Experimental Cargo Ebike Trip, Vancouver to SF Maker Faire

[justin_le]

Be warned, this is an example of a build project where a lot of things seem to fall perfectly into place. It doesn't usually work out that way!

When we were initially involved in the Edgerunner frame design last year, it came up a few times that we all wanted to have an easy integration of the Stokemonkey drive as one of the electric assist options, as it is most eminently suitable for cargo bikes. Previously, Sam Wittingham of Naked Bikes had done this on some custom handmade longtails by including a 2nd set of horizontal dropouts on the chainstay tubes just for a stokemonkey motor. This is one that we see now and then around Vancouver.


It's a simple and genius approach, since no multipiece clampset hardware is required and the motor is held super rigid in position. And if you don't install the stokemonkey motor, then the only offending hardware are two small additional dropouts.

However, this approach only works if the chainstays are split in two at the bottom bracket and spread wide enough for a motor to fit between them. In the Edgerunner frame, it was decided for several reasons that there should be just a single boom tube that goes back from the bottom bracket, rather than a pair of tubes. It also wasn't clear if the stokemonkey motor would even be available again, and so it didn't make sense to design permanent frame features around something that might not exist.

So instead, a generally uncommitted area was left open in the zone between the 20" rear wheel and the front of the Xtracycle 'V' rack tubes. And the thought was that with all the tubes in this general area it shouldn't be too hard to devise a way of clamping in a Stokemonkey or other mid-drive system:

 

[spinningmagnets]

Justin, I would like to suggest that the horizontal tube just above the word "Here" (boom tube?) be substituted with a section of rectangular cross-sectioned tube, rather than round, with the widest side facing up.

I understand the build-friendly strength vectors and stress distribution of round tube, but I am certain that this section is under tension, with no bending, twisting, or compression loads. As such, a factory provided flat surface in that spot will create a wide variety of options for customers. A variety of non-hub drive systems are being developed right now, and although I agree the SM should be an option, if you make other options difficult to apply, it can hamper a wider group of builders from embracing this awesome frame.

 

[justin_le]

Justin, I would like to suggest that the horizontal tube just above the word "Here" (boom tube?) be substituted with a section of rectangular cross-sectioned tube, rather than round, with the widest side facing up.

That's a really great idea, not only for the mount of other mid-drive options, but also to facilitate using this space to strap on random battery packs too. The edgerunner frame is in the course of some retooling with a new mfg. for the next production run, so I'll see what can can be about changing this tube spec while they are at it, (although it's mostly all Xtracycle's call at this stage). Part of the reason for the current tube shape was to match the Big Dummy and hence be compatible with the existing KickBack stand, which needs to rest against that tube. But that only dictates the profile of the lower half, the top half could still be flat, making something of a sidways 'D' shape to get the best of both worlds.

Anyways, back to the stokemonkey on the current frame incarnation. We met up with Todd of CleverCycle's at interbike last fall at the Xtracycle booth to see what options would work best for providing a stokemonkey mount. The idea that appealed to me the most was hanging the stokemonkey bracket from the cross tube between the chainstays, like so:


At the same time, we were also exploring ideas of implementing a right side drive option for the stokemonkey, which could look cleaner, and could also enable the use of a THUN torque sensing bottom bracket. With a normal stokemonkey setup, the torque of the motor goes through the spindle and so wouldn't be differentiable from the the pedal torque.

Todd had experimented with this previously quite a bit, and had examples like this from a Big Dummy conversion, where the chain does a 'U' turn around the motor freewheel and then does another 180 around an idler.


