Total ozone has been monitored with high accuracy at many sites for a relative long period of time. This is not the case for UV. The main reason is that UV is difficult to measure accurately. Extensive monitoring started first during the 1990s. However, there are some data series still in operation that were initiated at earlier dates. This paper will discuss one such series from Norrkoping, Sweden that began in March 1983. The instruments used to measure UV radiation have some characteristics producing an output signal that is not directly related to the UV. Here, these problems are discussed and corrections are applied. The goal is to produce a homogenous and an accurate set of UV data as is possible. An analysis of the various sources of uncertainty is done and summarised as total uncertainty. This was found to be about 6% for the yearly values of CIE weighted UV. However, the precision is much better at about 0.6%. Finally, some results are shown for yearly and seasonal UV. The trends for the full period 1983-2003 are tested for significance. The spring, summer, autumn and yearly data showed a significant increase in UV of respectively 0.8, 0.3, 0.8 and 0.5% per year. To get some indication of the cause of the variation of global radiation (indicator of cloudiness) and total ozone measured at the same site were also processed. A small, but insignificant, decrease of total ozone was found, whilst global radiation showed an increase. It seems that cloudiness is the major factor for the observed variation in UV for this period. The data are available on the internet.