Posted on

A New Chapter for Kelly Industries

Wow, I’m not too good at writing blog posts regularly am I. But I thought I should do it now, at the end of one chapter of my life and the beginning of another. I suppose making blog posts was one of those things that sounded exciting at first, until you get into the everyday reality of it. Like running an online store.

Kelly Industries started a year and a day ago. I had set up a website to sell a select fire Rapidstrike circuit board I was working on, and then my friend Phil’s high crush cages, released under the name Open Flywheel Project, were taking off, but nobody without a 3D printer had any way of buying them. Another friend, Ray, had a 3D printer, and I had a website, so I made pages for the cages and posted about it on Reddit. The first day I got an order for 12 screws, and then a couple hours later an order for one OFP cage and one Barricade cage of my design. For a couple weeks before I got my own printer, we would meet in the snow halfway between our dorms and hand over bags of cages and plastic filament. I didn’t have much experience with 3D printing so I didn’t know it at the time, and nobody had much experience printing cages, but the quality back then was awful, and I would like to thank and apologize to the people who bought cages early on. Once I got my own printer, I was able to continually improve the quality, and I like to think that my print quality right now is quite good. Ray designs some cool 3D printed Hammershot and Quadrant cylinders and is actually starting his own Nerf company soon, keep an eye out for Ray Blasters once he gets his 3D printers fixed.

Well I’ll just come out and say it. I wasn’t very good at running an ecommerce store. The printers would frequently break down, I would get several weeks behind on orders, and not read or respond to the email inbox full of people asking where their orders were. I also lost track of several customer orders. I was spending many hours each day printing and shipping orders, but wasn’t bringing in enough money to hire someone to handle the shipping and customer service for me so I could focus on what I was really interested in, developing new ideas. Selling 3D printed parts took to much time, and having backlogs of orders is terrible because everybody is wondering where their order is. I would have liked to have stock of all the printed parts, but for the cages alone, between all the different blasters and cage spacings, and the versions with an without Morpheus guides, there are 58 different cage designs to keep in stock, and that’s not even counting all the variation in colors. I am truly sorry to all the customers whose emails and messages are still sitting unread. I plan on going through and respond to them all this weekend.

For the last year, I’ve splitting my time between printing and shipping orders, repairing my printers, developing new products, and designing a better 3D printer from scratch. I realized that the Nerf business wasn’t bringing in enough cash to live off of long term, so the plan was to switch to making 3D printers. When I realized that the 3D printer wasn’t going to be done for a while, I started applying for mechanical and electrical design jobs. Today I got a job offer, and so I am writing this blog post.

I am shutting down the online store of Kelly Industries. All currently outstanding orders will be shipped on Monday. I am still going to be designing new products, as that is my favorite part of this hobby, and will be licensing out my 3D printed designs to other stores in the community. Today I’m publishing the files for my Stryfe/Rapidstrike flywheel cage, 3D printed wye that actually feeds full length darts, Mediator horizontal pump grip, and Rapidstrike switch mounting plate. I have some non 3D printed parts in the works as well, such as high concavity flywheels called Comet Wheels, MOSFET circuit boards, and a Rapidstrike switch plate circuit board that makes the wiring super easy, which will be for sale soon, both from community stores and shipped from Amazon. Maybe I’ll even finally finish that Rapidstrike select fire board that started this whole company. I also have a ton of stock flywheel blasters to mod and sell, and maybe I’ll even finish one of my 3D printed blaster designs.

This is the end of a chapter and the beginning of another one, both for Kelly Industries, and for me personally. I’d like to sincerely thank all the people who have supported me in this endeavor. I hope you will stick with me through this transition, and that Kelly Industries can move to bigger and better Nerf projects. Drop in HPA Longshot kit anyone? Maybe I’ll even build a springer.

Posted on

Beyond Unboxing: Years of Parts

One of the things blogs do that I find incredibly useful is taking a look inside consumer products. Things like Charles Guan‘s Beyond Unboxing, series, AvE’s BOLTR videos, and EEVBlog and Stefan Gotteswinter‘s teardowns provide insight into the inner workings of that device which is only possible by having the product in hand and taking it apart. It helps educate readers on how they work, and informs them how good quality they are, how easy they are to hack, and ultimately whether to buy them.