One thing about the mid-drive done on the same chain loops as the rear wheel is that the chain to the left of the motors has the combined human + motor tension in it, while the drive chain on the right side of the motor only has the tension from the human input. So in principle a small idler with a load cell on the front portion of the chain can detect the rider's pedal force. Here's how that could look, in a manner that requires a left side southpaw freewheel on the motor:


With a conventional right handed freewheel, it's not quite as tidy looking, but it would mean you wouldn't need motors made with lefthand freewheel threads:

 

[justin_le]

This was all conceptual, and it wasn't clear how all the mounting bracket and idlers etc. would look or install on a real bike and the worry was that things could quickly become complicated. But then the edgerunner bikes finally arrived, and we realized that the idler pulley for a right side chainwrap could actually fit right in the middle of the existing and proven stokemonkey bracket design, which was great:


We also had a customer show up with a bike that had the metropolis patterson 2-speed crankset on it, which in addition to giving dual speeds on the front also has built in freewheeling characteristic, which would meaning running of the motor does not force the pedals to turn. That had been the single biggest 'complaint' we'd heard of the original Stokemonkey setup.

Freewheeling cranksets are a frequent point of failure in mid-drive ebikes, but the hope was that since this one was engineered as a crankset it should hold up. In general, the patterson crank is designed for 2 speeds and the freewheeling is incidental, but for us the freewheeling was key and the 2-speeds was the nice 'extra'.

The build for this went together pretty well, with the stokemonkey bracket clamped to the front bridge tube between the seatstays.


Where things get interesting is in supporting the bracket so that it doesn't twist with the chain tension. In this layout, you can see that pedaling puts tension on the bottom chain which pulls the base of the bracket forwards and rotates it in a counter clockwise direction around the pivot point:


However, when the motor runs, it only puts tension in the left half of the chain, and where it comes out of the idler pulley the resulting force vector is almost inline with the pivot and hence produces little torque at all on the mounting bracket.

If indeed the idler was positioned so that the chainline was perfectly inline with the pivot point of the bracket, then we'd have a situation where human pedal force on the chain produces a significant torque twisting the bracket counterclockwise, while the motor force on the chain contributes nothing. And hence with a strain sensor measuring the static torque on the bracket we'd be able to sense the exact rider power input completely decoupled from what the motor is doing. And with THAT, we'd be able to accurately measure human watts without a torque sensing bottom bracket, and unlike the THUN sensor we'd be picking up both the left and right pedal strokes.

So we had the stainless steel stokemonky bracket somewhat redesigned so that included a lever arm against which we could restrict it from twisting and also measure the force, here is the prototype with just some plate metal welded in:


After a few revisions on ways of sensing the torque, some of which had way too much flex for my liking, we came up with this, machined from aluminum plate stock.

The lever arm of the bracket pulls downwards the metal in a way that results in a uniform shear force inside the dogbone cutout, which is then easily measured with a single shear sensing strain gauge bridge.

A few days of CAD and CAM and then finally CNC machining:


It's always a happy moment when something comes off looking exactly like the 3D computer models:


And here is a closeup showing what it looks like with the strain gauge sensor glued in place. Strain gauges are delicate things, but once glued to the solid aluminum substrate and covered with potting compound they're as robust and reliable as you could hope.

 

[justin_le]

Here's what the resulting clampset looks like installed on the bike. The controller manages to fit perfectly between the two left and right mounts, hanging from the battery rail so all of the battery/controller/motor wiring is just a few inches long and completely out of sight:


You can see too that the rear chainline passes right through the pivot point of the clampset, so motor torque shouldn't result in any force on the strain sensor:
View attachment 5

With the Xtracycle rack and bags installed, all you can see of the entire electric drive system is just a hint of the red motor protruding, and there is nothing in the visible front section of chainline to clue in of funky chain action.


The strain gauge output is only a few mV of full scale, so in order for the CA to read that it has to be amplified to a useful 0-5V range, which we did with the INA126 instrumentation amp and a few potentiometers to play with the gain and offsets:


With the V3 CA, you can set the exact gain ratio in Nm/volt for any torque sensor. So we calibrated it by standing on the pedal and reading the voltage output. With the 170mm cranks, the resulting torque should be trq=m*g*0.17, which in my case was 125 Nm with all my weight on the crank. The strain gauge output had a swing of exactly 2.6V between my weight being on or off the pedals, so the torque scaling was 125 Nm / 2.6V = 48 Nm/V, and with that set up you could push the cranks and watch the torque reading go up nicely in proportion.


