Kiwi-drive omni-wheeled go-kart, basic calculations

Hanna Lin and I are going to make kiwikart, a kiwi-drive go-kart. We are going to make two versions, a brushed motor 80-20 version and a brushless motor bamboo version.

Research so far:
Previous omni kart:

Previous bamboo go-kart:

Now for motor and gear ratio (sprocket) selections.

We chose 3x 6” wheels and 3x 4” wheels (for the cheaper brushed 80-20 version). We chose aluminum for the first one because it seemed more durable, and plastic for the latter because the Al version only holds 80lbs per wheel while the plastic version holds 120lbs per wheel. We picked kiwi drive instead of four-wheel drive because it’s much cheaper. (each brushless motor+wheel+controller assembly is upwards of 200 dollars). (plastic 4” dualie) (aluminum 4” dual omni wheel)

We also were trying to decide between a 3:1 and 4:1 ratio. Later we found out that for the 4” wheels at least we can’t get a 4:1 ratio because the sprocket is bigger than the wheels.

Max acceleration calculations:

[K_t = (K_v* frac{2 pi}{60})^{-1} [Nm/A]]
 where $K_t = 236 ; rpm/V$ from hobbyking motor specifications (
[K_t = (236* frac{2 pi}{60})^{-1} = 0.0405 ; Nm/A]
[tau_{max} = K_t I_{max}$ where $I_{max} = 50 ;A]
($I_max$ comes from the controller we selected,
Additionally, we know that

[tau_{max} = F*r = m*a*r] where $r$ is $radius$
[accel = frac{tau_{max}}{mr} = frac{2.03; Nm}{3 ; inches 200 ; lbs}]
where the radius of the wheel is 3” and the cart we estimate to be about 60 lbs with a rider weight of 140 lbs. Plugging into wolfram alpha or google we get
[ accel = frac{2.03}{0.076*90.7} = 0.29 m/sec/sec]

Finally, we account for the fact that, in a kiwi drive, while going forward we are only using 2 of 3 motors and the forward direction of motor force is only 70% ($sin 60 = sqrt{3}/2$). We also need to take the 4:1 or 3:1 gear ratio into account.

$accel = raw-accel * 2 * 0.7 * 4 = 1.6 ; m/sec/sec$ or about $1.6 / 9.8 = 0.16 ; g’s$ of acceleration. In other words, 3.67 mi/hr/sec or going from 0 to 60 mph in 16 seconds. Brisk but not award-winning, but should feel plenty fast on a low-to-the-ground go-kart.

Max speed calculations:
[Omega_{motor} = V_{sys} K_v [rpm]]
The motor can handle 37 volts so we picked 36 volts for the system voltage. As before, the $K_v$ is 236 rpm/V.
[Omega_{motor} = 36 * 236 = 8496 rpm = 141.6 ;rev/sec]
[v_{ground} = frac{141.6 rev}{sec}* frac{2 pi r}{rev} * frac {1}{K_{gear}} * 0.7 * 0.7]
where the 0.7 is for the forward efficiency again. We do another factor of 0.7 for cruising speed instead of no-load speed.
[v_{ground} = frac{141.6 rev}{sec}* frac{2 pi * 0.076}{rev} * frac {1}{4} * 0.49 = 17.5 ; mph]


That’s it so far. We’re using the same motors as chibikart, which are the lower $K_v$ in the SK3 50 mm class at 236 $K_v$. The wheels as mentioned before (100 dollars each for 6” wheels), #25 chain, and sprockets to be determined. We’re using kelly controllers KBS series that can handle 50 A 36V at $150 each. Whew. Okay, a lot more information to come, but a quick braindump for tonight.

research: best practices in online education [WIP]

1 to 2 hour lectures may be better for those driving to work

Yet as consummate professionals, all of you have conducted copious research, applying sound methods and appropriate metrics, which demonstrates the many positive academic benefits of high-quality technology-enhanced education for all learners. And your efforts have produced tremendous innovation in both technology development and the learning sciences.
The problem is that, for the most part, we are sharing that research with each other at conferences and in publications created specifically to advance the e-learning field. So without verifiable data to consider, our more traditional colleagues are still making instructional decisions based on personal experience or professional bias, political expediency, or just because everyone else is doing it.
To read! many links.

Firstly there is the issue of digital mimicry.  The Coursera platform, alongside rival Stanford start-up Udacity and the non-profit venture ‘edX’ from Harvard and MIT, currently hosts courses that are broadly conservative in terms of online educational practices.  All of these MOOC platforms appear to justify their status by promoting curricula that are equivalent to campus-based courses, with a strong focus on content delivery and an emphasis on the rigor and formality of their assessment methods.  However, some of the most interesting and innovative practices in online education have emerged by challenging these very ideas; loosening institutional control of learning outcomes and assessment criteria, shifting from a focus on content delivery to a foregrounding of process, community and learning networks, and working with more exploratory assessment methods – digital and multimodal assignments, peer assessment and group assignments, for example.


So we are keen to avoid both the over-celebratory fetishizing of the teacher associated with some MOOCs, and the tendency to see the technology as allowing us to write the teacher out of the equation altogether. We want to explore how a MOOC pedagogy might work with a construction of the teacher that has an immediacy that can succeed at scale.

The possibility of the ‘online version’ is overstated. The best online courses are born digital.

‘Best practice’ is a totalising term blind to context – there are many ways to get it right.

Vending Machine Update

As per my post on open-manufacturing,
“I’m part of a student-run shop, MITERS, at my university, and we considered stocking things like breadboards and arduinos (and other emergency project materials), but that’s really not in our budget as a club. We sort of have a “if the person who knows where the secret cache and can deal with money” system set up, but I wanted to make a countertop vending machine. Used vending machines on ebay are hundreds of dollars and gigantic (we’re tight on space). I decided that the relative portability/stealibility would be fine at MITERS since everyone that comes in is a student.”
From January 13th 2013:

i “finished” my vending machine prototype from mas.863 all in one go, it took ~a day to find all of my supplies (i swear, i had to think like a squirrel to figure out where i buried all my microcontrollers). by finish i mean only that the dispensing aspect works without being plugged into a computer. lasercut wood, buttons and spring off the internet, servos modded to be continuous rotation, arduino uno, and a wall wart. toward the end you can see some of the issues: i made the spacings large to accommodate larger items but… it doesn’t actually work that way, etc.

Here is a video of it failing:

One thing I did learn is that wall warts _will_ reset your arduino uno continuously if you are trying to draw current from them.

a 9V wall adapter: 5V to 23V…

I ended up adding a giant (1000+ uF) cap to smooth out the huge voltage swing, and that solved my stranger errors. I made a short video on this topic.
“300uF (not shown) may not solve your external power supply issue, but 3300uF will! 
In other words, if your arduino sketch runs on usb power but not external power (i.e. crappy wallwart), add a large smoothing cap across the rails and that might fix the issue x.x
Embarrassingly, I had this issue with a side project at my internship last summer and didn’t know how to fix it.”

In progress pics:

drill out hole to inner diameter

I actually got some that lit up, but only at 24V or something silly

Yea, even at 9V (I think I had a 9V wall wart) they are just barely lit up.
The buttons were from my trip to China, but nowadays you can get arcade buttons from places like sparkfun and adafruit.
The vending machine is set to be dismantled. Maybe one day I will build one from scratch.
The status of the money accepting problem, as per my post on open-manufacturing as well:
“I made this prototype (uh… keep in mind this was mostly made over 48 hrs in final project crunch mode, it was a cardboard prototype that looked like this 24 hrs earlier) but my budget was more like $50 than $200, so I got stuck figuring out how to accept payment, because you really can’t mechanically accept bills. (Maybe some sacrifice can be made where I make a bill reader that’s not $$ because it doesn’t have to check for counterfeits. But then I think something this flimsy is a bit sketchy with lots of cash in it. Hmm. I think for our own use probably an RFID reader + student ID would do the trick). It also has so many mechanical issues (but yay learning things from prototypes). “
eric hunting replied:

That laser cut box shape looks like it could be endlessly customizable. For instance, one might make a front panel with engraved graphics and extend its edges beyond the bounds of the box to accommodate any sort of decorations or stick-on sign graphics.
 It’s interesting how many vending machines today use those spiral/coil carrier mechanisms. When looking into this I found you could buy these as more-or-less generic modular units made in China that could be used in any cabinet design–though they’re still a bit expensive. So, apparently, there is more of a food chain of standardized parts for larger vending machines today akin to that of general electronics. It seems to me that the popularity of this kind of mechanism may come from its combination of greater reliability and flexibility. Vending machines with these are definitely more ‘solid state’ than those of the past and accommodate a wider variety of items in the same machine, with the ‘tuning’ between product types a matter of software/firmware. Different size item, different spacing in the coil, different number of turns to dispense. I wonder if the pro machines have modules pre-wired for a certain number of turns relative to the product size or if they are programmed by other means. The compromise, of course, is that the machines of the type tend to be rather generic in design because they’re relying on the packaging of the items to attract customers rather than design and graphics on the machine itself. The Japanese recently took a new angle on this, though, with the addition of display windows that double as touch screen displays. So the machine acts rather like a video bill-board when not in use then turns more transparent when you approach it with the window serving as a control panel for choosing items. 

