I was in the process of reassembling my long disassembled '75 Norton 850 and found that one of the rocker spindles was a sliding fit in the ol' RH4 head. Another one was almost as loose. They're supposed to be more of an interference fit to the point you need to heat the head to get them installed. Among the host of issues that loose spindles can cause is allowing the spindle to rotate (it's supposed to be held in position by an absolutely poorly designed stop plate) and will allow a lot more oil to enter the rocker cavity. This, in turn, swamps the valve guides allowing oil to get sucked in and burned which results in massive amounts of smoke and plug fouling.
One of the better fixes is sold by RGM in England (https://www.rgmnorton.co.uk/buy/one-piece-rocker-spindle-locating-plate_4062.htm). It solves two problems actually, it'll keep the spindle from rotating, and it allows you to really lock it in place with the grub screw which should help keep it much more stable over the long run.
Since I am impatient, had some stainless bar and a CNC mill, I decided to whip up my own version of these. I think they came out pretty good.
And here's the final result. I think they're going to work very well.
I a massed all the raw stock material from OnlineMetals.com
One aspect of this project was to make a "run", i.e. more than one or two parts. I started with the Petal Support member. Each "flower" has 5 of these "petals" so that meant I needed 60 parts. I ended up making a few more in case I screwed anything up later and also to have some spare parts.
I made a jig to hold the blanks and used the rotary to get both sides. It took about 15 mins per part, but after several hours I had 64 done.
Next I had to mill a slot at the pivot end, so I had to make another fixture and run a other op.
I had this idea. It's about data visualization.
My basic proposition is that most folks don't like analyzing data, don't really understand graphs and don't have the patience to figure it out. Along with this is a literal tsunami of data coming at us from sources such as the internet of things (IOT). So how does a normal person "grok" the information signal buried in the data?
Let's say you're monitoring your household electrical consumption, and you happen to have a PV Solar system too. Let's also imagine you have a data logger such that you are capturing how much you're consuming, how much you're generating and, by subtraction, how much you're purchasing from the grid. Now you could generate a graph to show the system performance.
However, that's sort of geeky. I decided to approach this problem in the form of a data driven sculpture. The idea would be to have the state - say what it looks like, sounds like, etc - attempt to convey the information signal in the data. So lets say you change the position of an element as one vector. And you can change the color, brightness and period (if you want a blink effect) of an LED as a second vector. So a happy system - i.e. one where the sun is shining, plenty of power is being made and you're not using too much sets the state of the sculpture in an "open" and bright color mode. And a sad system - i.e. sun is shining, but very little power is being generated (maybe due to a failure of some type) the state is "closed" and the LED is off or glowing dull red. So the casual observer could potentially figure out the meaning of the two states.
What I've chosen to do is to make a kind of flower where the petals can open and close and the bud area is illuminated with 5 RGB LEDs. I plan to make a dozen of 'em. I started working on this idea during Dec of '16 and it's been a pretty slow roll. I guess I'm in no hurry. Here's a few pics of the prototype:
My good friend Rob McDonald got a hold of me a few days ago. Apparently the rear door handle broke on his funky Indian Tuk Tuk. A Tuk Tuk is a 3 wheeled sort of motorcycle/car. Used extensively in countries like India, Thailand, and the Philippines.
Rob got the idea to use one to sell wine. Well not just any wine - but St. Mayhem. I guess when you name the wine St. Mayhem, you have to expect things like broken door handles.
Basically the handle was a poorly constructed cast piece that had seen it's days of hard use. It had incorporated a lock mechanism, but that looked like it has been non-functional since the days of British rule. My basic repair concept was to turn the OD of the broken area down, create a sleeve, then glue and pin it together.
When I looked the parts I found this - a drywall screw that was used in a previous repair attempt.
First I found a chunk of stainless steel tubing that had the right OD (0.625"), but needed to be bored out. I decided a wall thickness of around 0.030" would be about right. After that I had to turn the OD of the two halves down to 0.565-ish to fit inside the sleeve. Hanging on to the handle side of things required a bit of creativity to make an aluminum fixture that could be used to clamp on the handle shaft.
After the 3 parts were created, I ran the broken parts through my bead blaster to remove any loose material and create a nice surface for the glue to adhere. I mixed up some JBWeld and epoxied them together.
I didn't take a picture, but after the glue set up, I drilled two cross holes around 0.050" and drove 2 finish nails through to create a couple of pins. Hopefully they'll keep it from separating and twisting. I think the repair looks pretty good. We'll have to see how it holds up to St. Mayhem!
