A New Design Direction (plus some excellent news)

A project log for NeuroBytes

Build your own nervous system!

zakqwyzakqwy 03/23/2017 at 15:462 Comments

Since its inception in late 2014, this page has focused on the technical details of the NeuroBytes project. I try to cover everything from component selection to scaling challenges to firmware development to failed tangents; while I undoubtedly leave quite a bit out, from the engineering perspective I consider this running log to be reasonably comprehensive.

Getting the product in front of our end users -- primarily high school students, but also younger (middle school) and older (college) folks -- has also been a priority, just one that we don't talk about quite as frequently. By the numbers, we've sold NeuroBytes prototype kits into a dozen high school classrooms and two college neuroscience departments, and @NeuroJoe and I (but mostly him) have brought the platform into many more formal and informal learning environments (as shown in the image above). These interactions have driven our iterations starting with v0.4 and continuing through the green v0.91 boards we built last year.

Stuff We've Learned from Users

This won't be anything close to a comprehensive list; I'm focusing on lessons learned that are immediately relevant to future product development plans.

  1. Mode switches are confusing. Using a single base NeuroBytes board to do everything -- integrate-and-fire neuron simulation, motor neuron functionality, etc -- is great for minimizing SKUs, but it's confusing to new users. Every time we teach the patellar reflex kit, a few students end up setting an upstream interneuron into motor mode by mistake. As motor neurons directly output a servo-ready PWM signal -- i.e. a 50 Hz, 5% duty cycle square wave -- downstream neurons see an extremely rapid series of pulses, saturating their membrane potential value and holding the LED in a constant bright white state. This problem will grow more pronounced as we add new operating modes.
  2. Detailed PCB art is worthwhile. Students look carefully at NeuroBytes and notice just about everything -- that means the details need to be physiologically accurate. The gold-rendered NeuroBytes logo catches the eye and suggests a relationship to neuroscience, but the stylized design isn't perfect from an educational perspective. Specifically, dendrites should be more branched and spindly; the axon terminal should be longer and clearly split to output connectors; we should illustrate myelination; etc.
  3. Cables are too short. The stubby cables overly constrain even the simplest circuits, while the long cables still aren't long enough to adequately separate battery packs or sensors from circuits.
  4. We need dedicated power connectors. I had to see this problem first-hand to understand its importance. We spend a good deal of time teaching directionality (i.e. information flows only from dendrite to axon), and then tell students that power connections can plug into any free terminal. I view this as a convenience, but students see it as an exception to a fundamental rule!
  5. Our ecosystem is novel and has educational value, but it's not comprehensive enough. This was a hard truth, and one that took a long time to really believe. NeuroBytes don't have a ton of replay value and can't teach much beyond basic neuronal function, simply because they don't do much beyond blink LEDs and twitch motors. While teachers have enthusiastically accepted the platform and students are interested, the product doesn't have staying power because it lacks the flexibility for truly free exploration. In other words, we need to provide more options for input and output modules.

Product Development Plans

[above: new Motor Neuron prototype, featuring three equal-weight dendrites; a 'scope/programming port; two servo headers; and a revised PCB outline and look.]

Short answer: it's time for a massive prototyping sprint.

We're going to design and build a number of dedicated NeuroBytes modules, including a few concepts that we haven't explored at all yet (such as a model of the cochlea). Generally speaking, the modules will be single-purpose and designed to closely emulate (visually and functionally) real parts of the nervous system. On July 10th, we're hosting a week long workshop with two dozen high school educators where we'll test concepts, develop curriculum, get feedback, and do a few more iterations in advance of a more formal Fall '17 product ecosystem launch. Many more details to come -- I will try to post specifics related to a few modules in the coming months.

Hey what about the excellent news?

We got our Phase II grant from the National Science Foundation. This funding will provide adequate R&D resources to take us through commercialization. In other words: NeuroBytes continues to be my full-time job, but now I can (once again) get paid.



Peter McCloud wrote 03/23/2017 at 18:27 point

Thanks for sharing the lessons learned, its always interesting to see how the user feedback affects the product design. Congrats on the Phase II grant!

  Are you sure? yes | no

zakqwy wrote 03/23/2017 at 18:34 point

Thanks! It's nice to get paychecks again, not gonna lie.

  Are you sure? yes | no