One star, several different tasks

Imagine a night-sky image: a faint star is barely visible, while two close stars nearby merge into a spot. Enlarge the image on screen by a factor of two. The spots grow, but the faint star sends no additional photons to the camera, and stretching pixels does not separate the merged pair. What should change in the telescope or observation?

Light collection determines how much signal enters; resolution determines which angular details the system distinguishes; sensitivity concerns how reliably signal stands out against noise. Field of view—the patch of sky included in the image—is another question. One advertised “magnification” does not describe all these capabilities. ESO: light collection and sharpness.

Let us follow starlight from the entrance aperture to camera numbers. Every number below is a teaching assumption, not a measurement of a particular telescope. We first compare ideal open apertures, then separately account for losses in a hypothetical reflecting system. This reveals why “dim” and “blurred” call for different checks.

The entrance opening intercepts photons

The aperture is the working entrance for light. In a simple telescope, it is often specified by the objective or primary mirror diameter. But a cell rim, clips, secondary mirror or supports may block part of that entrance. The outside diameter of the housing is therefore not the collecting area.

Compare circular openings of 100 and 200 mm. With equal incident photon flux per square metre, band, exposure time and downstream efficiency, the second intercepts four times as many photons. It is twice as wide and twice as high: area scales as diameter squared. NASA: why collect light?.

Twice the diameter, four times the area
D = 100 mmD = 200 mm

Same scale, without central obstruction. Areas: 78.54 versus 314.16 cm². This is entrance geometry, not a guarantee of equal efficiency or brightness in each pixel.

This comparison concerns the total flux from the same point source. It does not yet promise four times the brightness in every pixel of a nebula. For an extended object, we must also fix the angular area of sky per pixel, band and optical parameters. Even starlight can spread over different numbers of pixels.

An open entrance is only the start of the budget

A photon can strike the mirror without producing a useful record. Some light is absorbed or scattered; filters and protective windows also transmit less than everything. The detector converts only a fraction of the photons into recorded electrons. Quantum efficiency characterizes that fraction. These coefficients depend on wavelength, surface condition and the particular detector; they are not fixed properties of the word “telescope”.

Professional verification measures the complete efficiency using known sources or standard stars, separating atmosphere, optics and camera. The Liverpool Telescope throughput study follows this approach; its measured numbers are not our teaching coefficients. Smith and Steele: modelling and measuring the whole chain.

For our budget, take a 100 mm reflecting system with a 30 mm central obstruction. At the entrance, after passage through the atmosphere, assume 10 000 photons per square metre per second in the selected band. Each of two mirrors reflects 90%, other elements transmit 95%, and quantum efficiency is 80%. Finally, we sum only a selected region around the star: suppose it contains 80% of its signal.

A teaching signal: from photons to extraction
  1. At the clear aperture71.47 photons/s
  2. After two mirrors and other optics55 photons/s
  3. After detector conversion44 electrons/s
  4. In the selected star pixels35.2 electrons/s

Hypothetical D = 100 mm, obstruction 30 mm; Φ = 10 000 photons/(m²·s) at the entrance, R₁ = R₂ = 0.9, τ = 0.95, q = 0.8, extraction fraction ε = 0.8. Not a measured telescope. The final rate is rounded to 35.2 electrons/s for the noise example.

Over 60 s in nine pixels

Star: expected electrons
2,112
Sky: expected electrons
600
Dark current: expected electrons
10.8
Readout: contribution to variance
81

Sky: 10 electrons/s over the extraction region; dark current: 0.02 per pixel per second; one read with rms 3 electrons per pixel. The SNR denominator adds variances, not standard deviations.

STScI · STIS 6.4 · Liverpool Telescope · Technical Note 1

The final step does not mean that the remaining electrons disappear from the camera. They lie outside our extraction region. We can enlarge the region, but then more sky and pixel noise enter the sum. The telescope’s entrance aperture and the photometric extraction region in the image are different things, although both may be called an aperture.

Branch · Area and efficiencyOptional · you can skip this

How to obtain 35.2 electrons per second

For diameter D, the radius is D/2, so A = π(D/2)^(2) = π(D)^(2)/4. If a circular central obstruction has diameter d, subtract its area:

A_(clear) = (π)/(4)((D)^(2) − (d)^(2)).

