Merging this blog with the company website:

We are merging the nordstromrobotics.com  company web site with the humanoidplatform.wordpress.com  company blog into the  humanoidplatform.com

humanoidplatform.com is a work in progress and is being updated daily.

When the migration is complete this site will be taking off line.

All of the blog’s content is being moved to humanoidplatform.com/construction/

AS5048A sending position data to the ARM based mBed – almost there:

 

Still working through the SPI issues with the AS5048A , but making progress.

 

Blue: Clock – Yellow: Position data

 

Selecting an optimal bus to connect the motors and sensors:

To limit the amount of wiring, the Humanoid Platform will only contain one power bus and one data bus. This simplifies the amount of wire that will need to snake around the various moving joints.  The power bus will be constructed of 10 AWG flex cable. This leaves the data bus.  I was considering CAN-BUS but now think a daisy chain Ethernet might be a better option. Ethernet is inexpensive and ubiquitous. Most existing industrial bus frameworks can run over Ethernet and Ethernet enables the use of a vast number of existing open source frameworks or I can create my own.  I was concerned about latency but a couple of papers suggest multiple network hops (daisy chain) will run in about 200us.  The ARM processor I am planning on using for the motor controllers comes with Ethernet; some even come with a two port Ethernet switch. If I use one without the switch I could add a simple IC two port Ethernet switch each of the motor controller boards.

Second casting attempt, casting with more precision:

This time I was more careful when producing the silicon rubber mold to try and produce a semi quality part.  As suspected it is very hard to get dimensional stability with soft silicon molds.  I knew I needed to move to a harder silicon or aluminum but wanted to get some hands on experience.  I also wanted to understand the silicon degassing requirements.

 

Getting SSI to talk to SPI is proving to be quite challenging.

Wiring an AEAT-6600-T16 encoder that speaks SSI to a microcontroller that speaks SPI is proving to be quite challenging.  I am working with the manufacturer, an electrical engineer and getting help from the Linked-In groups but if the integration takes much longer I may just move to an AMS AS5048A that natively supports SPI. I was hoping to avoid the move as the AEAT-6600-T16 has 4x the resolution per rotation of the AS5048A. But in the platform’s design the chips are swappable and in the future AMS may produce a 16Bit model or Avago may support the SPI interface.

Wiring the Avago AEAT-6600-T16 to the mBed (ARM Cortex-M3) via SPI:

 

Reflowing, soldering and breadboarding the Avago AEAT-6600-T16 absolute magnetic encoder:

In search of the right plastic and manufacturing process – Part 2:

After speaking with a number of vendors, various people in the Linked-In plastics groups (Plastic Experts, Polyimide & High Performance Polymers, Polyurethane Illuminati, Polyurethanes and the Society of Plastics Engineers, Inc.) and Michal Zalewski author of “Benchtop CNC manufacturing tutorial for robot builders, model makers, and other hobbyists”.  It looks like two part liquid polyurethanes with optional short or long glass or carbon fiber reinforcement is my best bet. This solution enables me to produce production parts with a simple direct pour process, e.g. no need for a large expensive press. I can use two part silicon or aluminum molds and when I start to scale production I can move to the RIM, RRIM and SRIM process.  This leaves injection molding to very large production runs.  Two part liquid polyurethanes can be purchased in liters, gallons, 50 gallon drums and by the truck load. Prices range from a few dollars to $120 a gallon.

 

First casting attempt, learning how urethanes work:

A simple, sloppy experiment used to gain some hands on exposure to the molding process, set time, temperature, part detail and part strength.  The reproduced gear is incredibly strong, when meshed with the original gear I am  unable to apply enough force with my hands to make the gear slip, bend or give. The part is lightweight and tough.  Naturally the production gear will be designed to use the minimum amount of material while providing the same strength.

PCB assembly – Low cost reflow option

I have the ICs, a couple of good PCB design software options and a PCB manufacture now I need a way to assemble the boards. Reflow looks like my best bet but reflow ovens are expensive and I was not looking for a DIY project (although the ones out there are very impressive) so after a bit of searching I stumbled on the C & A Scientific DB-1 Stainless steel hot plate, $140 with shipping from Amazon.

I will post details once I have had the chance to try it.

Temperature Range: 100c – 250c
Power Requirements- 300W
Heating Surface- 7.8″ X 6″
Overall Size- 14″ X 8.5″ X 4″
Shipping Weight- 8lbs.