A shiny mirror can still give a poor image

A circular glass blank has become smooth, received a shiny coating and now reflects the room. Yet a star remains blurred in the telescope. What was missed between a piece of glass and an optical instrument?

Making a telescope mirror means separately establishing the required curvature, removing fine roughness, producing the correct optical figure and testing it. A coating increases reflection; it does not correct the wrong shape. Grinding, polishing, figuring and acceptance therefore have different jobs. Stellafane: the complete manufacturing route.

We will examine a documented amateur route and explain what Foucault and Ronchi tests reveal. The numbers and test patterns below are synthetic. We checked documentation and a computer model; we have not physically made, coated or tested a mirror.

The test should guide the next operation

In his 1859 memoir, Léon Foucault considered a glass mirror metallized after working and described surface testing near the centre of curvature. A straight-edged screen in front of the eye turned differences in ray directions into light and shadow; the side from which it intercepted the bundle mattered. Foucault: original, pp. 197, 203–204.

In the July 1923 issue, Vasco Ronchi published a paper on two new methods for studying optical surfaces and systems. Its later digital publication date is not the method’s date of origin. Metadata of Ronchi’s original paper. The test now bearing his name uses a grating of alternating clear and opaque bands.

Stellafane’s modern instructions recommend combining methods: Ronchi is convenient for initial setup, the sphere and the edge; Foucault zonal readings provide quantitative control of the transition to a paraboloid. This is the authors’ chosen workflow, rather than proof that one band pattern establishes every kind of quality. Stellafane: why use two tests.

A blank and two different tools

The blank determines diameter, available thickness and substrate material. For a first small mirror, Stellafane recommends a thick borosilicate blank, with thickness around one sixth of its diameter. A thin substrate is lighter but depends more strongly on support during working and use. This recommendation belongs to one route; it is not a universal thickness for every professional mirror. Stellafane: blanks and tools.

In this route, the hard grinding tool has tiles on a plaster substrate. Wet abrasive works between it and the glass. The concave glass and convex tool gradually develop matching surfaces. Polishing needs another tool: a separate rigid base with a pitch layer that conforms to the curvature. Turning your only grinding tool into a pitch lap makes returning to fine grinding after a deep scratch more difficult.

Our numerical example selects D=150 mm and f=1200 mm, or f/8. It is our metric example, rather than an exact copy of a 6″/48″ kit. It describes geometry. The actual blank grade, tools, thermal conditions and critical fits must come from documentation for the real set.

Each stage changes a different property

  1. 1BlankDiameter, thickness, material
  2. 2Rough curvatureHard tool, coarse wet abrasive
  3. 3Fine grindingSmaller pits, matching contact
  4. 4PolishingPitch lap, fine polishing compound
  5. 5FiguringTest → small correction → retest
  6. 6Coating and cellReflecting layer; support without changing shape

An operation schematic based on Stellafane, not micrographs or a time schedule. Profiles are exaggerated; exact sagittas are calculated in the text. Polishing and figuring may use the same pitch lap but have different completion criteria.

Each stage changes a different property

Coarse grit sets curvature; fine grit removes its traces

Rough grinding removes glass quickly and generates the main hollow. Fine grinding successively replaces large pits with smaller ones and improves tool–glass contact. The instructions give, among others, an 8″ example: silicon carbide #80→120→220, then aluminium oxide 32→20→12→5 µm. This is one possible sequence; a grit number must not be read as its diameter in micrometres. Stellafane: fine grinding.

Move on once the larger pits of the previous grade have disappeared. Inspect centre, middle and edge under magnification; a finished centre does not establish a finished edge. Monitor sagitta too, so fine grinding does not carry the curvature away from its target. The method varies the balance of mirror-on-top and tool-on-top work, retaining random rotations for rotational symmetry.

Changing grit also means cleaning the whole process. One coarse particle returning from a tile seam, fingernails, clothing or a container can produce a new scratch. Open only the abrasive in use and keep the others sealed; clean glass, tool, workspace and water dispensers. Store fine supplies where coarse grains cannot fall into them. Stellafane: abrasive handling and storage.

