Unlike previous posts, there will be no gasoline splashing in this episode. This one is all about CNC goodness.

There is one final part to fitting the throttle bodies that I’ve not touched upon yet. The stock carbs mount to the intakes using short rubber couplers. The throttle bodies have a slightly larger diameter intake than the carbs, so they don’t fit. Something had to be modified.

Initially I had just considered machining down the intake ends of the throttle bodies to fit the rubbers, but that would have left very little material. Instead, I figured I’d go half-way and take out half from the throttle bodies and half from the rubber couplers.

This posed somewhat of a dilemma. I’m used to milling aluminum but I’ve never tried milling rubber. It’s relatively hard rubber, and after reading various people’s reports on the internet, I thought there was a decent chance of it working if I just cooled the rubber down so it got hard. At least it seemed worth a try.

Another problem was how to mount both of the pieces. I figured I could hold the throttle body in the vise by make a set of soft jaws and clamp the airbox end of the throttle bodies in them.

This is very cool, actually. I had already modeled the shape of the throttle bodies quite accurately to make sure everything would fit, and then you basically take a square block and ask Fusion to cut away the shape of the throttle body. What you are left with is a “mold” that should hold the throttle body quite securely.

Once I had the basic shape, I had to tweak it a bit to make sure that it was actually possible to make it in the mill. In some places I had to use quite small ball endmills to cut some contours, and there were some interference problems since you can’t have those small endmills stick out very far. But eventually I worked it out.

The finished set of soft jaws that will hold the throttle body for machining.

I didn’t really care about the finish, since this was strictly a functional part, you can clearly see the traces of the endmills. But the throttle body fit perfectly and were held very securely.

The throttle body mounted in the soft jaws, with the intake end aligned vertically for machining.

The part that needs to be machined down is basically the top end down to where the “wiggles” end.

To mount the intake rubbers, I used the same technique. I had modeled them as accurately as I could measure in Fusion, and then made a “mold” with the body of the intake rubber removed.

This is the fixture for the intake rubbers, matching the shape of the intake on the engine. The notch prevents the rubbers from sliding off.

The rubbers were also a perfect fit.

Here’s one of the intake rubbers snapped into the fixture. At this point, it can be clamped with the stock clamps.

I had to decide on what shape to cut the machine the two pieces into. I wanted some notches, like there are in the stock ones that prevent them from moving, but I didn’t really have enough material to make them as large. Instead I just made two small “sawtooth” rings, each half a mm tall. It’s not much, but hopefully it’ll be enough.

The first thing to make sure was that machining the rubber pieces was actually going to work. There’s no point in destroying the throttle bodies unless that will work, so that’s what I started with.

I zeroed the mill position on the fixture with the rubber and a jar of ethanol in the freezer. Once I was all set up, I took the stuff out of the freezer, snapped the rubber into the fixture and clamped it, hooked the chilled ethanol up to the mister, and pressed go. The actual cutting didn’t even take a minute, not counting the tool change since I had to switch to the double-angle chamfer mill to cut the radial grooves.

Since this was a time-sensitive operation, I didn’t pause to take any pictures, but it worked surprisingly well. The rubber cut without a problem and left a very smooth surface. Success!

Proceeding with the throttle bodies, I clamped them into the soft jaws, zeroed the mill on the intake bore, and pressed go. This was also a very short program, and while it’s always nerve-racking to see the mill cut into these pieces, it also worked very well.

The mill is zeroed on the throttle body bore and we are ready to go. Since I want to leave positive ridges on the part, i can’t cut with a normal endmill, so I used the key seat cutter instead.

The bore has been machined to size, now the chamfer cutter is making the triangular ridges that will hook into the rubber piece.

The end result is pretty good. The “grooves” that hold the rubber in place are definitely much smaller than the stock ones, but the rubber piece snaps fairly securely into place.

The intake rubber in place on the throttle body.

So that’s one thing less to worry about. We are now at the point that it should be possible to physically mount the entire throttle body assembly and the airbox onto the bike!

