Antimatroid, The

thoughts on computer science, electronics, mathematics

aBox: Electronics and Databases and Everything Else Inbetween

with one comment


Back in January, a local Boulder company had a special promotion to give away $100,000 in electronics goodies. Naturally, a lot of people caught wind of this and it was the incentive I needed to take a look at electronics and physical computing. While I didn’t receive any free goodies, I did place an order for a number of parts and components anyways and drove up to their office and picked up my order one day during lunch. The key component I ordered from them was an Arduino Starter Kit (DEV-09284). The Arduino (specifically the Duemilanove) is a simple piece of hardware built around the ATMega328 microcontroller that’s easy to program over USB using a free IDE from the Arduino website. After playing around with the components and learning a bit about electronics, I began to think more critically about what I’d like to make.

In addition to the starter kit, I also got a 20×4 LCD (LCD-00256) and a piece of hardware that attaches on top of the Arduino allowing for network connectivity called an Ethernet Shield (DEV-09026). From these components I began to picture a little box with a simple text display, a series of buttons, maybe a LED to show status and an Ethernet connection that I began to call a(rdunio)Box. In a nutshell: have a piece of hardware that is able to exchange messages with a server and display the output of these messages to the LCD. Interacting with the buttons would change which messages were sent to the server. Part of the message exchange would involve getting a configuration of which messages the device can send to the server. Given the device, it made sense that it would be fun to have a web interface to setup configurations and different views that could be displayed on the device and data providers that operate behind the views. Which, of course, hints to the fact that there would need to be a database behind the web interface that the server component of this design would talk to as well. Below is a diagram of the general architecture of the project.

In terms of technology decisions, the client side is comprised of C and libraries provided by the Arduino platform- pretty standard embedded environment. The aBox and HTTP Server is written in C# using the .NET framework running as a multithreaded process. The .NET stack is fantastic and makes it easy to get things done quickly. The Web Interface is straightforward XHTML, CSS, XSL and JavaScript. No libraries were used on the client-side as I wanted to learn JavaScript a little better. Finally, the server points to a MySQL Community Server edition database. The aBox Client-Server protocol is a simple TCP (most easily accessible protocol in the Arduino Ethernet library) message exchange that is somewhat RESTful. The exchange between the Web Interface and HTTP Server is a clean cut AJAX exchange following RESTful principles. Server-Database communication is the de facto TCP exchange.

In terms of deployment, the aBox and HTTP Server resides on a Windows XP laptop. The database on an Ubuntu 8.04 virtual machine running inside VirtualBox on the XP machine. I went with Ubuntu rather than XP for the database as MySQL is easier to manage under that environment. The aBox Client lives on the Arduino hardware. All of these are on the same Ethernet network.


The Arduino environment comes with a suite of really great tools that work effortlessly out of the box. Fritzing is one of those tools; it is a simple diagramming tool that comes with a preloaded set of widgets that can be arranged as though you had the breadboard in front of you. Given the functionality that I wanted I knew that I needed a handful of discrete components: at least one button to switch between views, a potentiometer to control the contrast of the LCD, a LED as a status indicator and for future purposes, a thermistor. In addition to these parts, I needed two 330Ω resistors to limit the current going into the button and LED and a 10kΩ resistor so that I can get accurate results out of the thermistor. The diagram at right is the initial setup I came up with in prototyping. The Ethernet Shield uses up pins 10-13 of the Arduino board. The LCD requires 4 pins for pushing data and an additional two for control. These are pins 2-7. Finally, a pin for receiving a digital input from a button and one for controlling a single LED used by pins 8-9. While the breadboard approach makes for easy assembly and rearranging of components, it’s a pain to continuously take it apart and put it back together. So, I decided to put together a soldered board of the components that were on the bread board. The Bill of Materials is the following:

