I just won the auction (on opening bid, no less) for a pair of throttle bodies supposedly from a 2008 Kawasaki KFX450R. Now, from what I can find, the KFX450R is a 500cc single, so it would seem unusual to have two throttle bodies, but then Kawasaki seems to have figured out a thing or two. Who am I to judge?

In short, this should be essentially half of the throttle body assembly from the 636. Rather than having to modify what I have, this is ready to go. I will need to fabricate an intake of some sort, but I have a couple of ideas in mind.

In the mean time, I’m still going to see if I can get a ride out of Buzz on New Years Day. That seems a great way to start the year….

There have been a few developments since my last update. I guess I’ll cram them all in one post.

I acquired a small oscilloscope. The jury is still slightly out as to whether it’s appropriate to the application, but so far, so good…

There are some limitations to the instrument. The LCD display is not particularly high resolution, so sometimes it’s hard to see the whole story at once. Likewise, there no external trigger input, so the missing pulse nature of the trigger wheel signal is difficult to sync. One has to adjust the sweep rate low enough and manage to capture an entire revolution with the memory function to analyze the finer points and the LCD resolution limits how much you can see at that level.

Even so, I was able to see a few things. First, the slightest eccentricity in the trigger wheel mounting is very easy to detect and could indeed be a contributing factor to my EDIS ignition problems. The VR sensor input ranged from 5V p-p down to 0.5V p-p. It’s difficult with the resolution of the display to configure the scope to show both ranges. That alone could affect the EDIS module enough to cause some problems.

Working on a related theory, I set the sweep rate very slow and watched the average amplitude of the VR sensor signal. I arbitrarily decided that low RPM was “normal”. As the RPM went up, the amplitude went up to about 5V, then began tapering off, fairly quickly, to “normal” levels. Since the engine would not rev beyond about 5500 RPM, I am supposing that the stock VR sensor may not have the frequency response needed for this application.

Putting a little math to it, with the one pulse per rev stock trigger, the signal at redline would be about 9500 Hz, or 9.5KHz. With the 36-1 trigger wheel, the signal would be more like 332KHz to 342KHz, depending on how you count it. Keeping in mind that there are radio signals at lower frequencies than that, it’s not surprising that maybe the sensor is not up to the task.

It is under this working theory that I decided to, at least temporarily, return the ignition to Yamaha stock. This means I lose my cute purple wires and those “wrath of Thor” lightning bolt sparks, but it could return me to fighting a single front war.

It turns out that it may be worth it, for now I can rev close enough to redline to cause the involuntary retraction of certain parts of my anatomy, and I’m not even riding at the time. There is some detectable miss in the rhythm, especially noticeable at low RPM. I may be suffering the famous Yamaha weak ignition problems, so I will still be searching for a long term solution, perhaps for Dyna coils or maybe driving the coils directly from MegaSquirt, maybe even with big enough drivers to run the EDIS coil. For now, however, perhaps I can march on with fueling issues and determine if I can even operate Buzz on these (quite likely too large) throttle bodies. Down that road are a few possibilities, from cutting my throttle body assembly in half and fabricating a manifold, up to and including restoring Buzz to stock and putting MegaSquirt on something else, probably the VW trike. My wife would like to see me get that thing on the road regardless.

I also shoehorned in an oil and filter change, which was probably overdue even when I parked Buzz last May.

Today, before I posted this update, I was explaining the issues to a friend when it suddenly occurred to me that the EDIS module itself may have a rev limiting “feature”. This could make sense because my module and coils came off a Tempo or Escort or some similarly pedestrian vehicle, RPM-wise. Web research, including in MSEFI forums, indicates that the jury is still out. Several people describe similar rev limitations, generally around 6000RPM, but some testing has shown 8000-9000 RPM as a limit.

I’m wondering if the VR sensor could *still* be the issue. Once one exceeds the sensor’s ratings, the module gets bad or no info about the crankshaft position and speed, so it stops firing. The engine RPM drops, the signal gets usable again and there it goes. Consequently, the Ford sensor may not work much better. Furthermore, the limit may indeed be a feature of the EDIS module itself, and that may explain why others *not* using Yamaha sensors on their EDIS modules are having similar problems.

One possible solution may be an optical sensor that may not would not be so frequency dependent. I will keep looking, but for now, I really want to get on with turning the fuel and maybe getting Buzz on the road, so it’s stock sparks for me.

All this, and I still haven’t properly investigated the throttle-failing-to-close-completely issue…

Well, the headlight issue was indeed a simple problem.

