One soft corner from misalignment

Lifted from commissioning the enlarger, whose A1 mirror null is the acceptance test named throughout and whose A3 reads the corners of a dry print; and from enlarger optics and design, which derives the alignment tolerance a soft corner is evidence against and is not re-derived here. Part XVII owns the exposure record that makes any interim correction repeatable.

What you see

One corner of the print is blurred, the diagonally opposite corner is as sharp as the centre, and the centre is sharp. That combination is the whole diagnosis: it is the only signature on this page that is asymmetric about the optical axis, and asymmetry is a thing a centred lens cannot produce.

Look with a loupe on a dry print. The corner must be blurred — edges spread over a visible width — rather than merely low in local contrast, which is what a flat negative, flare or uneven illumination look like at arm’s length. Then measure the printed rectangle with a rule: a tilt prints a trapezium, so the edge nearest the soft corner is longer or shorter than the one opposite it. Wall’s 1912 dictionary states the requirement and the consequence in one sentence — the easel must be parallel to the negative, otherwise the picture will be distorted.

Three further observations belong to the appearance rather than to the diagnosis.

• The negative is innocent. Put it on a light box under the loupe: the corner is sharp there. A camera-side fault is on the film and travels with it; this one exists only in the projection.
• It gets worse as the enlargement gets bigger, and it can be invisible at 2× and obvious at 6×.
• It improves markedly when you stop down two stops, and does not go away.

What the dry print shows	What it is	Where it goes
One corner soft, the diagonally opposite corner sharp, with a measurable keystone	Misalignment, tilted about a diagonal — this entry	A1, then A3
One whole edge soft, the opposite edge sharp	Misalignment about a single axis — this entry	A1, in the other axis
All four corners equally soft, recovering strongly two stops down	Field curvature, or the lens outside its rated scale range	Rodenstock’s Recommended scale range column
All four corners soft and darker, worsening to the extreme edge, not recovering	The lens does not cover the format	Rodenstock’s Maximum film format column
A soft patch or band that lands somewhere different on each print	The negative not flat, or moving in the beam	buckled-sheet softness
Everything soft, centre included	Focus, or diffraction at an effective aperture of N(1 + m)	Refocus at full aperture
Corners dark but not soft	Falloff, source, or a mechanical vignette	uneven enlargement

Likely causes

In the order they turn out to be the answer on an enlarger that has never been commissioned — the commissioning page’s own expectation is that most such enlargers are out, with one plane clearly off:

1. The easel or the baseboard. A warped board, a dried splash of fixer under one easel foot, a chip of paper under a corner, or a masking easel whose blades hold the sheet proud along one edge. The cheapest fault on the list and the one most often found by wiping the bench.
2. The negative stage. A carrier that does not seat all the way home, a bent stage, a filter drawer left half in and holding the carrier up on one side, a scrap of film or a grain of grit in the rebate. This is the plane whose tolerance is tightest, for the reason below.
3. The lens board. A burr on the mounting face, a squashed or missing washer, a lens screwed hard into a board that is not square to the axis.
4. The column and its lock. A head that is not square to the column, or a locking collar that grips at one height and slips at another. This is the fault that passes A1 on the day and is gone a fortnight later with nothing having been dropped.
5. A glass contrast filter askew in a below-lens holder. A flat plate in the converging beam shifts the focus; a plate that is not perpendicular to the axis tilts it.
6. The negative lifted in one corner of the carrier — a curled corner, a scrap of tape, grit under the mask. Not the machine at all, and the rotation test below separates it in one sheet.
7. A decentred or damaged lens element. Rare, and it is the only optical cause that can be asymmetric. Treated last because everything above it is more likely and cheaper to exclude.

What is happening: the chemistry and physics

A tilted plane makes a tilted image. The image of a tilted object plane is itself a tilted plane, so if the negative stage is out of parallel, the surface on which the projected image is actually sharp is not the baseboard: it crosses the paper along a line and runs above the sheet on one side of that line and below it on the other. The paper can only record what reaches it, so a band of the print is sharp and the rest is defocused in proportion to how far it lies from that band. It is the same geometry the optics page draws, and the reason it produces a corner rather than an edge is that a tilt has two components: a small tilt about each axis combines into one tilt about a diagonal, and the corner farthest from the sharp band is the worst point on the sheet. The centre sits on the pivot, which is why it stays right no matter how bad the tilt gets — and why a print that is soft in the middle as well is a focus fault and not this one.