So here’s some interesting internal photos I’ve collected over the past couple years:

First up, some pneumatic components. This 3V1-06 is a cheap Chinese knockoff of the original, which seems to be made by Airtac. They’re available on ebay under $5, and are rated for 10 bar. G1/8″ (BSPP) threads on the Common and two Normally Closed ports, #10-32 threads on the Normally Open port on top, 1/4″ spade terminals for 12V positive, negative, and ground. It’s an unbalanced direct acting solenoid poppet valve, which means that the bore is very small (1.2mm) so that the solenoid can overcome the pressure, and therefore the flow rate is fairly low at 140 L/min at 7 bar.

Next, the 3V1-06’s competitor, the Clippard EV series. This is the 2-way EV-2-24 variant I found in a box at Noisebridge, but if I were to buy one, I’d get the 3-way EVO-3-12 ($30 from Clippard). This one uses a steel diaphragm shaped poppet with flex joints instead of a piston, and the solenoid is axial flux instead of radial flux, with a flat coil instead of a hollow one. Other than that it’s pretty similar to the other one, an unbalanced direct acting solenoid valve with 10-32 ports. If it were the 3-way variant, it would have a hole in the center of the coil for the Normally Open port for the top of the diaphragm to seal against when it moves up. Clippard lists the poppet at an amazingly tiny 0.007″ of travel. It’s a bit more compact than the 3V1-06, but the airflow is much lower, 17 L/min at 7 bar.

The last pneumatic component I have is another ebay knockoff of an Airtac part, the HSV-15 manual slide valve. They come in G1/4″ through G1/2″ variants, meaning they very conveniently screw onto the back of QEV valves like the BQE-02 and BQE-04 for a simple semi auto air gun:

Unfortunately my QEV is too tight to be disassembled, but the slide valve is quite simple. Only two parts, the body and the sleeve. The sleeve has two O-rings that seal against the outside of the body, which has a wall in the middle blocking the air, and a set of holes on either side of the wall. When the sleeve is forward, both sets of holes are between the O-rings, allowing air to flow between the body and the sleeve, and move between the two sides. When you pull the sleeve back, the front O-ring passes behind the front set of holes. The rear chamber is now sealed, and the front chamber is vented to the atmosphere. Simple, and effective. Just don’t leave the front O-ring sitting on top of the front set of holes or you’ll continuously vent air.

Here’s a great set of drawings from dewey-1 on Spudfiles for this slide valve, and the QE-02 QEV:

The other thing I have some decent photos of are brushless motors and controllers.
Here is a now-discontinued Hobbyking XK2860-B-2700KV, a 4 pole brushless inrunner. As you can see, it has an iron stator which the wires are wound around, pretty standard.

But I was very confused when I opened up this also-now-discontinued B28-57-15L 2 pole inrunner, to see no iron core at all. The copper windings are arranged in a caret (^) pattern, and are held in place with nothing but epoxy. This apparently allows for more copper mass, allowing more power to be put through the motor with less heat, but at the sacrifice of low-end efficiency and torque because there’s no iron core to direct the magnetic fields. They also operate more smoothly at very low speed, because there are no iron slots to create cogging. You can use this property to tell whether a motor is ironless or not without opening it up.

I really wish manufacturers would specify which type each inrunner is, but they do not. It is up to us hobbyists and reviewers to document which motors are which. Every motor in the XK series that I’ve used has had an iron core, with this B series motor being the only ironless motor I’ve encountered.

This motor is also a 2-pole motor, meaning for the rotor they can use a hollow cylinder of Neodymium, radially magnetized. For 4-pole motors, it’s impossible to make them from a single magnet, so they use 4 rectangular magnets attached to a square steel body, and held on with a carbon fiber wrap against the centrifugal force.

For speed controllers, the cheap and medium sized option is the Red Brick 70A. I have a number of the v2 iteration, first seen September 2015. The v3, which is almost the same with one pair of pins swapped, was first seen February 2017. Here’s a RB70A-Opto and RB60A-Opto side by side, and they appear to be absolutely identical, except for the label. They later seemed to drop the 60A from stock, which makes sense. Both have 18x  Ubiq-Semi M3006D N-channel DPAK FETs, rated at 30V 5.5mOhm. The input capacitors are three Chongx 470uF 35V 105C, the linear regulator for the onboard logic is the venerable 7805, and the buck regulator on the non-opto versions is the almost as well known LM2596S. Contrary to the name, the “opto” versions, like most “opto” ESCs, don’t actually have an optocoupler on the input, they just lack the buck voltage regulator. They have a single board for both signal and power.

