylinder head can propagate all the way around and make the entire head fall off.

On the Aerovee (and on most non-aircraft engines I’ve seen) the studs go through the entire head and hold the head in compression, which means it can’t just crack completely through and fall off the way a Lycoming/Continental head can (and frequently do). I’m not sure what the worst case scenario is, but it seems that a serious crack would be noticeable as bad compression long before it becomes a safety issue.  I’ve asked Sonex what their opinion is, but hopefully we won’t have to buy new heads just yet.

After mounting the stainless steel intake runners, I did a few test runs to adjust the AeroCarb needle. I’m not going to get into the results from those runs right now since they’re incomplete, except to say that it seems like when you plot the EGTs vs fuel flow, it doesn’t matter how you adjust the needle. This is indeed exactly what you’d expect, since the needle setting just changes what fuel flows you’ll get at different mixture settings. This is good, because that means we can make sure we get data on the EGTs for different fuel flows, decide what fuel flow gives the desired temperature drop from peak EGTs, and then adjust the needle to get that fuel flow at full rich.

However, as I was making a test run, the engine surged when I pulled back from full throttle, died, and wouldn’t restart. It seemed like it was lean, so I immediately thought it had sprung a leak somewhere in the intake system. I couldn’t find it, though, so I took everything apart again, took the plenum home, and pressurized it. No problem.

Eventually I realized I had run out of fuel. Classic… It turns out the fuel meter is really poorly calibrated down at the bottom, such that it went from saying I had 7 liters to 1 liter in less than a minute. I looked at it when I started the run, concluded I had plenty for what I wanted to do, and didn’t look again. Well, now I know that…

As I was trying to figure out why it didn’t run, I turned the engine over by hand several times and noticed a hissing noise coming somewhere from the area of the #4 intake. I couldn’t quite hear if it came from the intake itself, meaning the intake valve was very leaky, or the head gasket, so I brought my mechanic’s stethoscope down to listen.

Turns out it was one of the #4 spark plugs that was not sealing, and when I attempted to tighten it, the thread stripped. Great. I removed the spark plug and turned the engine over again. It was still hissing, so I also found a leaky, stripped spark plug on the #2 cylinder, and then one on the #1.

So, three stripped spark plug threads. No big deal. I remember when we reassembled the motor after Bob’s valve job in May that I thought at least one of the spark plugs felt fishy when I tightened, so this was not a big surprise.

It’s also apparently common enough that the Sonex Builders Foundation has an Aerovee spark plug Time-sert repair kit that they will lend you, so you don’t have to buy the $160 kit and use it once. A great service provided by Mike Smith at the Sonex Foundation (who had to repair his spark plugs…)

The IKEA airplane mechanic’s umbrella came in handy. Standing out on that ramp in the Sun a whole day is just a killer. The umbrella made it bearable, and it only blew over once.

After spending all of Sunday down on the ramp (in the shade!) taking the heads off and bringing them home, I removed all the other spark plugs and discovered that fully 5 of the 8 spark plug threads have already been repaired!

How this came about is a bit of a mystery. Builder and co-owner Bob has no memory of ever repairing a spark plug. The heads have been rebuilt twice by the local VW shop, but those bills don’t say anything about repairing spark plugs. And while some heads came from Sonex already with inserts in place, our heads are not in that S/N range.

Some of the repairs look like Time-sert style inserts and appear decent, but a couple are absolutely terrible.

This partial, rusty insert was sitting in one of the spark plug holes. The hole itself appears to be oval, so it’s as if someone tried to save it by jamming half a thread insert in there!

The partial insert above, and a really bad Helicoil, were also not square with the seat, so the spark plug washer appears to not have been making contact across the entire face.

It’s obvious that the washer on this spark plug has not been making full contact with the seat, so the repaired hole must not be square to the seat. No wonder it leaked.

Le sigh. What the hell kind of hack job is this? On an airplane engine no less! I came across the accident report of a fatal accident in a Sonex where the engine lost power after takeoff and the pilot attempted to turn back to the airport and spun in. Cause of power loss: a botched spark plug repair on the Aerovee engine. This is not some abstract theoretical thing!

