In September 1991 a protocol for quality control of large shaped irradiation fields was started in our department. In vivo dosimetry with semiconductor detectors was used to measure the absorbed dose and patient positioning was checked with portal films weekly. First, we set a computed dosimetric system yielding dosimetric values in real time and allowing their easy storage. Then, we calibrated the diodes and determined the correction factors for each of them outside standard conditions. Entrance dose, exit dose and midline dose were measured in 62 patients undergoing supradiaphragmatic radiation therapy for Hodgkin's lymphoma. The exist dose was measured weekly to assess treatment repeatability. High agreement was observed between measured and calculated doses; repeatability was also high, since only 6% of exit dose measurements exceeded 5% of the first determination. In 33 patients portal films were obtained in the first treatment session, and thereafter weekly, to assess mispositioning relative to simulation (reproducibility) and from one session to another (repeatability). A small systematic error was detected in both longitudinal (x = -3 mm; SD = 3.7 mm) and transverse (x = -2 mm; SD = 3.4 mm) directions. Statistically significant errors (> 6 mm) were observed in 14% of patients. Reproducibility was excellent. The protocol reported on in this paper not only helps avoid systematic dosimetric and/or positioning errors in the patients, but also helps identify the main causes of uncertainty and thus remove them.