09/18/08 Thursday The weather instrument has outlived its useful life and has been removed. 09/13/08 Saturday The weather instrument was brought partially back to life last night just in time for hurricane Ike. Of the external instruments only the anemometer and temperature were coaxed into working. The base station would still not talk to the slug. Data for the hurricane was recorded manually. The anemometer faded in and out before dying for good during the second pass of the eyewall. This data can be seen at http://www.randomuseless.info/ike/index.html 02/18/08 Monday The system went down during a thunderstorm and now the slug can't talk to the weather base station anymore. The base station also appears to have problems talking to the sending unit at times so it may be fried. The whole thing will be down until at least next weekend. 01/02/08 Wednesday The entire system went down 12/26/07 which wasn't noticed until 12/31/07. Rebooting the slug fixed the problem for several hours but it started missing uploads then stopped completely. Another reboot tonight got things working again but no problems were found with the system. Who knows how long it will stay up this time. 09/10/07 Monday Saturday afternoon DNS quit working so data wasn't being uploaded. Hopefully it's fixed for good this time. All the missing data was re-uploaded. 08/28/07 Tuesday The IP address changed about 2:50 AM and the slug wasn't getting DNS info- properly. Rebooted the modem about 10:40 PM and things started working again. Uploaded the missing data for the day to Weather Underground. 08/18/07 Saturday The anemometer worked for a while then completely stopped registering wind speed. Today it was replaced with a backup unit around 11:00 AM. It turns out the bearing for the anemometer had become stiff so it would not rotate except in strong winds. 08/01/07 Wednesday Something odd happened to the weather station during a storm this afternoon. Even though the high-low alarms weren't set they started going off when there was lightning in the area. In an attempt to shut the alarms off the entire system got unplugged losing several hours of data including all the rainfall data for the day. The anemometer may have been damaged as data from that sensor seemed to be causing the alarms. 07/19/07 Thursday The changeover of Road Runner to Comcast resulted in a change in DNS servers that wasn't propagated properly to the slug. As a result it has been unable to upload data since about 1:15 yesterday morning. A reboot of the cable modem got everything working again and the missing data from yesterday and today was re-uploaded. 07/07/07 Saturday The slug got unplugged during a lightning storm yesterday. It was plugged back in this morning. As a result the data between 07/06/07 11:10 AM and 07/07/07 02:05 AM was lost. Modified the scripts so that the slug will save the weather data if it has to reboot for some reason. Since the July data is already messed up due to the data deletion on the 4th and the gap last night, the unit was taken down for about an hour today to modify the UPS to remove its alarm buzzer and to clean up the cabling. 07/04/07 Wednesday The humidity sensor seems to have worked fine at least most of the time over the last few months. The weird readings may have been due to something temporary (cold temperatures? bugs?). The readings have been added back to the data uploaded to www.wunderground.com. If they freak again they'll be removed. The upload script for Weather Underground was modified to change the precision of some of the variables. Apparently they expect wind speed, wind direction and humidity to be integers. As the old script sent floating point values with one digit to the right of the decimal, Weather Underground was truncating the data instead of rounding. As a consequence, each of these values was being reported low by one unit about half the time. The script now uploads all data with the precision that is expected by the site. The data is still stored internally on the slug to the full precision. Also changed the uploads to round the time to the nearest minute. Weather Underground apparently ignores the seconds so 10:24:52 is recorded as 10:24:00 instead of 10:25:00. This should minimized the number of time stamps that are "off" 1 minute. Deleted all the July data off Weather Underground and uploaded fresh data with all the problems from the previous paragraph fixed. Unfortunately this resulted in all of the data from 7/1-7/2 being deleted from the monthly totals etc. 02/11/07 Sunday The humidity sensor appears to have gone nuts. It has been reading much too low for at least a couple of days. It will probably have to be replaced. 