It’s unclear whether these previously-repaired holes are repairable. I’ll give it a try, but if it turns out to be sketchy, we’ll probably just get a new pair of heads. Luckily parts for these VW engines are fairly cheap (extremely cheap by aviation standards!) and replacement heads are only $325 each (and rumor has it that the current manufacturer used by Sonex has much higher quality workmanship than the previous heads, so we’d probably end up with better heads, too.) But it’s not like we’re itching to waste upwards of $700 on that, there are other things that need fixing, too…

That’s the current situation. The Time-sert kit should hopefully show up in the mail tomorrow and then we’ll see what happens. I’m really hoping we’ll have the engine together so we at least can taxi the plane over and display it at the Annual Hilo Airport Aviation Day and Fly-In on 10/21. More later.

The last post ended with a test run with the cowling on, almost melting the temporary Nylon 3D-printed intake runners. It was time to get to work fabricating the real ones.

Once I had established that the engine ran well with the plastic ones, I had ordered four 90-degree, 2″ radius, 1.25″ OD mandrel bends in 18gauge 304 stainless from Burns Stainless. These come with 4″ and 12″ legs on the bends.

From the design in Fusion360, I knew the exact angles and lengths of the four runners, so now I just had to cut the bend to the right angle and weld the appropriate length straight on.

Speaking of cutting… When I welded the slip joints on the NC30 exhaust, I cut the stainless tubing with my Evolution RageII saw using the blade designated for stainless steel. This worked, but the blade only did maybe 10 cuts before it basically stopped cutting and started melting its way into the stainless. I returned the blade to Amazon and got another one, and talked to the Evolution people who said that as long as you don’t force the cut and make sure the stock is securely fastened, it should last much longer. Well, being super-careful about cutting, it didn’t really.

Given those blades are over $100, I gave up on using that saw for stainless. Instead, I ordered a benchtop band saw from LittleMachineShop.com (the guys I bought the mini mill from) It’s $250, and the band saw blades are $10. It cuts much slower, but there’s already proof that the blades last much better than those for the Evolution saw and the cuts are very nearly as clean as those made with a sharp Evolution blade.

One of the intake runners mounted in the band saw for cutting.

Setting up the cuts was a bit tricky since they were all at different, totally arbitrary angles, like 51.7 and 71.4 degrees. I used a cheap digital angle finder from Home Depot to adjust the angle of the vise on the saw against the blade. It took some fiddling, but the cuts came out to within a few tenths of degrees which basically is the accuracy of the angle finder anyway.

The four intake runners tack-welded together and ready for test fitting.

With the cuts done I tack-welded the straight sections on and took the whole thing down to the airport to test fit. Everything fit OK, although the #2 and #4 ones could have been a few mm longer for a perfect fit. That’s well within the slop of the silicone couplers, so no big deal.

Then it was time to weld them. I had just replaced my second Argon bottle that I use for purging the inside of the pipes. Apparently there’s a nationwide shortage of Argon and that little bottle cost over $200 to replace. Expensive hobby, this TIG welding thing…

I did some practice welds since this is 18 gauge (0.05″ or 1.2mm wall thickness), which is the thinnest material I’ve ever welded. They say it’s very critical that the tubing fitup is good, you should basically not be able to see any light through the joint. Any gaps will vastly increase the chance that you blow a hole.

The other problem I’ve always had when trying to weld stainless is that it’s really hard to see what you’re doing. When welding aluminum, everything is shiny and bright and it’s easy to see what’s going on. With stainless, you use much less current and it just seems less reflective. On my practice welds I managed to weld way off the joint because I couldn’t see it at all.

I came up with a few tricks to help this. The first, and obvious, one is to make sure you have really good work lighting. The second one was to tape a piece of fabric to hang down over the back of the welding helmet to avoid reflected glare from lights and windows behind you. Finally, I discovered that if I before starting the weld closed my eyes for ten seconds or so and then opened them and immediately started welding, they would have adjusted to the darkened helmet better and it was much easier to see.

