DIY autocollimator

[OAPOli]

An autocollimator is a bit like an SLR. Except that 1) the mirror is semi-transparent; 2) instead of a film or sensor there is an illuminated high-contrast reticle. The device is a combination projector and telescope.

If the lens on the "SLR" is focused to infinity, placing a reflector in front of the lens allows the observation of the reticle's reflection. If the reticle and the VF eyepiece graticule are perfectly centred on the optical axis of the lens, this is can be used to measure minute tilt angles of the reflector. Useful to align rocket stages

If the reflector is placed on the focal plane of a second "testing" lens, AND if the testing lens is focused to infinity, one can also observe a sharp reflection of the reticle. This is really neat because one can inspect the true focal plane of a lens directly on film, which does reflect enough light (up to f/8 approximately).

I've been working on a DIY autocollimator to test the focusing accuracy of my cameras. Here is the prototype.

1. Flashlight bracket and USB flashlight ($10);
2. Target reticle (dark field crosshairs, $30) and diffuser;
3. Beam splitter unit;
4. Eyepiece: Peak 15x loupe, coated 4-element, focusable with graticule ($100);
5. 3D-printed lens tube plus tripod bracket;
6. Lens helicoid from M42 200mm lens ($20), optics removed;
7. Kenko no.3 coated achromat (f=333mm, $40)

The beamsplitter unit was given to me and I'm not sure where you would find something similar.

The lens helicoid has 58x0.5mm threads for the front lens cell but I was able to use some step up/down rings to attach the 52mm close-up lens. I can use the native diaphragm as an aperture stop; using the full aperture of the lens is inadequate. I also tried a Chinese helicoid but those have way too much play.

The lens tube was made short so that the infinity position of the objective is close to the maximum extension of the helicoid. This allows the collimator objective to be retracted, which creates a virtual image of the reticle at a finite distance. This is useful if your testing lens can't reach infinity.

To calibrate the device, focus the lens to a distant subject via the eyepiece. I used the green masking tape shown above to indicate the position. Then place a mirror in front of the lens and play wih the helicoid to find the sharpest reflection of the reticle. I needed to raise the eyepiece via shims to colocate the focal planes.

I've tested a few cameras so far.

For 135mm film cameras, the film floats between the inner and outer rails but the offset from the film rails is small. There is more variability with 120 film. The Kowa Six has the film slightly behind the film rails like on the 135 cameras. With the Mamiya Press and Konica Rapid (6x7), the film bulges slightly in front of the rails.

 

[albada]

I built a DIY autocollimator in the 1990s, but yours is more sophisticated.

Like yours, it's mounted on a tripod.

The adjustable eyepiece was taken from an old binoculars. Achromat on the left, and two baffles on the right because the incandescent lamp (far right, not shown) splattered light all over, creating much flare without the baffles.

Instead of a graticule, I scratched up a piece of polycarbonate with sandpaper. The mount for the polycarbonate target is adjustable within a few mm, which I used to adjust infinity focus. I don't remember where I bought that semi-silvered mirror; Edmond Optics?

Here's the view through the eyepiece. I put film in the camera (not a mirror), so this is the image reflecting off the surface of the film. It's instructive to rotate the camera, and watch how this image degrades from center to corner of the film. Although I built this to adjust infinity focus on cameras, I found it's also useful for assessing quality of lenses.

Mark

 

[OAPOli]

Hi @albada, your design is pretty much the only one I found online. What's the focal length?

It's true that by angling the collimator vs. the optical axis of the testing lens one can see the degradation of the reflection. Same by changing the f/stop. Some lenses give a fainter image at the same f/stop of another lens. Not sure what happens there...

 

[ic-racer]

These are nice projects.

 

[ic-racer]

I put film in the camera (not a mirror), so this is the image reflecting off the surface of the film. It's instructive to rotate the camera, and watch how this image degrades from center to corner of the film. Although I built this to adjust infinity focus on cameras, I found it's also useful for assessing quality of lenses.

Mark

Yes, infact this is mentioned in the Pearl/National Camera instructions:

 

[reddesert]

Hi @albada, your design is pretty much the only one I found online. What's the focal length?

It's true that by angling the collimator vs. the optical axis of the testing lens one can see the degradation of the reflection. Same by changing the f/stop. Some lenses give a fainter image at the same f/stop of another lens. Not sure what happens there...

Speculating, this could be due to pupil diameter or pupil location mismatch between the lens under test and the collimating lens. In principle if the collimating lens and the lens under test are both at/near infinity, then pupil location shouldn't matter since you're sending a parallel bundle of rays from the collimator into the lens under test and back. However, the brightness of the image may be correlated with the aperture diameter of the lens under test rather than the f-number , since the diameter determines how much of the collimated bundle actually enters the lens and comes back. IOW a 100/2.8 lens might return a brighter image than a 50/2.8. I don't know if this is the case for a real system, but you may be able to figure that out.

