RazErBlades: Completion and Deploy

I am satisfied.

To me, there’s no more magic in ultra-small mobility devices. Building the Death/RazEr/murderblades has shown that a handful of brushless DC motors and a lithium polymer battery is fully capable of pushing a passenger around at 10-12mph, which is a reasonable sidewalk speed. Extreme stealth is also possible –  RazEr was a miracle of packaging, but the skates beat it out by far in terms of volumetric component density. Additionally, I’m very satisfied with the way that the Skatroller turned out in terms of user interface – so much, in fact, that I might start making more vehicle throttles wireless.

Now that Otakon is over with, I’m going to start collecting more test data and video with other people who can actually wield the ‘blades and not look like dumbasses, unlike me. Anyways, here are the events of the last week or so.

After I left the remaining skatemotors on a shelf to set, I began completing the left motor driver board. It was fundamentally “done” weeks ago, but I never added motor connectors because of the lack of motors. On this board, I elected to discontinue using side-bullets and on-board mounted sensor connector. It was too difficult to maneuver the cabling in the tight space, and the large power wiring directly pulling on the board traces worried me. So this board uses sensor and power pigtails that extend outside the aluminum frame.

I put together the sensor pigtails using the same headers and 5 (originally 6) pin cable. A few globs of Goop sealed the electrical connection.

I had to flip the direction control pin on the left board also, since the motors were mounted oppositely to the right side and therefore wanted to run the left skate backwards. Unfortunately, there wasn’t an easy was to do this since the DIR pins were connected to logic ground through a circuitous (hurrr) trace that also hit several other pins.

Solution? Cut off the mating contacts of the signal header. I then hardwired the DIR pin to the DECs’ onboard 5 volt rail using 30 gauge wrapping wire.

After the board was assembled and installed, I gave both skate units a full charge. It was the first time the left skate had been charged period, since it was never used before this.

skatroller revisited

It was time to get started on updating Skatroller. For this version, I wanted to have control over the DECs’ output enable lines. The output causes the motor to brake if the commanded speed is less than the current motor speed. Now, the DECs aren’t actually regenerative controllers  or anything, but enabling the outputs while commanding zero speed acts as a dynamic brake. It’s just another added convenience if I want to limit speed, as this kind of braking will never actually bring the motors to a full stop.

I decided to use another sensor input instead of trying to calculate or design ranges of inputs for one FSR. Above is my spare FSR mounted in the original “wrist forward” position. The whole thing now acts sort of like a large rocker switch.

Additionally, I’ve elected to pitch an Arduino Nano on the board just out of convenience. It’s easier to read in two sensors and do whatever I want with the data using a microcontroller and some software (as much as I hate to admit it).

…but first I had to pre-cut all the fusible links in the Lilypad breadboard because they are the most annoying things ever.

I scolton briefly thought about connecting all the circular perimeter holes to alternating terminals of a METALPAKKK such that the sheer short circuit current would blow all the inter-pad links, but the setup for that would have taken longer than just going through them with a knife.

To facilitate alignment, I used a machine parallel as a straightedge to guide the +1 Exacto Knife of Trace Gouging.

I got a Lipower board from Sparkfun as a more efficient means of powering the Arduino and XBee stack. The Lipower takes the 3 to 4 volts of a single lithium polymer cell and outputs 5 volts. The Arduino uses the 5 volts directly and the XBee board internally down-coverts it back to 3.3 volts.

For simplicity, the Lipower board was double-sided-taped to the 900mAh lipoly cell through a layer of heavy heatshrink. On the other side of the battery, I stuck a piece of Velcro. The battery never sees any loads on its pouch surface, which is what I want.

I found these board standoff headers to connect the XBee board with the Arduino & glue circuitry below.

The arduino board  houses connectors for the 2 FSRs and the battery. It also has a small input RC filter for each FSR and another RC filter on a PWM pin of the Arduino to bring the square wave PWM back to an analog voltage. This is all shoved between the two header rows that the Arduino sits on.

Here’s the back of the signal processing board, showing my usual tactic of point-to-pointing with 30 gauge wrapping wire.

After verifying that the circuit worked, I coated the whole bottom side in a healthy layer of hot glue. After the glue set, the self-leveled surface got a chunk of Velcro attached to it so I can mount it to the wristpad.

And now here’s the whole rig mounted to the wristpad once more! No obnoxious LEDs this time around – the only indicator of power is the three little red LEDs, one on the Lipower board, the Xbee board, and the Arduino. The assembly is almost excessively stealthy, especially when hidden under the cavernous sleeve of my modified fuzz jacket.