This project started as the result of a failed earlier project. I had attempted to build a motorcycle logger - i.e. a platform to record as many things as I could manage during a ride. I had rpm, gear, head temp (left and right), oil temp, oil pressure, speed, GPS location and ambient air temp. I was hoping to add the AFR. Although I got it all working, it would crash often. I tracked it down the way I was capturing rpm - by using the coil negative. I was getting some sort of noise, either RF from the spark plugs or maybe back EMF from the coil or something. I could figure it out and it made the project unstable.
Roll forward a few years. I ran across some info on this and found a reference to someone having a similar problem. They solved it by making sure the ground from the coil negative interface circuit went back to the battery, not the circuit board as the voltage spikes from the coil collapse and subsequent ringing would cause the processor to crash.
I wanted to play around with this, but I didn't want to do it on the bike. So I thought I would create test bed for running ignitions.
I came up with an Arduino that drives a DC motor. The motor is connected to a shaft that has a taper bored in one end to simulate the end of the cam. On the other end I mounted a magnet on a cross shaft. A hall effect sensor triggers when the magnet goes by and I use this to determine "TDC". I then cobbled together a couple of stock Lucas 17M6 coils, plugs, a motorcycle battery (as I need about 3W of 12V to drive the ignitions) and my trusty Tektronix TDS2024 scope.
I ended up using a circuit published by MegaSquirt for ignition pickup - basically using a 1N4001 and a 4N32 (optocoupler). It’s simple but seems to work pretty well. I used a simplified version of this circuit (no “John” Zener, no C31, C11 or C12. I may add the C11 and C12 in as I do get a bit of noise):
This circuit seems to work well. I think the key is making sure XG1goes back to the battery, not to the ground plane the uC uses (i.e. don't jumper to XG2).
The basic procedure is to mount an ignition, set it up so the spark happens before the magnet triggers the hall effect sensor and run a program I wrote on the Arduino. The program uses PWM to increase the motor speed by 1 (the Arduino allows 255 steps) and then determine the rpms and the amount of time between the spark event and the TDC event. This is converted to an angle and is the advance. By running from around 600 rpm to 6000 rpm and taking advance measurements, you can plot the curve.
I happen to have a small collection of ignitions I've collected over the years. A Boyer MicroMKII, an Old Britts PowerArc, an older and newer Trispark, and an ancient Lucas AB11. I have a decent AAU, but setting it up was a pain so I didn't test it.
A note about the reported values. The RPMs are crankshaft RPMs. The Advance angles are crank angles. You also must realize the advance angles are somewhat arbitrary as they are offset to the manufactures spec. In other words, when I run my setup I really don't know where TDC is in relation to the spark. (I didn't have an easy way to accurately determine this in the test rig.) I just set it up to be somewhere before TDC using my scope to verify. After the test run, I take the raw data and look for a "calibration" point - i.e. I know the ignition should be set to 28 degrees BTDC at 3500 rpm. So I simply find my closest 3500 rpm point, calculate how "off" my raw number is from the spec and then subtract that from my raw data. This seems to work pretty well.
HOWEVER - as DynoDave would probably point out, it's more meaningful probably to use cam shaft angles and absolute advance data. This effectively eliminates this attempt at aligning the advance with TDC.
Here's the data I used to set this up:
Lucas 31 degrees @ 5000 rpm
Boyer 31 degrees @ 5000 rpm
Trispark 28 degrees @ 3500 rpm
PowerArc 35 degrees @ 3600 rpm (taken off the Old Britts Web page).
So in order of evolution:
Lucas AB11 - this is an old analog ignition. What is interesting is the coil is on - i.e. energizing - almost the entire time. The only time this isn't true is when a spark happens and it's briefly grounded to get the field to collapse. I would guess this uses the most power due to this. In the following image the yellow trace is my hall effect sensor. When it goes low, that's TDC. As you can see about 2.7ms before that on the blue trace is the coil negative line. It goes low for 0.1-0.2 ms.
Next up is the Boyer MKIII. Another analog unit. It's coil on/off times are better, probably something like a 50% duty cycle (I forgot to take a picture).
What's very interesting to me is the Lucas and Boyer curves are almost identical.
I then tested two Trisparks. This is a digital ignition. I had an older one and a new one. (Apparently it was redesigned to better tolerate the heat in the timing cavity with more robust components.) They ended up with basically the same curve. You can clearly see the idle stabilization effects in the 1000 - 1500 rpm range. The Trispark seemed to manage the coil on times to minimize current load, but maintain a good spark. It seemed like they were trying to keep the on time to around 8ms.
Last up was the Old Britts PowerArc unit. It used a completely different pick up method - optical instead of magnetic. It also used a very different approach to managing the spark. This unit actually sparked three times for each cycle. In addition the coil on times were managed for each segment of the advance curve. In this picture you can clearly see the 3 pluses in the blue trace.
You can clearly see the digital nature of the ignition in the following graph. The increases in advance are very discrete. Since there 3 spark events per cycle, there are 3 advance curves. Four, if you count the coil on event.