For D = 0.100 m and d = 0.030 m, this is about 0.007147 (m)^(2). Without obstruction it would be 0.007854 (m)^(2): 9%, rather than 30%, of the area is blocked. Thin supports are omitted here.

Define optical effective area specifically as A_(eff,opt) = A_(clear)R_(1)R_(2)τ. It is approximately 0.005500 (m)^(2). Account for quantum efficiency q and extraction fraction ϵ separately:

s = ΦA_(clear)R_(1)R_(2)τqϵ.

Here Φ is photon flux per area and time at the entrance; R_(1),R_(2),τ,q,ϵ are dimensionless fractions; s is the expected number of recorded electrons per second in the extraction region. Multiplication gives s ≈ 35.198, rounded to 35.2. Photons/(m²·s) times m² gives photons/s; q provides conversion into electron counts. A photon and an electron are different particles.

As a check, when every coefficient is 1 and d = 0, only s = Φπ(D)^(2)/4 remains: doubling D quadruples the result. A real broad band requires integrating wavelength-dependent flux and efficiency; one set of coefficients is a simplification.

Even an ideal star has an image profile

A distant star may be effectively point-like for our instrument, but its image has a finite width. Light is a wave; after a limited aperture, waves interfere. An ideal circular open entrance produces a central spot and rings: an Airy diffraction pattern. The PSF, or point spread function, describes how a system images a point source.

A star does not become a mathematical point

Intensity / its own peak

00.51-3-2-10123

Angle from the star’s centre, arcseconds (″)

  • D = 100 mm
  • D = 200 mm

Ideal unobstructed circular apertures, λ = 550 nm; linear intensity scale. Each peak is separately set to 1: the graph compares width, not photon counts. First zeros are about ±1.38″ and ±0.69″. Atmosphere, defocus and pixels are absent.

NIST · Mielenz, eq. 7a · DLMF 10.2.2

Both profiles use the same angular scale. Their peaks are deliberately equalized to compare width; this graph does not show absolute light collection. A larger ideal opening produces a narrower spot at the same wavelength. If two close stars have broad overlapping profiles, screen enlargement changes neither their angular separation nor their optical width.

Branch · Angular resolutionOptional · you can skip this

What 1.22 λ/D actually means

The Rayleigh criterion compares two point-source patterns: the centre of one coincides with the first minimum of the other. For an ideal unobstructed circular pupil at small angles,

θ_(R) ≈ 1.22(λ)/(D).

Use the same units for λ and D; their ratio gives an angle in radians. One radian is approximately 206 265 arcseconds. At λ = 550 nm and D = 100 mm, we obtain 6.71 × (10)^( − 6) rad ≈ 1.38′′; at 200 mm, 0.69′′. OpenStax: circular apertures and the Rayleigh criterion.

This is a reference for a defined model, rather than a universal boundary for every kind of data analysis. Source contrast, pupil shape, noise and estimation method change the task. The plotted profile is calculated as [2J_(1)(v)/v]^(2), where v = πDθ/λ, with a limiting central value of 1. J_(1) is the first-kind Bessel function of order 1; you need not know it to see the spot narrowing. NIST: the Fraunhofer–Airy formula, eq. 7a.

Check: doubling the diameter halves the angular width; doubling the wavelength doubles it. The central obstruction in the preceding budget changes the PSF. This open-aperture profile is not presented as the exact profile of that reflecting tube.

Atmosphere, focus and field impose their own limits

On the ground, turbulent air changes the wavefront. A short exposure may show a moving spot, while a longer exposure records its smeared sum. This is atmospheric seeing. Even without clouds, transparent air does not necessarily produce a sharp image. Defocus, optical misalignment, incorrect surface shape and vibration also change the PSF. ESO: atmosphere, surfaces and supports.

Adaptive optics corrects some rapid wavefront distortions under particular conditions. It does not replace good focus, accurate manufacture or thermal control. System acceptance checks both the recorded signal and the stellar profile: a shiny coating does not establish correct surface geometry.