When I removed the stock ignition coils, I neglected to secure two ground wires that had been held by one of the ignition coil bolts.

One little bolt and the lights are back on!

Now I just need to fix everything else. I’m looking for the ground wire that will fix the throttle body issue… Wish me luck.

Ok, maybe it’s not that bad. One step forward, 1.273 steps back…

As suggested by a XJ-lister with MegaSquirt experience, I set about reinstalling the butterflies for the subthrottles, the intent being to set them to restrict the air intake. My theory, verified by his experience, seems to indicate that my throttle bodies are probably way too big for this engine. The butterflies restricting the overall air flow into the engine may serve to widen the throttle range a bit. I fear the only real option will be to change them out for smaller unit or perhaps build a manifold to let me use only two of them. Much remains to be discovered, but I need to start somewhere.

I removed the air filters and was able to reinstall two of the butterflies without difficulty. Of course, the first one I tried on the “dark” side of the bike (the side away from the lighting in my workspace) crossthreaded and twisted in half in the removal attempt. After warming the area with some rather colorful language, I removed the throttle bodies, drilled out the broken screw, installed the two remaining butterflies, LockTite’d all of them and finally secured the assembly at the recommended 9 degrees from open.

I reinstalled the throttle bodies and started the bike up, only to find that now the throttle doesn’t close against the stop. I can gently force it against the stop, but it springs back open, reving high the whole time. The only thing that comes to mind is that maybe when I put thread lock on the screws, a drop fell in on one of the throttle butterflies, dried there and is now holding them open. I suspect it will be #3 because it seems like it’s always #3.

Well, with the bike trying to idle at it’s maximum attainable RPM of about 3800, the air restriction does not appear to have been too much. The advance is still crazy up there and it surges like it’s hitting a rev limiter, but at this point, I have done no adjustments to the ECU, just added the subthrottle butterflies back into the intake tract. That, and broken it such that the throttle is cracked open pretty much permanently.

Just this moment, I’m far too frustrated to safely pull the throttle bodies back out to troubleshoot the stuck open issue, so I will call it a night.

As suggested by an XJ lister, I checked over my meter and there is indeed something flaky about it. I turned it on and with nothing connected, it showed 38V. One sound tap and it settled down. I found that I could spin the selector switch a few times and make it screw up a bit, but after running through the ranges, it settled down. This meter has bounced around in tool boxes, floorboards, saddle bags and my driveway for about 10 years now, so I think it’s due a little crotchitiness, especially since it’s almost always sitting on the 20V range

In any case, with it settled down and reliably measuring 12v and 5v from a known good PC power supply and 9V from a fresh 9v battery and 15.5 from my laptop power supply, I now feel I can trust it.

So, I started measuring voltages on Buzz.

Before I started, the battery was at 12.6V. I turned on the key and saw it drop immediately to 12.4V then slowly ‘drain’ down to 12.1V after 3 minutes of just sitting there with the key on. At that point, running lights, instrument lights and the ECU would be the only current draw.

While cranking, it dropped to 10.5V. As is the current state of the project, it was hard to get him started and running, but once going and warm enough to idle and run, I had enough hands and eyes to watch the meter. At idle, battery voltage is about 12.8V, at 2000RPM steady, it’s 14.1. As I recall, those seem to be reasonable figures to indicate that the alternator is at least working.

The headlight is still quite dim, but it does fluctuate with reving. It still never gets above what I would subjectively call about 40% brilliance. More on that subject later.

The battery itself is still slightly suspect, because the cranking voltage drops pretty deep, so I will probably replace it anyway. I have not checked the spark while cold cranking the engine lately. The low cranking voltage is probably contributing to the trouble starting.

With the battery voltage question shelved for the moment, I set about trying to extract a little better running condition. Another one of those little things bothering my subconscious about the fuel system is the tendency of the fuel filter to run “dry” when the fuel pump is running. The fuel pressure seemed to be ok, so I put it out of my mind. Yesterday, I was browsing some of the MS forums for stuff about power loss at certain RPMs. At least one post suggested that fuel starvation could be a factor. I rev Buzz up and, as usual, it sounds like the engine hits a rev limiter at about 3200 RPM, but this time, I watched the fuel pressure gauge during the event. It bounced down from the usual steady 42psi to about 35, jumping roughly in time with the stutter. Hmmmm.