Of course, being digital means you can have more than one mapping, so Old Britts supplies a "sport" curve meant to be a bit more aggressive.
It looks like the real difference is the 2nd and 3rd sparks are more advanced in the Sport version.
Here's an overlay of the two. Sorry about the jumbled legend ( I couldn't get Excel to sort it...)
I though it would be fun to overlay them all. I used the first spark event for the PowerArc curves, but as you can see - it looks a bit fishy. Perhaps someone with more experience can suggest a better way to compare apples to apples.
Well it's pretty clear to me which one is best. Although the analog units have been around for a long time and seem pretty reliable, they really don't offer the best curve. It's well known the Boyer has issues with low battery voltages (I even saw this during my tests when my test battery went low.). The Lucas is no longer being made so it's more of a curiosity I suppose. The PowerArc I found to be very fiddly getting it set up (i.e. getting it timed correctly and fooling around with the optical wheel). You have to set it's version of TDC to your bikes. However when you strobe it, there is no easy way to adjust this and it's very easy to be off a couple of degrees. I also found mounted on the bike, it was susceptible to kickback when starting - there is mention of a start procedure on the Old Britts site you'd better follow if you want your ankles. This leaves us with the Trispark. I think this is an awesome ignition. It's easy to install, easy to dial in, makes the bike run great, seems to have a nice curve and you get idle stabilization that literally transforms your bike.
Here's some links to information on curves I found around the net on this:
http://atlanticgreen.com/boyerexposed.htm (also has Lucas Rita!)
One of the last things I had to do was weld together a tail light/license plate bracket. I had to wait for the light to show up so I could figure out the relationships. Here's what I managed. (Note the last time it was registered - 2002!)
I got the bike on Nov 1, 2015 and it's pretty much done as of this week - Aug 7, 2016. I guess that's about 10 months. Since there were long stretches where I didn't do much, I guess that's not bad. The bike turned out OK. If was planning on making this my bike I would have gone further, but I was primarily interested in spending as little as possible to get it back on the road with cafe styling. My inspiration was something like these, but they have $1,000's in custom paint and body work. Maybe next time.
Mine turned out a lot more stock looking. It runs well, and seems to drive well. I need to get it registered and do some little tweaks and adjusting.
I probably should have written this in a more realtime fashion so I can remember all the bits. Once the painting ordeal was complete, the reassembly went reasonably quickly. I had been collecting bits over the months, so I had a lot of the parts I needed once I dug in. First up was respoking the wheels. I used Buchannans SS spokes and nipples. I also stripped and polished the rims. Things turned out pretty nice.
Next up was to get the engine back in the frame - without scratching things too much! I was working by myself, so I ended up putting the engine on it's side and then lowering the frame around it. I then replaced the lower mounting bolts, turned it upright and it worked great. It was simple matter to then bolt on the transmission.
I then mounted the swing arm and set up the wheel bearings. BMW's use a tapered roller bearing stack that needs the preload set up. I'd never done this before, so it was a bit fiddly, but in the end things seemed to go together correctly. I then mounted the tires and balanced the wheels.
I had decided to make a custom battery holder from the start. I used a blanking plate supplied by Boxerworks to eliminate the stock air cleaner and in it's place, on top of the tranny, I mounted the battery. I thought it came out pretty well. I also decided to put the crank case breather inside the starter cover. I hope it doesn't spew oil in there. I wanted to use peashooter mufflers, but couldn't find any in my budget. I ended up going with Dunstall replicas. Although they didn't look exactly like what I was after, they ended up sounding great.
I went through a bit of hell with the front brakes. The original set up had the master under the tank and was actuated by a cable from the front handlebars. Since the original master was toast (rusted and pitted) and the throttle control was trashed (corroded, teeth worn, etc) I needed to replace them. I also had read this set up suffered from a lot of issues (leaking fluid under the tank, poor braking due to the cable, etc. ) that I changed to a later style. This used a handlebar mounted master and simple junction block under the tank. I also upgraded to SS lower lines. It took a bit of fiddling, but came out OK.
I then started in on the wiring. I was able to use the original harness without much modification. All that was really changed was the ignition as it had a Dyna III and aftermarket coil. (I also had to fool around with the turn signals, taillight and horn.)
I was getting close to completion. Here's a shot just before turn signals, taillight, horns, grips, etc. I had originally considered painting the body work, but frankly ran out of interest and budget. Also the original paint cleaned up well enough I decided to use it. Unfortunately, the battery color didn't work. Oh well.
I have to wait for a few bits, so I can't start it up and try to ride it. I'm hoping the transmission (which I didn't do anything to) is OK!
