Modern Vespa

That's is how a sensor works on a electric wheelchair seating for tilt, with a smart seat motor. It has a long flat potentiometer. Readings from both ends as a check and balance. Such as a 10K pot would have 5k on both ends when centered correctly.

Some electric beds have a round potentiometer connected to a gear that does the same thing. You can control how much variation you get with gear sizes.

Wonder if you could use a magnetic proximity sensor at each detent, with a rare earth magnet on the arm? Magnet might activate sensor in one of the adjacent gears also.

So good news is that the throttle pot is going to work just fine as-envisioned. The cap is almost a perfect fit into the throttle pulley, so I can just glue it in there with contact cement and declare victory, then fabricate a bracket which will attach via one of the cable bracket bolts. Assuming I can fit it all in there...

It might be time to get serious about cleaning up my wiring, especially for the speedo. I have some bonus wiring for turn signals, CHT power, CHT light, and a few other things in there, too, which were mostly added piecemeal for maximum mess.

and for the gears, a pot mounted to the selector box cover and a blade tacked to the selector pin for it to interface with is going to win the day. I could even probably use contact cement and attach the knob, same as the throttle, if I didn't want to weld (yeah, right! )

I made less progress than I expected today, because I wound up spending a decent chunk of time just cleaning up the workshop. It really wasn't deliberate, more of just cleaning and straightening every time I couldn't find what I was looking for, which was pretty much all the time. Eventually, I swept the floor and it all looks a heck of a lot better as a result.

So while I might not have moved the ball forward all that much, I definitely made some progress overall.

for the gear channel - why not use engine speed and vehicle speed to determine gear ratio and call it done? It will read shit when changing gears, but otherwise your gear ratio can easily be used to call out which gear you are in.

That's not a bad idea at all. So long as I knew if I was in gear or not, it work work just fine.

It won't let me identify a missed shift vs. worn gears or cruciform, though, I don't think. Given that it's a pure software solution, I may code it up too and see how the two results compare, other than yours probably being more reliable.

yeah i figured you already had those pieces of information. we used to use something similar to determine manual gear selected in our ECU logic when I worked at a large OEM.

The obstacle to using gear ratio & speed is also (one of) the challenges I'm having right now, namely getting a good tach read. My coupler worked fine the last time I tested it, which was when I was still on the breadboard prototype.

Now, it's reading either super-low or super-high, like off by an order of magnitude in either direction. It's consistent on each test run, but unpredictable over all. I don't know if that's because I'm overloading the bus or because of some external factor. I need to re-test with some of the other boards removed and see what happens to try and isolate the issue.

I will probably also revisit my Hall Effect sensor and see what that gets me. That's probably the ideal option, assuming I can get the noise under control. I can apparently use a small capacitor to generate better edges on the pulses, but I'll have to get that going, plus I don't have any capacitors yet (I ordered them over the weekend, but they're not here yet).

I gain new respect for electrical and hardware engineering types every time I work on this project.

btw you don't even need to know when you are in gear if you use the speed/speed idea, just say that anything that doesn't fit into a band around your known gear ratios is obviously neutral. So if you are within 0.3 of a known ratio (pick a band), you are in that gear.

if you put a strain gauges on the axle you could read the torque at the rear wheel, from there you can also get the power while you ride.

Strain gauges are a new one to me, but seem to be frequently immersed in liquid when used as accelerometers. I have an accelerometer which I have not yet begun to calibrate, but I'm struggling to understand how a strain gauge would connect with the axle.

John, the more I think about it, the more I'm liking your suggestion for figuring out gearing and/or, at the minimum, clutch status.

oh, detecting clutch slip would be a huge plus

strain gauges are a thin foil that is you epoxy to the metal you want to measure. the resistance changes as the metal stretches or shrinks with strain. for torque they are arainged at 45°



the hard part is making the pickups to read the gauge.

That strain gauge is probably beyond my skills for the time being, especially given the tight physical constraints, because I'd have to figure out how to fit a slip ring for the wiring harness along with the strain gauges in a pretty tight area.

Maybe something interesting to play with in the trashed cases, though. I'd have to build it inside the gearbox, between the gears and the bearing. I'd like to be able to use an IP68 through-bore slip ring, but I wouldn't be able to get it onto the axle.

If a pair of contacts onto copper would rings be adequate conductors for that level of voltage, then it could work. I'd just need to epoxy a couple of foil rings onto the axle along with the strain gauges and use a couple of contacts against those, run the wires out and good to go...

mmm....torque gauge...

(Don't you like how this post went from "I can't do that," to "here's how I could possibly do that..." ?)

Could you put strain gauges into the motor mounts instead of the rotating shaft? "For every action there's an equal and opposite reaction."

I don't even begin to know what that would look like. Plus the buffers in the motor mounts and everything else involved making it all MUCH harder, even compared to stuffing a sensor in the gearbox, at least for me.

