(In Czech: popis měření viz PVmereni.htm, obrázky zařízení viz sensor_img.html, širší komentář viz např. html verzi textu Noc a den kdysi a dnes.)
Old files are renamed yymmdd.png or yymmdd.txt (the very data on PV panel voltage, in volts), dd is the morning date of the night. Overviews of individual months are named like 0608.htm etc. In the first two graphs from June, 0606.htm, the PV panel was a different one, with just two wafers in a series instead of eight ones. I simply assume its voltage to be four times lower.
The current PV panel sized just 1 dm2 is attached to a white concrete cube with a 20 cm side (see sensor_img.html). This is a way to overcome quick temperature changes sun—cloud. However, daily course of temperature is still very apparent, as the cube gets warm by sun and air. A reading for a hot panel can underestimate the true amount of light up to four times (and a cold panel can overestimate even more, when the sun appears suddendly), if a temperature-independent formula for conversion of voltage to illuminance is employed. Jan 04, 2007, I've introduced temperature into the formula in a simple way, the estimates of illuminance during full daylight are better now (unfortunately it's not the sensor temperature, but a value from a 40 m distant thermometre). The graphs, for which temperature data were available, have been replotted.
For low light, causing hardly any warming of the PV panel, the calibration should hold to within 30 per cent, for moonless or overcast sky. In such a situation, much of the light comes from low angular heights, where the PV panel is much less sensitive than a cosine detector. For direct light, with Moon over 20 degrees high, the computed illuminances tend to be up to one half larger than the true ones. One reason might be, the PV panel gets more sensisitive for such angles, the other (more probable?) that Moon has higher albedo in IR (which the panel registers too, about 1 micrometre) – see ev. graph at the bottom of Lunar Calibration – Lunar Irradiance Model page.
Voltage of the panel is measured by a METEX ME32 multimeter, which is able to switch ranges and which is connected to a serial port. An adapted programme Multimeter by Martin Hammerschmid is used to record the readings on my linux computer. Originally, it was each 20 s, since Aug 29 each 5 s and since Sep 22 noon each 1 s. Such maximum sampling frequency was inspired by lightnings catched even with 5s sampling rate in the evening Sep 19 (the whole night afterward, the readings were an order of magnitude larger than they should be... galvanic voltage due to acid rainwater on one of Al-Cu contacts?).
A reference curve, shown by blue colour, plots the data on sunlight from clear sky, as taken from Paul Schlyter's Radiometry and photometry in astronomy. The red curve does the same for moonlight. Total natural light from clear sky is the sum of the two, plus deep night light (airglow, zodiacal light, all stars but Sun and Moon).
Air temperature and dew point, as well as precipitation amounts are included since Aug 15. Inclusion of pressure and wind since Sep 4 (within *_m.txt, one-hour precipitation has been then changed to wind direction – a new heading is meteo_head.txt; precipitation sum there is 3.48× larger than real, due to the collection opening being 3.48× larger than the original one). Plots of all available meteo data since Aug 15, and of all light data since beginning of measurement, are accessible through all.htm thumbnail overview.
Clouds can add light, when they are adjacent to Sun or Moon (their luminance being larger than that of the clear sky) or when they are illuminated from below by man-made light (this is the case in Brno at night). They reduce light when over Sun or Moon, and would reduce light always in nature on deep moonless nights (they are dark against the surrounding clear sky there).
(I got the very idea to measure night amounts of light cheaply, by a PV panel, from my Austrian colleagues, in 2003. Follow-ups are available at our poster PV_measur_Posch.pdf within a directory on Cancer and Rhythm conference, and in my letter “monitoring light by PV power plants”. Just in June 2006 my local collaborator (from Masaryk Uni in Brno), Tomas Miler, helped me to start continuous measurement really. So it's a common project of the University and my Observatory.)
(On 2006 Nov 1, I've solved the daylight saving time problem within the graphs somehow. Just one graph is peculiar, that from Sat/Sun: meteo data from 2:00 CET onwards are shown as if they were CEST, an so there is a glitch between 2 h and 3 h.)