How much blur a given tilt costs. The cone converging on the paper has an effective f-number N(1 + m), so a point displaced a distance δ from best focus spreads into a circle:

blur = δ ÷ [N (1 + m)]

Defocus blur at the easel

Turn it round and you have the depth of focus the optics page derives, T = 2cN(1 + m), where c is the largest circle of confusion you are prepared to see — that page uses 0.05 mm and says so rather than pretending the number is standard. At f/5.6 and 4× it comes to 2.8 mm at the paper, and converting each plane’s depth into an angle over its own half-width gives about 0.28° at the negative stage for a 35 mm negative, four times that at the easel. Over a 100 mm base 0.28° is a rise of half a millimetre — less than the width of the line on a spirit level’s vial, which is why alignment is tested as a null and not as a reading.

Why the negative stage is the plane to align first. Longitudinal distances scale as m² between object and image space, so a tilt at the carrier costs m times what the same tilt at the easel costs: four times at 4×, ten times at 10×. The same relation appears, from the other direction, in the rules W. Piper published in 1901 and Wall reprints under Distortion, for deliberately tilting a negative and a copy to straighten converging verticals: perfect focus is secured when the two are inclined to angles proportional to their respective distances from the lens. Read against the course’s own derivation — and this reading is the course’s, not Piper’s, who was solving a different problem — that proportionality is the same statement as θ_easel = m θ_negative. A photographer in 1901 was using on purpose the geometry that spoils your corner by accident.

Why stopping down helps and does not cure. Depth of focus is proportional to N, so two stops doubles the tolerance while the corner’s displacement δ stays exactly where it was. The corner comes back inside the tolerance without the plane having moved, which is precisely why recovers on stopping down is a signature rather than a repair. Piper wrote the same limitation down for his own case: while a small stop will do a great deal, we cannot expect it to produce perfect focus if negative and copy are inclined in the same direction. His restriction to the same direction is his problem’s, not ours — he was tilting both planes deliberately, whereas a misaligned enlarger has one plane out and the other flat — but the point carried over is that a stop buys tolerance and never restores the coincidence of two planes. And there is a ceiling, because the aperture that governs both depth of focus and diffraction is the effective one, N(1 + m). Rodenstock make the point in their own manual about the Apo-Rodagon-D: at a scale near 1:1 the effective aperture is already about two stops smaller than the nominal marking, so stopping past nominal f/8 gives visible blur from diffraction. Stop down far enough to hide a tilt at 6× and you have traded one soft corner for a soft print.

!Why this cannot be the lens, and the maker's own data is the proof

Rodenstock publish, for every lens they make, modulation transfer as sagittal and meridional curves at 40, 20, 10 and 5 line pairs per millimetre, along with distortion, longitudinal chromatic aberration and fall-off in illumination against the 1 − cos⁴ reference. Every one of those charts has the same horizontal axis: relative image height.

That is not a presentational choice. A centred lens has no preferred azimuth, so its performance is a function of distance from the axis alone and is necessarily the same at all four corners of the frame. Whatever a lens does badly, it does it four times over, symmetrically — which is why “all four corners soft” appears three times in the table above under three different lens-side causes, and why one soft corner appears under none of them. A single soft corner is a statement about the geometry of the machine, and the only optical exception is a lens whose elements are no longer centred, which is a damaged lens rather than a lens working at its limits.

Diagnostic questions

• Which corner, and is the diagonally opposite one sharp? If the opposite corner is soft too, stop: this is not the entry you want.
• Is the printed rectangle a trapezium? Measure opposite edges with a rule rather than judging by eye. The keystone is the tilt’s second signature and it does not depend on the print being dry.
• Does it recover two stops down? Depth of focus is proportional to N. A corner that does not improve at all is coverage or a mechanical fault, not a tilt.
• Is it worse at 6× than at 2×? The tolerance tightens with magnification, so a plane sitting just inside its limit at small enlargements falls outside it at large ones.
• Is the negative sharp in that corner on a light box? If not, the fault is in the camera and this page has nothing for you.
• Did it appear after the head was moved, or after a lens, carrier or filter was changed? A fault with a date has a much shorter list of causes.

!The confirming test: refocus on the soft corner

Change exactly one thing — the focus setting — and watch what the rest of the field does.

Put the grain focuser in the soft corner, at the working aperture and the working height, and refocus there.

• The corner comes sharp and the opposite corner or the centre goes soft. The surface of best focus is tilted with respect to the paper. Nothing else on the list of causes can do that, because nothing else on the list tilts a plane. This is the diagnosis.
• The corner never comes sharp at any setting. It is not a tilt. Go to the lens — its rated scale range and its maximum film format are both printed in the maker’s table — or to the negative’s flatness.