Item SFE Part # Unit Price Units Ext. Amt.
Basic LED – Green COM-08532 $0.35 1 $0.35
Break Away Female Headers PRT-00115 $1.50 8/40 $0.30
Momentary Push Button Switch – 12mm Square COM-09190 $0.50 1 $0.50
ProtoBoard – Square 1″ Single Sided PRT-08808 $1.50 1/2 $0.75
Resistor 10k Ohm 1/6th Watt PTH COM-08374 $0.25 1 $0.25
Resistor 330 Ohm 1/6th Watt PTH COM-08377 $0.25 2 $0.50
Thermistor 10K SEN-00250 $1.95 1 $1.95
Trimpot 100K COM-08647 $0.95 1 $0.95

Given these materials, I set out with the above board layout. Layout of the board was done using a greedy algorithm. I started by first placing the button and associated resistor using as little space as possible- sat back and compared it to my breadboard to make sure my logic was sound. Once I was convinced, I repeated this process by placing the potentiometer, LED and the thermistor down on the board until I ran out of space. Designing the board and actually soldering the board turned into a rather interesting set of lessons. Applying the right amount of heat, making sure that parts were added in the right order and making sure that the polarity and orientation of the components was correct going onto the board required constant, conscious effort. The board was completed after a couple nights with a handful of little modifications. Notably, reducing the dual seeding of components and minimizing wiring distance between components beneath the board.

In terms of writing the code that would be placed on the Ardunio, I spent some time writing a handful of modules: dealing with dynamic array and C styled strings, interfacing with the button, LCD, LED, Ethernet Shield, logging, configuration, message exchange, handling and parsing, and of course the application itself. A lot of time was spent debugging memory allocation and deallocation issues, working on timing and response issues and making sure that final product was able to fit into the allotted 30k of memory.


Before I jump into the Server and Web Interface, I want to run down the basic data model that is used. At a high level, there are devices, views and providers. A device is a collection of views, a view is an instance of a provider with supporting arguments and a provider is an interface provided by the Server that can be setup with parameters. The entity relationship model for all of these items is summarized below:

To give some concrete examples of the above consider a provider that’s an interface into the Netflix Service that retrieves specific information based on the provider parameter- say the contents of a queue or the status of what movies are in the mail. A view is then created on top of this provider to specify a concrete realization of the provider- we might have one view called “Netflix Queue” and one called “Netflix In Mail” that can then be added to a device. The device is what is sent down to a specific Client. Right now, the first device found is what is sent down to the hardware, but in the future it would be ideal for a Client to have a configurable identifier (say a DIP Switch – 8 Position (COM-08034)) that would specify which device to get from the Server. Alternatively, an additional button that would cycle between devices similarly to how the view cycle button works.


Getting the server put together took the least amount of time in comparison to the Client and Web Interface as there weren’t any unknowns to research. Both the aBox and HTTP server have listener threads in place that then delegate requests to queues to make sure the listeners receive requests in a timely manner. Each queue then looks at the request and attempts to find the appropriate handler to produce a response. If no handler is found, then (in the event of the HTTP server) the contents of the requested file are returned. If all else fails, the server returns an error message in the appropriate protocol. The aBox portion sets up a TCP socket to pick up requests coming to port 8888, the HTTP portion utilizes the .NET class HttpListener under System.Net that picks up standard TCP port 80 traffic. I decided against implementing the the HTTP part as a TCP socket because the value from doing so didn’t justify the time to implement the protocol. Alternatives to HttpListener include ISAPI on IIS, mod_aspnet on Apache http and WCF AJAX Services on IIS. Each one has it’s pros and cons; ultimately, I choose HTTPListener as it reduces the complexity of the system at the end of the day and exhibited the fastest conception to completion.

The aBox exchange is a simple name-value list that first describes the verb and noun, followed by any supporting data. As an example, when the Client starts up, it will send Get:Configuration, and the server will query the database views table and send something like Post:Configuration|Time:0|Weather:1|News:0 which the Client will put in memory and default to sending Get:Weather until the End User pushes a button and the view selection increments to News and results in Get:News being sent.

