[macTak]

Non-pentax versions of the lens have demonstrated focus problems.

As Class A has pointed out so well, isolated cases do not indicate a rule. Here's what lenstip (testing on Canon) had to say: 'When it comes to the accuracy of the autofocus, our assessment is very positive. The number of misses in studio conditions reached less than 4% which is a very good result. What’s interesting, most of these misses were registered at 18 mm."

[UncleVanya]

As Class A has pointed out so well, isolated cases do not indicate a rule. Here's what lenstip (testing on Canon) had to say: 'When it comes to the accuracy of the autofocus, our assessment is very positive. The number of misses in studio conditions reached less than 4% which is a very good result. What’s interesting, most of these misses were registered at 18 mm."

I was totally unaware of all of these problems until reading the review here - but then my searches on the web revealed multiple problems with Canon tests (and possibly Nikon but I didn't read details on any yet). I will not have time today to dig them back up, but as your own point is that a single sample doesn't indicate a problem - so too it doesn't indicate a lack of problem. The fact that multiple copies of the lens across multiple brands have had the problem is troubling - the fact that other lenses from Sigma have had the problem also is even more so. The in depth review on this site also wasn't of a single copy but two - the second supplied by Sigma after the initial failure of the first sample to focus in Phase Detect.

At this point I would be very very cautious of this lens until more is known.

[AquaDome]

I was totally unaware of all of these problems until reading the review here - but then my searches on the web revealed multiple problems with Canon tests (and possibly Nikon but I didn't read details on any yet). I will not have time today to dig them back up, but as your own point is that a single sample doesn't indicate a problem - so too it doesn't indicate a lack of problem. The fact that multiple copies of the lens across multiple brands have had the problem is troubling - the fact that other lenses from Sigma have had the problem also is even more so. The in depth review on this site also wasn't of a single copy but two - the second supplied by Sigma after the initial failure of the first sample to focus in Phase Detect.

At this point I would be very very cautious of this lens until more is known.

When I first looked into getting this lens a year ago (for Canon EOS), reports were that IQ was beautiful, but AF wasn't accurate. It was recommended to wait for the USB Lens Dock so focus calibrations could be made, so I waited a while. During that while, I switched to Pentax but I did not stop wanting the lens. Now, thirteen months later, the 18-35 is available for Pentax, the lens dock is available, and the focus problems persist across platforms. [Sigh]

Where does PDAF happen? I assume it happens in the camera's processor, which then controls the focusing motor in the lens. What information does the camera send to the lens? What information does the lens send back to the camera? How is it that the camera sees the image and tells the lens what to do, yet releases the lens from its focusing duties with the image out of focus? Is something in the optics spoofing the sensor to believe it has achieved focus when it is only near focus? Is the camera releasing the lens in-focus, but the lens motor keeps spinning (run-on), stopping passed where focus should have been? Is it just a timing issue between the two? Is there a protocol that lenses use when communicating with the camera? Someone at Sigma knows the answers to all of these questions. My guess it they are keeping tight-lipped about it while they attempt to pump the water out of their intermittently-leaky boat.

[luftfluss]

I think if you want the lens, just buy it from a store that has a good return policy, like B&H, Amazon, etc.

[Outis]

A lot of people have mentioned that this lens has autofocus issues with other camera brands-- would someone mind linking to one or two good writeups of this?

[MadMathMind]

It doesn't surprise me that this lens has problems with AF. AF systems are good to f2.8, With my 43ltd i find I can only get the AF accurate at mid/long distances or short but not both. Take that issue into a 1.8 zoom and your going to have a problem.

I don't know about that. I don't have issues with my FA77 or FA50 f/1.7 or F50 f/1.7. Ok, there are some issues with that last one, but they're pretty rare.


I always shoot the Sigma with the autofocus point selected by me and with enlarged AF area checked. Anyone else? I did check it with auto selection of the focus point and wasn't having trouble.

My lens does tend to require EV compensation about half the time, however. Sometimes it's as many as -1, which I bump in post, so it's really more like -1/3 or -2/3s. But other times it gives not trouble at all.

