How can a long focal length fit inside a much shorter tube?
In a Newtonian reflector, the primary gathers light and a small flat mirror simply turns the beam sideways. Now replace that flat with a curved secondary mirror. It changes the convergence of the rays, allowing a larger image to form behind the primary. A camera or eyepiece can sit at the rear, while the tube can be much shorter than the effective focal length.
A classical Cassegrain uses a concave paraboloidal primary and a convex hyperboloidal secondary; a Gregorian uses a paraboloidal primary and a concave ellipsoidal secondary; a Ritchey–Chrétien uses two hyperboloids, a concave primary and a convex secondary. The secondary’s position matters too: before or after the primary focus. This apparently small change determines the beam’s direction, housing length, aberrations and methods of surface testing. University of Arizona: Reflective Systems, pp.22–25.
We will examine layouts that return the secondary beam through a central primary bore. Other exits and off-axis variants exist. These three basic systems are reflective; they should not automatically be called catadioptric. A Schmidt–Cassegrain with a corrector plate is a different optical design. Lenses in an eyepiece do not turn the two-mirror objective itself into a Schmidt–Cassegrain either.
The exact rays below use our own hypothetical prescriptions. The author has not manufactured or physically tested these telescopes. We will distinguish mathematical checks, published documentation and checks that a real workshop would need to perform.
A proposal, a working instrument and a refinement are different events
James Gregory described his proposal in Optica Promota in 1663. Making the required mirror surfaces was difficult. Newton’s letter published in 1672 refers to that book and to Gregory and the craftsman Reive’s unsuccessful attempt. This records the manufacturing difficulty of the time, rather than proving that the optical layout cannot work. Newton Project: Philosophical Transactions No.83.
The same issue includes a translation of de Bercé’s account of Cassegrain’s proposal: a concave primary with a hole and a convex secondary. It was published on 20 May 1672. Describing a proposal does not establish that an instrument had already been made. Newton disputed its advantages, but his polemic cannot replace a modern optical calculation.
A Royal Society meeting record for 5 February 1673/74 describes Hooke demonstrating a reflector with an erect image, following the principle of Mersenne and Gregory. The dual year is retained as in the source: calendar notation can produce different dates in short retellings. We have a specific record of a working instrument, without needing to claim absolute worldwide priority. Reprinted record, Early Science in Oxford, vol.VII, printed p.417, PDF p.37.
Ritchey and Chrétien addressed another problem: good images over a wider field. The history of the Paris Observatory optical workshop associates their work there with 1926/27. That supports a local history of implementation, rather than dating all the preceding theoretical results. Observatoire de Paris: laboratory history.
A working classical system can also be traced in Ira S. Bowen’s own technical account of the 200-inch Hale Telescope: a paraboloidal primary with a 16.5 m focal length and a Cassegrain output with an effective focal length of about 80 m. Those are values from that historical description, not an exact prescription to copy. Ira S. Bowen: The 200-inch Hale Telescope, p.1.
The Cassegrain intercepts the beam before the first focus
Imagine the paraboloidal primary on its own. After reflection, parallel axial light would converge at its focus F₁. Place a secondary before that point, with its convex reflecting face towards the primary. It intercepts the still-converging beam and sends it back. Its curvature is chosen so that the rays now meet behind the primary.
The beam has not yet formed a real intermediate image before reaching the secondary. F₁ marks where it would have gone without interception. The required secondary profile is part of a hyperboloid: its geometry converts this convergence into a new focus. A flat mirror would not provide that increase in image scale.
The primary bore transmits the narrow returning bundle. The hole itself does not focus or correct anything; it provides a light passage. Too small a bore clips the bundle, especially away from the field centre. An unnecessarily large bore reduces the useful primary area. The secondary’s shadow, holder and baffles must also be checked, rather than relying on two catalogue diameters.
The Gregorian lets the rays cross first
Move the secondary beyond F₁. Rays from the paraboloidal primary now meet, form a real intermediate image and diverge again. A concave secondary intercepts them. An ellipsoid provides the appropriate axial conjugate pair: light from one focus reflects towards the other.
This explains both differences from the classical Cassegrain. The secondary has the opposite curvature sign because it receives a diverging bundle. The housing is longer in this comparison because it must accommodate the primary focus. The extra ray crossing changes image orientation: without additional image-rotating components, a Gregorian gives an erect terrestrial image where a classical Cassegrain inverts it.