So what’s left? I need to add the harnessing for the injectors and the throttle position sensor, which shouldn’t take long. There are two large, unsolved problems remaining:

• First, making a replacement for the fuel tap that also has a fuel return line, and routing the fuel hoses so the fuel pump can prime properly.
• Second, figuring out how to connect the throttle linkage between the four throttle bodies. They are meant to be lined up in a line right next to each other, but now the two cylinders are much further apart, and some sort of linkage needs to go between the front and rear cylinder pair.

I believe I’m seeing the end of the tunnel!

As mentioned in the last post, now that the injectors are mounted in the fuel rail and there are no leaks, it’s possible to measure the injector performance.

A fuel injector is essentially a very fast-acting and well calibrated valve. It has an electromagnetically actuated plunger that opens when current flows through the solenoid, letting fuel flow at a precisely calibrated rate, and closes when the current is turned off.

At least that’s the idea. In practice, of course, it’s not that simple. It’s a mechanical device and hence is bound by the laws of physics. When a voltage is applied to the injector, it takes a finite amount of time for current to build up and for the magnetic field to start pulling the plunger. The plunger then has to travel a finite distance during which the valve is only partially open. Once fully open, the flow rate through the injector depends on the fuel pressure and on the manufacturing tolerances of the orifice that lets the fluid through. Then when the voltage is turned off, the same thing happens in reverse; it takes a finite (but smaller) time for the plunger to return to the closed position.

The end result is that while an ideal injector delivers a fuel mass m = r * t, where r is the mass flow rate of the injector, in the real world there is a “dead time” such that if you send an electric pulse of time t, the valve is actually open for a smaller time t – dt, where dt is the “dead time” of the injector. On top of that linear model, there are additional effects for short opening times due to the fact that the injector then is partially open for a significant fraction of the pulse, but the simplest useful description of a fuel injector’s performance is

m = r * (t – dt)

Oh, but there’s more. Since the dead time depends on how hard the electromagnet is pulling the plunger, it depends on the voltage applied to the injector. This means that at very low voltages, like when you’re trying to start the engine, you need to apply longer pulses to get the same amount of fuel.

So there’s a lot of information you need to capture in order to know how a fuel injector performs. Remember that any mismatch between how much fuel the engine controller thinks that it’s delivering and how much is actually delivered will show up as a deviation in the fuel/air ratio. If you don’t have a good description of the injector, that fact will feed back into your assumptions for how much air you think the engine is pumping and will make the tuning harder and less predictable. And if the injector to the different cylinders don’t deliver the same amount of fuel, it’ll be impossible to get all the cylinders to run at the correct air/fuel ratio.

If you have an unlimited budget, you can buy injectors from places like Injector Dynamics, that measure and match sets of injectors that perform very similarly. I don’t have an unlimited budget (although sometimes I wonder if that’s true in practice) and even if I did, they don’t sell them for 100cc cylinders that make 15hp. 1000cc cylinders that make 150hp is more like their target market…

When I tried to find injectors, I browsed the awesome collection of injector data collected by Stan Weiss and looked for ones that would have a suitable flow rate. The fuel flow rate needed is only determined by two things: the maximum amount of power the cylinder makes, and the air/fuel ratio it runs at. Since sportbikes have very high horsepower for a given displacement (since they are usually very high revving engines) they actually need higher flow rates than you might think given how small the engine is.

Since the NC30 makes ballpark 60hp, i.e. 15hp per cylinder, the required injector size comes out to be about 85-110 cc/min, depending a bit on the air/fuel ratio and the margin you want. Since gasoline has a density of about 0.72kg/l, this corresponds to about 60-80 g/min of fuel.

Alternatively, you can do the a simple calculation without any assumptions at all: At an RPM of 12,000 (power peak), a 100cc cylinder pumps 6000*0.1 = 600 l of air per minute. Air has a density of 1.2 g/l, so that corresponds to 720g/min. If we want to run at a lambda of 0.85 at full throttle, that corresponds to an air/fuel mass ratio of 14.7*0.85, so that comes out to be 720g/min/(14.7*0.85) = 57g/min. However, you don’t want to run the injectors at full duty cycle. If we demand a maximum duty cycle of 80%, we need injectors that flow 72g/min of fuel. (It’ll actually be a bit less than that because the volumetric efficiency is less than 100%, so each engine cycle doesn’t pump the full 100cc of air.)

