Rich Nelson

Fetch Machine Part 1: Launch Mechanism

Fri, 22 Jul 2016 22:00:00 +0000

My dog Gus has a ton of energy, and while we love playing with him he can be a handful and we’re not always around to keep him entertained. I’ve been wanting to build him a fetch machine so he can play and burn off some energy while were out. The launch mechanism is the first step in the project to be followed by an enclosure, sensors and logic.

#Materials:

DC Motor, Power Supply and Belt from an old printer.
Wheels, Bearings and some bolts from a longboard.
Frame and idlers were 3-D Printed.
A few bolts and screw posts from a hardware store.

The launcher should look familiar if you know what a pinching machine looks like. I started with the gap between the wheels, I knew I wanted it slightly smaller than the ball diameter so it will squish between them when launching, I made the gap adjustable by 20mm so that that the prototype could be fine tuned.

The rest of the layout is constrained by the length of the belt (the ‘path length’ dimension is SolidWorks really came in handy).

After the main dimensions were laid out the rest was just filling in the details.

The bed of my 3-D printer is 8”x8” so the frame had to be made in a few parts, fasted by machine screws.

Everything came out great! As you saw in the gif at the top Gus already loves it. Next I’ll start working on an enclosure and some sensors to automate the operation.

Raspberry Pi Case

Tue, 12 Jul 2016 22:00:00 +0000

I use one of my Raspberry Pis to runs the @_StormWatch twitter bot as well as some other automated tasks. For a few months it’s been sitting on my desk bare so it was long overdue for a case. I took a look at the available cases and nothing jumped out at me so I designed my own.

I spent some time thinking about interesting geometries, something functional that isn’t just another rectangular box.

The model of the Raspberry Pi 2 B is from GrabCad.com. I used it as a reference for the size of the case and the location of ports and mounting holes.

Luckily I had a nice combination of filament colors on hand. The top is white Inland PLA from my local (1hr away) MicroCenter, and the bottom half is the Clay Red 5 Dollar Filament from ProtoParadigm.com. They have a great choice of colors, and the price makes it easy to grab a variety without having to buy full 1kg spools.

I’m super happy with how it came out.

Spyder Colorimeter Design Concept, Part 2: Analysis and Improvement

Sun, 21 Feb 2016 20:00:00 +0000

This is a continuation of my Spyder Colorimeter Concept Design Project. After the initial prototype I did some some injection molding analysis in SolidWorks Plastics. The analysis and further research shed light on some areas that could be improved upon.

I ran the simulation with default setting of 100MPa injection pressure limit, 230ºC melt temp, 50ºC mold temp, 4 sprues, and generic ABS plastic. The result was a fill time of 3.2 seconds and an average cooling time of 16.86 seconds.

The plot of the temperature after cooling highlighted some potential problems, the structure has walls much to thick which would likely lead to shrinkage not to mention extra cooling tome, and wasted material. After seeing the simulation and doing some research into the design principles of injection molding I decided to design a physical prototype of the RGB vendiagram concept (from the previous post).

The goal of the second design is to improve upon the weaknesses of the first prototype, specifically focusing on the following:

Keeping wall thickness to .15” or less, assuming a plastic like ABS will be used
Avoiding vertical surfaces and using a draft angle of 2º or more
Using fillets for a stronger transition where features meet

The design went well, all surfaces are 1.5” or thinner except for the square area on the top half. Making the blinder channels deeper may negatively affect the function of the device, so the likely solution would be hollowing it out from the outside and using the logo to cover it up in production. The protruding supports, pins and holes all have a draft angle of 2º or more, and a thickness of 50-80% of the shells 0.1” thickness.

The fill time on the bottom was 2.36 seconds, about a second faster then the bottom section of the first prototype, though the side walls are lower so it’s not an even comparison. The average cooling time was 14.09 seconds, over 2 seconds faster than the first prototype.

Big improvement here, the max temperature is over 40ºC cooler and overall the temperature is much more uniform with presumably thanks to the tinner walls. That last thing to do was print it and see how everything fits together, and how it feels as a physical object!

Spyder Colorimeter Design Concept

Sun, 21 Feb 2016 20:00:00 +0000

I set out to design a new enclosure for the Datacolor Spyder Colorimeter. The first step was to research past designs, to see how they’ve changed over time.

My goal was to create a design that was familiar and consistent with the look of previous models, but also brought something new to the table. Looking at the past designs there are some common themes that stick out such as the triangular nature and radial symmetry present in most of the designs. The most current design has moved toward a more compact, rugged design.

While looking for inspiration I focused on patterns and geometry that were present in the previous designs, while also incorporating influence from photography and color systems to reflect the target customer and function of the product.

I decided to move forward with these two concepts.

After doing the rough mock ups in SolidWorks I preferred the look of the aperture design. So, I completed a more detailed model before working towards a physical prototype.

Handheld Wireless Colorimeter

Wed, 10 Feb 2016 14:34:25 +0000

Last Friday I took apart my old Spyder2 monitor colorimeter to see how it worked. Apparently it uses 7 colored filters with a light sensor behind each one, by comparing the amount of light absorbed by each filter it can determine the color on screen. I was fascinated by how it worked so I decided to take a shot at making my own.