 

[reddesert]

Beamsplitters are available from optics suppliers such as Edmund Optics or Thorlabs. Cube beamsplitters (beam splitter prisms) tend to cost around US $200, but it is possible to get a plate beamsplitter (effectively a half-coated mirror I think) for around US $40-50. There may be other sources - actually there seem to be small inexpensive cube beamsplitters available on ebay (I have no idea of the quality, but I don't think it's incredibly critical here).

 

[OAPOli]

Speculating, this could be due to pupil diameter or pupil location mismatch between the lens under test and the collimating lens. In principle if the collimating lens and the lens under test are both at/near infinity, then pupil location shouldn't matter since you're sending a parallel bundle of rays from the collimator into the lens under test and back. However, the brightness of the image may be correlated with the aperture diameter of the lens under test rather than the f-number , since the diameter determines how much of the collimated bundle actually enters the lens and comes back. IOW a 100/2.8 lens might return a brighter image than a 50/2.8. I don't know if this is the case for a real system, but you may be able to figure that out.

Thanks, that makes sense. The faint reflections are coming from 45-47mm f5.6 lenses so you're probably correct.

 

[OAPOli]

A few years ago I built a DIY panoramic camera. A Wirgin stereo body with a 24x92mm film gate and a Schneider 47mm f5.6 lens. Under the autocollimator, no reflection was found from film, until I retracted the collimator objective almost all the way. The lens was previously collimated to the film rail plane.

In this position, a virtual image of the reticle is approximately 5m from the objective. This means the film is about 0.5mm behind its infinity position.

Here's the camera:

Looking at the pressure plate, one can explain the observation.

The pressure plate has ridges for the original stereo pair of frames, allowing the film to bulge inwards in the central section. In addition, the pressure plate was significantly bowed towards the door. I shimmed the central section with slick tape, and rectified the plate. Now the film sits where you would expect it.

 

[ic-racer]

Nice work figuring out that pressure plate issue.

 

[albada]

Hi @albada, your design is pretty much the only one I found online. What's the focal length?

I don't remember exactly, but I think it's at least 300 mm. The longer it is, the less effect a calibration-error in the collimator will affect infinity-focus calibration in the test camera. Thus, longer is better.

Mark

 

[ic-racer]

Yes, for example the NatCam 6400 has a 200mm FL, but the 'upgrade' model 6800 is 300mm.

 

[OAPOli]

That's good to hear. Although when testing a lens with a mirror, you can quickly see red/green fringes when turning the collimator helicoid.

I've tested most of my 120-film cameras. Many of them (P67, Rolleiflex, GSW690, Kiev 88, Kowa) have the film pretty close to the film rails. The press cameras appear to have a bit of a bulge which required a small readjustment of the RF (I had collimated them to the film rails before). I'll do some film tests, although shooting at f/8-11 can mask a lot of issues.

I've noticed that there is some frame-to-frame variation in the focal plane. But I've been reusing the same roll (up to frame 4-5) and it may have lost a bit of tension.

The P67 and Rolleiflex seem to natively focus past infinity. It's hard to tell but the focus screen appears to match.

 

[OAPOli]

With the autocollimator you can locate with excellent accuracy the focal point of a given lens on film. But for reflex cameras, how can you check if your focusing screen is at the correct height?

First, one has to assume that the reflex mirror is at 45 degrees. As we've seen in another thread one can use the autocollimator in its native function, that is to measure small angular deviations. One needs a special fixture that tilts the camera at precisely 45 degrees w.r.t the film rails. Then, with the autocollimator setup vertically, the reflex mirror will be perpendicular to it and the reflected target reticle will remain centred and coaxial, within some tolerance. This won't work with Hasselblad type cameras where the film rollers are inaccessible.

I'm not equipped to check the mirror angle so I'm assuming it's correct. Back to the screens.

Traditional screens will have a Fresnel plate contacting a ground glass, or just a piece of ground glass The ground glass works by diffusing light in all directions. Some is diffused towards your eyes forming the image, and some is diffused back to the mirror. The ground glass thus yields a faint reflection back to the autocollimator, which allows to check its position.

Modern screens have a fine, regular etched pattern which doesn't reflect any light coherently. Instead, you can place a light source above the screen. The view from the eyepiece of the autocollimator will give a highly magnified view of the screen pattern through the viewing lens. Because the pattern has some depth, it's difficult to find the exact plane of focus.

This seems to work with a microprism spot as well. With split prisms I'm not sure.