Grind wet: water is part of the working mixture and reduces airborne dust. It does not replace workshop rules and material safety documentation. Handle hot pitch, power tools and waste according to their specific instructions; this article supplies neither a home chemistry recipe nor a pouring procedure we have tested.

Branch · CurvatureOptional · you can skip this

A millimetre of depth and hundreds of nanometres of difference

Let a=D/2=75 mm be the aperture radius and R=2f=2400 mm the vertex radius of curvature. Sagitta s is the distance from the surface vertex to its rim plane. For a sphere and paraboloid sharing a and R:

s_(sphere) = R − √((R)^(2) − (a)^(2)) = ((a)^(2))/(R + √((R)^(2) − (a)^(2))), s_(parabola) = ((a)^(2))/(2R) = ((a)^(2))/(4f).

The first equality follows from a right triangle in a circle; the second form of the same spherical equation is better suited to accurate computation. The parabolic formula is exact for the given parabola and approximates a shallow sphere. Stellafane: exact and approximate sagitta.

We obtain s_(sphere) ≈ 1.172161242 mm and s_(parabola) = 1.171875 mm. The difference is about 0.000286242 mm, or 286.24 nm. An ordinary feeler gauge helps reach the millimetre-scale hollow but cannot establish the required nanometre-scale figure.

This compares surfaces with a common vertex and vertex radius of curvature. Those 286.24 nm cannot automatically be called the finished telescope’s peak-to-valley wavefront error: the reference wave, pupil and focus setting still need defining. Coarse geometrical depth and optical acceptance are different measurements.

Polishing removes roughness; figuring corrects shape

The pitch lap conforms slowly to the surface. Channels and contact facets help it work; a very fine polishing compound, such as cerium oxide, acts on its surface. Contact depends on pitch condition and temperature, so jerky motion is not cured by randomly applying more force. Stellafane: polishing and contact.

Check completion over the entire surface through residual scattering from pits; remove contamination first to avoid mistaking it for incomplete polish. A smooth sphere is still not the paraboloid required for starlight. Their difference is explained in the Newtonian reflector article.

Figuring is a small, controlled change to an already polished shape. Choose a correction after testing, perform a limited amount of work, restore contact and test again. Record conditions, action and result. Continuing simply because a stroke “should make a parabola” means abandoning feedback. Stellafane: the figuring and testing cycle.

The bench source is nearby, so the focus condition changes

A star supplies almost parallel input; a paraboloid focuses it near f. On a simple bench, a small source is placed near the centre of curvature, about R=2f from the vertex. For a sphere, a source exactly at its centre sends rays perpendicular to every surface region, returning to the source. The sphere therefore has a common null here, despite its spherical aberration for a star. Stellafane: why test near R rather than f.

The bench needs a stable mirror support, a source, transverse insertion of a knife or grating, and measurable axial movement. Source and observer need separate paths so they do not block one another. Moving along the axis and inserting the blade across the bundle are different motions.

Our numerical explanation uses a fixed point source optically on the axis at R, with a mask in the return channel only. A beam splitter transmits some light and reflects some. An ideal splitter arrangement shows this channel separation. It explains directions rather than specifying a particular homemade tester. In Stellafane’s real Ronchi tester, both illumination and observation pass through the grating; our variant’s synthetic patterns are not calibrated templates for that instrument.

Separate illumination from the returning bundle

Full path near a spherical mirror

R=2400 mm

Enlarged region near the tester

12345
  1. Side light source
  2. Ideal beam splitter
  3. Mask: knife or grating
  4. S₀: equivalent axial source and spherical return
  5. Onwards to the observer

Our ideal configuration: D=150, R=2400 mm, source S=(2350, −50), splitter x+y=2350, equivalent S₀=(2400, 0), mask K=2380 mm. The upper drawing compresses the longitudinal scale threefold relative to the transverse; the lower has equal x/y scales. The lower drawing crops the path towards the mirror. The open S₀ circle is not an opaque object. Splitting losses, thickness, splitter aberrations and diffraction are omitted; symbols are schematic. This is a separate teaching arrangement, not a Stellafane tester drawing.