02/01/07 Thursday The total rainfall is sometimes not being reset at the start of a new day in the weather instrument. As a result the rainfall from the previous day is often carried over to the next day. This may be happening because communication with the base station sometimes fails. If it fails during the single call to reset the rain total each morning just after midnight then the total will not be reset until the next day. To minimize the chance of this happening the script has been modified to reset the rainfall three times five seconds apart instead of just once. 01/06/07 Saturday Put the temp/humidity circuit board back in its case to minimize the chance of condesation on the circuit board. It was removed originally because the complete case wouldn't fit in the original fan aspirator. The modified version has enough room, however. There's a slim chance this might fix the anemometer problems but even if it doesn't it should protect the circuit board better than it had been protected. It will also slow the response time an unknown amount. During the change the base unit lost the connection with the outdoor unit for several hours so some of the day's data is missing. 12/14/06 Thursday The anemometer obviously has a problem with rain/fog. Wrote yet another filter to try and keep it from reporting bogus wind speeds until it can be replaced. This filter works like an older one that keeps all gusts below 10 MPH and removes any others that are > 3 times the 5 minute average. This version seems to be working properly. 12/12/06 Tuesday Overnight the anemometer really freaked out reporting a lot of bogus wind gusts. Those were all removed manually this morning. The original software filter that removes any wind values coming from 337.5 degrees has been reinstalled temporarily. There have also been a lot of connectivity problems with Weather Underground of unknown cause so a lot of the data for the last two days is missing. The simple filter didn't work, there was still one bad data point. Removed the filter yet again. Modified the weather script slightly in the hopes that it may be a little more robust. Also modified the watchdog script to help recover network connectivity when the IP address changes in the future. Finally, uploaded the missing data that didn't get uploaded because of the connectivity problems early in the day and fixed some bad data from 12/1 and 12/9 caused by misdirected uploads of later data. 12/09/06 Saturday The weather instrument was offline from 6:00PM to 10:30PM while the router was being used for a science fair project. It is now back up and the missing data has been uploaded. 12/08/06 Friday There was one bad point where the reported barometric pressure was much too low. This was removed manually. Maybe the bad wind reading the other day wasn't due to the anemometer cable but was a random glitch the same way this pressure reading was. 12/07/06 Thursday The software filtering had a bug in it that prevented any wind readings faster than 10 MPH. This only happened for about an hour before it was discovered. The filtering was turned back off. If there's any more bad wind data it will just be removed manually at this point. 12/04/06 Monday The new anemometer cable substantially improved the noise problem but did not totally eliminate it. To trap the bad gusts, filtering was added back into the software. Temporarily, if a gust is greater than 10 MPH and more than 3x the 5 minute average it is considered bogus and is written to a debug file. In the future this will be probably be improved to only flag the above gusts if they are from the NNW as before. The cutoff will probably be tweaked as well. 12/03/06 Sunday Replaced the anemometer cable with a shielded version with larger gauge wire to try and avoid the problem of occasional bogus wind readings. Removed the software filter that had been used to ignore these readings in the past. The filter ignored any wind data from the NNW (all the bogus wind data was from this direction corresponding to 1111b). Northerly wind data should now be a bit more accurate than before matching the accuracy of wind data from other directions. 12/01/06 Friday Replaced the averaging technique for wind direction with something a bit more rigorous. Before this value was a simple average of the angles which results in bogus values when the wind direction is bouncing around north. The method used is explained in the "software" section. 11/28/06 Tuesday At 3:00 AM the cable modem IP address changed which apparently the NSLU2 didn't like so data upload stopped. Rebooted about 7:30 PM and uploaded all the missing data for the day. 11/27/06 Monday The 1st rain greatly increased noise on the anemometer line resulting in bogus high velocities. Modified software to temporarily filter out bad points until the cable to the sensor can be replaced. Also deleted bogus wind gust values from today's data on Weather Underground. Effect of the bogus gusts is still in the average wind speed data but data after today shouldn't be affected. The rain rates and totals were unrealistically high due to a software bug. Basically instead of setting the rain value equal to the current value every 5 minutes it was being summed every 10 seconds over the period resulting in values approx. 30x too high. Fixed the bug then replaced today's data on Weather Underground with the old values divided by 30. Data after today should be unaffected. 11/26/06 Sunday Setup complete. Started uploading to Weather Underground.