Sigma focusing anomaly

[Argonaut]

I have an old and trusted Sigma 400mm APO Macro. A couple of days ago I noticed that my a55 was locking focus when the lens was NOT in focus. This happened with center focus only. If I used area focus, lock was good but the a55 always found an off-center area to focus on. Since I use center focus for birds a lot, this is bad news.

This happens in AF, or using the lens and camera in MF mode and focusing until the center sensor lights up. If I go to live view and focus manually, the photo comes out nice and sharp.

All my other lenses are working fine. On my Minolta KM5, no problems with any lens.

???

[David Kilpatrick]

Technically it's not compatible with the A55. So I guess odd results may be expected.

David

[Argonaut]

Oy. I thought I had beaten the rap there - it *did* work, and now the problem changes a bit every time I try, but basically, it won't AF. Question - is this lens similarly incompatible with the other SLTs? I'm looking for a second body, and if any one of them works with this lens, it would make a big difference.

Thanks - A

[David Kilpatrick]

Almost certainly not - it's just a very old lens.

David

[Argonaut]

Welp, Sigma just confirmed the bad news. The 400 is too old - no parts, no shoes, no service, no SLTs. In a way this makes my decision between an a65 and an a580 a no-brainer. I ordered an a580 yesterday.

It's a shame that a really good piece of glass is now semi-obsolete. You'd think they could sell a lot of these lenses for half the price of a 70-400G, but one assumes their marketing department knows what it's doing. Oh, right, I forgot about the SD1.

[Heidfirst]

In a way this makes my decision between an a65 and an a580 a no-brainer. I ordered an a580 yesterday.

For your sake I hope that it is but I wouldn't assume that it will be compatible with the A580 either.

[Greg Beetham]

We pretty much know that computer makers depend on OS vendors and software builders to dream up new ways to overload your existing computer, slow it down to the point where you have to spend money on a new one, besides the OS makers stop supporting your OS sooner or later anyway.
Now in the age of digital cameras the OEM company can do stuff to create a captive audience and force people who buy the latest cameras to think about upgrading lenses preferably with OEM ones of course, because even the late model third party lenses might have compatibility problems. With Sony of course it’s hard to tell sometimes if that is what is going on for sure because as time goes on their equipment seems to be getting on average more unreliable and unpredictable, just looking at the amount of threads with issues about new cameras.
Greg

[Atgets_Apprentice]

In a way this makes my decision between an a65 and an a580 a no-brainer. I ordered an a580 yesterday.

For your sake I hope that it is but I wouldn't assume that it will be compatible with the A580 either.

I've used a number of older Sigmas, although not this particular lens, on my a580 with no compatability problems.

[Greg Beetham]

Well that’s good Atgets_Apprentice that you don’t have any problems with your Sigma lenses, so far, but what’s going to happen with new Sony models that will come out shortly? Also do your Sigma lenses work 100% for sure with ADI and SSS?
I’m not imagining things with the amount of problems I notice with new Sony models, things that don’t work properly or some kind of compatibility glitch, over at dpr (I have a glance through their forum at least once a day) lately there seems to be at least one a day, sometimes more than one.
Greg

[David Kilpatrick]

It's possible to 'repair' the lens by ordering 400mm f/5.6 chipped contact array from James Lao (or others). What is wrong in the Sigma is a small chip connected by a flexible lead to the contact group. You can't replace the chip, and nor can Sigma - but Lao can programme a new chip with the latest protocols, to be 400mm f/5.6. If you can then fit this in place of the old one (removing the contact module, cutting the cable to chip, and installing the new one) the lens will work. This is what I did with my older-version 400mm f/5.6. You can almost certainly do exactly the same with the later one.

David

[Argonaut]

Can anyone explain exactly what the problem is with this (these?) Sigmas? My layman's mind cannot understand why the SLTs have a focusing problem. Isn't focus determined by analysis of the image? As long as the sensor can see the image, why does it lock at an out-of-focus position? I can understand why a camera could not read chipped data like f-stop and focal length (which, in fact, is NOT a problem with this lens), but this seems to be a mechanical problem.

[jcoffin]

Focusing is done by analyzing the image -- but the camera uses information from the ROM to figure out how to translate what it analyzes in the image into physical movements in the lens.

I can see (at least) two obvious possibilities for problems. One is that there was information present in the older lenses that simply wasn't used by the older bodies. Since Sigma figured out the information by reverse engineering, chances are pretty good that they simply didn't include any analog to information that wasn't used at the time.

Another possibility is that some information was used, and Sigma included it, but didn't guess quite correctly about how that information was really used. They came up with a value that worked well enough within the limits of cameras at the time, but was never really correct.

For example, let's assume the camera finds that there's an offset of 5 pixels between the two images seen by the phase detectors. It looks in a table in the lens' ROM to find how many times to rotate the motor when the picture is that far out of focus. Let's assume that value was really supposed to be in the range -50 to +50. Being a binary value, that's going to be stored in 7 bits, which can hold values from -64 to +63. Sigma may have incorrectly guessed that the entire range from -64 to +63 was valid, and used (say) -60 for a particular lens.

Now, one design of computer might use what's called "saturating arithmetic", which would turn -60 into -50, so the entry of -60 would work just fine. A different computer might use modular arithmetic, in which case the -60 would "wrap around" and be treated as +10.

In the first case, everything works fine -- the -60 has been treated as if it was a valid value of -50, and is used exactly as if it the ROM contained that valid value. In the second case, however, when the camera moves the lens and re-checks focus, it finds that instead of being nearly in focus, the image is actually further out of focus than it was previously. Although it probably won't give up immediately, if this happens repeatedly the processor in the camera will probably (correctly) conclude that something's wrong, and quit trying to focus at all.

This also points to how two lenses that appear similarly designed can still focus differently. For example, Tamron might put extremely conservative numbers into that table so the numbers are all in the range -10 to +10. A nearly identical lens with a Sony badge might use (at least closer to) the full range that's allowed. With the more limited range, the camera ends up going through more cycles of checking focus, moving the lens, and re-checking focus, so the focusing is slower but since the values are always in bounds, it does always work. By using the full allowable range, the Sony version can reduce the number of cycles of checking, moving the lens, and re-checking. At the same time, they know the exact limit, so they never go beyond it and cause problems.

[Argonaut]

Jcoffin, if I read you correctly, that 'spin' algorithm would focus incorrectly, but wouldn't the two phase detectors *still* have an OOF situation? Seems to me that this kind of error would force never-ending search for focus. What happens in my case is that the camera believes it has focus lock and displays the little illuminated square to show me it has. Also, I get the focus confirmation signal when manually focusing the lens, at the same OOF point as in the AF cycle.

[Argonaut]

Well, just now received my a580 and the Sigma 400 works perfectly on this body.

[darius_fkrois]

Gald to hear that!

My two main Sigma lenses (50mm f1,4 and 70-200 f2,8 dg os ex hsm) both made in 2011 were not AF compatible with my new Alpha 77 so I had to send them in for repair....for a firmware upgrade. Selling and living with only Apple products it was quite a shock for me. How can something not be compatible?! Basically I thought "What does that even mean??".

It has been 14 days....and still waiting. grml