A real intermediate focus is useful when field or thermal loading must be controlled there. For example, the off-axis Gregorian layout of the DKIST solar telescope sends light onwards to the Coudé system and instruments. Its decentered path also shows that central obstruction is not mandatory for every member of the family. This is a professional solar instrument; its purpose is not an instruction to observe the Sun with our home-scale model. NSO: Follow the Photons.
One aperture and the same effective focal-length magnitude
Classical Cassegrain
A convex secondary intercepts convergence before F₁. Dashes extend the path without a secondary; they are not extra light.
Gregorian
A concave secondary follows the real intermediate F₁: rays cross and diverge before reaching it.
Ritchey–Chrétien
A convex secondary precedes paraxial F₁. The hyperboloidal primary alone does not bring every axial ray to one point.
All three systems in the drawing have a 200 mm diameter and a 5500 mm effective focal-length magnitude. The Gregorian is longer than the other two despite the same image scale. That follows from the secondary’s position, rather than an arbitrary illustration size.
Ritchey–Chrétien changes both surfaces for the field
A paraboloid brings a parallel axial bundle to focus well, but that does not promise good images far from the axis. A classical Cassegrain retains coma: different pupil regions give displaced intersections for an off-axis point, producing an asymmetric spot. Refocusing alone cannot remove it.
In a Ritchey–Chrétien, RC, the primary is a concave hyperboloid and the secondary a convex hyperboloid. Together they can cancel third-order spherical aberration and coma. The primary alone need not be a perfect axial paraboloid; the two surfaces are assessed as a pair. Lallo 2012, Section 4.2, Eqs.1–2.
“Third-order” limits the claim. Higher orders remain. RC does not automatically eliminate astigmatism or field curvature either. With astigmatism, two mutually perpendicular bundle structures have different best-focus positions; with field curvature, the best image surface does not coincide with a flat sensor. The RC name therefore cannot replace information about the corrector, field and camera size.
| System | Secondary and primary focus | What the forms provide | What to check off axis |
|---|---|---|---|
| Classical Cassegrain | Convex, before F₁ | Exact axial conjugacy of paraboloid and hyperboloid | Coma, astigmatism, field curvature |
| Classical Gregorian | Concave, after real F₁ | Axial conjugacy of paraboloid and ellipsoid | Coma, astigmatism, field curvature |
| Ritchey–Chrétien | Convex, before paraxial F₁ | Joint cancellation of third-order spherical aberration and coma | Astigmatism, field curvature, higher orders |
Branch · Conic surfacesOptional · may be skipped
Deeper: define a surface beyond the word ‘hyperbolic’
The x axis points from the primary towards the sky; the primary vertex is x=0. Incoming light travels towards decreasing x. For a surface with vertex v, transverse radius r and conic constant K, choose the branch near the vertex:
(r)^(2) − 2R(x − v) + (1 + K)(x − v)^(2) = 0, x − v = ((r)^(2))/(R(1 + √(1 − (1 + K)(r)^(2)/(R)^(2)))).
In this convention R₁=+2000 mm. R₂ is positive for the convex Cassegrain secondary and negative for the concave Gregorian secondary. These are signs in an explicit equation, not a universal rule for all optical software. K=−1 gives a paraboloid; − 1 < K < 0 a prolate ellipsoid; K < − 1 the two-sheet hyperboloid whose required near-vertex branch we select.
| Our prescription | Vertex separation s, mm | K₁ | R₂, mm | K₂ |
|---|---|---|---|---|
| Classical | 800 | −1 | +4400/9 | −169/81 |
| Gregorian | 11600/9 | −1 | −4400/9 | −81/169 |
| RC | 800 | −123/121 | +4400/9 | −1631/729 |
All use D=200 mm, an Ø80 mm secondary, an Ø50 mm primary bore and a final plane at x=−300 mm. These are ideal centred surfaces for explanation, without completed mechanics or manufacturing tolerances.
The classical hyperbola has foci at x=1000 and x=−300 mm; the distance difference is a constant 900 mm. The Gregorian ellipse has a constant distance sum of 16900/9 mm. Checking this geometry and the constant axial path independently checks the ray tracer beyond its picture. For RC, we also checked the conic constants against Lallo’s equations and the scaling of residual aberrations.
Exact profiles: enlarge the sag to see the curvature
Classical Cassegrain
Gregorian
Ritchey–Chrétien
“Hyperboloid” does not universally mean “deeper than a paraboloid”. A comparison must fix R, aperture and coordinates. At the same R, our fourth-order sag term depends on 1+K: the equation determines the sign and size of the departure. Similar silhouettes cannot identify or certify an exact surface.