Once I could see where I was supposed to weld, it was much easier. Purging the inside of the pipes is kind of a pain when they’re this small because everything gets really hot, the purge hose falls out, etc. The regulator on the purge bottle is also the really shitty regulator that came with the welder, and on the first pipe I didn’t realize that it had totally stopped the flow of gas after a little while and I ended up with the “sugary” carbide precipitation on the part of the inside of the weld that is exactly what you want to avoid by backpurging. Maybe it’s worth investing in a better regulator; given that a little bottle of Argon is $200 it might pay for itself in a little while to know exactly how much gas flow you have…

The final welded runners. The welds aren’t awesome but they’re decent. There are a few places where I melted through enough that there’s a bulge on the inside but most of the joints are good. With the exception of the one where the purge gas stopped, of course…

In any case, the welds got done and they’re not total disasters. They’re at least air tight which is the most important thing…

After mounting it all, I recorded a short video to better show how all this looks on the engine:

So that should hopefully complete the fabrication part of the intake project. Now it’s time to tune the carb so we get the right mixture and then see if we can run it with the cowling on without melting the plenum…

Since the last test run had the engine running quite rich, I adjusted the carb a bit leaner and tried again. It was definitely leaner, but the differences are getting small enough that I think I have to let the EGTs settle for a longer time to get a stable reading. This means I’ll basically have to do a run at full throttle with a fixed mixture setting, go down to idle and let it cool, do another full-throttle run with a slightly different mixture setting, and repeat until I get enough data points to get something reliable.

I did also decide to try a run with the cowling on, to see what the effect on temperature was. The cylinder head temperatures were overall a little bit higher but showed less spread than without the cowling. In both cases #4 was the first to overheat. Without the cowling, the CHTs when #4 hit redline (set at 400F, or 204C) were 189C, 167C, 182C, 204C. With the cowling on, they were 203C, 185C, 195C, 204C. So the spread without the cowling was 37C but only 19C with the cowling on. I guess this indicates that the cowling actually does help force the air through the cylinder cooling fins, making the air flow more equal.

However, the thing that stood out the most was that temperature sensor I had mounted on the cylinder #2 intake runner where it passes near the exhaust header. With the cowling off, the temperature at that point rose moderately, hitting 50C as I pulled back from full throttle. With the cowling on, it just kept rising: 60C, 70C, 80C… As it went into the 80s, I pulled back the throttle to idle, since that’s far higher than what the nylon they’re made of can take. With the cowling on, however, the temperature just kept rising, even after I shut the engine down. When it hit the 90s I jumped out and quickly took the cowling off to vent the hot air. The runners were noticeably soft when I squeezed them.

So, there’s no doubt that the runners need to get replaced with stainless tubing. I’m in the process of cutting and welding those right now. I also fabricated two exhaust shields from 24-gauge stainless that I bent into a cylindrical section and welded tabs for hose clamps on. These will prevent the headers from radiating directly onto the runners that exit the plenum. I’m really hoping this will be enough to keep the temperature down enough for the plenum to survive.

The last ditch solution would be to use exhaust wrap to keep the headers from radiating all their heat inside the engine compartment. I’d prefer not to have to do that since this raises the temperature of the exhaust pipes and can lead to premature failure. People often use this with cars, but cars don’t spend a lot of time at full throttle, whereas airplanes do. We’ll see what kind of temperatures we see with the heat shields.

I’m hopeful I can weld up the stainless intake runners and get another test run in this weekend. Stay tuned.

With the vacuum leak fixed, and with manifold pressure and temperature sensors, I was excited to try another test run. Unfortunately there is no video this time.

The short story is that it runs really well. Static full-throttle RPM is a tad higher than it was with the leaky plenum, about 3230, but more importantly it now idles perfectly fine. Rather than running lean as it was with the leaky plenum, it’s now quite rich.

Let’s look at some data:

This plot is like the one from the first test, showing the exhaust gas temperatures for cylinders 1 and 4 (the #2 EGT gauge is having a bad day and need to be replaced, so I’m using #1 instead) along with the fuel flow. The x-axis is time in seconds where zero is approximately onset of leaning, so negative times is full rich.

The solid lines show the current, non-leaky plenum, and the dashed lines the leaky one. It’s obvious that the current run has much higher fuel flow at full rich mixture than the leaky one, and correspondingly lower exhaust gas temperatures.

When leaned, both runs end up at similar fuel flows and EGTs. This isn’t surprising since you lean until the engine starts running rough and then go a bit richer, so you should end up at the same fuel flows. (Those fuel flows happen at quite different mixture settings in the two runs, however.)