[Class A]

Where does PDAF happen?

In the camera.

What information does the camera send to the lens?

Requests for AF adjustment values (and potentially other information, such as MTF data) and commands for the lens AF motor (if one is present).

What information does the lens send back to the camera?

Relevant for our discussion are the AF adjustment values.

The lens EEPROM contains a matrix which contains AF adjustment values that depend on the subject distance and the focal lens. In theory, they could also depend on the target aperture (for lenses with focus shift), but I haven't seen any evidence for the presence of such data in Pentax lens EEPROMs. I have read, however, that even Pentax service cannot fix a lens that shows focus shift by just changing the EEPROM data. I do not know whether that's due to the lack of corresponding software or due to the fact that the EEPROM does not contain respective data.

BTW, all modern lenses have many AF adjustment values stored in their on-board memory. The Sigma dock only enables convenient access to them. For some Pentax lenses, the locations of the respective AF adjustment values have been reverse-engineered and you may fine-adjust these lenses by changing the lens EEPROM.

How is it that the camera sees the image and tells the lens what to do, yet releases the lens from its focusing duties with the image out of focus?

There are two cases:

1. Either the camera (incorrectly) believes it has achieved focus, or
2. the camera is unable to achieve focus and goes with what it can get.

There is a feedback loop that ensures that the camera keeps trying to achieve focus while it has not obtained focus yet, but there is also a time-out, i.e., the camera will not try to obtain focus forever.

Is something in the optics spoofing the sensor to believe it has achieved focus when it is only near focus?

The reasons why PDAF can yield misfocus on the sensor are manifold.

Just imagine the PDAF module sitting a miniscule fraction too far back. This will produce a constant misfocus as the module then calculates the correct focus for a sensor that would be set back by the same amount.

There are also a number of optical properties of lenses, e.g., focus shift (the shifting of the focus plane due to a change in aperture caused by spherical aberration) that can misguide the PDAF module.

However, all other such issues that I know of, are constant when you fix the focal length and the subject-camera-distance (and the aperture). In other words, one can correct them with appropriate AF adjustments (the Sigma dock allows you to do that for a number of distance/focal-length pairs).

Potential reasons as to why the reviewer did not succeed to address the AF issues with the lens include

1. he didn't try hard enough,
2. he used inadequate procedures to obtain the AF adjustment values (I'm referring to the AF adjustment values that are to be transferred to the lens via the Sigma dock),
3. the error is consistent but depends on the aperture (however it would then be fixable for one aperture choice),
4. the lens is so screwed up that the number of available AF adjustment values is not sufficient, or
5. the AF errors are completely erratic, i.e., cannot be addressed with the constant offsets that the AF adjustment values represent.

I cannot speculate which of the above applies and it is possible that I'm missing something. I'm hopeful that we will get to the bottom of this mystery issue (which seems to affect some other users of this lens as well, but certainly not all of them).

Is the camera releasing the lens in-focus, but the lens motor keeps spinning (run-on), stopping passed where focus should have been?

We know that this is not the case because CDAF is reported to work excellently.

Hence, the problem is not in the enactment of the motor commands.

Is there a protocol that lenses use when communicating with the camera?

Sure, but Sigma knows at least the relevant bits of it and I highly doubt that there is any problem here.

Someone at Sigma knows the answers to all of these questions. My guess it they are keeping tight-lipped about it while they attempt to pump the water out of their intermittently-leaky boat.

This is assuming that they actually have something to cover up.

The latter, however, has not been proven.

It is very unfortunate that the reviewer received two copies which he could not get to focus with PDAF properly, but I'm close to 100% certain that either the AF adjustments were not performed properly or the individual lens copies have some issue, but that there is not a systemic problem of the lens design (only the latter would justify giving the lens a "4" for AF performance).

Just for reference: DPReview looked at three copies of the DA* 55/1.4 for reviewing and all of them had decentering issues. They also experienced a lot of focus inconsistencies. This, despite the fact the lenses had been ordered from Pentax UK specifically for reviewing.

