Introduction

This web page describes a low cost temperature recorder built around the Texas Instruments, ez430-F2013 development kit. Addition of a 32.768 kHz watch crystal and a small LiON cell and the device records temperatures for up to a week. Once the data is dumped using the USB debugger, reloading the software readies the device for the next use. It is useful for documenting office temperatures or the profile of a shipment.

Background

At $20, the ez430-F2013 development kit is a terrific value. Shortly after getting mine, Texas Instruments proposed a contest to see what projects could come out. This project is both a training and hobbiest entry in the contest to create a low-cost temperature recorder.

Temperature Recorder System

The following photograph shows the three key elements:

Across the top from left to right and down:

Software

The development system uses an IAR Kickstart system that includes a "c" compiler, loader and debugger.

Application Software

Interrupt driven, after power-on initialization, the software checks the first free FLASH word to see if run-length encoded data is present. If not, the watchdog timer is configured as a watch crystal controlled interval timer and each 8 second interrupt measures the temperature. But if FLASH has run-length encoded temperature data, it dumps the temperature time and values to the USB debugger terminal window and log file.

The 16-bit ADC measures the temperature using a built-in thermister device. The built-in 'short' interface is used to provide a relative zero. By happy accident, the low-power, 32 oversampling values are a simple offset and shift from 1/10th degree, farenheiht values. However, the low order three bits suffer from thermal noise and are shift out.

To smooth the data, a five-element, Gaussian filter is use on each data point. This minimizes noise yet preserves most local peak events.

The data from the Gaussian filter is stored in a three level, run-length encoded array. The three levels are the high, middle and low temperatures saved as a byte value and a byte counter. When a new high or new low temperature reading occurs, the last array element is saved as a run-length word and the temperature arrays shifted. The run-length word is written to FLASH and the part goes to sleep again.

Source Code

Operation

Initialization

Operation requires initial loading of the software which also clears the FLASH memory. Thereafter the part is stored, unpowered until use.

Begin Recording

To use the part, plug it into the battery. It takes 40 seconds to initialize the Gaussian array and the LED flashes on and off three times. Unplugging it before the last flash keeps the FLASH unused and the part still ready for use. Put the part in a suitable container and let it run.

End Recording

When the FLASH is filled, the LED will light. But recording can be stopped at anytime by unplugging it from the battery. If the last temperatures are important, change the temperature and wait about 30 seconds for the run-length buffers to flush to FLASH.

Dump Data

To read out the data, plug the MSP430-F2013 into the USB device and bring up the IAR system. Make sure the "FET Debugger" is configured to 'preserve memory contents' and start the debugger. When "run" is activated, the terminal window will list on each line the seconds and temperatures. Save these to a file or the log and load them into a spreadsheet.

Analyze Data

Each timer tick is 8 seconds. Each temperature value is 1/10th of a farenheiht degree. Set the Excel spreadsheet to calculate time from the start and divide by 10 to get the temperature. The following excel spreadsheets are examples of an ice water bath and boiling-to-ice test.

Accuracy

The following graph shows the results of an ice-water bath:

This graph shows the results of a boiling water filled thermos to an ice-water bath:

The initial calibration constants were developed using the 3.5V powered USB device in an ice-water bath. However, the 3.15V battery power apparently resulted in a 28F reading instead of 32F. This data suggests Vcc effects on the thermister need to be quantified but for measuring office environments, it is not needed.