Scanning through that very large table there are a few injectors with flow rates in the correct ball park, and after also searching for the corresponding model numbers on ebay, I hit upon the Bosh 0-280-155-965 (used in a 1.2L Opel engine) which has a posted flow rate of 68g/min and ordered a used set of those. That was a year ago, and it’s not until now that I can verify whether those injectors actually are going to do what they are supposed to.

To test this, I took the assembled fuel rail and throttle bodies, connected up the fuel pump and filter, and pumped gas from a gas can. The injector is hooked up to the Microsquirt, and in test mode you can tell it to fire the injector for a specified number of times for a specified length. Then I measured the resulting mass of fuel on my 0.01g resolution scale.

The injector test setup. Fuel from a set number of injections is meterd into a glass jar, and the mass of fuel injected measured with my very sensitive 0.01g scale.

I have to point out that this is a somewhat sketchy operation. The fuel injectors don’t just dribble liquid into the jar, they atomize the fuel and a significant part of it gets into the air. I had the garage door open, the fire extinguisher standing by, and wore a respirator the whole time since the garage positively stank of gasoline.

To get enough mass of fuel to reliably measure it, I pulsed the the injectors anything from 250 times to 3000 times (for short injection times.) This resulted in 5-10g of fuel in the jar, easily measured on the scale. Being a good experimental physicist, all measurements were repeated 3 times to get an estimate of the variance.

This plot shows effective mass flow rate for each of the injectors after accounting for dead time. My best estimate of dead time is 0.95ms.

The result is shown in the plot above. This is a slightly processed plot, in that the quantity plotted is (fuel mass)/(injector pulse time – dead time), which allows you to estimate the dead time. If your estimate of dead time is correct, the curves will be mostly horizontal. Now, they aren’t exactly horizontal, since the injectors aren’t perfectly linear. My estimated dead time is 0.95ms, but at times less than 2ms, the curves drop like a stone. This is an indication that the injector doesn’t have time to open completely in that time.

For times longer than 2ms, though, the behavior looks pretty good except for the cylinder 2 injector in green. I think this is an experimental artifact. In the interest of taking the measurements as fast as possible, I initially only waited a very small time between each injection, like maybe a 7.5ms injection pulse every 10ms. This seems to overestimate the flow rate, maybe because the fluid flow through the injector never really has time to come to a complete stop. For the other cylinders I decided to wait at least as long as the injection pulse, and never less than 20ms, between each injection. Those curves look more linear, so I’m going to redo the cylinder 2 measurements, too.

The other thing to note is that the flow rate is more like 75g/min, not 68. However, the 68 is specified for 3bar fuel pressure, and my fuel pressure regulator apparently is set more like 3.5bar. The flow rate through an injector goes as the square root of the pressure, so taking 68*sqrt(3.5/3) gives you 73g/min. Close enough for me.

Another thing that stands out is that the cylinder 3 injector, in red, flows less than the three others by about 1.5%. This is a bit suboptimal, because it means one cylinder will run leaner than the other. (If two injectors were lower, I could put those on the same cylinder bank and tune those to be on a bit longer, but the Microsquirt only has 2 injector outputs so they have to be wired two and two.) However, a lambda error of 1.5% isn’t bad, I don’t think that will dominate the air/fuel ratio error.

I still have to measure how the dead time depends on voltage. That means repeating this exercise for a bunch of different voltages (this was at a voltage of 12.9V as read by the Microsquirt), but I think I’m just going to assume that all four behave similarly in this aspect and just measure one of them. But that’ll be the topic of a future post.

With the fuel rail parts completed, it was time to assemble it and see if it worked.

Step one was to replace the seals on the injectors. I’d ordered a replacement seal kit from “Mr Injector” a while ago but didn’t want to replace the existing O-rings until I was mostly done with all the test fitting. I can’t make that excuse any more.