I cannot rule out that Sigma has some problem with the lens, but the anecdotal evidence collected so far does not prove that the lens does not work in K-mount, nor that Sigma has a QC problem that is worse compared to other manufacturers.

I hope the reviewer and/or Adam are willing to contact Sigma about their findings and that Sigma will provide a response.

---

Last edited by Class A; 11-05-2014 at 03:30 AM.

[JimmyDranox]

A lot of people have mentioned that this lens has autofocus issues with other camera brands-- would someone mind linking to one or two good writeups of this?

Reading through the net, you can think that all those lenses has problems. In reality the situation can be totally different, because those people who has good lenses don't feel the need to write about that. You can read hundreds of posts complaining about one lens, or camera, but it is relevant? Maybe those hundreds are just 1% of the users.

So, the best is to try it, and to return, if you don't like it.

[Class A]

Reading through the net, you can think that all those lenses has problems.

+1

It is worth reading Roger Cicala's (of LensRentals fame) article
"This lens is soft" and other myths. Many user reports about soft or front/back-focusing lenses turn out to be not a reflection on the lens.

I'm not saying that this forum's review of the Sigma 18-35/3.5 should be dismissed. I just want to argue that "I've read about other users mentioning issues" is not a strong point unless it is backed up with something more substantial.

[Rondec]

+1

It is worth reading Roger Cicala's (of LensRentals fame) article [url=http://www.lensrentals.com/blog/2008/12/this-lens-is-soft-and-other-myths]
"This lens is soft" and other myths. Many user reports about soft or front/back-focusing lenses turn out to be not a reflection on the lens.

I'm not saying that this forum's review of the Sigma 18-35/3.5 should be dismissed. I just want to argue that "I've read about other users mentioning issues" is not a strong point unless it is backed up with something more substantial.

I think it is interesting that most of the people talking in this thread, both for and against the Sigma 18-35 haven't actually used it.

My experience with Pentax lenses has not shown any inconsistency with regard to PD auto focus, even with the DA *55. If I focus on a spot, defocus and refocus, the lens/camera will focus the same. I have calibrated my lenses, but having a fast lens should no preclude accuracy with regard to focus.

[Class A]

My experience with Pentax lenses has not shown any inconsistency with regard to PD auto focus, even with the DA *55.

Other users have had different experiences:

"Well, I finally gave up on the DA*55/1.4. After testing 5 different copies, I was unable to find one that focused accurately and consistently on my K-7, and one of them even failed on a K-5."

DPReview also had issues with focus consistency (and decentered copies).

The reality is that off-kilter lens copies do exist from any manufacturer.

Some here believe that "Sigma QC is dreadful" or "Sigma reversed engineered protocols, that's why their lenses cannot AF properly" but I don't believe for a second that such claims would hold up under scrutiny.

[Rondec]

Other users have had different experiences:

"Well, I finally gave up on the DA*55/1.4. After testing 5 different copies, I was unable to find one that focused accurately and consistently on my K-7, and one of them even failed on a K-5."

DPReview also had issues with focus consistency (and decentered copies).

The reality is that off-kilter lens copies do exist from any manufacturer.

Some here believe that "Sigma QC is dreadful" or "Sigma reversed engineered protocols, that's why their lenses cannot AF properly" but I don't believe for a second that such claims would hold up under scrutiny.

Sure. I don't know what was up with DA *55 lenses in the past, or Sigma 18-35 lenses now. I have an n=1 lens and it is OK with regard to the DA 55. I certainly am not defending Pentax though. They have had plenty of other issues, including SDM problems on multiple different lenses. I just don't hear much at this point about focus inconsistency with the 55.

How it relates to the Sigma 18-35 issues is uncertain, as well.

[panoguy]

I think it is interesting that most of the people talking in this thread, both for and against the Sigma 18-35 haven't actually used it.

That lens is Sigma SA mount and I use it on an SD1 Merrill. Great lens, and it's one of the few lenses that can stand up to the Merrill sensor. I've had no focus problems other than the usual hiccups from the really poor, aged PDAF of the Sigma DSLR. No CDAF on the SD1, either, since it has no live view!