Most importantly, they both have an Atmega8 microcontroller and run the amazing Simonk firmware (another topic for a future post). However my pull request with the rb70a3 config file hasn’t been accepted yet, cough it’s been 3 months Simon, so you need to grab it from my copy of the repo. These also don’t come with a bootloader on them, so you’ll need the incredibly handy pogo pin tool and USBasp for the first time you program it, or just buy one with Simonk already loaded.

I can’t seem to find my pictures of them, but another solid speed controller option is the F-60A. Separate power and signal circuit boards means it takes up more space, but they use 18 of the better quality Ubiq M3016D, which has 4 mOhm Rds(on) instead of 5.5mOhm.

My favorite, however, is the ZTW Spider 60A. Though it is a double board design, it’s significantly smaller than the other two, because of its higher quality International Rectifier H8318 FETs, with a 3.1 mOhm Rds(on) rating. For some reason they only populate 18 of the FETs, even though the board has spots for 24. It’s also a little strange in that the auxiliary voltage output is 12V instead of 5V, but I just cut that off. Input capacitors are a pair of 35V 330uF Rubycons. Unfortunately they’re out of stock, but it appears that you can buy the boards with a different firmware under the Aerostar brand. Aerostar also makes an 80A version, which would appear to be the same board with all 24 FETs populated.

Well that’s it for now, I’ll do this again when I have some more interesting things to take apart.

Posted on

Another attempt at a Blog

Perhaps I’m just nostalgic, but I miss the engineering blogs and build logs that seem to be so scarce now.  A lot of it has moved to short form, ephemeral social media, which doesn’t give readers the same depth of insight into the engineering process, or have the long term reference value of a blog post. Vlogs are similarly easier to digest, but lacking in searchability and the ability to carefully curate your words. I’ve tried blogging before, but I’m serious this time. Even if they’re not in depth as I would like, I’ll write weekly updates on what I’m working on, or explaining some physics or electronics concept that I think hasn’t been covered properly, or I just find interesting. If a post hasn’t come out by Sunday, please bug me about it.

This week, a status of my projects. As most of the people reading this will know, I’ve gotten pretty involved in Nerf over the last couple months. As you can see from this website. I love how this hobby is both very technology heavy, and not so advanced that one person can’t push the boundary of the state of the art. It makes for a great hobby for engineers and makers. I got a Tevo Tarantula 3D printer about a month ago now, which will be the subject of another post, and have been selling Open Flywheel Project flywheel cages, a few of my own designs, and reselling some other parts, on this site.

First up, the last project I actually completed, a Rapidstrike flywheel and pusher motor cover set for 180 motors. I think they look pretty good, and fit into the Rapidstrike aesthetics well, but they haven’t been selling for some reason.

The flywheel cover is also unique in that it’s the first one to use the screw from the stock Rapidstrike attachment rail to hold itself on, and so doesn’t require any adhesives. Fasteners are always better.

Also, barrel attachments. I prefer functionality over aesthetics, so wouldn’t actually use the Barrett 50 cal muzzle brake, but some people are into that. The smooth and fluted attachments look pretty good in my opinion, and are very practical in that they hold the Rapidstrike battery door on without needing any screws. There’s also a ring version for people who want a lighter and cheaper option and don’t care about aesthetics. Thanks to LordSquiggles and Eli Wu for the battery door retention idea.

I’ve been trying out PETG for flywheel cages, as there are some new motors, like the Titan Hyperions, that might run too hot for PLA cages. As many people have said, PETG doesn’t print as well on overhangs and bridges, as you can see here on the bore of a Hyperfire cage.

This happens a bit with PLA, but is much worse with PETG, and the long, deep bore makes it difficult to clean up with an X-Acto knife. That picture was with flat top to the bore, which came out a bit worse than the round bore.

It was alright on the front and back, where it can bridge across the flat spot, but since there are cylindrical cutouts for the flywheels, there’s no way to bridge across the center section. A teardrop shape, which was commonly used in early 3D printed parts for this reason, works better, but the bottom edge of the bridge in the center of the cage is still unsupported.

