The readout noise of a H2RG HgCdTe NIR detector from Teledyne is measured at a temperature T=110K. It is shown that a Fowler mode with n = 240 allows to reach a noise of 2.63e (single read). A description of the power spectrum in terms of 3 parameters reproduces the variation of the noise as a function the number of Fowler samples, as well as its dependence on the periodicity of the sampling. The variance of the noise decreases with frequency with an effective power of 0.62 in our measurement domain. The behaviour of the detector under different experimental conditions can then be predicted.
THE APPARATUSThe measurements described in this paper were carried out in a dedicated setup built to evaluate Hawaii 2RG (HgCdTe) detectors from Teledyne. The detector was on loan from LBNL in view of the evaluation of its performance when used in a spectrograph for the JDEM project. 1 In the following measurements the cryostat is operated at a temperature of 110K with fluctuations smaller than 0.1 K. The polarisation of the substrate was chosen as V sub − V reset = 0.4 V in all the data analysed in this work. Additional experimental details are provided in a previous paper 2 where the conversion factor of our setup at T = 110K was evaluated to be 2.042 e/ADU for a single pixel. The goal of the present measurements is to lay the ground for a determination of the frequency distribution of the noise, so as to be able to predict the behaviour of the detector in different experimental conditions (Fowler samplings, interval between groups, etc...). The merit of the power spectrum analysis is also the link it provides with the physical processes in the CMOS semi-conductor, such as the trapping and detrapping of electrons with a distribution of time constants.
THE MEASUREMENT METHODIn the 2Kx2K H2RG detector, a window of 31x31 pixels has been selected. The clock frequency is 100khz, and the time needed for the readout of 1 frame is δ = 17.5ms . The non destructive frame readouts are organised into groups of 250 frames, read at the clock frequency of 100kHz and separated by a time interval which can be tuned, with the clocking of the pixels stopped. Taking into account the readout of the frames, the periodicity of the readout of a given pixel between 2 consecutive groups varies from 5s to 13s. All the frames are stored on disk for further analysis. The noise is characterised by two variance measurements:• the frame to frame variance, with the readout interval of δ = 17.5ms for a window frame 31x31 as considered here, is convenient for a calibration (conversion gain) of the detector and its readout. It was studied in [2];