Kipp&Zonen BSRN Scientific Solar Monitoring System - Manual (Page 20)

8
2.0 Sampling Frequency and Accuracy Requirements for BSRN Stations
2.1 Sampling Frequency
2.1.1 Sam pling Frequency of Radiation Measurem ents
The BSRN requires that all radiation variables be sam pled at 1 HZ with an averaging tim e of one
m inute. The final output for each variable should consist of the one-m inute m ean, m inim um ,
m axim um and standard deviation. T his specification is based upon the typical 1/e response tim e
of first class pyranom eters and pyrheliom eters being approxim ately one second. Although som e
instrum ents require the m easurem ent of m ore than one signal for the calculation of a specific
radiation elem ent, the archived data will consist only of the m ean, m inim um , m axim um and
standard deviation of the radiation elem ent.
W hen an elem ent requires m ore than one signal to be m easured or the conversion from the
signal to the final value of the elem ent is non-linear, difficulties arise in providing a single true
sam ple standard deviation for the one-m inute m ean value. This can be accom plished if the
m easurem ents are stored each second and the calculations done later or the data acquisition
system is capable of calculating the param eter each second.
The m ost com m on radiation observation m ade in the BSRN that requires m ultiple signals is
infrared irradiance, where between 2 and 5 m easurem ents are m ade each second, depending on
instrum ent type. There are two m ethods of data handling that provide the exact standard
deviation for the flux and two m ethods that provide an estim ate of the standard deviation if the
standard deviation cannot be calculated.
(1) Observations can be m ade of each of the required signals once per second and stored.
Using this data, the one m inute average can be calculated by applying the appropriate
instrum ent responsivity to each voltage m easurem ent and the appropriate effects of the
case and dom e tem peratures. The standard deviation can then be calculated from the
individually calculated flux values. The prim ary drawback of this m ethod of signal
processing is storage requirem ent associated with collecting one-second data.
(2) W ith the increasing com putation power of data acquisition equipm ent, the determ ination
of the infrared flux can be m ade following the m easurem ent of the appropriate signals.
This would require the conversion of the instrum ent therm opile signal into a flux, and
between 1 and 4 therm istor resistance m easurem ents into tem peratures and then the
equivalent blackbody fluxes. This m ethod requires the em bedding of the therm opile
responsivity into the data acquisition system . Many scientists are unwilling to include
such inform ation in the acquisition stage of an observation because of the risk of error
and the difficulty of correcting the problem when discovered. To reduce the potential of
this type of error, while m aintaining the capability of calculating the standard deviation,
the m ean, m inim um , m axim um and standard deviation of each of the raw signals can be
stored along with calculated infrared irradiance.
Alternatively, given the difficulty associated with observing, storing and calculating the exact
standard deviation of the infrared flux, the standard deviation of the flux can be reasonably
estim ated based by the standard deviation of the therm opile signal. This estim ate assum es that
over one-m inute the tem peratures of the case and dom e rem ain nearly constant and therefore do
not affect greatly the overall standard deviation of the flux. This assum ption is substantiated by
reference to Figure 8.1 that illustrates that a 5% change in therm istor resistance alters the overall
flux by less than ±1.6 % over an extended tem perature range.
The observation of tem perature using therm istor technology is illustrative of a non-linear
conversion from resistance to tem perature using the Steinhart and Hart equation (see Sec. 9.2.3).
In cases where the one-second data is not stored or the conversion of resistance to tem perature
is not accom plished within the data acquisition system each second, the standard deviation of the
tem perature should be estim ated based on the positive standard deviation