High-Precision Photometry with a scientific CMOS Camera: I Lab Testing of the Marana camera

TL;DR

This paper presents a comprehensive laboratory characterization of the Marana CMOS camera for high-precision photometry, measuring noise, dark current, PRNU/DSNU, QE, and window transmission under HDR and FFR modes following EMVA-1288. It demonstrates low read noise (~1.57 e-), high dynamic range (≈93 dB), and strong linearity, while highlighting mode-dependent artifacts such as the HG–LG transition and amplifier glow. The authors provide detailed performance metrics, compare against CCD benchmarks, and discuss calibration strategies, including Anti-Glow and spurious noise filtering, to enable precise photometry. The work lays the groundwork for on-sky time-series photometry comparisons at NGTS, aiming to establish the CMOS Marana as a competitive alternative to traditional CCD detectors for high-cadence astronomy.

Abstract

Scientific CMOS cameras are becoming increasingly prevalent in modern observational astronomy. We assess the ability of CMOS image sensors technology to perform high-precision photometry with a detailed laboratory characterization of the Marana 4.2BV-11 CMOS camera. We characterise the camera in the Fastest Frame Rate (FFR) and High Dynamic Range (HDR) modes. Our evaluation includes read noise, dark current, photo response and dark signal non-uniformities, quantum efficiency and window transmittance. The read noise is found to be 1.577\,e$^-$ for the FFR mode. For the HDR mode the read noise floor is measured to be 1.571\,e$^-$ for signal levels below approximately 1800\,e$^-$. The bias level shows dark signal non-uniformities with values of 0.318\,e$^-$ and 0.232\,e$^-$ for FFR and HDR mode, respectively. Pixel well capacity reached 2366 e$^-$pix$^{-1}$ for the FFR mode and 69026 e$^-$pix$^{-1}$ with a dynamic range of 93\,dB for the HDR mode. The camera demonstrates good linearity, yielding linearity errors of 0.099\,\% for FFR mode and 0.122\,\% for HDR mode. The uniformity across the image arrays show a photo response non-uniformity of 0.294\,\% for the FFR mode and 0.131\,\% for the HDR mode. The dark current displays a noticeable glow pattern, resulting in mean dark current levels of $1.674\pm0.011$\, \eps\, for the FFR mode and $1.617\pm0.008$\,\eps\, for the HDR mode at a constant temperature of -25\,$^\circ$C. We measured the quantum efficiency across the visible spectrum, with a peak of of >95\,\% at 560\,nm. Our tests indicate that the Marana CMOS camera is potentially capable of performing precise photometry.

Figures (25)

• Figure 1: Optical bench setup for the Marana camera. The black cylindrical component is the light tube, which houses the light source with a diffuser mounted at the opposite end. A spare diffuser is also visible in the setup.
• Figure 2: Bias level, light source signal and temperature for the Marana camera over time. Top: Average bias level value of a series of images at minimum exposure time and at dark environment. Middle: Average light source signal of a series images at a fixed exposure time. Bottom: Temperature values measured from the camera.
• Figure 3: The optical bench of the QE measurement of the Marana CMOS camera, including the UV light bulb. Each instrument is labelled with yellow coloured text.
• Figure 4: Photon transfer curves for the Marana camera at -25 $^{\circ}$C. Left: The FFR readout mode. Right: The HDR readout mode. The solid blue line is a best fit linear model. The data point with the highest variance is marked with a blue star, and indicates the FWC value. A zoom-in panel illustrates the transition region.
• Figure 5: Top left: An example sub-frame showing the illumination difference across the LG and HG transition region. Top right: the binary image is highlighting pixels that exceed the threshold noise value. Also shown are three example pixels processed with the LG, HG channel and in transition, indicated with the red square (Y,X = 310, 125), blue triangle (Y,X = 140, 130) and green circle (Y, X = 215, 130) respectively. Bottom: The ADU counts for the three example pixels as a function of frame number, with the RMS calculated for each pixel.