Before replacing the O-rings, I actually hooked them up to the “test port” that I used for leak checking and flushed some carb cleaner through each of them. Since there’s some volume inside the part, I could undo the air hose, fill the injector holder with carb cleaner, reattach the air hose and then pulse the injectors to spray the cleaner through. A couple of them did not particularly want to do anything for short pulses, but after blowing a couple of loads through them they all seemed to behave better. I hope I’m not going to have sticky injectors after going through all this…

Replacing the seals was mostly painless except to get the fuel strainer cups out I had to screw in a sheet metal screw and pull it against a pile of washers. According to the instructions, it should work to just pull them out by hand, but mine did not budge when I tried that.

Here are the injectors, complete with new Viton O-rings, filters and retainers.

Step two was to hook up the compressed air hose that I used to leak-test the O-ring ports to the entire fuel rail and see what happened. One thing I noticed was that I have a tad too much space between the fuel rail and the throttle bodies. It looks good when first put together, but when you pressurize the fuel rail, the injectors are pushed downward and actually move far enough that the “tab” (on the top right in the picture above) almost ends up below the guides I machined into the fuel rail. The space between the throttle body and the fuel rail is what is recommended in the data sheet, but they also assume that the injectors will be held in with little clips. It doesn’t leak, but it’s far enough out that you might be able to rotate the injectors which isn’t supposed to be possible.

In hindsight, I could have machined slots for those clips into the fuel rail, and then the injectors could have been locked into the fuel rail. That would have precluded them moving, and it would also have been easier to test them than now when I have to tie them up with wire to prevent them from shooting out when the rail is pressurized. Oh well… I’ll think of that on my next EFI conversion.

The other thing that was immediately obvious when the fuel rail was pressurized was a hissing sound. A leak… now what could be leaking, the injector O-rings, the AN-port O-rings, or my welds?

If you guessed the welds, you were correct. There was a corner in the weld to the square tubing that looked a bit low and incompletely fused and, sure enough, that’s where the air was coming from. After dismantling everything again, I added a few quick beads there and after reassembling the hissing sound was gone.

Step three, I now decided, was to go ahead and attempt to pump gasoline through and see what would happen. Alongside all of this I’ve been attempting to find a good spot for the fuel pump. Because there is just no space near the tank, I had settled on the tool kit pocket on the right side of the tail. I knew this wasn’t optimal, but it was just very difficult to fit the inline fuel pump, with inlet and outlet on opposite sides, in the battery area; there’s a fair amount of space there but it’s not long enough in any of the dimensions to easily fit the pump.

The problem with putting it under the tail is that the fuel line from the tank has to do a 180 (because the fuel tap outlet is in the forward direction) then go back to the pump, and then make another 180 before getting into the pump. This is uneven enough that there’s air trapped there, and it turns out that while the pump can get enough suction to suck the air through and fill the line with fuel when there is no back pressure, when it attempts to do that against the fuel pressure regulator it doesn’t work. It just sits there and spins air. This will require a wholesale redesign, but I’m leaving that for another post.

To make progress now, I manually primed the pump by disconnecting the hose going to the fuel rail and running the pump until it had filled it with fuel. Then I reattached it and restarted the fuel pump, and now  it did build pressure.

I was now greeted with numerous small puddles of gasoline forming on the table. Apparently there were more pinholes in the welds; too small to hear air coming out but apparently not too small for fuel to go through. It was actually pretty cool, in one area there were numerous super-tiny streams of fuel escaping.

See the three tiny streams of fuel coming from the lower corner of the injector block?

On the other rail, there were no streams of fuel but gas was seeping out and gradually covering the top surface. It was hard to see where it came from, but I was pretty sure I knew which weld was the offending one.

Ok, time to dismantle again, although now everything had gasoline in it, so that was a messy ordeal. After wiping everything up as best as possible, flushing alcohol through the rails and blowing it out with compressed air, I left the garage overnight to air out and to deal with my gasoline-induced headache…

The next day I ground the welds down, cleaned them as best I could, and gave them another pass. This time I tried to make them on the hot side to make sure the welds penetrated.

I wasn’t about to try to reassemble the whole thing again, so now I just put the injectors and the fuel pressure regulator back, hooked up the compressed air line, and dunked the whole thing in a bucket of water.

