Suggestion on use of USB dock accessory for Sigma Art 35/1.4

<p>Having just purchased the Sigma DG HSM 35/1.4 Art prime for my D750, I found performance to be without question the sharpest wide angle or standard prime that I've ever tested for Nikon (including Leica M). Sharpness well exceeds the capability of the sensor at f/2.8 and smaller. Even at f/1.4, central resolving power is 64 lp/mm and MTF contrast is 70% at 40 lp/mm (using Norman Koren's MTF targets). I suspect it will likely exceed the capability of the D810 sensor. The only thing as good may be the 14-24/2.8 in a wide angle. My only point of this post is simply to point out that the USM dock is a useful accessory in setting up AF fine tuning by re-programming the lens rather than using up one of the 12 slots Nikon as allocated in the body for AF fine-tuning. Simply use "0" for the camera's AF fine tuning setting and customize the lens's focus scale setting with the USB dock for sharpest focus. In my case, I only had to set it up for +5 points out of +/-20 points.</p>

<p>thanks for sharing. i think the USB dock is a gamechanger, would love to see Nikon offer one for its OEM lenses. </p>

<p>Wow - your 35mm behaved that well? When I tried mine without the dock (on my D800e), it was essentially unusable - it swung from one end of the tuning range to the other depending on focus distance. It was tack sharp on live view, but incapable of reliably taking sharp images with phase detect. With the dock and on a D810, mine behaves much better, but it still swings wildly from one setting to another throughout the AF range. I thought the 50mm Art behaved somewhat better, but even so I was struggling taking truly sharp images with phase detect last week.<br />

<br />

Assuming (and I suspect this is a big assumption or it would have happened) that nobody has a patent that is blocking everyone from doing this, Nikon should:</p>

<ol>

<li>Provide automated AF fine tuning by racking the AF range on the camera and cross-referencing regions of the image against contrast-detect results (this would either only give one answer or be subject to field curvature, so ideally you'd be wanting to try multiple images). Bonus marks for cross-referencing with every image taken as a way of dynamically calibrating.</li>

<li>Provide as many sample points as possible throughout the focus range - Sigma provide four (plus a number of zoom settings), which is arguably not enough.</li>

<li>If they want to provide lenses with this functionality for use with older bodies, they should offer a utility that drives the camera tethered over a USB socket and can simultaneously update the lens settings (ideally with a second USB socket rather than something proprietary). Critically, unlike the Sigma solution, they should do this while the lens is still mounted to the camera, because the "try a setting, take the lens off, walk back to the dock while covering the mount and rear element to avoid dust while you've got the camera pointing at a distant tree, fiddle with the dock, tweak some settings, return to the camera, iterate" approach is massively frustrating and error-prone. It's nice that the functionality is there at all, but given a dock or a micro-USB port with a rubber plug covering it, I'll take the less faffy option.</li>

</ol>

 

<p>I'm glad I have a dock - it made an almost unusable lens into a usable one, and I commend Sigma for innovating. But it's not the pinnacle of phase detect fine tuning development - it's just a step in the right direction. If Nikon don't get this sorted out, they're eventually going to struggle compared with the accurate (if currently slower) autofocus offered by on-sensor phase detect and contrast-detect solutions. They may well have ideas for new hardware, but the dock shows there's plenty more that could be done with current phase-detect technology. (And by "current", in Nikon's case, I mean "D3". Roll on D5...)</p>

<blockquote>

<p>And by "current", in Nikon's case, I mean "D3". Roll on D5...</p>

</blockquote>

<p>do i detect snark? careful or they'll retroactively discontinue the 200-500.</p>

<p>Unfortunately Sigma doesn't provide much guidance in their instruction sheet but I learned by trial and error to apply Sigma's distance scale adjustment program for AF fine tuning, not for more accurate distance scale display. Avoid changing Nikon's AF fine tuning and just leave it set for "0". I initially made the mistake of fine tuning the lens' AF with Nikon's fine tuning control and then used the Dock's distance scale adjustment to more accurately display focus scale distance. This wound up throwing off the Nikon's AF fine tuning for the lens and led to significant AF mis-focus at wider openings. Just like Nikon's fine tuning, "+" for Sigma shifts the lens focus towards infinity and "-" shifts the lens focus towards 0.3 meters. I also kept the same incremental distance scale change in all four groups in the program. Everything I've stated is predicated on the camera's AF to be correctly calibrated from the start.</p>

<p>Example of sharpness of the lens at f/2.8 with a very tightly cropped image of the full moon.</p>

<p>Let's try this again. Full moon at f/1.4. It was well inside the AF square and atmospheric condition was slightly hazy. All crops are at 100%.</p><div></div>