Here’s a peak at the finished product

##Constraints:

Only use components from around the house
Finish it in roughly a weekend
Ideally small enough to be hand held

##The Sensor First I tackled the most import part, the sensor. I had some color gels used for camera lighting, I chose to use red, green, and blue (the additive primary colors) for even coverage across the color spectrum.

Unfortunately I only had one photo resistor so I used a servo to take measurements through each filter.

##Calibration After assembling the sensor and the next step was to collect test data for calibration. An Arduino was used to collect the analog light measurements form the photo resistor and to control the servo motor. I limited the scope to measuring hue, there are 1560 hues in the 24 bit RGB color system, I wrote this color picker web app in AngularJS, mounted the sensor onto a calibrated monitor, and setup a python script to iterate through the 1560 hues and record the readings.

With a database of calibration measurements, we now have a set of known values to compare new measurements to in order to find the closest match.

##Finishing Touches The last step was to create an interface and assemble all of the components into the case, for the “start button” I mounted the sensor on springs, with another spring and screw arranged so that a small amount of pressure on the front will make them touch and complete a circuit to trigger the measurement function. An LCD screen displays the RGB and hex values of the measured color, and an RGB led lights up in the measured color. After testing across 3 different screens that colorimeter is over 97% accurate.

##Colorimter Mark two update 3/7/16

Even though the first version worked well, there was definitely some room for improvement, the most obvious improvement would be eliminating the moving parts and having a sensor dedicated to each filter, but I still only have one photo sensor so I’m sticking with the rule of just using parts that I already own.

The focus focus of these improvement is to optimize the programming to eliminate unnecessary hardware. The original program referenced the entire database of calibration measurements when determining new colors, the database alone was about 45kb. So I did a polynomial regression analysis of the calibration measurements to establish a set of equations that can be used to determine colors.

The data naturally split into three intervals each section spanning the range in which one of the primary color channels was at 100% or 255 in our 24 bit color system. After finding the representative 3rd order polynomial equations we no longer need the Raspberry Pi computer to do heavy lifting with the database, and the whole program will fit onto the 32kb of memory on the Arduino’s ATmega328 micro controller.

I rewrote the code to get everything running on the Arduino, Also ditching the USB power bank for a simple 9v battery. Ditching the two most expensive components brings the total cost of materials to about $10, not including the case. It also starts up much faster now, ready to measure within about 2 seconds of powering up. The accuracy seems to be about the same as before, but testing is required to compare. I’m considering making a new case for it, but if I do that I would probably want to get some more sensors to eliminate the need for the servo to make things more compact and reliable. also I saw 16x2 LCD screens online with an RGB backlight so the function of these screen and led could possibly be combined.

Text Message Controlled Dog Treat Dispenser

Fri, 25 Dec 2015 19:34:25 +0000

Video of the device in action

How It Works

This project uses a phone number from Twilio. When a text message is received it triggers a webhook HTTP post to a python flask server that’s running on a Raspberry Pi that’s inside the treat dispenser.
The server checks the content of the message, if there are no specific commands it wiggles the servo motor a few times to make some noise and get the dogs attention.
At this point the Camera takes a picture and starts uploading it to Dropbox, while the servo moves a piston to dispense a treat.
After the Dropbox photo url is received, the picture is text messaged to the sender through the Twilio API.

It has been working great. To the left of the camera there is a motion detector that is hooked up, but with no software running for it. While we’re house training Gus he stays in his crate during the day, but when gets free range of the house during the day the motion sensor will be used to make sure that he’s in front of the camera before it takes a photo.

Vacuum Fixture for Machining Longboards

Mon, 15 Jun 2015 20:00:00 +0000

Normally parts can be secured or clamped from the outer edges for machining. Unfortunately this isn’t possible with a longboards, since the entire perimeter is cutout the board would come loose.

In the past we secured it from the edges, then after the truck mounting holes were drilled we would fasten the board to the mount through those holes. This required extra time, and even using rubber spacers the boards would sometime get marked up.

It was time to replace the bed of the CNC machine so I decided to incorporate a vacuum system.

The system uses PVC piping underneath the machine and hooks up to a 6hp ShopVac. In order to mount the boards the fixture needed to match the contours on the board. We make 12 boards that are pressed on 7 different molds, so a fixture was made to match the contours of each mold. In addition to the surface it also needed channels to for better airflow.

Notice the fixture also have bolts on each end, these are used to align the board on the mount, they match up with holes in the presses.

The fixtures work great! As intended they reduce setup time and prevent the boards from being blemish by mounting bolts.

Board Design Overview Videos

Mon, 15 Jun 2015 14:34:25 +0000

A series of videos giving some insight into the design and features of the 2015 Nelson Lineup.

Run down of the changes and updates for the 2015 lineup

Siphon 35 & 38

Prime

BatRay

StingRay KT36

StingRay DH

MantaRay

Drop Series

Nelson Longboards Manufacturing Process

Sun, 10 May 2015 20:00:00 +0000

Paint Booth

Sat, 07 Jun 2014 20:00:00 +0000

I designed a paint booth to fit our manufacturing space. Since our machining and finishing takes place in the same room, we needed a clean environment for applying clear coat.

We use a water based clear coat, so fire proofing was not necessary like it would have been for solvent based coatings.