Now, with the whole shebang hooked up and underwater, I’m pretty sure there aren’t any leaks. Except maybe from one of the injectors…

It’s apparent from this exercise that you have to tighten the steel AN flared fittings quite a lot to get them to seal. They warn to not tighten aluminum ones too much, for obvious reasons, but there the sealing surfaces are softer and you probably also don’t need to. In the end, I could pressurize the whole thing to 3.5bar, when the pressure regulator starts to open, and it was all tight except occasional bubbles coming from the end of one of the injectors. Maybe I need to squirt some more carb cleaner through them…

While I’m thinking about how to redo the fuel tank plumbing, I’m going to hook up the injectors again and measure their flow rates and dead times (how long it takes the injector to open after the current is turned on). While I know what the flow rate is supposed to be, knowing these quantities accurately is necessary to be able to calibrate the fuel delivery in the Microsquirt.

Doing these measurements should be a simple albeit perhaps tedious exercise. I’ll squirt a number of pulses of various lengths into a container and weigh how much fuel is delivered. It’ll be just like a physics lab, complete with plots, except with flammable liquids and noxious fumes! Stay tuned

In the last post, the first half of the fuel rail, for cylinders 2 & 4, was made. Now it’s time to complete the other half. This will hold the injectors for cylinders 1 & 3 (the rear ones), plus the fuel pressure regulator.

A rendering of the fuel rail for cylinders 1 & 3, with the fuel pressure regulator on the right.

I fiddled a lot with the fuel pressure regulator to get it to fit. As you can see in the rendering above, I settled on mounting it to the rear, with the fuel return going basically straight down and the manifold pressure port going out to the right. This fits, but forces the fuel return line to go down between the cylinders and then probably come back up between the two rubber boots. There’s just no way the hose can make the >90 degree turn needed to go immediately toward the rear. The manifold pressure line is just a small silicone hose that’s very flexible, so it should be able to handle the quite small radius turn it will need to do to avoid running into the tank.

Unlike the fuel rail on the other side, which only consisted of two pieces welded to the square tubing connecting them, the need to mount the pressure regulator (FPR) further down than the fuel rail itself necessitated making that side of 3 separate pieces: one for the injector mount, one for the FPR mount, and one little piece connecting the two, with holes crossing at 90 degrees for the fuel connection.

I’ve already welded the injector mounts to the square tubing in the middle here, so now it’s time to weld the two pieces on the right.

Given my welding skills, I wanted to avoid welding inside corners, because I know from past experience that those rarely come out pretty. For that reason, I made sure to make the “seams” between the pieces on sections that were straight. It also seemed like a lot of mass with those big pieces hanging out on the right side, so I had some fun designing the 90-degree piece to be as light as possible. It’s basically just a skin around the two holes, with a thin layer of material in the center. It looks pretty cool, but the CAM work is certainly not perfect.

Here’s the welded piece. Most of the welds came out OK, but the weld on the left in this picture was close enough to the corner that I couldn’t prevent the arc from occasionally wandering off to the other side.

Since welding the other rail, I went through and tightened all the Argon gas connections on the welder, which actually seems to have helped. There seems to be less black junk in the welds now, but it’s certainly not anywhere near perfect yet.

The back side was flatter, which is always easier to weld. These don’t look too bad.

The final result, with injectors and fuel pressure regulator mounted.

A closeup of the fuel pressure regulator and the injector for cylinder 3. It’s tight, but everything should fit.

The end result came out pretty nice. Apart from one little interference between the crankcase vent connector on the airbox (the black round thing just above the fuel rail in the rendering above) and the square tubing in the middle, it all fits. When you tighten the screws holding the rail, it’s very solidly held.

I also mounted the AN O-ring ports and made up the short piece of Aeroquip hose that connects the two fuel rails.

This is the short piece of hose that connects the two fuel rails to each other. It fits pretty much exactly like the CAD drawing says it would. It does touch the screw holding the throttle position sensor, but I don’t think this short piece will vibrate very much.

So that’s pretty much the fuel delivery hardware. It would now be possible to mount all the injectors, hook up the fuel pump, and test that there are no leaks. Before I do that, though, I want to clean the injectors and replace their O-rings and filters, so that’s what’s going to be next.