The carrying case for the ‘blades was a recommissioned Facebook swag bag obtained at a career fair many moons ago. The skates sort of fit in perfectly side-to-side, and the wristpads filled the rest of the space. So here’s the final shot of them ready to travel.

I ended up not having time to replace the wheels with the harder compound K2 wheels, but that will happen later.


The Otakon deploy wasn’t particularly exciting since the convention center specifically banned wheeled vehicles inside and there was no planned photo shoot for the Air Gear series anyway. There wasn’t a good reason for me to just stay outside the convention center and orbit around on the streets more than once – which is what I ended up doing. So, the ‘blades were under power for maybe 10 or 15 minutes, then stored  back inside the rest of the time. They’re certainly stealthy enough such that I didn’t rouse too much attention. A few people “got it” because they were familiar with Air Gear (though not hardcore fans). Overall, in terms of outward appearance, the ‘blades aren’t very impressive. The engineering is mostly hidden inside.

Pics or it didn’t happen? I’m sure someone had pictures, but I kept my camera with me at all times, so I don’t have any voluntary pics of the whole setup. Overall, what the con showed me is that I’m too lazy to actually cosplay and too focused on building props and gadgetry that nobody at a pop culture convention or without at least some familiarity with engineering would care about.

But, as I said originally, now that the RazErBlades are fully functional, reliable, and require no particular explanation of quirks to operate, I’m going to lend them out to friends and peers who have a better handle on inline skating and possibly put up a few more demo videos later on. For now, however, I’m going to call this version finished. It’s always open to improvements in packaging, power, and usability, but it’s August already and Dragon*Con and my annual robot party are fast approaching.

Soo… what did I end up inventing? Air Gear or Jet Set Radio?

Hub Motors on Everything, Part I: The RazErBlades Contingency Plan

Reason #1 to not engineer things at 5 in the morning: You think that putting hub motors on inline skates is actually a good idea.

Reason #2 to not engineer things at 5 in the morning: You forget how many magnets each hub motor needs, and like a total dumbass, only order half the number you need.

Well, guess who is guilty on both counts. Late in June, I put in a reorder of the custom arc-segment magnets that I got for the first two skatemotors from SuperduperfabulousMagnetGeorge. Each motor takes 7 “north” magnets and 7 “south” magnets, where the designations just describe which pole is on the inside face of the arc. So, I ordered 16 magnets in total, 8N and 8S, so I have 2 spares in case I break something.

wait, what?

If you’re keeping track, the left RazErBlade has 2 motors. That means I only ordered enough magnets to make 3 wheel drive skates. By the time I discovered this minor oversight, it was already two weeks ago, so I hurriedly put in an appended order. The custom magnet service has a minimum turnaround time of 3 weeks, and there was (at that point) 3 weeks left until Otakon. Now there is one, and I’ve been informed that my appended order will ship next Thursday.

You know, when I leave for the con. Clearly, this was not going to work at all.

And so I deployed the backup, pain-in-the-ass-but-it-would-get-them-moving plan, and dropped some more dimes on a set of rectangular magnets for the left side motors.

Using GoBrushless’ excellent rotor magnet placement calculator, I discovered that SMG’s stock 20mm x 5mm x 2mm magnets were a good fit for the can if I doubled them up side-by-side. They would require some spacing games, but I was used to playing that with RazEr anyway.

And here they are! I got the shipment notice 90 minutes after I entered my order – that’s essentially on par with McFaster-Carr. Due to the miracle of express shipping, they were in my mailbox the day after.

I printed out the generated magnet placer graphic to use as an epoxying guide. Step one is to put in the “keystone” magnets, the first 14 of alternating poles. Trying to jiggle too many magnets next to eachother, I have found, always results in unsatisfactory placement and a dent or two in the workbench from my forehead.

After the first 14 magnets set in each motor, I crammed their complements in next to them. Putting 2 magnets of the same pole orientation next to eachother means they tend to force themselves apart. To combat this, I wedged little plastic spacers into the horizontal gaps as I placed each new magnet.

I made the spacers using a handy-dandy sheet metal notcher tool and some strips of thin unknown plastic.

new rimz

Usage reports from friends who actually are good at skating have told me that the 72-78A durometer scooter wheels are too soft to perform most skating maneuvers effectively, such as sliding or otherwise breaking traction.  So I wasn’t totally crazy when I thought the ‘blades handled like bricks – they actually do!

