For the start of the new year, I decided to make a few changes to the layout of the weather data. For a while I had been generating and uploading graphs of the wind speeds recorded by the station but I had not got around to altering the code for the pages to display this extra data. The time had come to correct that omission and while I was doing that, I also altered the graphing code to produce charts of the wind speed against direction. I decided too, that any notes would be placed on a separate page, this one.
During the stormy weather of December I had noticed that there were quite large discrepancies between the recorded wind data and that displayed on the weather station console. I had at times thought about what was recorded for each ten minute period, was it the maximum gust during that time or the most recently read data? The conclusion being that it was the latter. In an attempt to improve on this I have switched from a ten minute save interval to a one minute interval. The most obvious result of this is the the WS3600 data logger now has only enough memory to store about 28 hours worth of data, necessitating an upload to the PC every 24 hours. Since the PC is not always on, and I don't want it to be, I needed to make some modifications to automate the upload. I now have the PC powered up by a timer. The weather software starts up, delayed slightly by Startup Delayer ( http://www.r2.com.au >) to give Windows time to get itself sorted out a bit (It is an old Cyrix 686 P133+ machine running Windows 98). After a further delay to allow the data to be uploaded from the weather station, Cobian Backup 7 first of all backs up the data and then shuts down Windows before the timer shuts off the power again. (Cobian can be found at http://www.educ.umu.se/~cobian/cobianbackup.htm or for version 8, http://sourceforge.net/projects/cobianbackup )
Increasing the sample rate for the data also led to an encounter with the graphing limitations of Excel. Normally a maximum of around 32000 data points can be graphed and a one sample per minute collection rate leads to roughly 44600 for a thirty one day month. The simple solution would have been to revert to the original one sample every ten minute but since I wanted to maintain the increased accuracy I decided to sample the data set instead. I have therefore reduced the amount of data to graph by using the total, average or maximum of every ten measured samples as appropriate.
The battery power of the thermo/hygro unit once again became an issue in November. Whilst generating the graphs of October data I noticed an unusual spike in temperature which led me to discover missing data shortly afterwards. For a couple of days until I could get time to replace the cells in the thermo/hygro unit I managed, more or less, to maintain the wireless data link by moving the display console closer to the sensors and syncronizinging with the PC manually. I also managed to do a reset on the display unit during this time which reset the sampling rate to once every five minutes! The main thing is that this has reminded me of a job I intended to do a while ago, use the cable link socket on the thermo/hygro unit to connect to a battery box inside the house so that it is easy to change them when the time comes. No looking for ladders and climbing up when it is almost always raining and windy. I do, however, need to establish which wires do what before I can do this.
Back to Weather for 2007