While doing the tuning rides on the NC30, I’d noticed a faint smell of gasoline from the seat area. While there shouldn’t really be a way for fuel to get out of the system, that’s where the fuel pump housing is located, so I figured the smell was making its way out somewhere. There was no visible gasoline anywhere.
Eventually I decided to take the seat off to check it out, and it was good that I did. When I pulled the Ballistic battery out, it looked like this:
The battery after being exposed to gasoline. The mottled pattern is from the texture of the foam that holds it in place.
The bottom of the battery had a consistency somewhat like chewing gum, and smelled of gasoline!
It turns out there is a small, almost imperceptible, leak in the fuel pump housing after all. It was small enough that there was no noticeable gasoline leaking out, but apparently enough that it eventually saturated the foam around the battery and partially dissolved the plastic battery body.
Now that I knew there was a leak somewhere, I did some persistent checking and eventually noticed a faint sheen covering the aluminum pump housing right in the corner where the weld had leaked before. It was so little that I could not see it on a piece of tissue after wiping the area, but the paper did smell like gas and after a little while, the sheen returned.
I think what happened is this: the tank has a check valve that lets air in but not out. This avoids pulling a vacuum as fuel is used up, but avoids letting the gasoline evaporate into the environment. After riding for a while, the NC30 gets very warm, including the frame and, eventually, also the fuel. This will pressurize the tank, since the expanding air has nowhere to go, so I suspect this increased pressure when warm pushed enough gas through the pinhole to be significant.
I had leak-checked the housing after welding it, but there must have been some porosity or something that eventually let go under pressure, because, when I put the whole thing under water again now and blew into the inlet hose, a small stream of bubbles came out of the inside corner weld… Oh well, I guess I’ll have to pull out the welder and see if I can fix it.
Also, does anyone else think it seems sort of sketchy to make a motorcycle battery out of plastic that apparently has no resistance to gasoline whatsoever? In contrast, the plastic battery box that the whole thing sits in, and even the foam holder, were totally unaffected by the gasoline. It doesn’t seem out of the question to think that something on a motorcycle would sometimes come into contact with fuel.
Over the past week, I’ve done a bunch of tuning rides on the NC30, logging data and tweaking the tables. Compared to the first run, it’s now fully rideable with only a few touchy points.
The main task when tuning the fuel injection is to calibrate the “VE table”. VE stands for “volumetric efficiency” and is the basic number describing how much air the engine is pumping. The entire function of the fuel injection system is to supply the engine with the proper amount of fuel for the air it’s getting, so calculating the mass of air is half the work. The other is calculating the corresponding mass of fuel and making sure it gets into the engine. That’s arguably the simpler part, once the injectors have been characterized.
The basic calculation is that, for every stroke of the engine, a cylinder sucks in an volume of air equal to its displacement. If the density of the air is known (which you can calculate from its pressure and temperature, both measured, using the ideal gas law) you know the mass of air pumped. Depending on the operating point, the air/fuel ratio changes from around 14.7 at low throttle to 12.5 at full throttle, but how to set this is pretty well known.
The complication is that the engine is not an ideal air pump. Especially at low throttle, it has to suck air past the closed throttle plate, and it also has to suck it past the intake valve. The intake system and airbox has some resonance behaviors that affect airflow. Because of valve overlap, some combustion products get pushed out into the intake when the intake valve opens, and so on. All these effects add up to the engine not pumping as much air as it ideally would, and this is the volumetric efficiency. The VE mainly varies with RPM and manifold pressure, so the Megasquirt has a 2D table of VE against these two inputs that it uses to look up the VE and calculate the length of every fuel injection event.
Since you can’t calculate the VE, it has to be measured by running the engine through all these operating points and adjusting the amount of fuel injected until the measured air/fuel ratio from the oxygen sensor in the exhaust agrees with the desired value.
In the end, you get something like this:
My current VE table (for cylinders 1/3.) The X-axis is RPM and the Y-axis manifold pressure (as a percentage of ambient pressure.) In general the VE table should be smooth, and there are still some bumpy spots here showing I have more to do.
This is my current VE table. How well does it work? Here’s a plot of the difference between the desired and measured lambda (ratio of air fuel ratios) on my last ride:
The difference between measured and desired lambda (air/fuel ratio relative to stoichiometric). Positive values mean the engine is running leaner than intended, negative values richer. X and Y axes are the same as on the VE plot.
Over most of the table, the values are within 0.05 and many are within 0.02, which is quite good. There are a few lean spots, at 4500 RPM and full throttle (Y-axis 100% of barometric pressure) and at 11500 RPM and >60% barometric pressure. The very large values in the lower right corner are not real, that’s when rolling off the throttle from a full-throttle run to the rev limiter, which makes the wideband sensor read bogus.
Here’s an excerpt from the log for that full-throttle run (you may have to click on the plot to enlarge it to see clearly).
An excerpt from the log file showing a full throttle run to redline. The top pane shows RPM in white, manifold pressure in red, and throttle position in green. The second pane shows the air/fuel ratios of cylinders 1/3 (white) and 2/4 (red) compared to the target values. The third pane shows injector pulse widths. Note the lean spot at 10750 RPM (just to the left of the vertical marker) where Lambda1 shoots up. The RPM trace also shows acceleration dropping off here and the engine runs too lean to develop peak power. Also note that once the engine hits the rev limiter, the lambda shoots up. This is because it shuts off the spark, sending unburned air out into the exhaust.
The log analyzer program, “Mega Log Viewer”, can calculate an updated VE table based on the information in plots like these, so then you update the VE table in the Microsquirt and go for another test ride, and repeat as necessary.
So that’ s the tuning basics. I’ve encountered a few subtleties beyond this, but I’ll leave them for another time.
On Sunday, I rode the fuel-injected NC30 for the first time! It’s great to finally be at this point.
Before I got to that point, there were a few final issues. After putting together, hooking up the fuel system, and verifying that all the sensors worked, I noticed that the throttle position sensor reported “20%” open with the throttle closed. I thought that seemed weird, since it should have been pretty well calibrated before, but I recalibrated it and moved on…
After filling the fuel tank, I was pleased to note air bubbles coming out of the tank connection, indicating that the air was making it out of the fuel pump housing, and when I ran the fuel pump it only took a few second for it to prime and start pumping fuel around. Glad that worked out.
When I cranked the bike it didn’t take many seconds for it to fire and the immediately rev up to the rev limiter (which I set to 8000 RPM for a cold engine). What the …?
Remember that 20% throttle? Turns out that the throttle cable end for the carbs is too wide for the GPz throttle body, and was wide enough to catch on a bolt so the throttles weren’t actually closing to less than… 20%!
Ok, whatever, I had hoped to not need to put a new end on the throttle cable but since it was also really too long for the adjustment to be able to take it all up, I cut an inch off the end and soldered a properly sized end on the cable. (I had planned for this and ordered a few cable ends of the right size before.)
With that fixed (and the throttle position sensor recalibrated again) I figured I would do what I really should have done in the first place and log the manifold pressure while cranking. This is the real test of how well the throttles are synchronized. When I did this, I was disappointed to note quite large discrepancies. I had hoped not to have to attempt to adjust the linkage on the bike since getting to the adjustment screws would be almost impossible, but no such luck.
While it requires some contorted fingers, it turns out to not be too bad to do this adjustment (probably actually easier than with the carbs) and after a while it looked pretty good, the cranking MAP is about 66kPa and there’s about 1kPa difference between them.
After putting the bike together again I was rewarded with a pretty nice idle. I put the seat and fairing back on and took it for a spin on our street. It at least pulled well enough that it was rideable below maybe 5000 RPM, so I went out on the super-wide road we have nearby and tested it.
It ran, but certainly did not run well. It didn’t want to rev past about 5000 because of lean misfires, and the front and back cylinder pairs registered quite different air/fuel ratios, by 10-20%. But from now on, it’s all about tuning. No more fabrication necessary!