The HTTP exchange is very typical, standard HTTP Request that comes in and an appropriate handler is picked based on the URL and Content represented in the request; the handler accepts the request and produces a response. The response will be an asset that exists on the Server, e.g., a CSS file; or the Server will return an XML document and function as a RESTful Web Service. This is preferential to the SOAP Web Service approach as it greatly reduces the design-by-committee XML boilerplate that is commonly associated with SOAP (not to mention that it is an order magnitude easier to work with in JavaScript). The Web Interface is then responsible for performing something with that data. The Server strictly yields data back to the client and does not generate any XHTML for the client to consume. The usual REST verbs are allowed where HTTP Get, Put, Post and Delete map to corresponding handler functions Get, Add, Update and Remove.

In addition to the main threads for dealing with requests, there is an additional thread that periodically runs data providers. As an example, a view that uses a RSS provider for a weather site will run at a specified period, requesting data from the specified web site and store the resulting data in the database that will be acquired by the aBox or HTTP threads to be shipped down to the client. The reason these are decoupled is that we do not want to go and query potentially expensive or metered resources every time a request comes in. This way, if the Server is ever used by multiple devices, a corresponding third party web service isn’t issued multiple requests a second, instead it will issue a handful of requests a minute or hour instead.

Web Interface

Dealing with the Web Interface took almost as long as the Client to implement, but the majority of the time spent was learning JavaScript and XSL to make sure that the Client ended up being the fat-client that I had in mind. The basic approach was to use HttpXMLRequest to issue HTTP Requests to the Server and then apply XSL files against the resulting XML response using the XSLTProcessor to produce XHTML fragments. I like this approach as it keeps the logic squarely in the JavaScript, the look and feel of data in the XSL and CSS and the data as XML. I found that once I had gotten a grasp on the nuances of JavaScript that I was able to implement the main page and make minor modifications with each additional page resulting in a large time investment up-front, but a low time commitment to bring on additional pages.

When I produce a website, I typically jump into Photoshop and start producing some mock pages of what I want the site to look like and start formulating how it will all work in terms of End User actions as well as polishing the front-end with user experience elements. I decided to go with a simple tabbed header menu and tabbed menu that produces a central details and appendages regions. Details are things like that entity name and description, appendages are things like the views that are associated with a specific device. The following is a sample of the devices tab:

I found it easiest to organize my assets by elements of the page (tab, menu, details, appendages) rather than by the main tab (devices, views, providers) as it allowed for an easier generalization of these regions as it applied to each of the tabs rather than trying to repeat most of the same code along each of the tabs. This process applied to the CSS, JS and XLS files. There was a single XHTML index page that then routes all of the options back through the document; this probably isn’t the preferred way to do things from an exchange standpoint, however, this is a single front end that won’t likely ever need to have its URLs sent around to go find a specific item. learning JavaScript was an interesting experience as it reminded me a lot of writing the client code for the Arduino. Very procedural, quasi ability to fake traditional OO design, but overall, much more of an exercise in keeping everything ordered and organized.


This has been a rather involved project over the course of the past two months working 2-3 hours during the week and usually a Sunday afternoon to produce the hardware, implement the code for the Client, Server and Web Interface and debug it all, as well as produce this write-up. It’s been exciting to get into hardware and spend some time getting a feel for the minutiae of electronics and microcontrollers. Spending time getting a better feel for some of the AJAX approach (especially for a guy coming from a ASP.NET/PHP background) has been very eye opening and enjoyable. I’m looking forward to spending some time in the future to revisit the hardware and think about how I would add additional functionality from a data acquisition stand point (sensor data) and end user experience stand point (more buttons, LEDS, using ICs and so on). Finally, once all the hardware is squared away, to actually sit down and produce an enclosure for it all so that I can keep the device running all the time. Taking my usual approach with designing Web Sites to designing an enclosure, I spent some time in Photoshop and produced the very basic design of where this could head.


Written by lewellen

2010-03-01 at 8:00 am

One Response

Subscribe to comments with RSS.

  1. Alex

    2010-03-02 at 2:58 pm

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: