A Small Sigma 180 f/2.8 OS Macro Issue

[Canon Rumors]

Issue with the new Sigma Macro
An issue with the Sigma 180mm f/2.8 OS Macro on the 5D Mark III and EOS-1D X has been discovered by LensRentals.com. If this has caused anyone to pull their hair out, the solution is pretty simple, just turn off illumination correction.

Below is an image taken by LensRentals.com showing the phenomenon.

Sigma 180 f/2.8 OS Macro Issue – Click for larger

The Findings

1. The phenomenon shows up on Canon 5D Mk III and 1Dx cameras only. T4i and 7D bodies with firmware upgrade do NOT do this, nor do any other older cameras we could test.
2. The Sigma 180 OS and Sigma 150 OS macro lenses both show the effect and it’s identical. No other Sigma lenses that we stock showed the effect, nor did any Tamron or Tokina lenses. Obviously I can’t test what we don’t carry.
3. If you turn off Illumination Correction in the menu the effect goes away. To repeat, though, Illumination Correction in the 7D and T4i, on or off, doesn’t cause the effect.

The issue probably doesn’t show up on the APS_C cameras because of the image is about the same size as the circular anomaly in the center of the above image.

Read More at LensRentals.com

cr

[Deathlens]

I actually noticed this issue with the Sigma 105mm 2.8 Macro as well...on the 5D MKIII.

[dilbert]

Does the problem manifest itself on all full frame Canon cameras?

Looks to me like the lens tells the camera it is a specific Canon lens and the Canon camera incorrectly applies correction based on that.

Isn't it time that Canon and Nikon grew up and made their cameras more 3rd party lens friendly?

[heptagon]

Why the heck are the new top line Canon cameras so broken?

[Random Orbits]

Why the heck are the new top line Canon cameras so broken?

Companies like Sigma reverse engineer Canon algorithms rather than license them.  I'm surprised it doesn't happen more often.

[dilbert]

Why the heck are the new top line Canon cameras so broken?

Companies like Sigma reverse engineer Canon algorithms rather than license them.  I'm surprised it doesn't happen more often.

This isn't because of reverse engineering.

The Sigma lens is telling the Canon camera that it is Canon EF-ABCE and the 5D3/1DX have a built in map to address light falloff for the Canon EF-ABCE lens and apply that to the image created with the Sigma lens without knowing that it isn't required.

[lol]

I wonder what does the body report as the present lens in the PIC menu when this happens?

For example, simply as they're to hand right now, if I put a Zeiss 50mm on a 5D2 (I don't have a mk3) the PIC menu says 50mm lens - correction not available. If I put a Canon lens on, it will give the full name and may or may not offer correction depending on the lens.

[YoukY63]

Actually that is nothing new.
I found the same issue last year on my 5DmII coupled to a Sigma 50mm F1.4EX.
The lens report the code of the Canon 50mm F1.2L, and therefore the body corrects the illumination while it is not (that much) necessary. 

[Random Orbits]

Why the heck are the new top line Canon cameras so broken?

Companies like Sigma reverse engineer Canon algorithms rather than license them.  I'm surprised it doesn't happen more often.

This isn't because of reverse engineering.

The Sigma lens is telling the Canon camera that it is Canon EF-ABCE and the 5D3/1DX have a built in map to address light falloff for the Canon EF-ABCE lens and apply that to the image created with the Sigma lens without knowing that it isn't required.

Sounds like it to me.  Canon wrote the software and knows how the mapping works and Sigma does not.  Sigma tries to find an entry that "works" without any adverse affects.  Sigma doesn't know what the correct code should be.  Sigma tests their new lens on existing bodes (pre-5DIII and 1DX) and thinks they have their settings correct.  New cameras come out with new properties for the existing table entries (i.e. for lens correction in camera) Sigma gets caught with an improper entry.  If Sigma had licenses, then they would have had the tables and possibly have their own lens profiles loaded into the newer cameras like Canon did.

[Mt Spokane Photography]

Why the heck are the new top line Canon cameras so broken?

Companies like Sigma reverse engineer Canon algorithms rather than license them.  I'm surprised it doesn't happen more often.

This isn't because of reverse engineering.

The Sigma lens is telling the Canon camera that it is Canon EF-ABCE and the 5D3/1DX have a built in map to address light falloff for the Canon EF-ABCE lens and apply that to the image created with the Sigma lens without knowing that it isn't required.

Sounds like it to me.  Canon wrote the software and knows how the mapping works and Sigma does not.  Sigma tries to find an entry that "works" without any adverse affects.  Sigma doesn't know what the correct code should be.  Sigma tests their new lens on existing bodes (pre-5DIII and 1DX) and thinks they have their settings correct.  New cameras come out with new properties for the existing table entries (i.e. for lens correction in camera) Sigma gets caught with an improper entry.  If Sigma had licenses, then they would have had the tables and possibly have their own lens profiles loaded into the newer cameras like Canon did.

The solution is fairly painless, turn off  illumination correction when shooting jpeg.  Use RAW if you can, and do not turn it on in DPP if you use it.
Canon is not going to reverse engineer or modify camera software to support lenses that they have no control over.  Sigma could pick a Canon lens with a similar properties to report to the camera. 

[mememe]

Looks like sigma always picks silly lens-IDs...
 
The illumination/vignetting correction doesnt work right at any non-canon lens...

[heptagon]

So Canon doesn't provide a proper way for other manufacturers to make EF lenses without paying horrendous licensing fees. Then it purpously breaks compatibility with existing 3rd party lenses with new bodies. That doesn't sound good to me. All Canon would have to provide would be a proper way to ID the lenses. A simple manufacturer ID + item ID would be sufficient.

[mb66energy]

This is a (more or less) funny example of our believing in data correction algorithms etc.
I see a tendency to suspend development for better lenses in terms of vignetting, chromatic aberrations, distortions because you can correct them easily AFTERWARDS. I have seen such tendencies just in experimental setups ... but the best way is always to get the best raw data you can and decide THEN if you correct the data by a smaller amount.

And it is a funny example for (more or less unexpected) side effects in a more and more complex world.
But on the other hand it is phantastic how good the stuff works in general despite the complexity of the equipment and its interactions.

[Kathode-Ray]

Actually that is nothing new.
I found the same issue last year on my 5DmII coupled to a Sigma 50mm F1.4EX.
The lens report the code of the Canon 50mm F1.2L, and therefore the body corrects the illumination while it is not (that much) necessary. 

Yep, the 50mm does the same here with my 60D. No big issue, unless you're a JPG shooter

Ray

[Kathode-Ray]

That's a bit blunt, it has nothing to do with QC. The lens ID happens to be the same as a Canon lens.