I did something similar on an engineering project I once worked on. It was easier (and much cheaper) to put a sensor in the stationary motor mount than the rotating shaft. However, that was at much lower rpm -- less than 100 rpm at most and I did have control over the design of the motor mount so it was easy to include the sensor. I can see your point that it could be more complicated in a scooter.

Good luck & have fun!

Edit: Forget it. On my project I could discount gravity. I don' t think you can because of the motor bouncing around over bumps etc. That would make the calculation of torque much more complicated.

Thanks! On board a single-cylinder 2t is also pretty much the worst possible operating environment for what I'm trying to do, too. Space-constrained, tons of vibration, electrically noisy. Oil everywhere.

Still, it's all good

The answer to rotational torque measurement is here, but is too expensive for your application:

https://magcanica.com/

They use it in F1 applications.

yup...they make exactly what I want, but I expect the the cost is solidly in the "if you have to ask, you can't afford it" range.

Perhaps you can find a used one on eBay.

but chandlerman, it's for science

As will be my body once my wife sells it to medical researchers after she kills me for buying one!

Today was throttle and gear position sensor install day.

I went with my plan of spot welding a blade onto the selector box pin and attaching a potentiometer to that and it worked well, but it's really tight down there. I had to cut down the dial on the pot to fit it in.

The video kind've sucks because the bottom of the terminal is cut off so you only see the response a few seconds after it occurred (it was updating once per second), but it works just fine. I defined ranges for each gear and then just check for if the output from the pot is within the range or not.

The throttle sensor install also went pretty well. I just used Goop, which is a thick contact cement, sort've like Shoe Goo, and glued the dial cap into the center of my throttle pulley, then wired it up. I still need to fish the sensor wire through the frame, but other than that, it's in good shape.

Next up will be adding a second CHT sensor and then figuring out the tach. I need to order a Hall Effect breakout board that will give me back a digital pulse, because the Pi doesn't deal will with analog inputs and using a pull-up or pull-down resistor doesn't seem to be working for me (I put this down as User Error).

I didn't do any further examination of why my Opto-Electric coupler seems to be mis-reading since it was working well the first time I tested it, so I'll also check that out, but the Hall Sensor breakout would certainly be smaller, even if the breakout board would most likely need to be mounted by the flywheel. More to come on that, obviously.

Oh, and while I was at it, I replaced my frayed throttle cable before it jammed up on me.

Ha! Pretty awesome.

Can't wait till you complete this to see how it all gets put together with visual displays.

I'll be looking for the parts list and instructions when its complete!

Me, too!

Wow... Period!
Awesome project. Been spending some time just trying to figure how to get the neutral light to work on my wife's Px. Thanks for the inspiration.

Me Three! Actually, I'd rather be your beta tester, after the alpha testers work out the kinks. In other words, a working system with most of the kinks worked out.

I think I'm going to have a winner with the Hall Effect breakout board as my tach.

I glued a magnet to my flywheel over lunch, so I'll be able to test how it works later this afternoon. My "wave a magnet back & forth in front of the sensor" testing looked good, though.

this was my setup for testing the hall sensor at my desk.

was chasing an overflow that happened every ~30 miles, stupid math.

I assume that was 78 RPM's * 3 reads per actual revolution, so effectively 234 RPM's? That'd take about 3 hours per crash with a 10" tire, right?

aaaaannnnd.....we have a tach.

This one turned out to be a little bit tricky, because everything looked good at low speed, but as soon as I tried testing it with the motor running, the measures went to crap.

Finally, it occurred to me that maybe my time wasn't high enough resolution, so I swapped over from calling to This changed my timer precision from more-or-less a millisecond or so of precision to a nanosecond of precision. Given that the interval between two spark firings could be down in the 5ms or so range, that seems pretty obvious in hindsight.

Once I did that, other than having mis-counted the number of zeros in my multiplier to get RPM's, the values were pretty much dead in line with my existing tach.

Also, on a whim I tried putting the sensor on the ignition wire just to see what would happen and it worked perfectly, so I'll probably just roll with that going forward.

The one change I'm going to have to make is that I added a 0.1uF capacitor on my output to clean up the signal, so I need to add that to my board.

Also, the rattling in the video is my flywheel cover being mostly loose.

Now, I just need to get the accelerometer going, which, critically, is about getting it calibrated to give me meaningful data. I'm also a little curious how it's going to do with the vibration of the motor, but I'll cross that bridge when I come to it. It's just breakout board on the main board, so it's all a software problem.

Next up will be getting the wiring inside the frame, figuring out where I'm going to mount the Pi itself. I'm still thinking under the seat, but we'll see.

Oh, and start gathering some telemetry.