Effective focal length sets image scale, not distance travelled
Take the classical teaching prescription. The primary focus is 1000 mm from the primary; the secondary is at 800 mm. It intercepts light 200 mm before F₁ and directs it to a focus 300 mm behind the primary. The remaining secondary-to-final-focus path is 1100 mm. To first order, the secondary increases the scale by 1100/200=5.5 times. The system has a 5500 mm effective focal length even though the mirror vertices are only 800 mm apart.
The meaning needs no matrices. Images from two nearby sky directions are separated by approximately F·α, with small angle α in radians. At F=5500 mm, one arcsecond corresponds to about 26.7 μm in the image plane. That does not mean the system resolves one arcsecond: resolution also depends on aperture, diffraction, quality, atmosphere and detector.
For an ideal axial bundle, the sum of path lengths from the entrance plane x=1400 mm to the output focus is 3300 mm. It is not 5500 mm. Effective focal length describes conversion from angle into image size, rather than the length of the folded path. The Gregorian with the same |F| has an axial path of about 4277.78 mm, again not 5500 mm.
Branch · First-order opticsOptional · may be skipped
Deeper: derive focal length from changes of slope
Let y be a small primary height and u a small ray slope. After each reflection, count longitudinal distance onwards along propagation, as if unfolding the axis. For a parallel axial input:
u_(1) = − (y)/(f_(1)), y_(2) = y + su_(1), u_(2) = u_(1) − φ_(2)y_(2), φ_(2) = − (2)/(R_(2)).
Here f₁=1000 mm, s=800 mm and R₂=4400/9 mm. Hence y₂=0.2y and u₂=−y/(5500 mm). Propagating beyond the secondary to the axis gives l=−y₂/u₂=1100 mm. Returning to physical coordinates, xout=s−l=−300 mm.
The same result for a nonzero input slope follows from the matrix:
(1 0; − φ_(2) 1)(1 s; 0 1)(1 0; − 1/f_(1) 1) = (A B; C D), F = − (1)/(C), l = − (A)/(C).
Each matrix has determinant 1, as does their product. Independently, the numerical derivative of image position with respect to a small input angle in the exact trace gives +5500 mm for Cassegrain and RC, and −5500 mm for Gregorian. The minus sign denotes opposite scale orientation under the same angular convention, not a negative tube length. The formal central ray used for this derivative is mathematically extended through the obstruction; it is not shown as physically transmitted light.
First-order optics explains scale and focus. It does not predict coma or astigmatism spot sizes; those require exact tracing or an aberration calculation.
A smooth surface and the correct surface are different achievements
In making a paraboloidal mirror, we separated rough curvature, fine grinding, polishing and figuring. The same causal sequence becomes more demanding here: both mirrors must fit one prescription, and secondary testing depends on its form and accessible conjugate points.
Start with the optical pair’s drawing and materials: diameters, thicknesses, bore, vertices, radii, K, seats and allowable deformation. Radius sets the basic surface power; the conic constant sets the departure from a sphere. Once curvature and smoothness are established, small controlled removals bring the asphere to shape. Measurements must show where and by how much the surface departs, giving the next polishing action a specific purpose. One attractive central spot does not measure the whole profile.
A concave Gregorian secondary allows testing arrangements based on the ellipsoid’s two real conjugates. For a convex hyperboloid, the usual knife placement at a real centre of curvature used with a concave sphere is unsuitable: that accessible return is absent. A different optical test configuration, calibration and agreement with the prescription are needed.
A concrete example is the Hindle test of Hubble’s backup secondary. Robert Parks’s report describes a large fast sphere returning the test beam to an interferometer. Setups were repeated with equipment repositioned, and calibration was separated from mirror data. However, this arrangement showed only 85% of the axial aperture, and its long air path added noise. The tested optic was a backup; its fine-scale defects cannot automatically be assigned to the flight mirror. Parks, 15 March 1991, summary and setup, PDF pp.2–3.
This shows why a test name is not a quality certificate. Aperture coverage, positioning uncertainty, calibration of auxiliary surfaces and an independent check matter. Coating and mounting must preserve the figured surface too: stress, clamping or a temperature gradient can change a finished profile.
Hubble combines mirrors, baffles, a truss and adjustment
Hubble is a specific RC implementation: a nominal 2.4 m primary and an effective focal length of about 57.6 m, or f/24. Lallo’s table gives an optical secondary diameter of 281 mm and 310 mm including the housing. A rounded “about 30 cm” does not justify replacing the optical diameter with the housing diameter in a beam calculation. Nor are Hubble’s rounded nominal parameters our exact 200 mm prescription.