Separate illumination from the returning bundle

Branch · Zonal geometryOptional · you can skip this

Why fixed and moving sources give different readings

The x axis points from the mirror vertex towards the bench; the vertex is at x=0 and the source at x=R. This differs from the preceding tube diagram’s coordinate system. At radius r, the paraboloid is:

x(r) = ((r)^(2))/(2R).

With a fixed source at R, the returned ray’s exact axial intersection lies a distance L from the vertex:

L_(fixed)(r) = R + ((r)^(2))/(R) + ((r)^(4))/(2(R)^(3)).

It was checked using direct ray–surface intersection and vector reflection. If source and knife move together, reaching each thin annulus’s axial normal intersection, the corresponding position is different:

L_(moving)(r) = R + ((r)^(2))/(2R).

For our R=2400 mm, these are synthetic L−R values. They describe ideal thin rings, not mask windows we cut or micrometer readings.

r, mmFixed, mmMoving, mm
150.0937520.046875
300.3750290.187500
450.8438980.421875
601.5004690.750000
752.3448941.171875

For a spherical mirror with its source at the centre, all these positions coincide at L=R. They vary with radius for the paraboloid. Real zonal readings are compared after removing a common origin; the carriage’s absolute zero need not be assumed perfect. Mixing these configurations between a table and software produces roughly a factor-of-two correction error.

An annulus’s normal, its axial intersection and its local meridional radius of curvature are different geometrical concepts. Calling something a “zone radius” in a practical test description does not remove the need to define what the carriage measures.

Knife side and grating position are part of the result

The knife progressively blocks returning rays. For a sphere with its source at the centre, the whole pupil darkens together at the common return point. Closer to the mirror, rays have not yet crossed; farther away, they have swapped sides. With a knife entering from the right, the first shadow therefore appears on different sides of the visible mirror. Without specifying blade side and axial position, it is easy to reverse the interpretation. Stellafane: shadow direction.

A Ronchi grating acts as many neighbouring edges. Near a sphere’s centre of curvature, straight parallel bands are visible; moving away from the return point increases their number. A paraboloid has a different pattern at this finite conjugate. Straight bands do not universally mean “correct parabola,” and every bend does not automatically indicate a defect. Stellafane: reading Ronchigrams.

Test conditions determine bands and shadows

Knife enters from the right

Sphere · K=R−20 mm
Sphere · K=R
Sphere · K=R+20 mm

Grating: 0.254 mm period · 100 LPI

Sphere · K=R−20 mm
Sphere · K=R+20 mm
Paraboloid · K=R−20 mm
Paraboloid · K=R+20 mm

The fixed point source is optically at R; the mask is in the return channel only, as above. Pupil coordinates are directions u=y/(R−x), v=z/(R−x), not camera coordinates. Uniform initial brightness; 161×161 grid. Black is blocked and white transmitted; grey marks an illustrative intermediate state of uniform extinction, rather than calculated photometry. Clear/opaque grating widths are equal, with a transverse quarter-period shift. No diffraction, interference, real source extent or noise. These are not photographs or acceptance templates for a double-pass grating.

Test conditions determine bands and shadows

These synthetic panels specify surface, source at R, mask at R±20 mm, its period and knife direction. Rays reflect geometrically; diffraction, interference, actual source extent, noise and roughness are not calculated. An illustrative angular pupil is shown, rather than a camera photograph. It explains the mechanism but does not accept or reject a real mirror.

Inspect a hooked band edge on both sides of the return point and in the context of the whole pattern. A gently closing loop in a parabolic pattern must not be confused with a local turned edge. A feature moving with air currents suggests something different from a static surface error; one observation still does not prove the cause.

