An ongoing research project in the area of radiation monitoring employing the Timepix technology from the CERN-based Medipix2 Collaboration profits greatly from optimizing the precision of the position and energy information obtained for the detected quanta. Wider applications of the Timepix technology as a radiation monitor also puts new demands on the precision and speed of the energy calibration. We compare the analog signal in pixel front-end electronics for different sources used during detector evaluation and energy calibration. We use the direct measurement of the analog signal from the pixel preamplifier and comparator to characterize pulse shape differences for different sources, e.g. internal test pulses, external test pulses, ionizing radiation, etc. and study their interchangeability. Accurate per-pixel energy calibration of the Timepix detector enables the direct measurement of the energy deposited by different types of ionizing radiation. The energy calibration process requires the application of a known charge to front-end electronics of each pixel. The small pixel size limits use of the radioactive sources. The 59.54 keV line from 241Am is commonly used as the highest point in calibration curve. The heavy ion dosimetry as encountered in the space radiation environment requires a considerable extrapolation to the energies in the MeV range. We have observed that for energies around and beyond 1 MeV the response of the Timepix's front-end electronics no longer follows the extrapolated calibration function. We have investigated this non-linearity and identified its source. We also propose both hardware and software solutions to suppress this effect. In this paper we show the impact on pixel calibration and the subsequent energy resolution for different detector settings as well as the resulting power consumptions. We discuss the parameter optimization for several different real-world applications.