The return from the regulator for my temporary fuel supply is a clear hose. It’s been handy because I can see flow in it and that gives me an operational check at a glance. Well, an extended bonk against the pseudo rev limiter revealed another symptom that simply hadn’t occurred to me and would not have been easily detected without the clear hose. Bubbles. Lots of bubbles in the return line, but only when the engine is stuttering at 3200 RPM. Hmmmm, ok.

I bypassed the fuel filter and tried again. No bubbles in the return line and while it still sounds like it’s hitting a rev limiter, now it does it at about 4300 RPM and there is no fuel pressure gauge bounce. There is still much tuning to do, but I now feel that fuel starvation is probably no longer a factor.

A tweaked around with the fueling settings in MegaTune, but didn’t make any huge progress. By that time, it was getting chilly enough outside that my fingers, especially on my mouse hand, were getting cold, so I wimped out and packed up and went inside.

The brain, however, doesn’t stop and I began trying to figure out why the lights would be dim when the voltage should be high enough for bright lights. Something to check came to mind this morning….

Alert readers following the blog will recall that when I needed to tap an ignition switch controlled voltage, clips on the stock fuse box were pretty much falling apart in my hand. I had a spare DelCity fuse block, so I replaced the stock fuse block. What occurred to me this morning is that I crimped those connections and perhaps I have a bad enough crimp to limit the current available to the headlight.

No, no election coverage here… we’re talking amps!

I’m surprised that I didn’t remember to do this until last night, namely measure the current draw of various components of the system.

For each measurement below, I first started and ran Buzz until warmed up, mostly so he’d idle without fiddling. Then I shut him off, pulled the fuse feeding the measured equipment and bridged the fuse connection with my ammeter probes. I started the engine and noted the values at idle, while reving and while holding the speed steady at what is currently the highest at which he’ll run smoothly, about 3200 RPM.

As expected from the start, the fuel pump may be the largest draw, coming in at 3.7 to 3.8A. This is also a fairly steady draw, since so long as the engine is running, the fuel pump is on.

The next big draw is the ignition system, coming in at 700mA at idle and 1.6A at speed. I suspect this value will climb at higher RPMs due to the nature of an ignition system. It’s essentially a PWM signal in which the pulse width stays about the same, but since the frequency increases, the apparent percentage gets larger, thus the current draw will go up.

Everything else is almost trivial, but may not be exactly accurate because the low range fuse on my ammeter is apparently blown.

The fuel injector rails draw 70mA each at all speeds. This will probably go up a little as RPMs get higher for similar reasons that the ignition system will draw more, but proportionally, it will still be a low value. Even at 100% duty cycle, each injector would draw a little less than 1A and with this small an engine, 100% duty cycle would pretty much be a hydrolocking degree of flooding.

The ECU itself appeared to draw about 30mA, but that seems low according to the documentation. I would expect it more like 100-200mA. The final measured value is the FIDLE output. One day, I will have a solenoid there, but for now it’s just an LED. The 10A range on my meter would probably not show it, but since the engine was warm, it wasn’t lit anyway so no reading.

All this measuring of current draw made me take note that the headlight was pretty dim while the engine was running. I checked the battery voltage and it was 11.58V, unchanging when either revving the engine or shutting it off. That is extremely unusual. With no charging system, I would expect the voltage to raise a bit when shut off, the load having been removed. If the charging system were working properly, it should raise when reving. I didn’t even see it change with my external charger connected. Hmmmmm

In any case, Buzz is to the point where I can begin tuning the engine in other RPM ranges, though I now also need to get this electrical issue resolved. No juice, no go.

Short version: It worked.

Long version: I varied the plan slightly, based partly on not having the right size drill bit. I didn’t have a small bit to use for plan A, so I used a 1/8″ bit and, since I happen to have a gob of them, a 1/8″ x 3/8″ steel rivet.

I centerpunched the appropriate spot and drilled a 1/8″ hole about 3/16″ deep in the edge of the wheel. I drilled another 1/8″ hole to intersect the first one. I inserted the rivet to the right depth, which was scant less than bottomed out. I used a center punch to stake the rivet in place. I cranked up the torch and filled the space in with brazing rod through the intersecting hole. Easier to do than to describe, I think.

After that, a little grinding, filing and wire brushing and it’s got another decent tooth.

I used what I’m sure will turn out to be temporary spacers and bolted the wheel to the crank. It took a couple of times to get it adjusted close enough to start, once it was going, Buzz had what was actually a pretty ok idle. All four pipes hot!