In order to use a torque signal on the CA3 though, it's also essential to have a pedal rotation sensor so that the CA knows your rpm cadence. Unfortunately the patterson crankset uses and external BB ball bearing system, and has almost no gap to install a normal PAS magnet disk. So we had to machine off a layer of the crank and also machine the PAS sensor disk to be a lot thinner in the mid section, and that seemed to do the trick:


Then finally, with all that in place it was testing time! That finally happened the day before yesterday after quite a few months of development. The 12 pole PAS sensor enables the assist to kick in almost the moment you start pedaling, (a noticeable improvement to the 8 pole THUN device), and since the chain tension senses your left and right pedal torque equally well, it's way more responsive off the line regardless of which side you start to pedal on. And, of course it is totally silent. The good quality Terracycle idler pulley played an important role in that.
http://t-cycle.com/idlers-chain-management-c-41/idlers-c-41_9/sport-power-idler-p-134.html
In earlier tests we had a cheaper idler without properly shaped teeth and you could hear/feel the chain travel over that when the chain was taught form the motor running.

The verdict from pretty much all who have tried it so far has been resoundingly positive. I had the CA3 torque assist factor set to 1:1 in proportion to human power, and at the end of every trip the total consumed human watt-hours and electrical watt-hours were always within 1% of each other:



The only down side is that the whole assembly has a little bit of flex with the chain routing like this, so when you stand with all weight on cranks there is like 8-10mm of deflection at the pedal. Some of this comes from chain stretch, some comes from the bracket and clamset flexing under the strain, and a fair amount comes from inside the Nuvinci hub itself. Overall it's not even noticeable when pedaling with the assist, but when you turn the assist off and ride hard it seems like the direct connection from your pedals to the wheel is a tad squishy.

 

[mr.electric]

Inspiring project. Very consistent testing and implementation.

 

[boostjuice]

My, you are a clever chimp Justin!

 

[tfahrner]

I can't wait to try it -- inspiring! Plugged you on our Facebook page: https://www.facebook.com/clevercycles/posts/10151557658029826

 

[Rassy]

That's a real slick setup. I Like mid-drives set up with the motor behind the crank, helping you pull on the chain and working through the gears. Of course I'm prejudice, since both of my trikes have ended up that way. I couldn't lower my motor into the frame as much, so I used more idlers to manage the chain.

I've never tried a pas type of system so don't have any feeling for it.

+1 on the Terracycle idlers, but I am using their return idlers (between the motor and pedals, so motor torque doesn't go through it) with no teeth, just a neopreme bed for the chain.

I know the StokeMonkey guys don't feel you need a crank that freewheels, but I set mine up that way to start with and didn't like it. Very happy now with a freewheeling crank.

 

[justin_le]

I've never tried a pas type of system so don't have any feeling for it.
I know the StokeMonkey guys don't feel you need a crank that freewheels, but I set mine up that way to start with and didn't like it. Very happy now with a freewheeling crank.

The PAS system puts a bit of a twist on this. Since the motor now automatically engages only when you are pedaling, you rarely wind up with a situation where the crank actually needs to freewheel. However, I'd be very uncomfortable running this system with a non-freewheeling crankset since an error offset in the torque sensor could cause a runaway positive feedback situation. The CA would think that there is torque and that you are pedalling so it drives the motor, the motor in turn keeps spinning the pedals and there would be no easy way to shut it off. But with the freewheeling chainring, you can always cut power to the system by stopping the pedals, since no cadence RPM = no assist power.

Anyways I just had a look at your tike built threads and you've also got a perfect mount location for a strain sensor on the arm that supports your forwards most idler, if you ever wanted to measure your pedal torque and setup a similar PAS system:

 

[beast775]

Great read,perhaps some 60 degree slight angle v bars at the bottom of the frame to the base of the hub motor might get rid of half of the squisheez felt while putting stress on pedal force? im interested in this post but need to draw in pencil what im thinking,cant use computer drawings :? .

 

[MattyCiii]

Daaaaaaaaam Justin!
You write great stories - just the perfect level of detail and pictures so all is crystal clear, and by the end I'm both educated and inspired.

