ABSTRACT
Trojan Technologies Inc., 3020 Gore Road, London, Ontario, Canada N5V 4T7 Contact: gfang@trojanuv.com ABSTRACT: The Draft USEPA Disinfection Guidance Manual (USDGM) [1] suggests that the spectrum of mercury arc lamps changes as the lamps age. Spectral measurements were made for 100-hour and aged lamps to determine the magnitude of the spectral change. Lamps were of two different lengths; some were aged with electronic ballasts and some aged with magnetic ballasts. Lamps were aged inside water-cooled sleeves and others were aged in air. No spectral change was found on some aged lamps, but some changes in the low wavelength region on other aged lamps were observed. However, little effect of the spectral change on water disinfection is expected. Prediction would suggest 0.3% change in equivalent dose, which is negligible. INTRODUCTION UV mercury lamp output will decrease over time. For medium pressure (MP) lamp degradation, shorter germicidal wavelengths degrade faster than longer wavelengths. However, little information was available on these spectral changes and the conditions under which changes occurred. Various observed depreciation of MP lamp in the UVC versus UVB region was reported, from 50% versus 20% within 800 hours [2] to 18% versus 15% for an average of 4 lamps in 4,000 hours [3]. Another ozone-free MP mercury lamp deteriorated by up to 22% in UV output compared with the visible output at half of the lamp life [4]. Trojan Technologies conducted an investigation using a number of Trojan MP mercury arc lamps aged under a number of different conditions. METHOD Lamps were of two arc lengths, 25cm and 61cm. Some of the 25cm lamps were aged using electronic ballasts in two different sleeve sizes. Some of the 25cm lamps were aged by the lamp supplier using electromagnetic ballasts operating at 50 Hz. All of the 6I cm lamps were operated by magnetic ballasts. Lamp outputs were measured in Trojan's Lamp Research and Measurement Laboratory. This facility uses a double-monochromator and photomultiplier tube to measure the spectral output of lamps operating in air from a distance of 3m. This system has been carefully adjusted to provide 1nm resolution. The distance from the lamp to the detector ensures that even long lamps subtend a small angle at the detector. The system is calibrated with NIST-traceable Quartz-Tungsten-Halogen and Deuterium sources. These sources have a stated uncertainty of approximately 3% and 5% respectively, with small variations with wavelength. UV output was also measured at the suppliers' facilities by similar spectroradiometers with I nm resolution. One of the primary concerns expressed in the USDGM is that sensors may provide falsely high readings if the short wavelengths decrease with lamp age and the sensor is overly sensitive to longer wavelengths. For this reason, it was decided to normalize each wavelength output reading in a given spectrum to the total energy in the 365nm peak (362nm to 368nm). This peak was chosen because measurement is less affected by humidity and other factors, and the calibration accuracy of the measuring instrument is better in these longer wavelengths. This peak also falls within the range of an unfiltered SiC detector, and so disproportional changes in germicidal wavelengths relative to this peak could affect the accuracy of unfiltered sensor readings used to monitor germicidal UV (200-300nm). RESULTS The spectral output of two 25cm arc lamps and the normalized spectra are shown in Figure I. The data are for a 100h reference lamp and for another lamp that was operated by an electronic ballast in a water-cooled sleeve for 5000h at a power level between 800 and 2650W, with half of the time spent at maximum power. The lamp output decreased with age. In order to clearly show the spectral change, the data were normalized by the energy in the 365nm spectral peak, using the sum of the values from 362nm to 368nm. A closer examination of the spectral change can be conducted by taking ratios of lamp outputs. Since a direct ratio would exaggerate small differences at individual wavelengths at low lamp output, wavelength ranges were used. The lamp output spectrum was divided into ranges that included major output lines and also had approximately equal energy output. Taking a ratio of the sums of each range normalized to the 365nm peak provides clear indication of the relative spectral output of the two lamps while removing the effect of a uniformly low output, or of power supply variations. The selected wavelength ranges, along with the ratios of the output, can be seen in Figure 1.