<p>At f/2</p><div></div>

<p>At f/2.8</p><div></div>

<p>At f/4</p><div></div>

<p>As you can see, sharpness is quite high as claimed by Sigma. Detail was too small to see in the viewfinder but only revealed at 100%. Here is the image uncropped to get an idea how small the moon was in the frame.</p><div></div>

<p>boy those crops are small. </p>

<p>It's a day past full moon and cloudy here, but I'll see how your Sigma arty lens stacks up against my Samyang 35mm f/1.4 on a D800 when I get the chance. I'm betting there won't be much in it. And I have no AF tuning issues with it. ;-)</p>

<p>I'd like to see what that Sigma lens does when the moon is stuffed into the frame corners though. Central sharpness on even a fairly average lens can be very good. It's the corners that tell the story.</p>

<p>I have to say that IMO Nikon's AF "fine-tuning" leaves a lot to be desired. It's plainly hopeless for zooms where the AF error varies with focal length - i.e most of them. I also have a couple of lenses that can't be accommodated by the rather limited tuning range. Plus if the AF module was truly positioned in accurate alignment with the sensor plane (and didn't rely on the repeated swinging of two mirrors!), then I don't see why there's a need for AF fine-tuning at all. Except in the case of wide aperture lenses that exhibit a focus shift on stopping down. However correcting for that would need a lookup table geared to the chosen aperture. Seems to me that the positioning of Nikon's metering and AF modules could be swapped over to great advantage.</p>

<p>BTW. I'm not sure why those Sigma USB docks aren't available for rent. They must make an ugly paperweight most of the time.</p>

<p>Corner sharpness and contrast was surprisingly very high, even wide open, although obviously not as high as central images. Mid frame was in between corner and center test images. The main advantage of keeping it over renting is for future software updates or fine tuning AF on another body, after all who keeps the same digital body much more than about 5 or 6 years?</p>

<p>I might regard the Sigma focusing dock as an admission of failure. Much as I like their optical performance I do distrust their AF.</p>

<p>Maybe Nikon should provide a mechanical or electro-mechanical user-accessible AF module position adjuster. That would appear to be the <em>real</em> answer to AF fine-tuning. After all, they don't seem to be able to get it right in the factory.</p>

<p>Is there anywhere online where I could find a video tutorial of the dock in use? I think I read the instruction manual when I got the dock but decided to wait until I had some experience with the lens before trying it. I didn't find any glaring problems with the lens on my D800E and never bothered with the dock. Perhaps there is a pleasant surprise awaiting me? </p>

<blockquote>do i detect snark? careful or they'll retroactively discontinue the 200-500.</blockquote>

 

<p>Polite encouragement. (Not that I'm saying the MultiCAM 3500 hasn't seem some improvements, nor that replacing the AF system is free from risk *cough* 1D3. Just that it would be nice to have a few more cross sensors.)</p>

 

<blockquote>...Samyang 35mm f/1.4 on a D800 when I get the chance. I'm betting there won't be much in it. And I have no AF tuning issues with it.</blockquote>

 

<p>There will be if you use the digital rangefinder...</p>

 

<blockquote>I also kept the same incremental distance scale change in all four groups in the program. Everything I've stated is predicated on the camera's AF to be correctly calibrated from the start.</blockquote>

 

<p>Unless I'm doing it very wrong, in my experience of the Sigma 35mm, that's astonishing. Don't let go of that lens!</p>

 

<blockquote>Maybe Nikon should provide a mechanical or electro-mechanical user-accessible AF module position adjuster.</blockquote>

 

<p>Drop the camera?</p>

 

<blockquote>Seems to me that the positioning of Nikon's metering and AF modules could be swapped over to great advantage.</blockquote>

 

<p>You'd have to get rid of the prism. By the time the light's hit the focus screen, you've lost the depth information that the AF needs. Of course, the F5 can use its middle AF point as a spot meter when you take the finder off (or mount a waist-level finder). Which is, frankly, an awesome bit of engineering.</p>

 

<blockquote>BTW. I'm not sure why those Sigma USB docks aren't available for rent. They must make an ugly paperweight most of the time.</blockquote>

 

<p>They probably are, although you're tempting me with a little earner. They're not exactly big, though - they look like one of the Android TV players. If you want to borrow mine, just ask nicely!</p>

 

<blockquote>Is there anywhere online where I could find a video tutorial of the dock in use?</blockquote>

 

<p>I've not seen one, but it's not complicated. The annoyance is just that you have to try the lens on the camera (IIRC I go about it the same way as in-camera AF tuning: I focus in live view, then adjust the AF tuning either way until I know the centre of the the phase-detect sensor position - but of course the units are different for the Sigma dock), then take the lens off and attach it to the computer (and wait for it to synchronise) before you can change things - tweak and iterate. And it's annoying at different focal lengths. Although given my efforts with my 50mm last week, it's possible I've not done it perfectly - I should have another go.</p>

 

<blockquote>Example of sharpness of the lens at f/2.8 with a very tightly cropped image of the full moon.</blockquote>

 

<p>Nicely sharp. Mine is too - my problem is just hitting focus. Of course, it doesn't help that it's usually dark when I'm using it. Out of interest, were those shots from live view, or phase detect?</p>

<p>Easiest thing to do is:<br /> 1. Set the camera's AF fine tuning and Sigma USB Dock distance scale setting to to "0"<br /> 2. Using a tripod aim the camera's viewfinder center AF focus sensor on the middle of a ruler at a 45 degree angle to the camera, focus and shoot at f/1.4 on a desired marking like exactly 6". Use a shooting distance of about 2-3 feet.<br /> 3. Look at the shot at 100% to see if the in sharpest focus where the sensor was aimed. This can be done on the camera's LCD at 100% crop.<br /> 4. If actual focus is closer than where aimed, add plus to the four groups on the focus scale adjustment with the Dock and reshoot. This will shift focus towards infinity. If actual focused further from the intended point of focus, add minus to the four groups to shift focus closer. The amount selected will be by trial and error but I woud start with +/-5 points on the Dock to see what happens since the maximum is +/-20 points.</p>

<p>All shots were taken on phase detect.</p>

<blockquote>

<p>"You'd have to get rid of the prism. By the time the light's hit the focus screen, you've lost the depth information that the AF needs."</p>

</blockquote>

<p>Durrr! Yes I was forgetting there was a focussing screen in the way. However it still seems a bit over-hopeful of Nikon to expect the mirror and sub-mirror to land in the same place every time within a few microns. My spreadsheet tells me that @ 2 metres subject distance with a 50mm lens, every 1 micron image plane displacement gives about 1.5mm focus error. 'Twould be better to take mirrors out of the AF path altogether - oh wait, that would be a mirrorless camera, right? So now I'm wondering if phase detect could be done straight off the sensor by simply obstructing the central section of the lens to form an annular prism? A simple flip-up lollipop affair behind the lens while focusing?</p>

<p>Alan: Thank you. That suggests good AF behaviour/calibration, as well as the "is the lens sharp when in focus" question!<br />

<br />

RJ: I'm not sure that annular is enough - you'd still have conflicting views from either side of the lens. I believe the phase-detect-on-sensor systems effectively block either side of the aperture (though probably not as precisely as a dedicated phase-detect system) in order to get differences. I can't say I'm awake enough to handle the logic of phase detect autofocus right now, though. (One day I might even understand why some Nikons offer two different max flash sync speeds...)</p>

<blockquote>

<p> I also kept the same incremental distance scale change in all four groups in the program. </p>

</blockquote>

<p>Alan, did you just take the one measurement and then apply it to all four ranges arbitrarily? Or did you take an individual measurement for each distance range and they all just happened to work out to the same +5 adjustment? Or does the dock only accept one change and it is up to the tester to decide which range to choose to measure initially? </p>

<blockquote>

<p> Use a shooting distance of about 2-3 feet.</p>

</blockquote>

<p>Why would you shoot at 2.5 feet if adjusting the lens to be accurate at any of the first three focal distance ranges available on the dock?<br>

Thanks.</p>

 

Yes, I arbitrarily applied +5 to

all 4 distance groups. Sharpness

is excellent at near and far

despite my determination of the

+5 setting at one distance (6ft).

I only used 2 to 3 ft as an

example of a test distance to

make it easier to see ruler

markings when angled at 45

degrees to determine focus in

front or behind the point of

focus. Any distance you choose

should work. I would suggest

trying my method first rather

than apply different adjustments

for the 4 groups as I think it

may be overly complicated.

<p>Gup - the dock does let you change all four values independently. My values (not that I've looked them up just now) are wildly different between + and - for both my art lenses, which correlates with the problems I saw when simply trying to use the in-camera AF correction on my 35mm: reliable focus at the 3' range leads to extreme defocus even at 10' if I don't correct. I probably need to give it another go, since I saw pretty poor phase-detect focus at about 20' recently, but it's hard to tell how much might be my miscalibration and how much is the limitations of the lens interpolating AF settings. Incidentally, the four distance settings at which you're expected to calibrate are different for the two lenses (and chosen by Sigma, not you). I don't have a 120-300 to try the additional matrix of calibration.<br />

<br />

Don't get me wrong - both the Art lenses I own are extremely good optically, and I'm happy with them. And they're much more useful with the dock to improve the autofocus behaviour than without. But they do appear to need it on two separate bodies (originally D800e, now D810) and the way the dock works is... not how I would have done it, having tried the user experience. Annoyingly, this process could be easily automated both by the camera and by software which used tethering (if it had not been necessary to unmount the lens to update it, and if nobody is blocking everyone from doing this with a patent or something - I've still not checked my suspicions there); we're stuck with a frustratingly manual process, but I'd rather have it in manual form than not at all.</p>