Abstract: The upper-air sounding data are one of the basic data applied in meteorology. The analysis and forecast of synoptic situation and weather system are directly affected by the accuracy of the data. Therefore, for the purpose of improving sounding accuracy, L-band electrical radiosonde has been employed broadly since January, 2002. Its data acquisition rate, accuracy and reliability, and the degree of automation have improved significantly since then. The new radiosonde has been implemented with carbon-film hygristor instead of Gold beaters skin hygristor to measure relative humidity. The measurement is made more accurately by the high sensitivity, short lag time, and fast response in low temperature of upper-air of carbon-film hygristor. In addition, the carbon-film has good consistency with device calibration line, which does not need to calibrate one by one. So the cost of production can be saved. The carbon-film hygristor, however, is significantly influenced by temperature. Data errors will be yielded by failure to calibrate it accurately. There are many researches on carbon-film hygristor performances but still without very encouraging results. Because of the complexity of upper-air observation, there is not a widely accepted standard to test sounding instruments' performances by the international community. Right now, the relative humidity data are corrected mainly by the radiosonde manufacturer. Whether the applied revising formula is reasonable or not, how is the corrected results, all need to be tested by objective analysis. However, a large number of the actual release record still shows a huge error in relative humidity sounding curve. Especially the unstable relative humidity data in the clouds which decrease strikingly with the increased altitude, and the decision of cloud location are affected directly. The accuracy of temperature detection and calculation of the geopotential height are thereby lowered. The process of correcting relative humidity data of the new L-band sounding system is described. Using the high-resolution humidity calibration equipment, many static-state testing experiments of the carbon-film hygristor are conducted at temperatures from-30 ℃ to 30 ℃, by which a substantial amount of data is acquired. Through the test results, the various properties of this element in detail are noticed. Also a mistake in correcting formula by the manufacturers is found because of their negligence to test the influence of paralleling 1 Megohm resistance in the circuit. Based on the analysis and calculation on these lab data, the correction equation of relative humidity is re-established. The comparison and analysis of correction on real upper-air sounding data of relative humidity indicate that not only the accuracy of relative humidity data is improved after correction, but also the detailed variations in relative humidity at the upper atmosphere during high humidity range can be clearly manifested. So the ability to judge vertical position of cloud layer is improved, so is the temperature measurement precision.