The instructions for setting the carburetor say that full rich should have about 90-100F (i.e. 55-60C) colder EGTs than the lean letting. In the present case, the difference is more like 70-100C, with the back #4 actually having a higher difference, so appearing to run richer than the front cylinder now. In any case, the engine is quite rich. At idle, it also had to be leaned almost to cutoff mixture to run cleanly.

The full throttle manifold pressure was about 97kPa, within a few percent of atmospheric pressure, so the air filter can’t be a significant restriction. After warming up, it would idle at a manifold pressure around 37-40kPa, which is quite low compared to the motorcycle which idles at about 60kPa. This is presumably the difference between having a nice big plenum to even out the pressure.

The manifold temperature was kind of interesting. When idling, it showed 28C, pretty much ambient temperature, but at full throttle the temperature dropped to 17C, more than 10C below ambient. This must be due to the evaporation of the fuel droplets in the plenum. This makes sense, since I had already noted that the stock intake pipe had condensation on the outside after running the engine. I’m just surprised it makes that much of a difference.

So this is very promising. The carb needle needs to be adjusted to lean the mixture a bit and then we’ll see if we can make a more thorough check of the mixture distribution.

With the vacuum leak fixed, I was excited to try another test run. But first, a small side track.

The “RDAC” box that connects all the engine sensors to the MGL Enigma comes in two flavors, one that has a manifold pressure sensor and one that does not. The one in N132EA does not, but from reading the manuals (and looking at the pitot-static connections on the Enigma itself) I had a sneaky suspicion that the MAP sensor used was the same Freescale sensor that I used for the MAP sensor on the NC30. The manual said it was a 0-250kPa range sensor, and coincidentally I happen to have a spare MPX4250AP. Opening the RDAC box up, sure enough, there were unused solder pads where the MAP sensor should be, although only 5 and not 6 pins like the MPX. However, I determined that the pinout was compatible (only 3 pins are used on these particular models) so I soldered it in, reconnected the RDAC and fired up the Enigma.

The RDAC, now with MAP sensor. I only had to solder the Freescale sensor in and drill a hole in the lid.

I was extremely pleased to see a MAP reading that matched the atmospheric pressure reading within a kPa, and sucking on the MAP hose made the reading go down. Perfect, now we have a MAP sensor (and the sensor is only $15 compared to the $65 or something MGL charged extra for the MAP version.)

After drilling a hole in the plenum (two holes, actually, I also mounted an LM355 temperature sensor so we can read manifold temperature) and 3d printing a small hose barb that I epoxied into the hole, I could hook the MAP sensor hose up to the plenum.

Re-mounting the plenum and connecting the intake runners didn’t take long, so it was time for a test run. However, when I pressed the starter button, all I got was a “click”… The starter relay was definitely kicking in, but the starter did not turn. I wondered what could be causing that connection to not be made when I noticed the main battery lead to the starter coming out of the lug at a peculiar angle.

The lug on the lead to the starter was not crimped real well, it was more “crushed” than a real crimp.

Apparently the wiggling of the plenum in and out had been enough to pull the lead out of the crimp in the lug. Now, this should just not be possible, I should be able to hang from a properly crimped #4 AWG lead, but the crimp was not a proper hexagonal crimp but rather looked like someone had stamped a cross in the copper lug.

A proper crimp squeezes the lug from all directions and actually cold welds the strands and the lug together into a single piece of metal. Luckily I happen to have a few lugs I got for the NC30 #6 AWG battery wire which actually fit the #4 AWG starter wire too. I also have up a cheap hydraulic crimper I picked up on eBay, so I could replace the lug easily.

The new, properly crimped lug. Note how the entire perimeter of the lug has been compressed into the wire.

After that small delay, it was now time for the test run. That’ll be in the next post.

Well, the good news is that I figured out where the vacuum leak is coming from. The bad news is that it’s the 3d-printed plenum itself that’s not anywhere near air tight.

I guess I should have checked this more carefully, since the fact that 3d-printed parts aren’t water tight is pretty widely known. I had assumed this was because of poor layer adhesion when printing with PLA, and that nylon would not have this problem since the layers adhere very well. Well, I was wrong.

To try to figure out what was leaking, I plugged up the outlets with plastic sheeting and clamped them. When I blew into the inlet on the bottom, the pressure wasn’t building up at all and you could hear hissing all over the place.