Where we left this project months ago was I had managed to get the engine back together after discovering shot rod bearings. Next was to put it back in the frame and get this project to a roller. I really want some of my shop space back! That necessitated painting the frame and other bits.
Let me preface this - I hate painting. It's not my thing. In the end I'm kinda aiming for a decent "5 foot" finish. (You know, looks good from 5 feet away or more?)
I've always wanted to try automotive paints and thought I would give a try on the BMW. I decided to use Eastwood paints - they target the DIY market and their stuff is a good compromise between cost, east of use, quality and variety. I chose the DTM Epoxy Primer and a Single Stage Black Urethane. But first I had to build a spray booth, sand down or media blast the parts, figure out a way to mount the frame so I could get at it, and acquire some safety gear (fresh air respirator, hood, suit, etc).
There's tons of info on the net about building you're own small booth. I went with the painters plastic over a PVC frame, cheap box fans with furnace filters, and fluorescent lights style.
What this is really for is two fold - 1) to contain the overspray and fumes, 2) keep contaminates out of your new paint finish while it cures. This style of booth is probably just OK at both, but better than nothing. I did notice there definitely was a positive pressure as there was overspray blown out all around at the bottom of the frame and floor joint creating yet another thing to try to clean up.
Here are some various bits masked off. I think this was the most enjoyable part of the whole thing. It's clean, easy and kinda looks cool when you're done.
Ready for paint. Turns out kinda. Well lets just say I learned a lot. The overarching problem was the fact I had too much going on in here. The parts were too close together, they were too high to paint comfortably, I had a lot of trouble getting all the sides (I had to hold and manipulate with one hand while painting with the other. I got lots of runs. I got lots of overspray. I missed some places.
It was horrible.
Primer on. Another mistake, don't get black primer if you're going to shoot black as a color coat. You can't see where you need to paint. I should have used a grey primer.
I ended up sanding the primer out on the frame sections (where there were runs and stuff) and respraying. (The other pieces ended up being good enough). This would have gone great had my stupid painting hood (which inflated due to the fresh air system and pressurizing and popping up the top a good 6 inches) not collided with my fresh 2nd coat. I'll have to sand that down and touch it up tomorrow. Sheesh!
Did I say I hate paint? The prep work. The set up. The space it takes. The mess. The clean up. The chemicals. The safety issues. The ease with which you can totally screw it up.
I think I've proven to myself, once and for all, that if I need to go beyond rattle cans - I'll send it out.
Here's the basic layout of the shop. The circle with an "x" in it represents a ball valve.
Below is a pic of the chart in ThingSpeak. The slopes of the different curve segments are calculated below:
So it appears the best the system can do is around 2 psi/hr. That's with as little of the system as possible in the mix, i.e. best case. That said ,there's still the compressor check valve, a water separator/filter, air line quick disconnect fitting, and 3 ball valves in the mix. It's possible the air line quick disconnect fitting could be the bulk of it, however the filter also has an auto drain feature which could be leaking too.
UPDATE: I've been opening up each end point separately and it looks like the Bridgeport is kinda a problem. I'm pretty sure it's the Kurt Power Drawbar IN/OUT switch - I knew it was leaking, but it's leaking more than I wanted. I think the rest of leaks are probably a collection of minor things I'll have to find by soapy water, etc.
A couple of years of ago I finally added hard lines to my air compressor throughout the shop. What a nice upgrade! No more hoses all over the floor. However, it seems my copper sweating skills are maybe not what they should be as it seems to leak a fair amount of air. After I switch off the compressor it takes something like 4 hours for the system to depressurize.
I felt I could do better.
But how to isolate the problem? And how to make it a complex project?
Well, that's easy. Just build a wireless network connected microcontroller with a pressure sensor, an SD card for logging, and a graphical display. Have all your data get collected in two ways - 1) local to the SD Card and 2) upload to ThingSpeak.com.
My thinking is to attach the logger to different segments of the air line. As I move away from the compressor - therefore adding more pipe joints, valves, equipment - I should be able to deduce from the pressure decay rates what the relative leak rates are per section. That should give me an idea of what to work on. Additionally, I can verify my repairs in quantitative terms.
So after a few days of cobbling things together
Here you can see it's attached to my air lines - running about 115psi.
Close up of the mess. The screen is an older Arduino TFT (now discontinued) with an SD card socket. The smaller chip above that is an ESP8266 WiFi chip. For $5 you can't beat it!
The ESP8266 is a 3.3v part which needs some juice when transmitting. I happen to have had lying around an old XBee shield that I had previously added a 2 channel level shifter to. It took care of both issues as it also has a robust 3.3v power supply.
Here's a link to the ThingSpeak page where I'm uploading the data : https://thingspeak.com/channels/6008
I'll have to update this as start logging the system and try to figure it out.