Tim e is critical with respect to the frequency of the m easurem ent sequence and the absolute tim e of the observation. This requires that the clocks for all observations be m aintained to within ±1%
of the averaging period used for the m ost frequent m easurem ents. For a one m inute average this equates to a tim e accuracy of 0.6 seconds. Because of the difficulty in m anually setting a clock to
better than one second, this tim e accuracy was relaxed to one second at the BSRN Science and Review W orkshop (Boulder, Colorado, USA, 12-16 August, 1996).
The autom atic determ ination of tim e to within one second can be easily achieved on portable com puters using one of three com m on m ethods: (1) tim e-synchronization with the G lobal Positioning System (GPS) satellites; (2) conversion of radio frequency tim e signals sent out by national standards agencies; and (3) through tim e updates obtained via the internet.
C om puter clocks can be synchronized to within 2 m illiseconds of UTC with sim ple and inexpensive GPS system s that consist of a sm all antenna (< 100 m m diam eter) and a decoder box that can be plugged into the serial port. More expensive bus system s can increase the accuracy of the synchronization to better than 1 m icrosecond. The need for an antenna and a view of the sky, m ay reduce the applicability of such system s in built-up areas, or where the unit is deep within a building com plex. In certain regions of the world (North Am erica, Australia, China) a sim ilar system has been developed, CDMA (indirect GPS) that operates through cellular networks and is m aintained by the cellular providers. CDMA, where available, is sim ilar to GPS, but uses a m uch sm aller integrated antenna and will work within office com plexes. T he tim e is kept to an accuracy of better than 10 m icroseconds.
Typical Meteorological Measurement Field Specifications
Measurem ent Resolution Uncertainty
Air Tem perature 0.1 °C ±0.3 °C
Dew Point Tem perature 0.1 °C ±0.5 °C
Soil Tem perature 0.1 °C ±0.3 °C
Relative Hum idity 1% ±7%
W ind Speed 0.5 m s-1 ±5% or ± 2 m s-1
W ind Direction 5° ±10°
Accum ulated Precipitation 0.2 m m greater of ±0.2 m m or ± 2% of total Precipitation Intensity 0.2 m m h-1 greater of ±0.2 m m h or ± 2% of total-1
Snow Depth 1 m m greater of ± 10 m m or ±1% of value
Atm ospheric Pressure 0.1 hPa ±0.5 hPa
Table 2.2. Recom m ended m easurem ent requirem ents for ancillary m eteorological variables.
Many national m etrology institutes (NMI) transm it tim e signals, based upon the national tim e standard, at short-wave radio frequencies. These radio tim e signals can be received thousands of kilom etres from the transm itter, depending on ionospheric conditions. W ith the proper decoding and correction for tim e delays, the accuracy of the tim e signal can be better than 1 m s. Like the GPS system , short-wave radio receivers require external antennas. W hile radio signals are as accurate and less expensive than GPS system s, and m ore accurate than m ost internet system s, the radio tim e signal is losing popularity over the ease of use of GPS and internet system s.
C om puter tim e-synchronization has advanced rapidly since the onset of the m odem com m unication and the internet. The tim e obtained in this m anner is not usually as accurate as the GPS and radio tim e signals, the need to synchronize a clock to better than one second can norm ally be accom plished through these m ethods. Many NMIs provide analog-m odem dial-up links and the associated software required to set a local com puter clock to UTC. The software can translate sim ple telephone codes that allow corrections to be m ade for the signal propagation delay. Using this m ethod, com puter clocks can be set to within several m illiseconds of UTC. New, high speed m odem s that use digital processing can add a variable delay of up to 140 m s and a 20 m s jitter beyond the delay due to signal propagation. Further uncertainties associated with the use of m odem s using digital technology. These telephone services are also degraded when
T hree locations where further inform ation on tim e synchronization can be found are:
9
(1) N IS T : http://www.boulder.nist.gov/tim efreq/service/its.htm (2) PT B: http://www.ptb.de/en/org/q/q4/q42/ntp/_ntp_main.htm and (3) T ime S ynchronization Server: http://www.eecis.udel.edu/~ntp/index.htm l.
com m unications are m ade via satellite links (long-distance services), but again, correction can be applied. Obtaining a true tim e via the internet is m ore difficult than with m odem s because of the increased variability in response tim es of the service. To overcom e som e of the variability associated with these delays, the Network Tim e Protocol (NTP) was developed. The advantage of N T P is its ease of use and its ability to be used on m ost com puter platform s. Both the NTP service and the software required to use it is freely available on the worldwide web, along with details of its operation . It m ust be noted that the NTP service will not operate correctly if the com puter9 being updated is behind a firewall, unless the firewall is set up to allow N T P packets through. NTP is a very effective m eans of standardising com puters on a local area network. For rem ote locations, a high quality GPS system associated with the network server and N T P can synchronize the LAN to better than 1 m s of the GPS tim e base.
T hese signals can either be incorporated directly into m ore advanced data acquisition system s or set on a daily basis for less advanced system s or externally controlled data acquisition system . The software operating m any PC Card Data Acquisition System s (DAS) uses the com puter clock for tim e inform ation. Therefore, m aintaining the com puter clock tim e as accurately as possible is essential. For a PC with a tim e gain of 10 seconds per day, the clock would require autom atic updating approxim ately once each hour. Com m unication with m any new external DAS can be accom plished through local area network protocols, especially Internet Protocol (IP) addressing that provides an easy m eans of ensuring that all system s, com puters and dataloggers, m aintain tim e precisely.
For data acquisition system s with internal tim e keeping, the sam e clock correction m ust be m aintained for the relative sam pling rate, while the absolute tim e can be corrected during data processing.
2.3.2 Data Acquisition System Accuracy
The specification for data acquisition system requirem ents for BSRN radiation m easurem ents is set forth in the report of the W CRP BSRN Im plem entation W orkshop, Davos, Switzerland, 6 - 9 August 1991. Uncertainty of the com plete system (digital voltm eter (DVM), scanner (m ultiplexer) and cabling) was set as ±0.01% of the reading or ±1 :V, whichever is greater. If the overall accuracy of the data acquisition system is greater than 10% of the accuracy required for the observation (e.g., 1 W m for an instantaneous uncertainty of 10 W m ), then a high quality-2 - 2 pream plifier should be used. If such an am plifier is required, it should be placed as close to the signal transducer as possible so that line noise is not also am plified, as it would be if the am plifier were associated with the DAS. C are m ust be taken to ensure the tem perature stability of the am plifier, either through am plifier selection or tem perature control, so that tem perature influences do not increase significantly the uncertainty in the m easurem ent.
Each instrum ent should be scanned at least once per second with the analog signals integrated to provide one-second values. On those system s where integration tim e is program m able, the shortest period to be used for sam pling of radiation signals is one power line cycle (PLC).
Sam ple averaging and special filtering techniques are required when em ploying m any new high-speed DAS system s, especially those that are directly connected to a com puter bus, to reduce uncertainties associated with electrom agnetic noise.
W hile the prim ary aim of the BSR N is to obtain accurate radiation fluxes, the accuracy of the data acquisition system used in the collection of ancillary data should be com m ensurate with the general aim s of the program . Therefore, ancillary m easurem ents should be sam pled and recorded following the sam e principles as applied to the radiation observations.