The leaks were concentrated in the regions that were printed horizontally. I suspect what’s going on is when the walls are close to vertical, the perimeters are continuous and form a sealed wall. On horizontal parts, however, the space inside the perimeters is “filled” with a zig-zag pattern. I have observed before that whenever the nozzle changes direction as it meets the perimeter, there tends to be a little unfilled hole. Since every layer has a different angle for the zig-zag fill pattern, and the parts were printed with 5 solid layers on horizontal surfaces, I thought the holes on different layers would not overlap, but apparently they do.

The infill pattern used on the mostly horizontal bottom parts of the plenum. These were the regions that leaked.

Needless to say, this was a disappointing discovery. I contemplated the alternatives (including burning all parts for this project in a ceremonial pyre) and eventually settled on attempting to coat the outer surface of the plenum with epoxy. Epoxy is resistant to avgas, so that would not be a problem, and by brushing it on the outer surface and using the shop vac to suck a vacuum inside the plenum, the epoxy would get sucked into any pinholes and cavities. I had already RTV’d the two halves together, so I only had access to the outside, anyway. It was worth a shot.

I mixed up two shots of West Systems and started brushing it on. I was a bit worried that the epoxy would not “wet” the nylon, but that turned out to not be a problem. In fact, the areas that were porous sucked in resin like crazy. I only applied a few, short, bursts with the shop vac; since the part was so porous I figured that if I left it on it would suck all the resin straight through to the inside which would not be the desired outcome.

The freshly coated plenum, waiting for the epoxy to cure. It sure is more shiny than it was before, and hopefully also air tight.

After letting it cure overnight, it was time to test it. First I had to sand the outlets, which had prominent ridges from the layers of plastic and did not want to seal against the silicone couplers. After plugging up the outlets, it was immediately clear that it was tons better. Blowing into it now, there’s a clear buildup of pressure. By dunking the whole thing into the bathtub again, it became clear that most of the leakage was at the plastic sheeting over the outlets. However, the part itself was still not completely sealed, there were (small) leaks from a few spots. I noticed only one leak from any of the areas that were previously coated, the others were either from the join of the halves or from right next to the threaded inserts where I didn’t really want to put epoxy for fear of getting it into the threads.

I’ll just have to mix up another shot and redo this once and then I think we should be in business. In the meantime, I’m working on adding a manifold pressure sensor…

Finally, almost 2 months after the initial intake experiment that indicated the need for a complete redesign of the AeroVee intake, we now have some data!

After assembling everything on the engine yesterday, I test ran the engine today and the results, while not entirely perfect, are very promising. A full throttle run shows that the mixture is much more even between the front and back cylinders than it was before.

Here is the updated version of the EGT plot I made after the first experiment. It’s a bit busy, so let’s go through it step by step. The three runs are plotted with different lines, today’s plenum test in solid, the stock intake in dashed, and the 3d-printed elbow in dot-dashed. Times are aligned such that at times before 0, the mixture is full rich and at time 0 the mixture is leaned until the engine starts running rough.

The EGTs of cylinders #2 (front) and #4 (back), in degrees C, are plotted in blue and green. The red lines are fuel flow and black ones engine RPM, scaled to fit on the same plot.

The fuel flows at full rich, just before time 0, are very comparable for all three runs. The engine RPM is also very similar, although the solid line is slightly higher than the others. Static RPM with the plenum was about 3220 while the others were around 3150, so that’s a good sign; whatever else is going on, the engine is making more power than it used to.

Looking at the EGTs, the blue and green solid lines, it’s obvious that they are closer together than they were before. At time 0, the EGTs are 675C and 700C while the stock intake run had them at 550C and 720C. The behavior when leaning is also very different; with the stock intake the #4 EGT (dashed green line) dropped while the #2 (dashed blue line) rose. With the plenum, they were almost unaffected, although there’s some indication that the green line dropped and the blue line rose slightly.

All in all, these data indicate that the #2 and #4 cylinders now run a much closer mixture, but also that that mixture is quite lean since both cylinders appear to be running near peak EGT with the mixture at full rich. The #4 cylinder appears to be about the same mixture as before, since its EGT is the same, while #2 is much leaner. Yet fuel flow is the same. How is this possible?