(Sorry, no pics or movies, I will attempt to make a recording later.)
When the last post finished up, I had the link piece for cylinder 3 left. Since it’s basically just a longer version of the one for cylinder 1, it was pretty quick to get that done.
Here’s the link for cylinder #3. The spring isn’t attached at this point.
With that completed, it was time to put it all together and see if it worked.
Finally, the complete throttle linkage assembly. The #1 cylinder is in the upper right, then going clockwise is #2, #4, and #3.
Testing revealed a few problems. Most seriously, the axes of the two throttle bodies were not exactly collinear. Not surprisingly, the central bracket that holds them had warped slighly when its two halves were welded together. This is a problem because the cylinder 1/3 linkages are slid onto the shafts and then linked together with a dowel pin around which the central part (the part with the adjustment screws in the top center of the picture above) can rotate. If these aren’t aligned, the shaft will bind, which it did.
To overcome this, I measured which way the bracket was warped and with a couple precisely placed squeezes in the vise managed to get the two much more aligned. I also decided that rather than having the dowel pin link both cylinder #1 and #3’s pieces, it would suffice to only let the center piece rotate around the #1 piece and let #3 be free.
The idea behind linking all of them with one long pin would be to ensure that, when the linkage moved, the throttles actually rotate rather than get pushed to the side slightly. Since the motion is transmitted between the pieces with the offset tabs, when the linkage rotates, these tabs transmit torque through an off-center force on the linkage. Unless they are securely held, this force will cause them to bend rather than rotate and not faithfully transmit the rotation angle to the next part.
However, it turns out that the parts mounted onto the shafts of the throttle bodies are very secure, so this is not necessary. Without a pin running through the entire linkage, it can accept the small misalignment better, so I used a smaller dowel pin to line up the center part with the #1 link and let #3 be free. That got rid of the binding.
Another small issue was the mounting of the links for the front (#2 and #4) cylinders. Unlike the rear ones, these throttle bodies don’t have shafts I can slide the links on (see pictures in the initial design post) but have tabs. I machined slots in the links to slide onto these tabs and fix them with a screw, but when I actually tried to mount them it turned out I didn’t have the geometry quite right. The slots weren’t deep enough and their plane wasn’t quite right to get the centerline of the linkage to line up with the centerline of the throttle bodies. This doesn’t matter so much for the motion of the front cylinders (as long as the link is parallel to the shafts it will rotate ok) but it meant that the attachment for the front-back link was not the right distance from the rotational axis. This would upset the precisely designed linkage geometry, since the front-back link has to be attached the same distance from the rotational axes at both ends.
Once diagnosed, this defect was pretty easy to fix. Since it required removing more material, I put them back in the mill and manually took off another mm from the slots.
The third issue has to do with the return springs. When I fixed the initial backward design that had the springs going in the wrong direction, I did this by mirroring the parts. The parts can be attached to the shafts in two directions, differing by 180 degrees (since they are held with the spring pin) but only one of those put the links in the correct orientation. Unfortunately, for both the rear cylinder links, this means I have the choice of the springs having barely any tension at all in the closed position or winding them up 360 degrees more which puts quite a lot of tension on them.
I first tried the less tightly would configuration, but then the springs did not have enough tension to firmly push the throttles closed. To avoid play in the front-back link, it’s important that it always be under compression. With the springs set up like this, they didn’t have enough tension when the throttle was completely closed to avoid the link going into tension. This play meant the front throttles had to open a bit before the back ones did, which was no good.
Winding the springs up another 360 degrees makes them quite tight. It solves the problem with the play, but now the throttle requires substantially more force to operate than the stock carbs do. I think it will be tiring for my weak wrists to handle, but it works well enough to give it a try. If it turns out to be a problem, I’ll remake the parts with the spring holding tab shifted 180 degrees to take half a revolution of tension out of the springs.
With all that, I now have a linkage that works! It’s going to be practically impossible to get to the adjustment screws once it’s mounted, so I’m doing my best to synchronize the four throttles before mounting it on the bike. Then it’s time to put it all together and see if it’ll run.
