Canon 13x zoom, 7.5 - 97.5 mm, f 1.4

[Janke]

Over 20 years ago I bought a Canon 13x zoom, 7.5 - 97.5 mm, f 1.4 at an auction held by a production house here in Helsinki, Finland. They let go all of their ENG cameras.

I paid only 20 euros ($22) for the lens, but found that even when adapted to the EOS M and using the 3x feature in ML to remove the vignetting, there was horrible aberrations; at f 1.4 it was totally unusable. To get any sharpness at all, I had to stop down to about f 11. How come it was usable by the previous owner?

Took me some years to stumble on the reason for it, when I finally found a technical paper explaining why it was so atrociously bad at full aperture:

"Offsets in back focal distances for television cameras with CCD sensors"

You can read the full paper here: https://tech.ebu.ch/docs/tech/tech3294.pdf

Well, this sleuthing told me that the lens was designed with these aberrations! In a 3-CCD camera, there is a block of color-separating prisms, splitting the light into red, green and blue components, each for a single black-and-white sensor.

Those prisms counteract the lens distortions, and provide a sharp image on the sensors.

So, I made a block of glass to fit inside the lens mount! That block consists of two binocular angle prisms glued together with optical UV-glue. By sliding the prisms' long sides along each other, I could adjust the thickness of the block between 30-36 mm, and the ends were absolutely parallel. Ideally, the block should have been 40 mm thick, but that much glass wouldn't have fit behind the lens, and in front of the camera shutter. I had to settle with 35 mm. Not perfect, but usable...

There still is some chromatic aberration that won't disappear when stopping down, and some spherical aberration (fuzzy halo around highlights) visible at apertures larger than f 4.

Here's an example, showing the wide zoom range, shot at f 4:

This certainly isn't perfect, edges are soft and "rainbowy", but for shooting concerts on a stage with spotlighted soloists, it's good enough...

[iaburn]

Very interesting, I'm sure you got a lot of personal satisfaction after finding the reason and being able to mostly fix it 

[Skinny]

cool, show us how it looks at 1.4! I think some rainbow effects are nice in certain situations

[Danne]

Very interesting. Do you have any images of the lens and how the glass was put in? Any before after the fix recordings?

[ML700D]

-

[Janke]

Very interesting. Do you have any images of the lens and how the glass was put in? Any before after the fix recordings?

The lens:



The two prisms can be 45-90-45 degree prisms, as seen here: https://en.wikipedia.org/wiki/Porro_prism

or, they can be of this, slightly narrower type:



Both types of prisms are found in binoculars, the narrower ones in "straight tubes".

The important thing is that they need to be attached with UV-curing glass-glue, so that the opposing sides are exactly parallel. Then the two prisms will behave as a single piece of glass. The nice thing is that the actual thickness can be adjusted by "sliding" along the long edges.

Here is a photo of how I mounted the 35 mm thick block inside a mount tube I turned on the lathe:



In fact, the block is attached inside the tube merely with hot glue! Placement is exact, since when gluing, I placed the EOS mount ring and the glass block on a flat surface = automatic parallelism! The other end of the tube is screwed to the back of the zoom lens which has a "back focus adjust" knob - quite necessary, otherwise I would have had to make the adapter tube to 0.01 mm tolerance... possible, but needing a lot of trial and error to get right!

Unfortunately, I didn't save any "before" videos, they were so dismal!

But here is a jazz concert, shot three years ago with the then already prism-modified 7.5-97.5 mm zoom lens plus ML 3x (the closeups), as well as the Canon EF-M 11-22 mm and 15-45 mm (wide shots), altogether 3 cameras (synced with an electronic clapperboard shown further below) - all set around f 5.6 - 8:

You can clearly see that the closeups aren't as sharp as the wide takes.

Home-made electronic clapperboard:

[iaburn]

When I think about lens construction and manipulation I think about millimetric laser-size precision, and then you mount all that with some weird shaped crystals and some glue and it works perfectly fine 

[Janke]

When I think about lens construction and manipulation I think about millimetric laser-size precision, and then you mount all that with some weird shaped crystals and some glue and it works perfectly fine 

Well, the only important dimension is actually the parallelism of the glass block, and that was simple, just a flat surface! The back-focus adjustment on the lens itself took care of the rest... The tube is made on my lathe, which ensures that front & back are parallel, too.

(Yea, I have had to do very exact "shaving" on some other C-mount lenses to get them to focus properly... e.g. a Canon 12.5-75 mm f 1.8 lens - in fact, the last adjustment was with a layer of magic scotch tape between lens and mount!)

cool, show us how it looks at 1.4! I think some rainbow effects are nice in certain situations

Here's the same subject as above, first shot at f 1.4, then stopped down to f 4.

As you clearly can se, it's no-go at f 1.4 !

But with 5 mm more glass, it certainly would be a lot better at f 1.4 - a pity such a clump wouldn't fit between lens and shutter...

[Skinny]

Interesting... I don't really know anything about optics, but is there any glass types (or maybe crystal) that can achieve the same effect with less thickness so it can fit?

[Janke]

Interesting... I don't really know anything about optics, but is there any glass types (or maybe crystal) that can achieve the same effect with less thickness so it can fit?

Maybe - but most glasses have pretty much the same refractive index (around 1.5-1.7). Maybe it could be done with a block of diamond (index 2.4)?  Have you got a spare piece to lend me?

3-CCD video cameras do not have a shutter, so the space between lens and sensor(s) is almost totally filled with glass. That's not possible on the EOS M, where you can't get closer than about 5 mm to the sensor.

(In fact, in many 3-CCD cameras, the sensors are "glued" directly to the prism... as seen here: https://en.wikipedia.org/wiki/File:A_3CCD_imaging_block.jpg - that looks just as "scientific" as my own hot-glued block, right? )

You can actually consider the 40 mm block of prism-glass as an optical element of the lens itself. Without that element, the aberrations appear. Being able to insert 35 mm of glass removed a lot of the problems, but not at full aperture. (I painted the f-numbers 1.4 to 2.8 red on the aperture ring...  )

For a description of such a prism, see: https://en.wikipedia.org/wiki/Dichroic_prism

[Skinny]

I see.. yeah, maybe Cubic Zirconia is a little bit cheaper
It is really cool that in the end you made this lens useable even though it is not perfect...
By the way I think it doesn't look that bad when you zoom in at 1.4. In the night scene it should look interesting, maybe it can be used as an like artistic effect.

For a description of such a prism, see: https://en.wikipedia.org/wiki/Dichroic_prism

I like how the light is perfectly split in this photo:



It is a prism from a very old and unique soviet camera