I think the answer is that the engine makes more power so, even though the fuel flow is about the same as with the stock intake, the fact that the engine is running at a higher RPM means that its pumping more air. More air, same amount of fuel means a leaner mixture. It’s interesting that, apparently, the more even mixture distribution means overall more power than before even though the cylinders appear to be running quite lean. (That also likely means it’s possible to pick up a bit more power by tuning the richening up the mixture, when I get around to that.)

So why would the carb not give more fuel with more airflow? I’m speculating a bit here, but the conventional wisdom in carburetor circles appears to be that a more pulsing airflow will make a carburetor give more fuel compared to the same average airflow without pulses. Something about the pulsing airflow giving higher maximum air velocities through the carb will tend to draw out more fuel. I don’t know whether this applies to the AeroCarb with its dirt-simple construction, but it’s at least plausible.

That’s the upside. The downside is that it won’t idle. I could not get it to run stably at an RPM less than 1600. Anything less and it would stumble and surge. After fiddling a while I found a throttle position where the RPM would oscillate like clockwork between 1100 and 1600 with a period of about 2.5 seconds. While doing so, the EGTs for the front cylinders also dropped out, so it was really only running on 2 cylinders. This was leaned a bit, at full rich all four EGTs were at least alive but it would still not idle stably.

This kind of surging idle is a textbook symptom of an intake leak. It doesn’t matter much at full throttle, since the pressure in the intake manifold is close to atmospheric, but at idle there’s a substantial vacuum in the intake and any leak means the engine will draw in air without a corresponding amount of fuel. The fact that the front cylinders both dropped out might mean that the leak is where the plenum halves are joined together. If the joint on the front side leaked, it might predominantly pull that air into the runners to the front cylinders and make them run lean.

When I joined the plenum, I just put a small bead of silicone RTV on the halves and put them together. It’s quite possible this was not entirely sealed, especially since the plenum flexes quite noticeably from the engine vacuum. During the test run I had the GoPro mounted on the firewall pointed at the plenum and intake runners, and you can actually see how the sides collapse slightly when the engine starts:

Look at the rear side of the plenum, to the lower left in the image, when the engine starts. I’ve included a few cuts back and forth between engine running and engine off to make it more obvious.

The static stress simulation I made in Fusion 360 to make sure the design was strong enough indicated that the sides of the plenum would flex about 1 mm under ~50kPa vacuum, but that did not take into account the fact that the wall is not solid but has a honeycomb infill, the exact material characteristics of Alloy 910 (I used Nylon 6), or the fact that the joint along the center will tend to pry itself apart as the sides flex inward and the edges of the joint rotate. (I’m reminded of the reason the SRB joints on the Space Shuttle leaked, leading to the Challenger disaster. It was a similar effect, the pressure on the inside of the SRB casing made the joint flex and unloaded the O-rings, letting gas blow by… but I digress.)

I will try reassembling the plenum halves using more sealant, and try “Aviation Form-A-Gasket” instead of RTV (which is not really fuel-resistant). Aviation Form-A-Gasket is very tacky and never really solidifies, so hopefully that will keep a seal even with some flexure of the joint. If worst comes to worst, I can glue the two halves together, because I’ve established that it is indeed possible to extricate the assembled plenum through the spaces in the engine mount.

So that’s where things stand as of now. I’m going to order the stainless tubing for making the permanent runners, and re-seal the plenum halves and see if it indeed is an intake leak. At least now we have a running engine to work with.

With the 3D-printing trials and tribulations overcome, I pulled the left half of the plenum off the printer.

This is the left half of the plenum as it came off the printer. It looks good and is plenty strong, but the shiny parts indicate that the outermost perimeters aren’t always fused properly. (I think the very shiny reflection essentially comes from total internal reflection against the inside of the round plastic string extruded by the nozzle. When they are fully fused, the part becomes much more transparent.

The two halves fit well together, so I proceeded to melt in the brass M5 thread inserts for the mounting bracket and the screws that hold the halves together. The plastic parts were now ready to be assembled.