Solution: Hop online and find some harder compound wheels. I decide to upgrade one step and go to 85A wheels. Finding 100mm wheels was actually pretty difficult, since the vast majority of inlines use smaller wheels such as 72 or 80mm. Then came the issue of filtering those 100mm wheels to find the ones which can be hollowed out to 2.5 inches on the inner diameter, which was a requirement not met by most.

I finally located these K2 wheels on skates.com and had a pack rush-shipped (By this point, I think express shipping has almost matched the cost of parts for this project).

These wheels have a glossy, blank white tread and a black plastic core. Very plain, yet functional, and I was impressed by the quality and finish.

No matter, they’re going on the lathe NOW. I made a quick mandrel to grip them by their bores, since the urethane was actually too slippery to grip with the outer diameter chuck jaws. A flying pass with a boring bar severed the spokes from the outer part of the rim.

Well, mostly. The bar broke through at a place that was not the outer diameter of cut, so now I have these spoke stubs to contend with. When the shops with bigger machines open again, I’ll just knock those out by virtue of gripping the wheel’s OD in a bigger lathe.

final preps

This weekend (and extending into next week, likely, due to laziness) I plan to re-engineer the Skatroller to allow for manual activation of the DEC modules’ electric braking. My spare force-sensing resistor will be hidden under the original wrist-forward trigger point such that it will detect two possible states – willful activation of braking and the palms-open-oh-shit-i-am-about-to-die faceplant mitigation position.

Which, mind you, may possibly be mutually coupled.

I’m also going to switch the analog op-amp circuit to an Arduino Nano based solution, because it’s much easier to throw some if() statements at the two FSRs than try to play the AND/OR/MAYBE game with logic gates and linear components.

Did I just advocate the use of software? Doom.

Past that, I’m going to refine the power system of the Skatroller to use a single lithium polymer cell with a Lilypad boost converter unit. This ought to net me much more efficiency and subsequently battery life, as well as avoid stressing out the XBee by running non-spec voltages.

Non-straightjacketed Agito, theoretically coming to an Otakon near you. Because I'm totally going to be able to stay upright while 95% blind, without the use of my arms, and with motors attached to my legs. Yeah.

Deathblades/RazErblades: On-the-ground Testing!

i’m still alive

A few days ago, I took the RazErblades into Boston proper in an attempt to gather real “in-city” usage data. Needless to say, as someone experienced in mounting wheels to other things, but a relative amateur to the concept of strapping wheels to yourself, I just barely survived the adventure, but emerged with valuable data and a laundry list of practical improvements to the controls.

Before that, though, I had to actually finish the new frames.

Last time, I showed the frames in their mostly complete state. I was able to make the last of the mounting blocks, so here’s a closeup of the mounting arrangement.

I discovered after trying to crank the mounting bolt with all my might that the thread I thought was metric (M6 x 1) was actually 1/4″-28.

How silly.

So, as the world’s worst and laziest machinist, I just drove a 1/4-28 tap right over the existing M6 threads. Because 28TPI and 25.4TPI (1mm pitch) are almost equal, there was a cool thread harmonic thing going on in the distal parts of the through-hole.

Transferring componentry over to the new frames…

The cavity being 1mm wider this time, the batteries slipped right into place. They seem to be resting on the horizontal T-nuts, so I added a layer of foam rubber between them and the frame for some modest level of shock protection.

A bit of stuffing later, and the ‘blades are ready again. I shouldn’t have problems with the whole thing falling off any more.

ground test

Boston is a city of hills.

And traffic, people, potholes, curbs, and don’t forget the broken-ass narrow sidewalks and whole stretches of ancient unmaintained cobblestone.

If I was going to die anywhere while testing motorized skates, it might as well be in Boston, firmly embedded into the front bumper of a T bus.

When a crew of Putzen decided to go shopping in preparation for Otakon, I tagged along – partially because I also needed material to work with, but mostly because I wanted to get in some IRL off-campus, urban testing of the ‘blades.

The test site was Newbury Street, a famed shopping district of Boston that combines literally all of the aforementioned challenges. It was 90 degrees and the day before Independence Day, so people were everywhere. The start of the street is a long downslope with alleyway curb cuts, seemingly random switches between asphalt and sidewalk concrete, and worst of all, brick pavement. Did I mention there were way too many people?

All my hallway cruising paid off, and there was not a single crash or ditch involved.

Well, technically the test site was along the MIT coastline and across the Harvard Bridge, which presented a lesser challenge to help me tune my own stabilizing loop. I only ended up skating one way – the group had dinner at a little Japanese noodle place.

Bad mistake – afterwards, I figured I could barely keep static balance, nevermind dynamic. So I called it a day then, and walked back.