 

[ic-racer]

Yes, my Rolleiflex 2.8F's original screen is somewhat coarse and has some depth just like you have pointed out, making the exact distance measurement a little difficult.

The focus screens on most 35mm cameras I have checked have less depth and it is easier to get the exact distance.

 

[OAPOli]

I'm finding it tricky to check.

Many screens (120 and 135) have this finely ground surface that's hard to focus on. Microprism or split screens have a definite depth and are a bit unreliable. The Rolleiflex is similar but a bit easier. For the Hasselblad, focusing on the Fresnel lines seems to work. But the ones on the Autocord are slightly closer to the lens; on the Canon they are significantly closer. And the P67 has a false reflection in autocollimator mode...

From what I've found in service manuals, the screen focus is checked by looking directly at the projected autocollimator target with a magnifier.

 

[ic-racer]

It will take me quite some time to go throught all the SPT journals I have. Anyway, I found this in SPT Service Notes Sept-Oct 1975. As a reference for people that find this thread and are curious.

 

[bernard_L]

It will take me quite some time to go throught all the SPT journals I have. Anyway, I found this in SPT Service Notes Sept-Oct 1975. As a reference for people that find this thread and are curious.

I find this document contains several errors / inaccuracies that may be misleading to the occasional reader that would rely on them.
Preliminary. Note that in the description, point X is purely conceptual and cannot play a role in a concrete alignment procedure. Plus, even conceptually, X is not clearly defined, being either the symmetrical of the reticle wrt the beamsplitter, or the image on the return path, or the object-side focal point of the eyepuiece.
Major. "When adjusting... until the image is sharp". Since the image at X on the return path is an aerial image (not formed on a ground glass) even if the collimator is misaligned (reticle not at focal point of objective) a sharp image can be seen by adjusting the eyepiece.
Plus, the eyesight (near-, far-) of the operator enters the equation. The writer actually acknowledges the problem "a clear crosshair... more than one operator"; yes, but then if the adjustment is operator-dependent, one should worry also about the original owner's alignment...??
Minor. The note to the right of the schematic "Distance OMR must equal distance OME" should read OMX instead of OME. But, then again, point X is not materialized and cannot serve in an alignment procedure.

 

[OAPOli]

That's true, the outlined calibration procedure is incorrect. One must calibrate either the plane 'X' or 'R' to the focal plane of the objective using an distant subject, then adjust the position of the other using an optical flat.

To calibrate 'X', you need an eyepiece with a graticule. Adjust the eyepiece focus for a sharp graticule. Place the eyepiece on the collimator and adjust its position or the objective focus until a distant subject (e.g. the moon) is sharp.

To calibrate 'R', you place the eyepiece graticule on the reticle plane, if possible. And proceed as above.

 

[ic-racer]

OMR equals OME, this is clearly stated.

 

[bernard_L]

OMR equals OME, this is clearly stated.

Look again. More carefully. R is in the focal plane of the objective. So is X (modulo a reflection on the splitter); E, he eyepiece is farther away by the focal length of the eyepiece.
In that respect, the text in the upper paragraph on the left is correct. The caption on the lower right (which I was referring to) is incorrect.

 

[reddesert]

Here is a trick for adjusting the collimation of a lens producing an aerial image without needing an object at infinity: use the parallax between a fiducial object and the aerial image. In the case of the SPT article, I would take the author's advice to insert a clear crosshair reticle (or just anything with a feature you can focus on, like a needle or stretched thread) at point X. Place a flat mirror under the collimator. Look through the viewing tube (you may not need an eyepiece). You should see an aerial image of of the reticle R, and the clear crosshair X. Now move your head side to side a little. If the aerial image is not focused on the same plane as X, you will see them move relative to each other due to parallax. Adjust the position of the collimating lens (objective lens) until there is no relative motion. Now X is at the location of the aerial image of R, and the objective lens is collimated. You can now set up the eyepiece to focus on X for your individual eyesight.

I was taught this parallax method by my thesis advisor on my first astronomical observing run, when we used it to collimate our spectrograph with a flat mirror. That was a little conceptually simpler since we didn't have a beamsplitter, so we just stuck a little corner of paper at the focal plane and observed the parallax between the paper and its return image (rotated 180 deg) while adjusting the position of the collimator.

 

[ic-racer]

Look again. More carefully. R is in the focal plane of the objective. So is X (modulo a reflection on the splitter); E, he eyepiece is farther away by the focal length of the eyepiece.
In that respect, the text in the upper paragraph on the left is correct. The caption on the lower right (which I was referring to) is incorrect.

Good eyes, thank you for the correction!

 

[ic-racer]

Here is a better diagram from "Camera Craftsman."