Now I want to go out and build a clone!

 

[JennyB]

I've never tried a pas type of system so don't have any feeling for it.
I know the StokeMonkey guys don't feel you need a crank that freewheels, but I set mine up that way to start with and didn't like it. Very happy now with a freewheeling crank.

The PAS system puts a bit of a twist on this. Since the motor now automatically engages only when you are pedaling, you rarely wind up with a situation where the crank actually needs to freewheel. However, I'd be very uncomfortable running this system with a non-freewheeling crankset since an error offset in the torque sensor could cause a runaway positive feedback situation. The CA would think that there is torque and that you are pedalling so it drives the motor, the motor in turn keeps spinning the pedals and there would be no easy way to shut it off. But with the freewheeling chainring, you can always cut power to the system by stopping the pedals, since no cadence RPM = no assist power.

Freewheeling may be useful offload where sometimes you want the power without the pedal strike. Onroad, not so much! I would imagine that with a non-freewheeling chainset and a slow PAS the effect would be like riding a road bike with a sticky freewheel. The derailleur goes taut as the motor pulls all the slack. Not pleasant.

I wonder if that could be mitigated in a system where the motor sprocket was close to the chainwheel so the slack had nowhere to go? I'm thinking of a chain puller using a small geared hub that a roadie might fit to their best bike or favourite hack without having to replace any components. Motor, controller, cadence sensor (on idler sprocket), bottle battery, switches, meter - all in one neat package that fits in the frame triangle.

Just:

• Remove the front changer (with 1:1 assistance your 52 ring will climb like a 26)
  Break and extend the chain.
  Fix unit in place
  Rethread chain
  Optionally attach front changer cable to select assistance on/off
  Program settings
  Done!

8)

 

[tfahrner]

So Justin, while I definitely understand the appeal of a pedal-sensing approach, and can hardly wait to try this implementation, I wonder whether you think a KISS variant without PAS, torque sensing, or specially-machined freewheeling crank would be viable as a product. Throttle only. Could have a much bigger ring than the Patterson for a more appealing top end. Probably several hundred dollars cheaper too? Are you running an auxiliary throttle to override the sensing system on demand?

 

[justin_le]

So Justin, while I definitely understand the appeal of a pedal-sensing approach, and can hardly wait to try this implementation, I wonder whether you think a KISS variant without PAS, torque sensing, or specially-machined freewheeling crank would be viable as a product. Throttle only.

There's KISS from a mechanical/electrical component perspective, and KISS from a user experience perspective, and usually they are at odds with each other. If you have a right side drive under throttle control without a freewheeling crank, then you have to be super mindful as a user to pedal ahead of and during throttling, otherwise you have slack chain buildup on the top of the chain which allows the chain to bounce off the sprockets. In our earlier testing when we had a sloppy derailleur that didn't have enough tension to pull the chain through the patterson freewheel mechanism, this was happening all the time. If the chain was tight all around and had no derailleur/tensioner in the back, then it might stay taught enough to not cause problems in the slack zone, but that precludes conventional multispeed rear sprockets, and requires horizontal dropouts or an eccentric BB or additional idler etc. in order to take up the chain slack.

I think that the non-PAS non-freewheeling design is best implemented with the left side stoker drive as per the original stokemonkey. Then you don't have any of these complications and component restrictions and have pretty much every drivechain option still available.

Could have a much bigger ring than the Patterson for a more appealing top end.

I have my patterson chainring sitting on a rotary table right now about to machine an adapter for a larger tooth count! I'm hoping that with enough interest we could have FSA tool up a few more front chainring options. They are easy to snap off and on for replacement.

Are you running an auxiliary throttle to override the sensing system on demand?