The signal is a solid 1.8V, so I gained quite a bit of signal power. I suspect the timing is still a little off because I can’t get him to rev over about 3500 RPM and by then the advance needle in MegaTune is swinging wildly. Sounds kinda like he hits a rev limiter that recovers kinda slow. The RPM is nowhere near the rev limiter settings.

Most importantly, I now seem to have reliable ignition, so next time I get to work on it, I think I will button up the ECU and EDIS module into their permanently attached locations.

As for the trigger wheel, what’s left is to solidify the connection to the crank. The wheel sits just about perfectly on two nylon spacers that I got at the hardware store. At this point, the plan is to get a dowel pin long enough to secure both spacers to the crank, use the timing light to get the wheel within the timing adjuster range of dead on, then use another smaller dowel or roll pin to secure the trigger wheel to the spacers. The roll ping will reside under the fender washer under the bolt that holds it all to the crank.

Let the tuning begin!

Sometimes, I am such an id10t.

I spent a couple hours last night making a new trigger wheel from a scrap 3/16″ angle bracket that was in the junk box that came with my Stires trike. I used a 3-1/4″ hole saw in the the drill press to (slowly) cut out the disk, the bolted the current disk to the blank and hand ground it to size. With the original still bolted in place as a template, I center-punched for each hole to drill. I took the original off and proceeded to drill all 36 holes. Once the holes were drilled, I fitted a 3″ cutoff wheel to a drill arbor and used that to cut slots into each hole. A little judicious grinding, wire brushing and LOTS of filing to get the teeth in pretty good form. They are not CNC clean, or even manual rotary table clean, but they are very nice for generating timing pulses.

Well, if I needed a 36-2 trigger wheel, it would be good.

My excuse is that by the time I was ready to remove teeth, my brain was on autopilot and somehow removed two teeth instead of one. I didn’t even realize it until I was showing Toni the fruits of my labor.

In the interest of avoiding a shopping trip, I have a plan that may be able to salvage this wheel, a plan that is very much like applying a dental implant. I will drill a hole in the edge where the extra missing tooth should be, use short bit of fluxless welding rod as a reinforcing pin and braze on a similarly drilled replacement tooth. Grind and file the tooth to shape and we’re back in business. Even with brazing, the replacement will still be a lot more ferrous mass than a missing tooth or the previous wheel. If it doesn’t work out, I’ll be headed to Norton Metals.

It’s Monday. I did get to work on Buzz for a while on Friday evening, but have only just now been able to sit down and bang out an update.

After giving up on the search for a local source for terminals and boots to customize my existing salvage spark plug cables, I elected to purchase a nice set of wires and hack the salvage coil connectors onto them. It actually worked out quite well.

Once I had the wires, it was a pretty simple matter to snip the old coil connector off and gently pull the connector out of the boot; it’s easy to bend and the brass won’t tolerate much bending. Then I carefully wedged the crimp apart and removed the old wire. I used needlenose pliers to bend the tabs back into a shape that looked usable, a little curve ready to bite the cable.

I cut the new wire to length, plus about an inch. Used the dielectric grease supplied with the wires to lubricate the wire and threaded it through the cap. I stripped off an inch of the insulation, folded the conductor back and manually crimped the terminal on. Crimping like this is pretty much by definition not as reliable as using a proper crimping tool, but they passed a moderate pull test. I saw now point in breaking it on purpose.

I routed the wires and installed them. Once the wires were in place, I secured the coil with nylon wire ties.

I put the rest of the system back together and tried to start Buzz up. It was a little cool out and as expected but not as hoped, I had trouble getting it to run well enough to even idle. I was able to get it to run enough to heat up exhaust pipes. Cylinders 1 and 2, warm. Cylinders 3 and 4, cold.

I played with the timing by finding TDC on #1 and adjusting the trigger wheel. Again. I toyed with the wheel and I think finally have a unifying theory of the problem.

Alert (obsessive) readers may recall that I voiced some concern about the signal voltage from the VR sensor and my thin little trigger wheel. I put my meter in bargraph mode and watched carefully as I cranked it. The displayed digits dithered between .484 and .486 volts, but the bargraph indicated a swing of .1 volts. I suspected I would know the cause of the voltage swing.

With no helper available, what I did next was kinda hard. I pulled the stock ignition fuse (well, technically my replacement of it), the fuel pump fuse and unplugged the ignition coils. I wire tied a wrench to the starter button and placed the meter where I would watch the meter and the gape between the trigger wheel and the VR sensor. I pushed the fuse in, which started the engine cranking. Sure enough, the gap between the trigger wheel and the sensor, as small an eccentricity as it may be, correlated with the swing in the output voltage.

