Announcement 2011-09-15

ERROR IN LASER CEILOMETER CLOUD BASE FILES, 2nd NOVEMBER 2010 - 10th SEPTEMBER 2011

The Problem
The Vaisala LD40 laser ceilometer cloud base altitude files contain an error in the observation times for the period 2nd November 2010 - 10th September 2011, inclusive. The recorded times were offset relative to the actual times by:

• +4 hours 00 minutes at the beginning of the affected period (i.e. 2nd November 2010)
• +3 hours 57 minutes at the end of the affected period (i.e. 10th September 2011)

the difference in offset being caused by a slow drift in the instrument's internal clock (see below). The plots of the latest 24 hours' worth of laser ceilometer data, which are available through the Facility's supplementary website http://www.mstrf.eclipse.co.uk/, have been free from this problem since at least March 2011.

The NASA-Ames cloud base data files, and their corresponding quick-look plots, were removed from the BADC on 12th September 2011. Data files for 11th September 2011 onwards are not affected by this problem.

Replacing the affected files
Replacement files, with corrected observation times, can easily be generated for individual days on request. However, this will rely on a manual inspection of the raw data files on a day-by-day basis. Owing to the underlying and ongoing problem of slow instrument clock drift, an automatic software solution is required in the longer term. All of the removed files will be recreated once the updated software is operational. A confirmation will be added to this page when the task has been completed.

Cause of the problem
Data messages produced by the instrument contain a date-time stamp which is generated by its internal, free-running clock. The latter

• is not very accurate - as demonstrated above, it drifts by around 15 s per month
• is non-trivial to resynchronise - requiring someone to connect to it with a laptop in the field

The data messages are captured by a program (written by a member of Met Office staff) which is run on a network time protocol (NTP) enabled computer. This software writes all messages which arrive on a given UT day (as determined by the computer rather than by the date-time stamps of the messages) to a single file. Consequently it is possible to determine the actual observation times to within the nearest minute (i.e. the interval between data messages) without having to consider the date-time stamps produced by the instrument. Clearly it would be more-appropriate to capture the data messages together with a date-time stamp generated by the acquisition computer. Nevertheless, the self-adjusting correction method will be adopted when the NASA-Ames file generation software is updated in the near future. At present the instrument's date-time stamps are corrected by a hard-coded offset.

When the current time error came to light, it was thought that an incorrect offset value had been applied when it was last updated on 2nd August 2010. On that occasion, the instrument's clock had been passively reset as a result of the electricity supply to the site being down for most of the day. However, further investigation revealed that the internal clock had again been passively reset at 09:14 UT on 2nd November 2010. This appears to have been caused by the first of three brief mains disruptions which were detected by the Uninterruptible Power Supply (UPS) units. The LD40 is not powered by a UPS. The clock reset went unnoticed at the time.