So to close, I'm pretty happy with where I'm at right now, which is all of my critical telemetry is now working:
- Tach
- CHT
- EGT
- Gear
- Throttle Position
- Speed (GPS)
- Position/Direction (GPS)

I have more Hall Sensors coming (they're super cheap, like a buck apiece off Amazon), so I might use one of those for the clutch. I can just put a magnet on the clutch arm and detect the clutch status that way.

I even have a linear Hall Sensor that I could wire it into my Analog-to-Digital converter and get *exact* clutch position as a % of throw, but that may be more information than I can actually take advantage of.

are you using interrupts for the tach signal?

One advantage to the arduino is the hardware interrupts, but have wondered if the SW interrupts on the pi are up to the task.

I'm using the GPIO library for Python, which is definitely good down to the millisecond level, if not further. And most of m measures are going to be take a few times per second, so it's not like I'm really pushing the board.

The one thing I did was to go with the hardware debounce for the tach, so I don't have to do (as much) software detection for that.

In my VERY limited testing thus far, though, everything seems to be doing just fine. This is on a Pi Zero 2, and people actually got this stuff going as far back as the original Pi's, albeit with a bit more effort than it took me. And we're talking 5ms as a minimum interval, which is forever to a computer.

Once I've got everything up and running, plus pushing out telemetry, then I'll really know if it's up to the task or not, but I'm not all that worried as of right now.

Today, I continued with mounting. Wires are almost all run, which isn't as much of an accomplishment as it maybe sounds since I only really had to pull one, for the throttle sensor.

The gear sensor's wire is too short to run inside the frame through the bellows hole, then back out. The only place I can mount the pi under the hump of my sport seat, and it wasn't long enough for that.

I have the EGT hooked up and have a bottom tap coming tomorrow from Amazon so I can install a CHT. I'm going to try and M6, then cut a slot into an M6 bolt and pot a sensor into that. I think it'll work well, just have to execute on the vision.

The magnet I had glued to the flywheel had flown off, so I went for something a little more aggressive. I don't think it's going anywhere now that it's inside the flywheel. I'm still having issues with the data from the tach, though. I'm getting bad readings, but had to bypass the capacitor I used on the breadboard because I was getting no responses from it. So that'll get some more exploration.

I also added a USB socket that runs off the scoot's battery so I can roll out once things are ready and ran the actual monitor a little bit, too.

And I found a basic Timestream visualization tool that someone built. It's a react app and has no persistence, to it's just for proof-of-concept grade stuff, but it's a tool and will let me start playing a little, which is more than I had. I may try to extend it to add some persistence to it, but haven't come to any decisions on that one way or the other.

So...progress, but also a really irritating setback with the tach. And frustrating because I'm getting really close to being able to generate some actual data.

Also, Patrick, I watched load on the system with everything running, it was at about 2% on a Pi Zero Two.

Sounds like you can up the refresh rates a bit maybe 500% 😄

LOL!

Maybe as a stress test, but first I need to get all the telemetry completed and accurate.

Critical path items right now are
- Installing the second CHT
- Figuring out wtf is up with the tach
- Proper mounting for the pi under the seat

Secondary tasks:
- build a longer cable from the gear sensor to the pi, run it through the frame
- Route the USB from the glovebox through the frame
- Sort out calibration of the accelerometer so the data from it won't be garbage

CHT is in

I drilled and tapped an M6 hole next to the existing CHT hole, then made a sender from an M6 bolt.

I cut a slot in the bolt for the sender, beveled the tip, then spaced it for the depth of the hole so it would seat correctly using a couple washers and a nut, then just screwed it in. I also cut a slot in the top for a screwdriver, which made a huge difference in getting it installed.

For the minute that I ran the motor, it was working like a charm. Seems to be reading about 10-15f warmer than the Koso, but it was accurate at room temp, so it may be that it has faster reaction, or it may be that the Koso is not at its best. Maybe both, because the Koso was about 10 degrees cooler than room temp before I started the motor.

Either way, another item off the list.

Try some thermal compound paste in there see if it makes a difference.

https://www.amazon.com/Silver-Thermal-Heatsink-Compound-Syringe/dp/B00M5G6AHY/ref=asc_df_B00M5G6AHY/?tag=hyprod-20&linkCode=df0&hvadid=309751315916&hvpos=&hvnetw=g&hvrand=4875965256553496821&hvpone=&hvptwo=&hvqmt=&hvdev=t&hvdvcmdl=&hvlocint=&hvlocphy=9019817&hvtargid=pla-761431552712&psc=1&tag=&ref=&adgrpid=67183599252&hvpone=&hvptwo=&hvadid=309751315916&hvpos=&hvnetw=g&hvrand=4875965256553496821&hvqmt=&hvdev=t&hvdvcmdl=&hvlocint=&hvlocphy=9019817&hvtargid=pla-761431552712

A layer of red ink from a Sharpie marker will make that bolt look like a $30 sensor.

I only have black Sharpies, but if I color it black, it'll be the new BlackCHT(tm), which I'm sure is a $50 sensor.