I cranked in another full day on the throttle linkage, and today went a bit more smoothly than yesterday. It turned out to be pretty painless to mirror flip the tabs holding the springs. Those were one of the first things sketched out in the design and often when you change things that far upstream you get tons of lost references in Fusion. This wasn’t too bad, though.
The CAM also had to be tweaked, of course, but by late morning the mill was back up. All 6 operations on the cylinder 1 part went smoothly.
With the cylinder 1 part done, I continued with the front-back link. This is a pretty simple part, because it’s prismatic it can be made in two simple ops.
This is the front-to-back link getting is final chamfer. It was a bit tricky to hold. There is a short straight section in the middle that could be clamped.
The ends are reamed 4.02mm so it’ll spin freely on the 4mm dowel pins.
A test fit of the front-back link. The 8mm dowel pin that will hold the linkage for the cylinder 1-3 throttles in line (on the left) is still missing, but it looks like it’s going to work.
Since the last post, which outlined the throttle linkage design, I’ve done some of the fabrication. Because the pieces are fully three-dimensional and not prismatic, it takes a bit of care to make sure that they can actually be fabricated. If you design some thing that has no parallel edges, it’s very difficult to hold it when milling subsequent operations, so I had to do some little “design for manufacturability” tweaks.
The first ones to be made were the cylinder 2 & 4 links. That went pretty smoothly, although the tolerances when the pieces were moved around weren’t great. Won’t affect the result though.
This is the almost-completed cylinder 4 link. The clamp on the left clamps to the throttle shaft, the two tabs on the right will make up the adjustment mechanism.
Here are the cylinder 2 & 4 throttle bodies with the linkage attached. The black and yellow screw is the M5x0.5 adjustment screw reused from the original GPz linkage. The 4mm dowel pin on the top is where the front-back link will ride.
The opposite side of the cylinder 2/4 linkage. The tabs on the end clamp onto the throttle shafts.
The only problem is that I did not make the cutouts in the clamps that hold onto the throttle shafts deep enough, so the rotation axis of the linkage doesn’t quite line up with that of the throttles. This could be solved by recutting them a bit deeper, but I think it will be easier to simply grind off some material from the tabs on the throttle shafts.
So far so good. Now for the back cylinder parts. There are three of them, because both cylinders 1 & 3 need to be adjustable in relation to the link coming from the front cylinders. Setting up the CAM and fabricating these parts also went pretty smooth. Perhaps suspiciously so…
These are the cylinder 1, on the left, and center, right, parts. The threads on the center piece on the right are the adjustment tabs for the two cylinders.
These parts slide onto the throttle shafts and are locked in place with spring pins. Luckily the 7.98mm reamed hole fit snugly onto the shaft, but when I mounted the cylinder 1 link, I was dismayed to find…
The cylinder 1 link fit perfectly onto the shaft and held the return spring as designed. Unfortunately, it held the spring in the wrong direction… Rather than pulling the butterfly closed, it pulled it wide open!
Each throttle body has a return spring, which is held by tabs on the part. I had carefully measured these, but had somehow gotten them mirror flipped! This is kind of hard to keep track of on the CAD, because you always flip everything around, and the springs are on different sides of the different throttle bodies, meaning both directions are used. In hindsight, it’s perfectly obvious when moving the mechanism in Fusion that the spring holding tabs are backwards on this part.
Suffice to say I was frustrated. However, I realized that I could actually use this part. Cylinder 4 needs a spring holder on its shaft, but it’s on the outside so doesn’t need to connect to anything. And because it’s on the outside, its spring works in the correct direction for this part.
Relieved I could actually salvage something from this part, I went ahead and milled off everything outside the end of the throttle shaft. Then I quickly hand-typed some G-code to put some chamfers on the outside so it didn’t have such sharp edges. Doesn’t look half bad.
This is the cylinder 4 throttle body spring, held by the remnants of the incorrect part. It doesn’t even look that hacked up.