Hubble: the optics need an entire structural frame
- Primary: concave hyperboloid, Ø2.4 m
- Secondary: convex hyperboloid, optical Ø281 mm
- Baffles reject unwanted light directions
- Graphite-epoxy truss maintains geometry
- Secondary actuators adjust its position
- Rear focal region and science instruments
The primary has a lightweight honeycomb structure. Reflective aluminium and protective magnesium fluoride layers provide the required spectral properties; coatings themselves do not correct the wrong asphere. A graphite-epoxy frame maintains mirror alignment, while secondary actuators permit adjustment. Materials and thermal control preserve geometry as well as withstand loads. NASA: Optics, Lallo: OTA construction.
In an ordinary centred layout, the secondary holder obstructs part of the aperture and its spider changes the diffraction pattern. Baffles have a different task: stopping unwanted light from reaching the detector without the required reflections, or after scattering. Their edges must be checked across the working field to avoid clipping useful bundles. A black interior helps, but cannot replace correct openings and screens.
Hubble’s primary error found after the 1990 launch was a figure error, not evidence against the RC family. The investigation identified an incorrectly assembled measuring null corrector; contradictory auxiliary tests had not been adequately acted on. The lesson is to check the reference that guides manufacture as well as the mirror. Later corrective optics in the science path compensated for the known aberration; the primary was not repolished in orbit. NASA: Optical Systems Failure Report abstract, 1990, NASA: optical correction.
Branch · SensitivityOptional · may be skipped
Deeper: why a small secondary movement shifts focus noticeably
For our classical prescription, hold curvatures fixed and slightly change vertex separation s. From the matrix above, the output focus position x=s−l has derivative:
(dx_(out))/(ds) = 1 + (M)^(2) = 1 + (5.5)^(2) = 31.25.
A 10 μm secondary displacement near nominal position therefore produces about a 0.313 mm focus shift in our physical coordinate. This local first-order estimate was checked by a numerical difference; it is not a universal specification for all RC and Gregorian systems.
Moving the sensor back into focus does not automatically restore nominal aberrations. Secondary tilt or decenter can spoil field symmetry; an incorrect K remains incorrect after moving the camera. Assembly therefore needs controlled axial and angular adjustments, stable seats and image checks both on and off axis.
An off-axis spot tells more than the central star
Imagine a flat-sensor camera. The centre is sharply focused, yet edge stars are elongated. Causes can include coma, astigmatism, field curvature, sensor tilt or a combination. Knowing the optical family narrows the search but cannot replace measurements across the field.
Below, the same three ideal prescriptions receive a point source at about 0.172°. Each dot marks where one transmitted ray crosses the chosen plane. A common scale makes the geometry comparable; it is not stellar brightness and does not include diffraction.
Spots on one flat image plane
Classical Cassegrain
Gregorian
Ritchey–Chrétien
The RC spot has not become a point here. It need not even be the smallest on every arbitrarily chosen flat plane. Third-order coma correction does not cancel astigmatic focus splitting or field curvature. A real instrument’s required performance is assessed with its corrector, wavelength band, sensor, best-focus position and tolerances.
The model separately checks the order of correction: doubling a small ray height makes the isolated comatic term of the classical system grow by roughly four. In RC it starts with a higher term and grows by roughly 16 in this check. This tests residual scaling; it does not claim RC photographs are 16 times better.
Choose the mirror pair together with the task and its test method
For a narrow field, an accessible rear focus and required image scale may matter most. A classical Cassegrain provides these with a paraboloidal primary but needs a precise convex asphere and a coma check. A Gregorian geometry is useful when a real intermediate focus is needed and a longer housing is acceptable. For wider fields, RC offers controlled third-order coma cancellation, while requiring two matched hyperboloids and checks of remaining aberrations.
Before manufacture or assembly, verify a complete prescription: definitions of radii and K; vertex separations; diameters of mirrors, bore, holders and baffles; required field and instrument space. Then agree on tests for each surface, the optical pair and mechanical adjustment precision. Acceptance separately evaluates axial quality, field spot shapes and their changes with focus and temperature. A laser returning to a mark does not measure the correctness of a conic constant.
Two curved mirrors offer more freedom than a primary with a flat deflector. That freedom works when form, placement, testing and mechanics describe the same system.