Yup, it's there, but you'd be surprised at how disinclined you become to ever touch the throttle. Whenever circumstances are such that I need to reach for the throttle, it feels like some great annoyance. I dunno why the PAS suddenly shifts ones perspective this way, but it does. -Justin

 

[tfahrner]

There's KISS from a mechanical/electrical component perspective, and KISS from a user experience perspective, and usually they are at odds with each other. If you have a right side drive under throttle control without a freewheeling crank, then you have to be super mindful as a user to pedal ahead of and during throttling, otherwise you have slack chain buildup on the top of the chain which allows the chain to bounce off the sprockets. In our earlier testing when we had a sloppy derailleur that didn't have enough tension to pull the chain through the patterson freewheel mechanism, this was happening all the time. If the chain was tight all around and had no derailleur/tensioner in the back, then it might stay taught enough to not cause problems in the slack zone, but that precludes conventional multispeed rear sprockets, and requires horizontal dropouts or an eccentric BB or additional idler etc. in order to take up the chain slack.

I think that the non-PAS non-freewheeling design is best implemented with the left side stoker drive as per the original stokemonkey. Then you don't have any of these complications and component restrictions and have pretty much every drivechain option still available.

OK I see all that. And you wouldn't have the mushy crank feel that the right side drive chain routing entails, either.

Are you confident that your approach here will work as well with, say, an 11-34 cassette as it does with an IGH?

Yup, it's there, but you'd be surprised at how disinclined you become to ever touch the throttle. Whenever circumstances are such that I need to reach for the throttle, it feels like some great annoyance. I dunno why the PAS suddenly shifts ones perspective this way, but it does.

Before I began working on Stokemonkey, I rode a PAS bike to destruction. Again, I definitely appreciate the idea of PAS. What I hated about it was the implementation, the specifics of the sensing and related algorithms. It trained me to ride in a manner that was hell on my knees, and didn't let me contribute as much power as I was ready to. It felt like a third of the time it wasn't helping enough, a third just right, and a third helping too much. So I was 33% happy with it, and hungry for the control of a 100%-throttle-based approach. That's why I'm so excited that your approach here is so highly tweakable through CA settings, but that you think you might have overcome the desire for too many algorithm overrides anyway!

Wait, aren't you leaving today? Godspeed!

 

[Rassy]

Yup, it's there, but you'd be surprised at how disinclined you become to ever touch the throttle. Whenever circumstances are such that I need to reach for the throttle, it feels like some great annoyance. I dunno why the PAS suddenly shifts ones perspective this way, but it does. -Justin

Thanks, just when I thought I had my trikes just the way I want them, you dangle this in front of me!

I'll probably seek a bike out to try before doing any trike surgery. Between the auto shift NuVinci and a good PAS system I won't have anything left to do but pedal, brake, steer, and enjoy the ride.

 

[justin_le]

Could have a much bigger ring than the Patterson for a more appealing top end.

I have my patterson chainring sitting on a rotary table right now about to machine an adapter for a larger tooth count! I'm hoping that with enough interest we could have FSA tool up a few more front chainring options. They are easy to snap off and on for replacement.

So that seems to have worked out OK. I pulled a 42 tooth middle ring with 4 hole spider that seemed to line up well with the teeth on the 28 tooth ring that comes with the patterson crankset. Both were steel so easily welded together, and the 42 tooth ring is stamped in a way that makes the teeth fit on the same plane as the 28 tooth ring when stacked together this way:




Going to a larger chainring on the front (rather than smaller on the back) has several advantages. It means that for the same pedal cadence, the chain is moving faster and hence the motor spins at a higher RPM, resulting in better power and efficiency. It also means less wear on the individual teeth, and a less stress / stretching of the chain since it is able to deliver the same power with less tension at a higher chain speed.

A consequence of this change though is that the original motor at 36V now no longer spins fast enough, and once pedaling over ~60rpm the available power drops right off. So I'll be swapping it for the faster wind stokemonkey motor, and running at 48V as well. All in all, a very nice way to increase the motor power capability by 50% with no real downside.

 

[justin_le]

Are you confident that your approach here will work as well with, say, an 11-34 cassette as it does with an IGH?

I see no reason why it wouldn't work even better, just haven't tried it yet. The terracycle's idler would guide the chain in regardless of how the angle and the cassette hub would have the squishyness of the nuvinci, so overall that would help it seem stiffer.