Recall that an enlarging lens has no focusing mount of its own: Rodenstock note that focusing is done with the enlarger’s bellows extension, which moves the whole lens along the axis. So focusing can slide the sharp band across the paper and can never untilt it, and the test above works because the only control you have is the one that proves the fault.

No published account of this particular test was found in this course’s corpus; it is offered as reasoning from the geometry the optics page derives. The published acceptance test is A1, the mirror null, and that is what the fix below runs.

Then three rotations, each changing one thing:

• Turn the negative through 180° in the carrier. A fault that follows the negative is the negative or its seating. A fault that stays in the same corner of the easel is the machine.
• Turn the easel through 180° on the baseboard. A fault that follows the easel is the easel, its feet, or what is under them.
• Swing the contrast filter out and refocus. If the corner improves, the filter is not sitting square.

And one displacement: raise the head 100 mm, refocus and print again. A soft corner that jumps to a different corner, or changes character, is a locking collar that does not lock or a column that flexes.

Corrective action

Work mechanical, then optical, then photographic, which is the commissioning page’s order and exists so that no measurement is corrupted by an uncorrected fault beneath it.

Clean before you shim. Wipe the baseboard, the easel feet, the carrier rebate and the face of the lens board. A dried splash of fixer or a chip of film is a shim you did not intend to fit, and it is the commonest single-corner fault there is.

Run A1, the mirror null, in two axes at both stages. The pass is the reflected image of the iris sitting concentric with the real iris to better than a fifth of its diameter. Two axes matters: a tilt has two components, and a first null that looks perfect while the second, at 90°, does not is the ordinary result.

Shim the plane that is out. Find the two fixings it pivots on, put foil under the low one, cut oversize and trim after, and tighten to the torque you will actually use — a joint aligned loose and tightened afterwards is not aligned.

Move the head, lock it, and null again. An alignment that does not survive being moved is a coincidence, not a setting, and it will be gone by the third print.

Then re-run A3 and read all four corners with a loupe on a dry print. Alignment is claimed on the evidence of a print, not on the evidence of a mirror.

If A1 passes and the corner is still soft, the fault was never the alignment. Check the lens against the two numbers its maker publishes — the recommended scale range and the maximum film format — then the filter, then the negative’s flatness in the carrier.

Stopping down is a mitigation with a stated limit, not a fix. Rodenstock’s own recommendations run from one to three stops from open aperture depending on the family — two to three for the four-element Rogonar-S, which is where they put the phrase sharpness up to the image corners, two for the Rodagon, one to two for the Apo-Rodagon-N — and the effective aperture N(1 + m) sets the ceiling above which diffraction costs the whole print. Stop down inside the maker’s range if a print must go out tonight; do not adopt it as a way of living with a tilt.

What you cannot do is burn the corner. Burning changes density; a soft corner is a defect of focus, and adding exposure to a blurred region makes a darker blurred region. If a print with a known soft corner leaves the darkroom, say so on the back, with the aperture and the magnification, using the exposure record convention so the next print of that negative starts from information rather than memory.

Prevention

Align, then record what you measured. An unqualified “aligned” on an instrument certificate claims nothing. “Reflected iris concentric to better than a fifth of its diameter at f/8 and 4×, in two axes, at both stages, re-checked after moving the head” is a claim that can be checked next year.

Keep the re-check after moving the head. It is the single step that separates an enlarger that is aligned from one that was aligned once, and skipping it is why a corner goes soft a fortnight after a good commissioning with nothing having happened that anybody would remember.

Focus at full aperture and print stopped down, and focus with the filter in place. Depth of focus is narrowest wide open, which makes the point of best focus easiest to find; stopping down then widens the tolerance around a setting you have already found rather than hiding the search inside it.

Buy the lens for the format and the magnification you actually use. Both numbers are published: Rodenstock tabulate a recommended scale range and a maximum film format for every focal length in every family — 2× to 8× on 24 × 36 mm for the three-element Rogonar at f/11, 2× to 10× for the 50 mm Rogonar-S, 2× to 15× for the 50 mm Rodagon. A lens worked outside its own table produces soft corners that no amount of shimming will cure, and distinguishing that from a tilt costs a sheet of paper every time.

Wipe the bench and the easel at the start of every session, and re-commission after any change to the head, the carrier, the lens or the board. Wall’s instruction of 1912 has not been improved on: the surface the picture falls on must be parallel to the negative, or the picture will be distorted.