Zonal numbers still need a justified interpretation

A Couder mask selects known annular zones with paired windows. For each zone, find the knife position where the relevant areas darken consistently and record the axial reading. Data reduction needs diameter, zone boundaries or effective radii, R, target figure and source configuration. Repeat readings, assess repeatability and check the readout mechanism. A 0.01 mm scale digit does not establish 0.01 mm accuracy by itself. Stellafane: measuring paraboloids.

Zonal reconstruction assumes a smooth, rotationally symmetric surface. Five readings cannot establish the absence of fine roughness, a narrow edge zone or astigmatism between measured directions. The method checks different diameters by rotating the mirror, including 0°, 45° and 90°. Glass support and the air path can also influence results; remounting helps separate these causes.

Surface error, wavefront error, peak-to-valley range and root-mean-square error must be named separately. Near normal incidence, reflection approximately doubles surface-height error in the wavefront, but that does not convert P–V to RMS. A “fraction of a wave” is incomplete without pupil, removed tilt/defocus and test uncertainty. Material, surface and wave are distinguished in the earlier manufacturing article.

Branch · Wavefront budgetOptional · you can skip this

A visible-light requirement, rather than a universal quality stamp

For a teaching example, set this objective: near the axis at λ=550 nm, retain at least 80% of the ideal image’s peak intensity for the same pupil. This is a target Strehl ratio S≥0.8. It means neither “80% of all light remains” nor a complete criterion for a wide-field camera.

For uniform amplitude and suitable small phase errors, use the estimate:

S ≈ exp⁡[ − (((2πσ_(W))/(λ)))^(2)].

σ_(W) is RMS wavefront error after the appropriate adjustment, in the same units as λ. For a Gaussian phase distribution, the exponential relationship follows from the mean complex field; amplitude nonuniformity requires additional treatment. Primary MagAO-X study: Appendix A, Eq. 10 and the discussion on p. 26.

For our chosen objective, that estimate gives:

σ_(W)≲(550 nm)/(2π)√( − ln⁡0.8) ≈ 41.35 nm.

If a hypothetical independent report gave 35 nm RMS with a confirmed upper error bound of 5 nm, we would check 40 nm too. If it gave 40 nm with the same bound, 45 nm would no longer provide the required margin under this model. This illustrates uncertainty; these are not our mirror measurements or certification from five synthetic zones.

Strehl does not contain every possible limitation: obstruction, scattering, field and atmosphere need separate descriptions. Why aperture and image quality cannot be reduced to one number is explained in the article on resolution and signal.

Apply the coating after checking the figure

An ordinary amateur glass mirror is usually aluminized in specialized vacuum equipment. Agree on substrate, dimensions, surface condition and required coating. Metallization follows the prepared figure; it is not ordered as a correction for the wrong paraboloid. A suitable cell and a check of the optics in its working orientation must follow. Stellafane: finishing the figure and next steps.

Other instruments have other manufacturing routes. Webb’s beryllium substrates were measured cold and corrected before final polishing and gold coating. That demonstrates why shape is checked at operating conditions; it is not a home substitute for glass. NASA: manufacturing Webb’s mirrors.

For XMM, ESA’s team described nickel shells produced on precise master mandrels and difficulties in release and support of thin replicas. Reproducing a surface did not remove deformations during integration. ESA’s primary account of XMM production.

Acceptance of an amateur mirror must also close the whole chain: complete polish, figure, edge and different directions, repeatability, post-coating condition and support in the telescope. A star test provides useful additional control in the working optics, where atmosphere, temperature, focus and alignment must be distinguished from the mirror’s own properties. We have not performed that physical test here.

An available next step is a protocol for one test: fixed or moving source; R and mask location; knife-entry side; definition of null; necessary zones and repeats; uncertainty allowed by the chosen objective. Even before a workshop exists, that protocol distinguishes “it shines nicely” from a testable optical result.