Today I’m trying some support material in the bore and see how that compares. Even though removing the strings only takes a few minutes, it will never leave as clean a finish as will come straight off the printer.

I also switched from the flywheels-down orientation that the cages were designed for to a flywheels-up orientation with support material inside the motor holes, with much better results. The bore in particular used to be fairly messy and needed cleaning with a knife, and now comes out totally smooth. I run my nozzle too close to the bed to compensate for not having automatic bed leveling compensation, which makes the bottom 2mm of my prints slightly bulged out. That used to cause a problem with the motor bearing holes being too small, which is better now that the prints are flipped. The bearing holes in the files are still about 0.8mm too small though, because so I have my own modified copy of each file with enlarged bearing holes, and a chamfer on the 20mm motor holes to keep the bulge at the bottom of the prints from being a problem.

OFP also released Cycloneshotgun cylinders a while ago, which I’ve been tuning in. Unlike the cages, where my printer prints holes too small, it’s printing the barrels too loose for a proper springer fit. Go figure. They have both full 3x 50 caliber shotgun, and 1/2 shotgun and half mega dart versions.

For my own designs, I have too many ideas and haven’t followed through on them. The Stryfe Battery Door I posted 6 months ago still hasn’t been test fitted and refined. The idea is to have custom sizes for specific battery sizes, so I made a poll to see what people use. Unfortunately they don’t fit flat in the battery compartment, so you have to turn them on their side, which makes the door stick out way farther. Maybe I can have the walls of this door overhang past the original door and make enough room to sit the batteries flat inside.

I’ve also been working on a fully 3D printed, scratch built bullpup blaster called Openfire. The idea is to make it modular, so people can easily swap in different flywheel cages or front ends if they want, and open source so people can easily design their own upgrades. I printed the flywheel cage and butt plate, and the design for the back half is pretty much done, consisting primarily of the Rapidstrike-style pusher and magazine well. However that piece is a 20 hour monolithic print, which I’m not sure I trust my printer with just yet. I’ll try to get auto bed leveling working first.

There’s a couple other ideas that have been floating around in my head, like an afterburner cage that goes on a Nerf barrel attachment point, and looks something like the Modulus barrel. I might call it the Long Range Barrel. The tricky part with this one is how to power it. One option is connecting it to the existing wiring of a flywheel blaster, which is a bit complicated to install, and leaves a wire sticking out of the blaster when you remove it. The other option is to have the battery and rev switch built into the foregrip on the barrel. That means it would quickly attach to any blaster, including a springer, but you have to remember to rev the attachment before you shoot, or you’ll jam the barrel. And of course there’s another barrel attachment point on the front, so you can do shenanigans like Coop. Let me know which version of this you would use. Here is the current state, without the sideplates.

Another project is the DIY chronograph. Put two infrared detectors and LEDs in a tube, measure the time between the pulses, and you know how fast it’s going. Simple right? You can even keep the hardware costs extremely low by using an audio input on a computer or phone for data acquisition. I started writing an Android app for this, before realizing that was stupid, and switched to a web app. It kind of works, but has some serious bugs. I also 3D printed a chrony tube that attaches to a Nerf barrel attachment, which is pretty cool. If the software ever gets working, I’ll get some circuit boards made with the infrared detectors and LEDs on them.

I was also thinking about a Rapidstrike SMG foregrip and front cover, but I have enough projects going on right now, as you can see.

And of course, there’s SmartRapidstrike. A select fire, ammo counter, OLED screen, lipo protection, burst fire, chronograph circuit board for full auto blasters like the Rapidstrike. I assembled one circuit board, the buck voltage regulator works fine, but the STM32F072 microcontroller doesn’t respond to anything. I may have killed it during my many attempts to get it soldered down correctly, or I may have messed something up with the circuit board layout. I’ll get to this at some point, but even once the electronics are done, the software is another huge task. Thankfully Josh Esbrook and Lachlan Sneff kindly volunteered to help out, though this was many months ago. We haven’t even really started, but I’ve changed frameworks like 4 times so far, and am currently planning on ChibiOS + uGFX + GoogleTest.

In other business news, I’m going to be stocking wire, battery connectors, and both MTB Hellverine and Foamblast Meishel 2.0 motors soon. Keep an eye out for them.

This ended up being a long post, but I did say it was long form content. It’s good to get my thoughts out there. Let me know which project you’re interested in, it helps me know what I should work on.