Wait, aren't you leaving today? Godspeed!

Yes! Give or take a few days. I actually ran into some passport problems (you'd appreciate that :wink: ) and wasn't 100% sure that the bike and gear would be ready. But look what finally came through yesterday afternoon:


Anyways, now that I have this in hand, I can finally get at the Touring and Hauling portion of the subject line. As mentioned on this thread here, we've got an exhibitor spot at the San Fransisco Maker Faire in a couple weeks, and are hoping to tour down there from Vancouver on this edgerunner bike while hauling along all of our stuff for the show. That should be a fun adventure and one way to put some new prototype gear through its paces. Just need to get this bike 100% in the next day or two if we want to get there in time.

 

[justin_le]

Todd's maybe not going to appreciate this as much, but we'll also be putting a direct drive hub motor on the front of the bike in addition to the stokemonkey on the chain, so it's a dual motor setup with one mid-drive going through a nuvinci which will help efficiently climb hills, and one fixed hub which will provide regenerative braking and efficient power on the flats and gentler inclines.

This front hub motor is the fruition of some work that started many years ago but is only finally all coming together. The goal was to take a standard issue 205mm 23 pole pair direct drive hub motor (nine continent, Crystalyte H series, etc) and see just how much weigh we could shave off without any real performance detriment, and in the meantime also improve the overall bicycle compatibility. We got a number of wound stators supplied from Nine Continent, but in the end endeavored to make everything else custom ourselves. The whole motor development process deserves a thread of its own. Here's part of the parts wall:


All the hoops at the top are steel magnet backing rings for the rotor, which we had machined in a range of thicknesses from 2.0 to 4.0 mm to see just how thin the steel can go without much too much flux leakage or performance reduction. On the bottom right are some of the finished side covers which we had tooled and die cast, while on the bottom left are various hand machined and CNC machined prototype motors. Not visible here are some stacks of rare earth magnets from N35 to N45 and, both wide enough for a full magnet fill or narrower so that there is a gap between them.

Interestingly, we ended up settling on just mid-range magnet strength. Although it was possible to get a bit more torque from the motor with the highest rated magnets and thicker steel rings, the corresponding cogging drag torque increased in by a much larger amount. So like a 5-10% increase in the torque output capability resulted in like a 50% increase in the cogging drag, which I found an unacceptable tradeoff.

Here we are about to start the process of gluing a set of magnets into the steel ring. The aluminum piece that looks like a crown is the magnet alignment jig that keeps a uniform gap between each magnet.


The gap between magnets not only makes the motor lighter with no real performance penalty, in our case it also helped solve a vexing problem of how to bolt the side cover plates in place when the actual steel rotor ring is far too thin to take a screw through the edge. Instead, we has some long and slender bolts made up that fit through one cover, right between the magnet gap, and then out the other side cover. During the assembly of the side cover plates I use set of 13g spokes through the eventual bolt holes to ensure that the side covers and rotor ring are all lined up.
View attachment 1

The finished motor has the same basic performance as a Nine Continent 2806, but weights only 3.85 kg, compared to 5.5-6kg for the equivalent from 9C (depending on if it's a model with a steel or aluminum spoke flange).


The lighter weight is great for a lot of reasons, but it has one downside that there is less metal mass present to absorb heat energy and delay the time it takes the windings to overheat. That's one of the reasons that I was interested in performing the thermal motor characterizations earlier this month, to see if a cooling strategy might be prudent for this trip. But for now I just have a thermistor inside the motor and will monitor that with the CA and implement a rollback if it gets too hot. And of course I have the stokemonkey motor to fall back on.

You can also see that the motor has a hollow aluminum axle. This is 20mm ID and allows the motor to be used with thru axle suspension forks, or with stub axles for supporting from a single side only. For normal forks, we just stick in a small insert to each end that makes it quick release compatible. And the hexagonal shape on the end of the axle is for an integrated torque arm. So there will be no spinout forces present on the fork dropouts.

Motor as shown on the scale is where it is at right now. Next step is to lace it up into a rim and get it on the front of the bike!