So that’s how things stand now. I need to correct the design for the cylinder 1 & 3 links, fabricate those, and finally the front-back link. It’s getting closer.
As mentioned in the last post, it’s time to start working on the throttle linkage for the NC30. The requirements for this setup is that it needs to be possible to adjust the throttles of cylinders 1,3, and 4 against the cylinder 2 throttle, which is attached to the throttle cable.
Since the GPz throttle bodies I’m using originally were designed for an inline-4, they have links between each other. These won’t work for my application, not only because of the need to link the front and back cylinders, but even the two throttle bodies that line up with each other are much farther apart on the NC30 (because the gears driving the camshafts are between the cylinders.)
It so happens that the back pair of throttle bodies have their “adjustment screws” on their now-inside sides. These are mounted on the 8mm butterfly shafts with spring pins and can be removed. This makes it possible to manufacture new parts and fix them to the shafts.
These are the existing adjustment screws on the throttle bodies, which mount to their 8mm shafts. By removing them, I can make new, longer parts and mount them on the existing shafts. The adjustment screw and the spring-loaded backer can be reused.
They also contain an M5 fine threaded adjustment screw and a spring-loaded backer that I can reuse for the new adjustment mechanisms.
The front pair unfortunately only have the mating adjustment “tab” ends of the throttle shafts facing each other. This makes it harder to mount something and ensure it is on the center axis.
The front throttles have the “tab” ends of their shafts facing inwards. These can not be removed and are a lot harder to mount something to securely.
There are a few tricky things about making this work. The first is that, because the throttle bodies are so far apart, a simple lever mechanism like the existing one won’t work well. The long lever arm means it’s much more likely to be pushed to the side than rotate. For that reason, I’m going with a linkage that will both have a side tab and a center pin. That way the links will always be on center. The potential drawback is that if the shafts aren’t lined up well, the mechanism may bind.
The second tricky thing is that the front and rear throttle bodies are turned 180 degrees relative to each other. This means the direction of rotation of the shaft when opening the throttle is opposite, meaning the linkage has to not just transmit the rotation but also reverse its direction. In practical terms, it means the pins pushing on the linkage have to be one on the top and one on the bottom of the respective axes. See the image below.
The linkage on top is not reversing and is completely linear; the two pins will rotate by the same amount. The linkage on the bottom, on the other hand, is reversing and nonlinear. Unless the two pins are infinitely far away, the rotation of two pins will not be exactly the same.
So what’s the problem? A linkage that translates rotation in the same sense can be made perfectly linear, such that a given rotation of the first axis translates to the exact same rotation of the second axis. A reversing linkage can not, it is always nonlinear to some extent. Since it is important that all four throttles open the exact same amount, at least when they are mostly closed, this is not good.
I was sufficiently puzzled by this effect that I had to pull out a notepad and do some trigonometry. The degree of nonlinearity depends on two dimensionless quantities, basically how the distance between the two axes of rotation and the length of the link relate to the length of the pins. It’s not that hard to write down the equation and solve it numerically. When you do this, you get something like this:
Plot of the linkage angle calculation. The x-axis is phi1, the angle of the first axis, and the y-axis phi2, the angle of the second axis, both in radians. The ideal is to have a 1:1 movement of the two axes. For a reversing linkage, this is only possible when the axes are infinitely far away. For any real linkage, phi2 will be a somewhat nonlinear replica of phi1. The cyan line in the center represents the geometry of the NC30 carburetor linkage, which is almost linear (compare to the ideal straight line in red) over most of the range. The throttle movement range is close to 90 degrees, so about 1.5 radians.
It turns out those guys at Honda apparently knew what they were doing! When I input the parameters of the NC30 carburetor linkage (which is that the distance between the axes is 6.17 times, and the length of the link is 6.00 times, the length of the pins, respectively) you get a linkage that is almost completely linear over the 90 degree motion.
Rather than reinvent the wheel, I decided to copy this geometry into my design. After a day of fiddling in Fusion 360 I now have something that clears any interference and looks like it works correctly. It will reuse the adjustment screws from the throttle bodies, use 8mm dowel pins to hold the links on the throttle shaft centerline, and 4mm dowel pins to attach the front-back link arm.
When you add all the joints in Fusion, you actually get something that looks like it’s going to work. I made a video of it moving:
Now I just have to fabricate it. It consists of 6 quite intricate parts, with reamed holes for the dowel pins that need to be lined up right, so it might be a bit of work. More on that later.
After fixing the fuel supply line, I finally got the fittings needed to hook up the fuel pump outlet line, too. With a 150-degree fitting after the fuel pump, it routed perfectly between the fuel pump housing and the battery.
Routing the return line was a little bit trickier. As shown in part 19 of this series, the fuel pressure regulator outlet faces downward, while the return inlet to the tank faces rearward. I had loosely assumed that it would be possible to make the hose do a 270-degree loop front, up, and then rearward, and come out reasonably aligned with the fitting on the tank. That turned out to be true, but the vacuum line to the FPR, which awkwardly faces straight out the side, runs right where the hose wants to go. It required a bit of jiggling while lowering the tank to not kink the vacuum hose, but it worked. It might be good to attach the hose somewhere to avoid it vibrating against the vacuum line, but for now it’ll work.
That completed all the required parts needed to start it the bike! First, however, I fixed a few small things.
The connection between the rear cylinder headers and the exhaust (the bent plate from the first exhaust post) had to be fixed. I straightened the plate in the vise, which wasn’t perfect but enough that it was possible to tighten. Then I replaced the rusted out flange nuts with brand new stainless ones, new gasket rings, and tightened it all up.
Another little fix was the heat shield between the rear headers and the rear brake cylinder. This had obviously been bent at some point, it was only attached at two out of the three bolt holes, and was also rubbing on the exhaust pipe. Some mild encouragement got this back in good enough shape to use all three bolts and not rub.
Now it was finally time to run it. I’m happy to say it started almost immediately when I was ready to try it, which was around 9pm on New Years Eve. I figured this would be the one day when no one would mind the noise…
The idle air valve seems to work as intended, and at full open it has plenty of air to idle the bike at around 1800 rpm even when cold. I’m going to call that a success!
The QuadraMap also works perfectly. Using the MAP logger on the Megasquirt tuning software you can sample the MAP value just as if you’d hooked it up to an oscilloscope, and you get four very nice dips from the four cylinders.
This is the manifold pressure log from the Microsquirt. The red is the manifold pressure, and the white lines represent the MAP sample windows when a measurement is taken. The graph has a MAP range of 50 – 100 kPa. For some reason, the #1 cylinder (the third dip on the plot) does not pull as low a vacuum as the others.
I had to play around with the injection settings a bit. This deserves a post all by itself, because it’s not very well documented exactly how the settings work, but I eventually got the bike running fairly well on “semi-sequential” injection. This means that the injection events are timed to the crankshaft position on the cylinders being injected, but that injection happens every crankshaft revolution. This means that one injection happens on the intake stroke and the other happens 360 degrees off, on the combustion stroke.
Since the front and rear injector pairs use share injector outputs, that’s the best we can do, but it fits quite nicely because the front cylinder pair is 360 degrees offset in the engine cycle. That means each of the two injections are appropriate for one of the two cylinders.
I captured a video of it starting and warming up. It’s not exactly riveting, but it shows how the Megasquirt tuning software works and gives you a chance to listen to a few minutes of V-4 idling…
As I mention in the video, the #1 cylinder had a tendency to not want to start and would die if the idle was lowered below something like 1500 rpm (1300 is factory idle). There are a few things that could cause this, basically bad compression or fouled plugs. It’s perhaps not coincidental that the #1 cylinder also can be seen to not have as good a vacuum on the MAP plot above.
I can’t check compression without taking the tank off, but I could get to all four plugs.
This is what the plugs looked like. Yeah, that’s pretty bad.
It turns out they all had a very dark appearance that strongly looks like carbon fouling. Since the bike has been running pig rich at least since I got it, that’s not too surprising. Luckily I had acquired four spares (the NC30 spark plugs are tiny 8mm ones that aren’t that easy to find) so I could swap them all out.
It seems that did the trick, it will now idle on all four cylinders, and I even got it to idle as low as 900 rpm without stalling. It’s not very happy doing that, but before one or even two cylinders would have dropped out and it would have stalled if I tried that.
I also had some problems with the communication to the wideband oxygen sensors dropping out. I was also seeing a lot of corrupted reads. Communication with the wideband controllers is handled by an Arduino that talks to them over I2C and sends the lambda values to the Microsquirt over the CAN bus. This was one of the first things done on this project and has worked solidly since then.
However, when I added the idle air control, I did a big cleanup of the Arduino code, and as part of that also changed the reading frequency from 10Hz to 100Hz, since that’s the frequency that the Microsquirt uses to read the value over the CAN bus. Maybe that was responsible?
Looking at the bus signals with the scope, it was clear that my 10k I2C pull-up resistors were on the large side. The size of those pullups are basically a compromise between signal integrity and power consumption but, with 10k resistors and the 100kHz I2C bus frequency, the signals barely had time to come up to full level. Changing those resistors to 3.6k made the signals look much better, and that’s still just over 1mA per line so there are no worries there.
I also noted that Alan To, the designer of the Sigma Lambda Controller I used, recommended a maximum polling frequency of 25Hz. I lowered it from 100Hz to 25Hz and with those two changes I have not seen any more hangs or corrupted reads.
I could play around more to tune the idle, but at this point it’s been proven to work so it seems to make more sense to take the throttle bodies off again and figure out how to set up the throttle linkages, which is the one remaining thing that needs to be done. Hopefully that won’t take too long to figure out.
One of the few remaining tasks was to hook up the fuel line from the fuel tank to the fuel pump housing. This is just a short length AN-6 hose, nothing complicated, but I was not entirely certain that it would actually be possible to get access to tighten the AN fittings.
After receiving the fitting a few days ago (I don’t know how many times I’ve ordered a fitting that “clearly” is the only remaining one I need for this project, but it’s getting a bit old) it seemed like a nice, easy vacation task for today.
After mounting the tank in its place, I marked the required length on the hose, cut it, pushed the fittings on and tried it… only to find that it was way (like 15mm) too short! Oh well, I must not have accounted for curvature or something…
After carefully cutting the hose to get it off the fittings, I cut a new length, 15mm longer, and tried it. Still about 5mm too short! Finally, I realized the problem: when the hose is pushed on the fittings it’s stretched radially, which shrinks it length-wise. By about 10mm on each side, apparently.
The third time I knew to add 5mm to the cut length of the old one, not the length as mounted, and got something that fit well. 104mm ended up being the magic number.
This is the fuel supply line going from the fuel pump housing visible on the right to the fitting on the tank (not visible in this view.)
I’m also pleased to note that it’s not going to be a problem to get access to the hose. The swiveling fitting on the tank means it’s possible to swing the tank upward and slightly rearward and get access to the fitting.
The swiveling fitting on the tank attachment lets the hose rotate down when the tank is lifted.
Here’s the hose, attached on both ends, with the tank lifted. There’s plenty of access. In fact, I think there’s more room than there was getting to the old fuel hose going to the carburetors.
The hose also maintains a nice slope between the tank and the fuel pump, so it doesn’t seem like there’s going to be a problem getting the air out.
I’m still waiting for one of the AN-4 fittings for making the hose going from the fuel pump to the filter do the 180-degree turn under the seat. Unfortunately it’s not going to get here before the new year, so that’ll have to wait a bit.
In the last post, I outlined the change to swap out the leaky clamp joints on the exhaust header pipes for double slip joints and got as far as cutting the old clamps off the exhaust.