I wasn’t in the mood to take video or to have video taken, so no video of this test! Sad, I know.


The distance: 1.97 miles

The battery usage, as determined by my charger afterwards: 0.52 amp-hours

I estimate that I was using the electric assist around 25% of the time – mostly because the region was hilly enough such that I didn’t think the motor would actually affect my mobility. Most of the traveled distance was cruising, controlling speed, or just straight out skating like they weren’t motorized or something. New concept, I know.

With this conservative estimate, the “fuel mileage” of the ‘blades is (.52Ah * 22.2v nominal ) / (25% of 1.97 miles) = approximately 23 watt-hours per mile. For reference, RazEr itself seems to hit around 25 Wh/mi.

I suspect that with more electric assist and less me-assist, this number will drop precipitously.

Some observations, lessons, and anticipated changes:

  • Speed control down a hill is a horrible bitch. I could only learn so much from watching Youtube videos, and having to do it myself is no fun… especially when dodging shoppers and tourists.
    • Consequently, I’m going to enable the motor braking on the DEC modules. All this entails is keeping the enable line high such that the controller tries to hold zero speed, fighting my motion.
    • To do this, I’ll have to add another sensor or two to the wristpad controller such that it can distinguish between when I want to accelerate, coast, or brake. I’ll probably make the brake position the old throttle position, which is wrist down, since not only does that trigger when I make the corresponding motion, but as mentioned in the post, also triggers in the palm-open position. You know, like catching an impending faceplant.
  • Electric mode is wonderful for crowd-mingling. I found myself not having enough space to actually keep up the kicking motion to move forward. This was when the electric mode shined – slow, walking pace mingling in close quarters. I definitely received a few weird stares for seemingly moving with no effort while next to someone for 100 or more feet.
  • Electric mode works when skating normally. The sensored motor control means that the controller never has guess where the motor is – it always knows, even if the speed is varying greatly. I haven’t definitively tested if actually using the motors while kicking contributes to speed that much, but it’s something I’d like to get to.
  • Cobblestone and brick pavement is a travesty to anything without pneumatic tires

    • Especially on a downhill section like the Boston end of the bridge.

The motors showing some battle scars from stopping and being accidentally run into a curbside or three!

So what about the new frames? They held up great. I had no issues with loose hardware or flexiness, so the T-nut design proves itself.

the day after

…was July 4th.

Every year, the giant barge of fireworks parks literally right in front of Killian Court, on the river, so MIT gets the best possible view without even trying. The problem comes when everyone tries to crowd the riverfront to watch – and this is hundreds of thousands of people on both sides of the river easily. Roads get closed and blocked off, and the police roll out in full force to make sure nobody drunkenly riots (or if they do, it’s not too outrageous).

Wait, did someone say “roads get closed”?

The three major streets that define the “MITmuda Triangle” were closed down to road traffic – limited to bikes, pedestrians, and…

…small electric vehicles!

We basically rolled out as many small EVs as we could find – the BWD, RazErblades, RazEr itself (before a mysterious controller malfunction grounded it), and the secret MITERS electric pocket bike.

I also unchained LOLrioKart from the ceiling and took it out. As usual, it was a total attention fairy, and I’m sure there are now a few hundreds of pictures of it floating around on Facebook with people making strange faces in them. The rollout was, all things considered, a success, and we even managed to squeeze in a trip to 7-eleven (in full EV regalia) for slurpies, which constitutes the strangest mutation of cruising I have personally witnessed.

We need more small EVs. I get hits from 18.*.*.* all the time, so I KNOW you MIT people are actually reading this site. BUILD MORE SMALL EVs. Swarm and destroy!

Here’s a short video of Shane taking the ‘blades out for a spin near the south end of campus. This is still with 50% motor capacity, by the way.

Once I’m able to get the left side motors magneted up, I’ll gather some people to take real running video. Hopefully, it will be interesting with all 4 motors, and unless something really dumb happens (like I blow all the controllers or… like… die or something), the thinly-veiled-ground-test at Otakon is on schedule.

Deathblades/RazErBlades Structural Redux

Alright, so the last test of the RazErblades ended in a mild disappointment as my shoddy metallurgy brittle-failed under load. Personally, I don’t blame it – I might as well have put the metal together with center punch dots the way I speed-brazed the thing together. No matter – a chance to rebuild the frame is also a chance to optimize the frame geometry for better component placement. As I was entering a dead week of waiting for parts, I took a few hours to think about the new frame design.

I decided to go back to using my classic T-nuts and slots method of building, just because everything goes together so quickly using it. For well-constrained structures with lots of fasteners, it’s also fairly strong.