It’s been a while since the NC23 figured in the list of projects. It hasn’t been ridden much over the past half a year, because it would leak gas from somewhere around the carburetors, and I was busy with the NC30 project. However, finally the state of Hawaii ensured the NC23 got some love, since it was time to get its annual safety inspection.

After removing the fairings and the airbox, it appeared the gas was coming from the pilot screw on the #1 carb (although it was hard to tell since they were so dirty…) Upon removal, the tiny O-ring that seals the pilot screw was found to be in sad shape. The O-rings for the float bowls were also hard and crusted with varnish. Since I had them out of the bike anyway, I figured it was time for a carb overhaul.

There are still O-ring kits for these carbs available from Honda. One set (16010-MN4-305, $20 each) has the internal O-rings, and another set (16040-MM5-601, $11 each) has the O-rings for the pipes that connects the fuel supply and vent lines between the carbs. According to the parts list, you need 4 of the former and 6 of the latter, making this a $140 project from O-rings alone…

Unfortunately I didn’t discover that there was a second set until I got the first one, so that meant waiting another week-ish for the second. Then I discovered that the second kit doesn’t actually contain all the O-rings the parts manual says it should, so I still didn’t have all the parts needed. Quite frustrating, to say the least.

At this point I found a company called LiteTek in Thailand that seems to specialize in manufacturing O-rings for motorcycles. They had a kit for the NC23, comprising both the Honda kits and the missing O-rings and new rubber caps for the choke plungers which Honda doesn’t have at all. Their O-rings are also made from Viton, rather than butyl rubber like the factory ones, and Viton stands up to fuel with ethanol in it much better than butyl rubber. On top of that, their kit was $37, including shipping. What a no-brainer!

While waiting for the parts, I had been cleaning everything out. I remember from cleaning the NC30 carbs how hard it was to get all the varnish off, so this time I looked a bit harder for a solution. Quite a few people on the web reported much luck with Berryman’s Chem-Dip, and luckily the local auto parts store had a can of it. (Chemicals like these cleaners aren’t something Amazon wants to ship over here, so your best bet is to find something locally.)

This stuff is amazing! After stripping off all the parts I could, I soaked the carbs in the cleaner for an hour or so. Check this out:

This is a before-and-after image of how the carbs came out after spending an hour in the Chem-Dip can. Amazing!

Similarly, the various jets and needles that were crusted with stuff came out looking if not quite like new, much, much better than before.

Here are the needle jets and the pilot screws before and after cleaning.

Once I got the set of Thai O-rings (they arrived faster than many mainland USPS shipments) it was assembly time. The O-rings looked good. I had already broken the packaging on the set of internal O-rings, but I took those out and replaced them with the Viton ones in the interest of longevity.

A selected set of carb O-rings. The green ones are the Viton seals from LiteTek, the black are the factory ones. (There should be no question which ones are the old ones.)

Compared to the long wait for parts and all the cleaning, the actual assembly just took an afternoon.

The assembled set of cleaned and overhauled carbs.

Here I ran into the first setback: when I hooked them up to the fuel supply, they leaked! There was gas seeping from the float bowls on all 4 of them. Clearly the Thai O-rings had some issues. Just to make sure it was the O-rings, I replaced one of them with the factory one. This did not leak.

After contacting LiteTek, they investigated and admitted that there was a manufacturing defect with the float bowl O-rings for these carbs. They refunded me and promised to send me a new set. It was nice that they owned up to the problem so quickly. However, I didn’t want to wait an unspecified time for them to manufacture a new batch, so I decided to go ahead and use the factory O-rings for the float bowls since I couldn’t return those anyway. I filled the carbs with gas and left them overnight just to be sure. No leaks.

Once I re-mounted the carbs, added fresh grease to the throttle and choke cables, and cursed the appropriate number of times about getting the carbs back on the rubber mounts and hooking up all the hoses, it basically started right up. Once I checked that it was running OK, I headed for the now-overdue safety inspection. Mission accomplished!

Now the only question is how I’m going to get the NC30 inspected…