I’ve watched a couple of videos on fabricating exhaust systems and they all make the point that the fit up needs to be exact. Any gaps in the joints to be welded drastically increases the risk of blow-through as well as makes the part warp more. Then they make a bunch of pieces, fast forward, and weld everything up. Well, it turns out that fast-forwarded part takes a long time. Everything needs to be aligned precisely for the slip joints to work, both the position and the angle needs to be correct. This is actually a tricky problem, because changing the position, by rotating a joint, also changes the angle, and vice versa.
Suffice to say, I spent a whole day tweaking two short lengths of header pipe, aligning joints by grinding on the surface until it was right, then making sure there’s even contact all around. People seem to use stationary belt sanders for this, but even our tiny belt sander from Home Depot actually was helpful in getting a perfectly flat surface.
In the end I had something that looked like it would work, so I tacked it all together for a trial fit. To get the angle right to cylinder 2 (front left) I had to cut the collector end and rearrange the bend a bit. Before, both the pipes coming out of the collector going forward were angled downwards. The right one wasn’t too bad, but the only way to avoid a sharp kink in the left one was to cut the bend near the collector and rotate it to a more horizontal angle.
The pieces have been tack welded together and are ready for a trial fit. Note the extra joint that was cut in the right pipe coming out of the collector. This made it possible to rotate the bend and aim it less downwards. The outer parts of the double-slips have not yet been attached to the headers.
The pieces successfully fit together on the bike, so that was nice.
As you can see, the headers aren’t perfectly parallel with each other under the bike. It’s not visible, in any case.
Since everything fit, there was no reason to postpone welding it all together. Based on some tips on the Burns Stainless web page, I used 0.035″ 308 filler rod and about 37A. I had made some practice welds on a piece of spare tubing I got for that purpose, and that went reasonably well.
Welding stainless is kind of cool, because of its low thermal conductivity it doesn’t immediately get all hot like aluminum does, and you can use low current and get a small puddle. It’s important to not use too much heat when welding stainless, it will darken and discolor and this also makes it more prone to oxidation later. If you manage to use the minimum amount of heat, the weld will have a nice, golden color.
Just like when I welded on the oxygen sensor bungs, I had to set up an Argon purge to get all the oxygen out from the inside of the pipes. If you don’t do this, you get huge balls of oxides protruding out from the inside of the weld, and the weld will be weak. This was kind of a pain, because the whole exhaust system is kind of large and unwieldy and has a tendency to roll around on you. The situation isn’t improved when you then add a ground clamp and the purge gas hose.
To purge the air out of the pipe, i attached the Argon hose to the large end and sealed up all four end with aluminum foil. To weld the headers, I simply removed one of the foil caps over the slip joints and slid the header pipe on. This was actually a lot easier because I could rotate the header in the slip joint as I welded around it.
It’s a bit stressful to hear the hiss of your precious Argon being continuously vented into the pipes as you struggle to find a good position, have to re-grind your tungsten, have to re-adjust your glasses, etc, etc. In the end I used about half of the little 60-bottle I got for doing purges. That’s about $50 in Argon right there.
It took a while, but there were no welding disasters. The worst part was when I started welding one of the slip joints to the collector and realized after having welded maybe an inch that the tacks had given out and the joint had opened up about a millimeter. I considered grinding off the weld and starting over, but I managed to clamp the joint back together and re-tack it. That’s probably not ideal and it’s possible this deformed the joint enough to matter, but it seems OK for now.
One of the headers after welding.
And the other. The welds aren’t stellar, but I’m pretty happy with how they came out.
And finally the collector end.
After letting everything cool down I was anxious to see if it would still fit together or if something had shifted. There was definitely more resistance now, especially in getting the slip joints to slide into each other. Any misalignment definitely makes them want to bind, but by putting everything in place first and then gradually tightening all the nuts, it came together. When the engine is started I guess we’ll really see whether there are any leaks, but by blowing into the muffler end of the exhaust system I could test that, after tightening all the nuts, the only place I could note air coming out was in the slip joints.
Now I’m just waiting for one more fitting and then I should be able to hook up the tank to the fuel pump and see if it starts!
