Thermal Tweeter is hacked together from a SparkFun thermal printer and the web platform internet development board. No computers are used to query twitter or run the printer, this is a 100% stand-alone Twitter appliance.
Keep reading to find out how it works.
You tweet, the web platform contacts Twitter and grabs the newest tweets, the tweets are formatted and output to the printer.
Is this wasteful? You bet. And the thermal chemicals on the paper may cause cancer – don’t eat it! Thermal Tweeter is an homage to the waste generated by checkout receipts. Unlike a store checkout where you always get a receipt, you’re in charge of the Twitter waste. Save the earth, don’t Tweet. Or whatever.
Setup is really simple. The thermal printer connects to the web platform with one data wire. A second wire connects the common ground between them.
A serial port assigned to pin IO7 on the web platform outputs serial data to the printer RX pin at 19200bps. See the web platform intro article for more about assigning hardware peripherals with peripheral pin select. See SparkFun’s Arduino code and the printer datasheet for the simple control commands.
The printer requires a beefy 5volt power supply for the heating element (1.5A+). We used a 1.5A 5volt wall wort.
The web platform can be powered from USB or a wall wort. Our best wall wort is powering the printer, and we want to use the web platform USB connection for debugging info, so we used the USB supply. USB connection is only for power and debugging info, it doesn’t contribute anything to the Twitter printer.
Firmware is built on the @tweet_tree source. It searches twitter for the latest tweets to @dangerousproto and prints them.
“created_at”:”Wed, 22 Jun 2011 16:19:17 +0000″,
“text”:”@dangerousproto Missed it =). Didn’t know ustream doesn’t work on the iPad.”,
Tweets, user names, and IDs are extracted from a simple JSON search results feed like shown above. JSON is a way to share data, like XML, but way more compact. The web platform stores the text in a giant array. After all new Tweets are processed, the text is sent to the printer.
The ID tag is especially important. The highest ID number is stored and appended to the next search URL. Twitter only sends tweets newer than the ID number so we only get fresh commentary.
The printer is controlled by 19200bps serial UART output from the web platform. Simple commands configure it. Using it is simple, send a line of ASCII text and print it with a line feed character (0x0a).
We used a 32 character wide row. Row breaks are inserted into the tweet by the web platform without regard for the location of words, words will often be split to two lines.
Server responses are dumped out the USB serial connection to help with debugging. The activity can be viewed on a PC with a serial terminal set to 115200bps. Even if you don’t have a thermal printer, you can program the firmware into your web platform and watch the twitter queries.
Twitter doesn’t like it when you hammer their server with requests. Thermal Tweeter waits at least 30 seconds between updates. This should be responsive, but stay within the 150 search/hour limit set by Twitter.
Don’t kick Twitter when it’s down. If Twitter doesn’t reply or responds with an error, Thermal Tweeter retries twice and then waits a few minutes before trying again. It’ll be easier for Twitter to get their act together if we don’t flood them with search requests.
Taking it further
Originally we wanted to add a keyboard for two way communications. Type a message and it is pushed to Twitter on enter, or every 140 characters.
Microchip’s non-open source TCPIP stack is a pain to work with because we can’t share all our code. We’d prefer to use the FreeRTOS port for the web platform instead.
Neither of these ideas made it into the final design due to our server battle over the last two weeks.