The welding of the left-hand intake, on the other hand, did not go so smoothly. The first runner went on fine, but when I attempted to weld the second one (where access to the area between the two pipes becomes difficult) I managed to blow a hole in the aluminum tubing. After a few attempts it was clear that filling the hole was not going to happen. After fuming about this failure for a bit I decided to just run with it and fill the hole and the area between the two intake runners with high-temperature JB Weld. This is good for 450+F and if the intake gets that hot something has really gone wrong…

Although it’s not used in a structural application, I’m not sure I trust JB Weld to hold up in flight. The first order of business, however, is to figure out if this setup is even any good, and it’s certainly going to be fine for that. If we decide that we’re not comfortable flying with a JB Welded intake we can ponder redesigns that will make it easier to weld.

Here’s how the whole thing is going to look, intake flanges, runners, and plenum.

Before we go and put this shebang on the airplane, I thought it would be fun to recap how the design evolved.

The progression of the plenum design, starting with the cylinder with four runners exiting upward on the left through the increasingly non-cylindrical shape we’ve ended up with on the right.

From the initial design, the plenum went from a cylinder to a much more tapered shape, and it also increased in volume substantially. The first drafts, before I realized how the plenum could be enlargened without interference with the engine mounts, had a volume that was significantly smaller than the engine displacement of 2180 cc, I think somewhere around 1500 cc. The final shape is about 2700 cc, minus the volume taken up by the internal bellmouths, so it probably comes out slightly larger than the engine displacement. This is good, the rule of thumb for plenum volume is that you want no less than the engine displacement.

Today I spent most of the day down at the airport (being baked on the black apron in the tropical Sun) putting all the parts on the engine. Apart from having to remove a mm or so of material from some spots on the rear cylinder head cooling fins where the front intake runners come down, everything fit well.

The new intake mounted on the engine for real.

There are a lot of clamps. We’ve gone from 6 hose clamps on the old intake to 18 on this one. This is not awesome, since it means many more opportunities for leaks, but there’s no way around that with individual runners. The only alternative would be to make either the plenum or the intake with very long, curved runners, which would in practice be impossible to fit.

The new setup moves the AeroCarb back by about an inch, which makes the fuel line too long. It’s fine for testing, but eventually I’ll need to shorten it. It also required some adjustments to the throttle and mixture cables, but they can be moved around more easily than the fuel line.

The first test run should be coming up rapidly, hopefully this weekend. Stay tuned!

The last post ended with the test print of the plenum.  This was a useful exercise as test fitting the plenum in combination with the “Lego” runners indicated some adjustments to the angles of the runners’ exits from the plenum. After fixing that, it was just a matter of printing them…

… or rather, that’s what I thought. I had test printed small pieces with Alloy 910 before and they had come out well. However, each of the plenum halves take more than 24h to print, and a few hours after starting the first half, it became clear that all was not well.

There are many things that can go wrong when 3d printing, especially large pieces. Since each layer is built up by extruding molten plastic on top of the completed part, and plastic shrinks when it cools, the part will tend to “curl” upward. As these stresses build up layer by layer, it’s possible for the part to pop off the printing bed entirely. That did not happen here, but on a smaller scale, printed overhangs, where the freshly printed section is the thinnest, did the same thing. Once it has started, this effect tends to reinforce itself since as the plastic curls up, the extruder will rub against it rather than go above it when depositing new layers. This repeatedly re-melts the top of the part, making it worse and worse.

The trickiest problem to solve was the curling up of convex edges. This closeup shows how the layers rise towards the front of the part. it also shows all the cooked plastic crud getting picked up by off the nozzle as the raised edge brushes against it. The brown line is from a nozzle travel move that always went in the same direction and bumped into the progressively more raised edge.

Once started, there are only two ways this process ends: Either the overhang angle of the part becomes more vertical such that the printed edge thickens and eventually becomes stable enough to hold its shape, or it builds up until the printer nozzle can no longer cross over it and the print head jams and loses position. The former results in a bad print, the latter in a complete fail.

Here’s a print that was aborted due to the upcurled edges. It’s obvious how the structure of the plenum wall towards the bottom is completely disrupted and the honeycomb infill no longer coherent. The part shown in the previous picture is the one on the left.

The edges on my print did not fail, because the edges in question were merged with other pieces of the part before it got that bad, but the result was that the wall thickness was severely reduced and the honeycomb infill destroyed in these areas.

Once the print was done, I proceeded to start prying the supports off and it rapidly became clear that something was seriously wrong. The supports came off easily; too easily, in fact. And as I was prying the supports off, the part itself started coming apart, making crunching sounds as it did so. I could push with my finger on an area and the plastic would make crunching sounds and collapse.

Here’s the initial print. As you can see, it’s cracked along layers just from being squeezed between my fingers. On the correctly printed part, I can barely make the edges move by pushing as hard as I can.

I’m not entirely sure what went wrong here. It’s possible that the 235C printing temperature is too low. The manufacturer recommends 250-260C, and cautions that too low temperature will cause poor layer adhesion. It’s also possible that the plastic did not handle sitting on the heated bed at 85C for 36h. That does not seem likely, because the Alloy 910 material is supposed to be able to handle those temperatures. Also, if sitting on the bed was the problem, the top section that printed last should not be affected, but the entire part is very weak.

This was pretty disappointing. I went back to basics and started testing a wide variety of temperatures, print speeds, fan settings, etc, on a small section of the plenum used as a test piece.

This investigation showed that both when printed at 250C and 260C, the recommended temperatures, the part was extremely strong. I also lowered the bed temperature to 45C, which they also recommend. This did not seem to cause any more warping and in fact made it somewhat better, presumably because the material cooled down faster.

The test pieces also showed that anything printed at 0.2mm layer heights quickly ended up with upcurled edges, while those using 0.3mm did not. Increasing the extruded width from 0.5mm to 0.6mm also helped decrease the upcurl significantly.

However, at 0.3mm layers, the space to the supports (there is at least one layer of empty space between the support and the printed surface) was large enough that some extruded loops of plastic on the bottom layers did not adhere to the rest of the part. Decreasing the layer thickness to 0.25mm seemed to be a good compromise. This still had much less upcurl than when using 0.2mm, but there was also no loose loops on the bottom areas, while the supports still came off cleanly. (Too little space to the support structures and they can fuse completely with the part such that they become practically impossible to remove.)

After printing some larger test pieces including a complete runner tube, which requires supports on the insides to hold the “ceiling” as its being printed in, also showed that as long as the bottoms of the supports, which sit on top of the part, had 2 empty layers, they came off cleanly and without too much work, even on the inside of a cylinder.

The sum total of all non-production Alloy 910 test prints. The parts numbered 1–14 used a variety of different temperatures, fan speeds, layer heights, and extrusion widths. The larger parts marked “F1” and “F2” tested the final-candidate and really-final settings. The almost-completed plenum on the right is not a test-piece, it ran out of filament with only maybe an hour left…

Armed with this wealth of information, I figured I now had the settings dialed in and printing the part would not be a problem. Not quite.

Even though I had successfully beaten back the upcurl on my small test pieces, it still came back when printing the entire piece. I think this is again a result of the fact that it takes upwards of  5-10 minutes per layer near the bottom when all the supports have to be printed, and the material has even more time to cool than in my test pieces (even though I printed them 4 at a time to make it take longer and more closely mimic the conditions when printing the whole part.)

The final piece of the puzzle was to make sure that all the convex, overhanging edges had some supports to “anchor” them to the base and prevent them from curling. The straight overhanging edges generally were not a problem, so by making a few “lines” of supports going out in the directions where the surface had the strongest curvature I was finally able to print the full 36h print and have it come out basically flawless.

The completed right plenum half, ready to come off the printer. This print took about 36h.

Then it was time to print the other half. You may have read in the previous post how I attempted to make sure the parts were well below 1 pound in weight since that’s how much is on a roll of Alloy 910. Well, I found out that Taulman also sells 1kg rolls, so I ordered two of those. Given that I now had a whole kilogram of plastic, I now let my guard down and didn’t monitor how much was left on the roll…

You can see where this is going, right? Yup, it ran out and I didn’t notice. The almost-completed part is to the right in the picture above. It only had probably an hour left of printing on top of the 24h+ it had beeen going… Frustrating, to say the least. At least enough of it was printed that I could use it to double-check that the runners were in the right angles by test-fitting both halves on the engine.

Luckily, I ordered 2 rolls… The left half is once again printing as I’m writing this, and should finish some time early tomorrow. I’m also finishing up the welding of the left aluminum intake flange, so with some luck we might be looking at a test run during the holiday weekend!