Similar to the "great SDM debacle," in this thread and in the review comments, it seems both whingers and apologists abound with no lack of typing skills. But like Digitalis, I rely on personal experience. My DA*50-135 is now 6 years old, I think, and the original SDM motor is cruising along with no problems... other than being pokey. That doesn't mean others haven't had problems with SDM, nor does it mean that the motor is overwhelmingly fault-prone.

As I wrote in the comments of the full 18-35 review, I went and tested my SA 18-35mm against a relatively new EF-mount 18-35mm on a Canon 70D. Very different PDAF modules and very different in-camera algorithms (Canon is way ahead!). Not scientific or comprehensive testing, but I was curious and decided to get personal experience. What I found was there was a mechanical difference between the two Sigma 18-35mm lenses. My older, personal lens would focus and the internal elements would stop moving immediately upon AF confirmation. The newer, Canon-mount lens would focus and the internal elements could be seen moving a bit after the PDAF confirmation (and the shot would be OOF).

My hunch is that this mechanical sloppiness is what's making some lenses not focus correctly, even though the camera PDAF module and communication protocols are spot-on. (Curiously, both Canon EF and Sigma SA lenses use the same communication protocol, but SA has a Pentax-like physical mount.) Meanwhile, CDAF in pretty much every DSLR is slower than PDAF, relying on feedback from the imaging sensor to stop the AF. If this drives the lens motor at a slower rate, whether waiting for feedback or correcting in small amounts when the contrast is being fine-tuned, then the mechanical issue might not be as severe.

All of that is just a guess on my part, since my 18-35mm is a fine performer, and if I thought Sigma QC was horrible I wouldn't be using an SD1 Merrill as my second system (with 5 lenses, all Sigma of course). But I am pragmatic, and if someone has a problem, I don't dismiss them because I haven't seen the same problem (or some other "expert on the interwebs" with a sample size of one also hasn't seen it). Stuff happens, particularly with mass-produced precision items, and even more so when a company cannot make enough of them to meet demand.

[Digitalis]

I think i'm qualified in my position on sigma lenses.




It is interesting to hear panoguy's opinion on the 18-35mm f/1.8 ART on its native mount. He raises an interesting point: Canon DSLRs are accustomed to a degree of overshoot from the USM motor used in Canon lenses - the motor itself has considerable mass* and therefore when it comes to a stop, there is some inertial movement. Canon DSLRs compensate for this inertial movement by tweaking the focus */- slightly. Nikon take a different approach, from what I have read ( I'll find citations on this) Nikon keep upper limits on the weight of the AF mechanisms to prevent the need for inertial compensation.

I'm speculating here with a hearty dose of conjecture - Pentax, having used screw drive AF for a number of decades are used to an AF drive system that doesn't have inertial "slop" to it. This is why I SDM is no faster than screw drive - Pentax is trying to keep unwanted inertial motion of the optical focusing groups to a minimum. In other words pentax DSLRs seem to be uniquely unsuited to lenses with hypersonic focusing due to the lower inherent AF tolerances caused by the hypersonic AF drive design.

* Especially in lenses that use focus by wire mechanisms like the EF 85mm f/1.2

[Class A]

I think i'm qualified in my position on sigma lenses.

I don't want to make a claim about your qualification either way, but in principle owning many lenses does not automatically qualify one for making educated statements.

However, given that you wrote earlier --

I value reliability, optical quality, and a decent repair/maintenance network - cost is of least concern. My customers have to pay a bit more if I have to buy a lens for a project, I have to recoup my expenses somehow.

-- the fact that you own so many Sigma lenses appears to be testimony to the fact that in terms of performance, according to your assessment, some Sigma lenses can compete with any other brand out there. Clearly, if other similar lenses were better, you'd buy them as cost is your least concern.

Interestingly, even the 18-35/1.8 is in your possession, so it seems that at least your copy does not have an AF performance that should be rated a "4", or would you have kept it anyhow?