Well, that’s all I have to say about vending machines for a long time. I need to dismantle it sometime this summer and free up my shelf space at MITERS.

Happy summer everyone!

RGB LED Graduation Cap Shenanigans, MIT Graduation Report

I present here my derpy graduation cap, based off of

thanks Jamison ( for the picture!

Here it is, just finished at MITERS and still tethered to a power supply. Later I electrical taped on a lipo battery,  I kid you not.

From the front it looks normal, which is unfortunate for the graduation video, but whatever, I tried.

Only if you get some height above me can you see the lights.

what it looks like from the side. yea I can’t believe it got through security too, especially after Jordan‘s kawaiicopter cap that she worked on all winter break was rejected :'(

The original plan was to have a spinning RGB POV display, ala the plenitude of persistence of vision clock displays out there.

I even milled out a board for it, based on the fabduino files.

cute non-RGB 7 LED pov board

I can upload the eagle files if anyone wants, but since I haven’t gotten it to work and I can’t be sure there’s nothing wrong with the circuit I don’t want to put the effort into putting it up. This was my first exploration of the autoroute tool, which seemed to work pretty well.

Anyway after pulling two almost-all-nighters working on that project, I gave up when I couldn’t get the arduino bootloader burned onto the atmega (couldn’t get it to program at all, actually — hit a Target not found error). The cap I actually ended wearing was very simple to make and took about 6 hours total, including time spent trying to waterproof it with hot glue and saran wrap. (Most of the letters are stuck on there with the original backing tape that came on the strip). I didn’t have black wires so I used thick brown but still single core wire. I also probably should have done a 6 LED strip for the “I” and just covered up the center portion with electrical tape. Instead, I just skipped the dot on top of the “i”.

I also found this tool supppeerr useful, way more useful than the helping hands, for holding thick and relatively unyielding wires together so that I could solder them:


If you look carefully there are also two burn marks on my cap. I just colored the cardboard underneath with sharpie.

I also would have taken off my cap and showed it to the camera as I was coming down the ramp after getting my diploma if I knew how graduation worked.

I should have sanded down all the wires because I think a lot of the joints were cold solder joints (this bit me when I was repositioning the letters), but it worked okay and is probably the most robust of all the projects I’ve done so far because the circuit is encased in hot glue. Sorry arduino mini! (I potted that into the breadboard). The circuit itself is super straightforward, 3 FETs directly driven by the arduino. See

This is what graduation looked like for the parents.

Graduation itself was simply wet and miserable. I was on-and-off shivering by midway through and the list of names seemed endless. I’m sure they were rushing toward the end.

APPENDIX: Milling circut boards on a Roland modela mill (in the IDC)

Ultimate lesson learned: Make your traces at least 16 mil, 12 mil does not work!
It’s a 5 LED font.

The fabduino I will be running (a waste to use it for just 7 LEDs, but I don’t have the time to deal with not just reusing other people’s POV code)
How to use the roland modela mill
Eagle libraries
modela@ubuntu:~/Documents/nouyang$ sh 
1/tmp/eagle-setup.4298/eagle-6.4.0/bin/eagle: error while loading shared libraries: cannot open shared object file: No such file or directory
ugh, let’s try troubleshooting this
sudo apt-get install libssl-dev
modela@ubuntu:~/Documents/nouyang$ lscpu
Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Uhmm… okay nope. Switching back to a windows computer. A shame. : / I really ought to switch over to something more happy running on ubuntu someday.

Pin mapping

FTDI headers:

How to create board outline in eagle:

Eagle MAS.863 documentation

Design rules:

5V 1A regulator is in a SOT-223 package
Fab lab inventory
input gnd output

hmm tangent about IC naming (78xx = vreg)

Hmmm eagle files for this vreg
gnd out in, so pin assignments are wrong, should be easy fix
MAke sure to download zip file! OTherwise file does not download properly (adds some html comments at the beginning)
Currently the Vreg-LM1117 package is
3 IN
4 OUT@1
which matches the datasheet 1:GND, 2:OUT, 3:IN.

However for the LM2940 I want
1:IN 2:GND 3:OUT

Steps ish (very incomplete)
Connect> Disconnect all of them
Create new device

Grid dimensions in autorouting:

Export dpi?

then export as a png (file — export — image)
settings should be MONOCHROME and 500 DPI – this will export a image with white traces.
Make sure the color mode is set to greyscale: image > mode > greyscale
then export Dimenson layer for milling the outside of the board
Remember: the Modela cuts out the DARK and leaves white!

Correction! It should be both the top and pads layer.
just type in
~$ fab

… upon trying to make path:
      distance 0.800000, 140873 exterior points remain
Segmentation fault (core dumped)

must run as sudo
~$ sudo fab