 

[tfahrner]

A consequence of this change though is that the original motor at 36V now no longer spins fast enough, and once pedaling over ~60rpm the available power drops right off. So I'll be swapping it for the faster wind stokemonkey motor, and running at 48V as well. All in all, a very nice way to increase the motor power capability by 50% with no real downside.

Well, except now you're talking about a kit that's so fun, er, fast and powerful as to present insurance/liability and maybe some ethics problems in filthy commerce, that requires more custom parts/fabrication, a generally more expensive battery, and a motor currently in shorter supply. Plus the "more efficient" part becomes sort of moot when the aerodynamics of >30kmh speed on a non-streamlined bike mean a much worse Wh/k, hence range, regardless of electromechanical efficiency optimization. But you knew that. So do you think a 36V variant using a standard Patterson and the higher winding is just not gonna hunt? I'll have to take your word for it since you'll be passing through town on more of a crotch rocket. With 2 motors, Doc Brown.

Re dérailleur in back instead of N360, using a standard Patterson with an 11T small cog at the wheel, I calculate a wimpy 75" high gear, while with a 16T on N360 the high is a decent 95". Also with a derailleur, the movement of the chain up and down the cog changes the vector relative to the pivot point, meaning it could mess with proper torque sensing.

 

[spinningmagnets]

You can also see that the motor has a hollow aluminum axle

Is the central stator core aluminum instead of the stamped steel I recall on the 9C? (AL core also found on the MAC)? If the core is aluminum and it attaches to the hollow Aluminum shaft, it sounds like you're creating a heat path from the stator core to the tube-axle without introducing outside air circulation (and the dust that could hurt bearings)...

edit: I really like the tube-mounted torque-arm with large hex interface, clearly that allows a full-threaded round steel shaft for the lugnuts to hold onto the drop-outs...

 

[justin_le]

All in all, a very nice way to increase the motor power capability by 50% with no real downside.

Well, except now you're talking about a kit that's so fun, er, fast and powerful as to present insurance/liability and maybe some ethics problems in filthy commerce, that requires more custom parts/fabrication, a generally more expensive battery, and a motor currently in shorter supply.

Sorry, I should have clarified, no real downsides to me
For general lower speed load hauling the initial configuration will be just fine for sure, and has a lot of production advantages. I'll be bringing a spare patterson crankset and can easily revert to that setup for you to test out. But in the meantime, each day of departure delay means the rig needs to move that much faster to get to the faire on time, so give me speed and I'll pay the kWhrs!

Re dérailleur in back instead of N360, using a standard Patterson with an 11T small cog at the wheel, I calculate a wimpy 75" high gear, while with a 16T on N360 the high is a decent 95".

Yeah, for this I was thinking more in terms of the Yuba Mundo's with their 26" wheels which is what we'll be trying it on first. For it to work on the edgerunner, we'd need a larger patterson ring.

Also with a derailleur, the movement of the chain up and down the cog changes the vector relative to the pivot point, meaning it could mess with proper torque sensing.

Yup, but the amount is very slight that it won't have much material effect on the measurements. I think it would be on the order of 5-10 watts that the motor tension could add or subtract with the resulting misalignment from one gear end to the other, and that is still better than the offset errors in the THUN sensor. I'll look forwards to trying it out in any case.

Also, we did another 2 sets of the strain sensing clampset over the weekend with some minor tweaks to improve the overall geometry and give more clearance between the motor and the bottom tubes of the edgerunner frame. In this one, we've also got the strain sensor amplifier circuitry in the machined cavity and shown here just before getting fully potted in rubber:


So that means I'll have an extra set to give you to play with.

-Justin

 

[Drunkskunk]

20mm through axle? I could stick one of those on the front of my Killer Tomato and have 2 wheel drive with the 5404 pushing and....
No. Just no. I must not give into temptation. I must not go over to the dark side.

 

[liveforphysics]

You raise the bar on ebike design my friend. I am drooling and thrilled at the same time watching your thread.

When can I buy one of these my friend? I am so loving how the stator hub bolts to the axle too. Such an awesome design!