Cameras for Astronomical Spectrographs

November 15, 2026

A spectrograph turns a star, a galaxy or a comet into a strip of light spread by wavelength, and the camera at the end of it decides how faint a target you can reach in a night. Long-slit, echelle and fiber-fed spectrographs on research and university telescopes ask different things of a detector than direct imaging does: long exposures, faint signals spread over many pixels, and calibration frames that must match the science frames closely.

This post is for astronomers and instrument builders choosing or replacing the camera on a telescope spectrograph. It uses Finger Lakes Instrumentation (FLI) cooled CCD and CMOS cameras as worked examples. For laboratory Raman and NIR instruments, see our post on cooled CCD cameras for spectroscopy.

What a spectrograph asks of its camera

In a direct image, starlight lands on a small patch of pixels. In a spectrograph, the same light is spread along the dispersion axis across hundreds or thousands of pixels, and across a few pixels in the spatial direction. Each pixel gets a small share of the photons, so read noise and dark current weigh more heavily than they do in imaging. Exposures run from minutes to an hour or more, and every science frame needs arc lamp, flat field and bias or dark frames to go with it. Massey and Hanson's 2013 review, "Astronomical Spectroscopy," is a clear guide to the whole chain.

Pixels per resolution element

Start with the image of the slit, or fiber, on the detector. That width sets the resolution element, and a common target is about two to three pixels across it. Fewer pixels undersample the line profile. Many more pixels spread each resolution element over extra pixels, and each one adds its own noise.

As an illustration, say the spectrograph camera images the slit to 30 µm wide. The FLI PL23042 uses 15 µm pixels, so the slit covers 2 pixels. The AR455, with 3.76 µm pixels, puts the same slit across about 8 pixels. Neither is wrong, but they lead to different binning and noise choices. Our post on pixel size and sampling covers the general math.

Format matters too. The PL23042 has a 2048 × 2048 back-illuminated e2v CCD230-42 sensor on a 30.7 mm square, a format FLI notes has been a standard for astronomy research since its release in 2012. The PL4240 has 2048 × 2048 pixels of 13.5 µm. The AR455 has a 36 × 24 mm sensor with 9568 × 6380 pixels, which can cover a long spectrum or many echelle orders.

Read noise, binning and faint spectra

For faint targets on a dark sky, read noise often limits the result. A CCD can bin charge on the chip before readout, so several pixels are read once and pay read noise once. Spectroscopists often bin along the slit, where the extra spatial resolution isn't needed. Ask our sales team which binning modes your model supports.

CCD readout speed also matters. The PL23042 lists 13 e- read noise at 500 kHz and 22 e- at 3 MHz, so slow readout is the usual choice for science frames, and the faster rate can save time during focusing and setup.

Aurora cameras list 1 e- read noise but bin digitally, because their Sony sensors do not support on-sensor binning. Digital binning adds read noise from each pixel in the bin. Summing 8 pixels at 1 e- each gives about 2.8 e-, which is still low. So the choice depends on your sampling, your sky background and how often you read out.

Cosmic rays enter the decision as well. Long exposures collect cosmic-ray hits, and a common fix is to split the time into several frames and reject the hits in software. Van Dokkum's 2001 Laplacian edge-detection method (PASP) is a widely used tool for single frames. Every extra frame adds another read, which favors lower read noise.

Dark current in long exposures

Spectroscopic exposures are long, so dark current needs care. For exposures of many minutes, FLI usually recommends the PL23042, which has lower dark current and is easier to calibrate, or an Aurora. Kepler cameras are designed for fast, continuous, short exposures, and FLI does not recommend single Kepler exposures longer than about 5 minutes. Stacking shorter frames is the alternative. For dark-current arithmetic on cooled CMOS, see our post on long-exposure dark current.

Red fringing and blue response

Thinned back-illuminated CCDs become partly transparent in the red and near-IR, and light reflecting inside the silicon produces interference fringes. In a spectrum, fringing shows up as ripples along the dispersion that change with wavelength. Deep-depletion sensors use thicker silicon, which reduces fringing and improves near-IR response. FLI offers deep-depletion MicroLine CCDs such as the ML4240DD.

There is a trade-off. Deep-depletion e2v CCDs cannot run in inverted mode, so their dark current is 100 to 200 times higher than that of inverted-mode parts. Plan cooling and exposure times with that in mind, and take flat fields through the same light path to divide out residual fringes.

At the blue end, UV-coated MicroLine models such as the ML4240UV extend response toward the ultraviolet, and the KL400BI model page cites excellent QE at 240 nm. On FLI cameras the F101 window transmits farther into the near-UV and the F116 window farther into the near-IR. Ask for the transmission chart, and choose the window early, because it is set before the camera enters our clean-room build.

Calibration frames and the shutter

Arc lamps and flat-field lamps are bright, so full well matters during calibration as well as on bright standard stars. The PL23042 lists a 150,000 e- full well and the PL4240 100,000 e-. ProLine cameras use a mechanical shutter, which gives clean, timed exposures and darks without help from the spectrograph. Aurora is not offered with a mechanical shutter, so take darks with the spectrograph's own shutter or dark slide closed, or with a blocked light path. To check gain, noise and linearity yourself, see our post on camera acceptance testing.

Mounting and control

Spectrograph cameras bolt to the instrument, not to a filter wheel, so check flange distance and mounting early. Our post on back focus and flanges covers the spacing. For control software, the FLI ASCOM driver supports FLI CCD cameras and filter wheels, and INDI drivers are available for FLI CCD cameras and filter wheels. For telescopes that run unattended, see our post on remote and polar observatories.

Questions to settle before you choose

  • How wide is the slit or fiber image on the detector, and how many pixels should span it?
  • What is the wavelength range, and does the red end call for deep depletion?
  • Will your faint science frames be limited by read noise, sky or dark current?
  • How long are your exposures, and will you split them to reject cosmic rays?
  • Do you need a camera shutter, or does the spectrograph provide one?
  • What flange distance and mounting does the spectrograph need?

Talk to our sales team

Send our sales team your spectrograph's slit or fiber scale, wavelength range, exposure plan and mounting, and we will suggest the standard ProLine, MicroLine, Aurora or Kepler configuration that fits. If an instrument needs a modest change to a standard camera, we are glad to discuss it. Request a quote or email sales@flicamera.com.

FLI cameras are Made in USA, and our sensor chambers are assembled in our 1,000+ sq ft ISO Class 7 clean room. See how on our in-house manufacturing page.