This is the updated rendering of the frame after a few hours of speedCADing, with few structural T-nuts designed in. Most of the changes are to minor aspects of the design that will have a substantial impact on the way the electronics sit inside the frame and the integrity of the motor mounting.

  • The bottom of the frame is now flat. Previously, there was a slight (less than 5 degree) slope downwards to the wheels. I’m not really sure why this was ever put in, but it impeded the sanding of the brazed frame surfaces and made machining setups much harder.
  • It also screwed me out of about 1/4″ of “cavity space”. This was 1/4″ that could have been (will now be) used to make sure I wasn’t stepping on the radios.
  • I changed the angle of the motor shaft slot such that it wasn’t directly tangential to the load path the wheel sees. There’s now a bit of downward “hook” to lessen the risk of ditching a whole wheel.
  • The boot mounting plate has become a mounting block. Instead of using the Weird Shady Nut, the center hole of 1/2″ worth of aluminum will be threaded to accept the M6 x 1 screw. This new mounting interface is two 1/4″ aluminum plates screwed to eachother, the bottom having a solid slot interface with the frame’s side plates. In other words, the mounting should no longer be the weak point.
  • The two notches in the middle don’t serve any useful purpose (wire passing, perhaps?) but I thought it made the whole rig look more like its namesake.

Now with more T-nuts! As mentioned above, the internal cavity is much taller. In fact, it’s almost twice as tall as the 6S battery now.

Oh, yes, one more thing. I’ve made the cavity 1 millimeter wider to let the battery slip in without requiring… you know, additional machining. The ends of the wheel shafts will be shimmed slightly to account for this.

There is now certainly the possibility of switching to an even bigger battery – 22.2v and 2.6Ah is already quite substantial to stuff inside, but I’ve been shopping for 3AH and larger packs on Hobbyking. Unfortunately, they all seem to be much wider than the 40-42mm range, which isn’t helpful. Laying the cells sideways would increase my “width capacity” to 46mm or thereabouts, so that’s a possible solution.

design for manufacturing

A day and sheet of 1/8″ and 1/4″ aluminum later.

Not everyone has abrasive waterjets on call at the tap of an ID card. Fortunately, optimizing of designs for waterjet machining (for myself, mostly out of laziness) is complemented by new online manufacturing shops that will take your flat parts drawings and lets you purchase the parts cut in different materials.  It’s one version of the PCB-fab-but-for-MechEs that I want to see more of, and believe will be key in aiding the United States’ manufacturing industry while encouraging the DIY technical arts, the TE part of STEM, and makerdom as a whole.

But enough soapboxing. I keep plugging Simon’s abrasive waterjet (and LASER!) service Big Blue Saw because he and I go way back and are homies. Isn’t that right, Simon?

The way I fasten all this stuff together is with square machine nuts. Überclocker and Cold Arbor are both chock full of 6-32 and 1/4″-20 square nuts.

1/8″ thick material is a difficult one to size up square nuts for. Even a 6-32 nut is far too wide to be useful – it can be done, but it’s excessive because of the amount of overhang at 5/16″ wide each. A 4-40 square nut is generally 1/4″ wide, which is far more palatable. But there was a problem: McMaster, my go-to guys for hardware, don’t stock any kind of 4-40 square nut. I was concerned, because that meant I was potentially spec’ing out hardware which didn’t exist. I mean, if McMaster doesn’t carry it, why should it exist?

A search through the Intergoogles took me to Amazon of all places (seriously, did you know they have a whole industrial components division?), but unfortunately all the 4-40 square nuts came in packs of 5000.

That’s alot of nuts. In fact, it’s 7 and a half pounds of nuts. I decided to press forward and bought a pack, figuring I would never, ever need to buy these things again.

hey mcmaster, you should carry 4-40 square nuts in packs of smaller than 5000.

The 4-40 square nuts are standardized to be 0.25″ wide and0.098″ thick, for reference.

And here are the new plates all loaded with square nuts, ready for fastening. All the screws are standardized to be 4-40 x 3/8″ long socket head cap screws.

And the “version 2″ frames assembled, but without hardware. I still need to modify the top 1/4” aluminum mounting blocks slightly – they need clearance holes for 4-40 cap screws to go all the way through to the second layer. Both layers also need the center hole tapped for a M6 x 1 thread.

Hopefully, my custom magnets from Super(custom)MagnetGeorge will arrive soon, so I can get to some 4 wheel drive testing. Adding 100% more motor will make a world of difference in terms of handling.