The camera also divides the image into pixels. If a spot falls on too coarse a grid, fine detail is undersampled; a finer grid helps only up to the limit set by the optical image. Field of view is another property: how much sky the sensor and optics encompass. A sharp small image and a wider image with poorer edges are both possible. A large field is not a large aperture. STScI: pixel sampling and field.

Subtracting the background does not remove its fluctuations

Return to our faint star. After rounding, it adds 35.2 electrons per second to the extraction region. Over 60 s, we expect 2112 electrons from the star. But the camera also collects sky light, generates dark current and adds readout uncertainty. The actual sum fluctuates from frame to frame.

Signal here means the star’s expected contribution after subtraction of the mean background. Noise is the typical random spread of this estimated signal. Even perfectly subtracting the mean sky does not undo the fact that a particular frame received randomly more or fewer sky photons. Similarly, dark-current calibration does not reveal every random electron. STScI: star, sky, dark current and readout.

In our example, the signal-to-noise ratio, SNR, is about 39.89 over 60 s. One 240 s exposure gives approximately 80.65: four times longer, but roughly twice the SNR. Four separate 60 s frames with the same total exposure give about 79.77, because readout occurs four times. The difference is small for these numbers; it may be more significant with another detector.

Branch · Measurement statisticsOptional · you can skip this

A simple SNR model and its limits

Let S = st be the star’s electron count, B = bt the sky count over the whole region, and K = n_(pix)kt the dark-current count. t is total time in seconds, and k is dark current per pixel per second. There are n_(pix) pixels and N independent reads with rms r electrons per pixel. All counts are numerical electron counts, rather than the camera’s raw ADU.

For independent Poisson arrivals, the numerical variance equals the mean count. Variances from independent sources add; readout contributes n_(pix)N(r)^(2). Thus,

SNR = (S)/(√(S + B + K + n_(pix)N(r)^(2))).

This simplifies the STScI equation for independent counts without binning. SNR is dimensionless. In our example, b = 10, k = 0.02, n_(pix) = 9, N = 1, r = 3, t = 60; the corresponding variances are 2112, 600, 10.8 and 81. Their sum has a square root of about 52.95 electrons rms; 2112/52.95 ≈ 39.89.

Check: without background or other noise, SNR = √(S). At constant rates in a regime dominated by photon statistics, SNR ∝ √(t). When a fixed single-read noise dominates, the time dependence differs. Increasing aperture may increase both star and background signals; the SNR gain cannot be declared from the area multiplier alone.

The model assumes a stable source, constant background, a linear unsaturated detector and independent noise. It neglects uncertainty in estimating sky from another region, flat-field calibration, correlations, cosmic rays, scintillation and systematic errors. Many short frames have advantages for rejecting defects and avoiding saturation; our three numbers are not universal advice on acquisition settings.

What to check when the result disappoints

If the star is dim but compact, first check clear area, transmission in the chosen band, exposure time, background and detector conversion. A cover, unsuitable filter, light pollution and strong scattering call for different actions. Measuring a standard star helps distinguish actual signal loss from a screen impression.

If the star is bright but smeared, a longer ordinary exposure is not the first remedy. Compare spot shapes at the centre and edge, different focus settings, elongation direction and short versus long exposures. Dependence on field position suggests a different class of causes from a common drift of all stars. This separates hypotheses; it is not a diagnosis without real data.

If the mean background is subtracted but the image remains grainy, estimate its fluctuations and camera noise. Increasing display contrast does not reduce measurement variance. Sensitivity comparisons need equal band, object, extraction region, time and adopted SNR threshold; resolution comparisons need angular scale and PSF shape.

Four quantities instead of one magnification

Ask about the instrument: what are its clear and effective areas; what is the angular image width under stated conditions; what is its field; what SNR does the required source achieve in a given time? These questions do not substitute for one another. More photons help measure faint sources, a narrower PSF helps distinguish close ones, and noise control determines the reliability of the inference.

A shorter introduction is available in “Why do bigger telescopes see more?”. This course’s next step examines how focus, eyepieces and cameras position an image that has already formed. Its formation mechanism matters more than a solitary magnification number on a box.