The information in the MegaManual indicates that the EDIS module doesn’t “arm” the counter until the signal exceeds .5 volts. I know that there could be some tolerance, but the tolerance could just as well require .6 volts as .4 volts. In any case, I can definitely see where the EDIS module could be missing pulses and screwing up the timing something terrible with this bad information from the sensor.

The best way to increase the signal is to add mass to the trigger wheel. I see metal chips in my future.

I discussed the cylinder 3 issue on the Yamaha XJ Owners list. I first wanted to verify that I was not making any assumptions about the actual firing order of the coils, whether the coils do indeed fire alternately, as opposed to sequentially. In other words, if the firing order was something like 1-4-3-2, it would need to fire the two coils in a left-left-right-right sequence, but with 1-2-4-3, it would be left-right-left-right. The EDIS module definitely expects to fire left-right-left-right and if the bike was expecting left-left-right-right, I would need to swap some plug wires around to account for the “change” in coil firing order. However, with all that was suggested on the list and my own understanding of things, I think the coils are firing in the proper order. Besides, it did run on three cylinders at one point.

Another suggestion was that the EDIS system may not work well with the higher impedence stock coils. One lister wrote:

What does the EDIS system expect the coil primary resistance to be? I think that the stock coils are about 3 ohms. When I looked up some replacement coils for an EDIS system from Accel, the primary resistance was 0.5 ohms. The higher resistance of the stock coils could cause the EDIS ignition to not produce enough current to fire the coils consistantly.

Intuitively, I would think that since EDIS module grounds the coil connection that it would power a 3 ohm coil as well as a 0.5 ohm coil. Upon analysis, the suggestion does have merit. Ohms law dictates that a 0.5 ohm coil would pull 28A from 14V, but the supply fuse for the coils and the module is a 15A. Yes, in brief pulses like the ignition system uses, the fuse could probably tolerate nearly double it’s rating, but it seems unlikely that they would do it that way. More likely, they are limiting the available current with an internal series resistor. That series resistor plus the already much lower current draw of a 3 ohm coil could combine to prevent the 3 ohm coil from getting enough current to make a good spark.

So, I stopped worrying about it until I could get a Ford coil to work with.

I stopped at a pull-a-part place on the way home Monday night. They close at 5:30; I arrived at 5:15 with warnings that I only had about 10 minutes. I found the coil on the very first car I walked up to. Four little bolts and two snips later, I’m at the counter laying down $27 for the coil, connector and plug wires. On the road home by 5:25.

So, I clean it up and put a compatible connector on it. I connected one plug wire and put a handy sparkplug (out of the Sportster trike) in it and pulled the fuel pump fuse. I grounded the plug on the engine and cranked. I got a teeny little unimpressive spark and an unhealthy dose of dashed hope.

Then I remembered that these sorts of coils like both plugs connected. The high voltage circuit is from one coil secondary lead to the plug, across the gap, through the block to the other plug, across that gap and back to the other secondary lead. So, I dug around and found another suitable spark plug. Lucky for me the Sportster is a twin. I connected both plugs to the 1/4 coil outputs and grounded both plugs to the engine. This time, when I cranked it, I got a nice fat blue sparks with clearly visible orange plasma. *AND* about every third spark or so, a 15mm long lightning bolt jumping BETWEEN the two spark plugs, as in from one plug’s electrode to the other plug’s electrode.

Faith is restored.

I have found that “Yamaha” is apparently Japanese for “hidden relays” and I will need to relocate another relay, the function of which I have yet to ascertain, one that once resided between the two stock coils. With the mystery relay out of the way, the Ford coil fits very nicely in the space where the stock coils were, even with all four plug wires attached.

On the car, the coil towers “faced” upward, looking like a square distributor at one end of the engine. I have them facing downward. It looks like it was intended to go there.

The remaining work will be wiring, both to the EDIS module and to the plugs. The Ford wires, besides being way too long, have molded on plug ends, I would need to replace these ends, which wouldn’t be too bad a job. They need to be radically shorter anyway, probably none longer that 30cm or so. I’ll have to get the screw on contact tips for my plugs.

The stock XJ550 plug caps appear to have 10K resistors in them, but the monster spark was off resistor wires, so I think I can use some sort of crimp on plug terminals and not worry about resistor caps. Hopefully, I can find a pair of 90 degree caps and a pair of some lesser angle, but straight connectors will work for all four if I must.