Tube Lens

A modern objective does not form an image. It collects light from a point in the specimen and sends it onward as a parallel bundle, a beam that never converges, headed for infinity. That is what infinity-corrected means, and it is why the space above the objective is called infinity space.

The tube lens is what turns that parallel bundle back into an image at the camera. And because the objective alone has no image to magnify, magnification is not a property of the objective at all. It is a ratio:

M = f_tube / f_objective

The 60 engraved on the barrel is a claim about a pairing, not about the part. It means the manufacturer built an objective of focal length f such that their own tube lens produces 60×. Nikon and Leica design around a 200 mm tube lens, so a Nikon 60× has a focal length of 200/60 = 3.33 mm. Olympus designs around 180 mm, Zeiss around 165 mm.

Put a Nikon 60× objective — focal length 3.33 mm — onto a stand with an Olympus 180 mm tube lens. The ratio is 180 / 3.33 = 54×, not 60. The objective does not know it has been moved and the engraving does not update.

Now follow it through, because the error does not stay at 10%. With a 6.5 µm camera pixel you calculate your sampling as 6.5/60 = 0.108 µm per pixel. The truth is 6.5/54 = 0.120 µm. Every length you report is 10% short. Every area is 0.90² = 19% short. Every volume is 0.90³ = 27% short.

Nothing about the image looks wrong. It is in focus, it is sharp, the cells are the shape cells are. A 27% error in nuclear volume passes review because there is nothing in the picture to argue with, and it survives into the literature attached to a number that was never measured, only assumed.

The same arithmetic applies to anything else that sits in the imaging path: a 0.7× camera adapter, a magnification changer, a beam splitter with its own relay optics. Each one is modifying f_tube, and each one makes the engraving less true.

Infinity correction is not an optical nicety. It exists so that things can be put into the beam path.

Slide a 1 mm slab of glass into a converging beam and the focus moves by roughly t(1 - 1/n); for ordinary glass, about 0.34 mm. Against a depth of field of half a micrometer on a high-NA objective, 340 µm is not a small shift, it is a different county. Every dichroic, every emission filter, every DIC prism would have to be compensated for individually, and swapping one would mean refocusing.

In a parallel beam that shift is zero. Rays go in parallel and come out parallel, displaced but not converged, and the tube lens cannot tell the difference. This is what makes a filter wheel possible, a multi-camera splitter possible, and a modern microscope a modular instrument rather than a sealed one.

It also explains why the distance between objective and tube lens is forgiving. In parallel space, moving them apart does not change focus or magnification, only the field, as the bundle eventually walks off the edge of the tube lens. That tolerance is what leaves room for everything else to live in there.

The fix takes ten minutes and settles the question permanently: image a stage micrometer — a slide with a ruled scale of known pitch — at every objective and camera combination you use, and count pixels between the rulings.

If 100 µm spans 833 pixels, your sampling is 0.120 µm per pixel, and that number is measured rather than inferred. Work backwards if you are curious: with a 6.5 µm sensor pixel, 6.5/0.120 tells you the real magnification is 54×, and now you know what you actually have.

Two habits worth keeping. Record the measured value in the acquisition metadata rather than in a notebook, because the analysis will read the metadata and will not read the notebook. And repeat it after anything in the light path changes — a new camera, a new adapter, a service visit — since all three silently redefine f_tube.

This is the cheapest calibration in the building and the one most often skipped, because the engraving is right there on the barrel looking authoritative.