Sunday, June 4, 2017

Mirrorless means Accurate AF, right?


This month features guest blogger Brian Ramage, whose dance photography was so impressive I wrote an article about him in an earlier issue of f2 Cameracraft (which you can read here for free - the article begins on page 25).  Brian wanted to know which of four different 85mm lenses for the Sony FE mount would give him the best real-world results for his portraiture work, and so he got his hands on them, examined the results, and was just a little annoyed at what he found.  His full article appears below after a few announcements.

UK Seminars

I'm leaving shortly for the UK where I'll be giving not one but TWO of the famous Friedman Archives High-Impact Photography Seminars.  There's still room to sign up but do it soon to make sure I have enough handouts for everyone.

Manchester  June 10-11  Learn more here
Edinburgh  June 17-18    Learn more here

There will also be one in Tacoma, Washington August 26-27th.  More info here

Are you a member of a photo club?  The seminars were originally designed to take the confusion out of digital photography for beginners (although experienced photographers love them too!), and photo clubs have been hiring me to conduct the seminar to bring their novice members up to speed so everyone can be on the same page.  Have your club president contact me for more details.

New Ebooks out!
In the pipeline are the Olympus E-M1 II and (maybe) the highly-acclaimed A9 if I get enough requests.  Email me at Gary at Friedman Archives dot com to be added to the notification lists!  All available ebooks can be found at http://FriedmanArchives.com/ebooks .

Next Time in f2 Cameracraft

This is every online photographer's worst nightmare: Paul Spencer saw that his image was being sold as giant framed canvases at home decor stores worldwide without his knowledge or consent, and had to mortgage the house in order to seek restitution with many different violators.  Learn more about his anguish and whether it was worth it in the next issue.  Subscribe to f2 Cameracraft today!

A99 II Videos 

The world seems to have forgotten about the outstanding A99 II ever since the A9 was announced; but I haven't. I still prefer using it in the studio due to its superior ergonomics.  There are so many things about this camera that are best explained by video that I had to create two of them.  Ignore the poor sound quality and enjoy. :-)





4 85mm FE Lenses and a Big Annoyance
I'm a portrait photographer. I make my living at 85mm and longer. When I switched to the Sony A7RII from Canon a year and a half ago, there weren't many lens options. I made the switch planning on using the Metabones IV adapter with my Canon EF glass, but I found it didn't perform anywhere near what I needed it to, and I shoot mostly static subjects. I got the 24-70mm f4 as an all around lens, but really needed to get back into my longer portrait lenses. Unwilling to pay the price of the Zeiss Batis, I decided to go with Rokinon.

While I had never shot using manual focus before, you just can't argue with an 85mm f1.4 lens with great image quality and sharpness for only $300. I picked up the 85mm and the 135mm f2 for $600, and it's those two lenses that I have been shooting with exclusively for a year. While manual focus is not the easiest thing to master, I have gotten a firm grasp on it, and my results are 95+% hitting critical focus, compared to a best of 70% I've gotten with auto focus - both on Canon and on Sony - over the past 17 years. Some people will try to make up that hit rate by taking more images, but in portrait work, you need the focus to match the expression. If the focus misses on that one perfect expression that really reveals your subject, then you have lost it. With manual focus, it's up to you if your shot is focused exactly where you want it. And let me be clear, I am not a "real photographers only shoot manual focus!" snob. I dream of affordable AF Sony lenses that have the image quality that I expect for my clients. And that brings us to this shootout.


First there was the Zeiss Batis 85/1.8 for $1200, then along came the Sony "G Master" 85/1.4 for $1800. The G Master line is Sony's forward thinking to accommodate 100MP resolution sensors - and this should be a huge hint to you about their sensor roadmap. That $1800 gets you sharpness and resolution worthy of the A7RII's 42MP sensor, and also 2/3 stops of more light than the Batis and slightly less DOF with slightly more background separation. It also gives a rounder bokeh characteristic compared to the Batis' "catseye" bokeh.

Then, just days before this shootout occurred, Sony released an affordable alternative, the FE 85mm f1.8 at only $600. While the online hobbyists and pixel peepers will always stick with the G Master for its reputation, photographers that answer to a bottom line want something that will allow them to deliver quality images to clients as well as let their business remain profitable. So this shootout is to see which AF lens gives me high critical focus hit rates and at least matching my Rokinon's image quality. If I am to move out of manual focus and into Sony's advanced auto focus features, such as eye-AF - where the camera recognizes your subject's eye and automatically tracks and keep it in focus -  I want to do it as economically as possible while delivering the quality my clients expect. In my FB group teaser for this shootout - before a single frame was shot - I was called a "Rokinon fanboy", but honestly, my goal here was to come out on the other end with the lens that gives me the best results. I have no blind allegiance to any piece of gear, except the one that allows me to perform my job and keep my business out of the red. And if I were a "Rokinon fanboy" only concerned with price, then I wouldn't have walked away from this test with a new lens in my bag. (And I did.)

The setup is simple, and it's real world. I won't be shooting test charts and resolution charts and measure sharpness across the full range of the lens - there's plenty of that already on the web, and besides, my clients don't care about that. I will be shooting both on-location and studio headshots with optimal lighting. And to be honest, I don't shoot headshots at 85mm for clients - it's too wide and it distorts facial features.  My 85mm only comes out for full body shots. But I chose headshots for this shootout in order to give the lenses and the AF system the best target to lock onto. I'm certain that at full body all of these lenses will give me subjects whose faces are all in focus, but I want to push them to the limits with inch-thin DOF and lock onto that important eye. So forgive me for using the lens against my typical use, but I'm confident the test is still fit for my purposes.

We started the shootout at the beach. We gave the lighting a workout by starting at about 11am, giving us a midday sun that periodically hid behind a healthy layer of clouds, but then would also come back out without warning, shining on the subject and scene with full harshness. I had the Godox AD600BM setup with a 48" octobox. While I prefer a 60" for most lighting situations, I tend to use the 48" or 36" for location shoots due to wind. My AD600BM is a manual HSS strobe, but they also make a TTL version, the AD600B. "B" stand for "Bowens mount", while the "M" stands for "manual". As for TTL versus manual, I shoot with my camera in manual exposure mode, and likewise I prefer to choose my own lighting exposure. No judgements, just my preference. I use this light as well as my other Godox strobes and speedlights with my X1T-S HSS wireless trigger. I'm working without a tripod as the vast majority of my shoots are handheld. I also want to be able to move the camera away and defocus between each shot so I can get a fresh focus approach to each. For the Eye and Face tracking, I didn't move away and defocus, but I did move the camera to force the AF to track. The AF-S shots were defocused between each, though.

The studio shots were performed the same way, but with a three-light setup - AD600BM key in 48" softbox, AD360 in a 12x56 strip box as an eyelighter, and TT600 illuminating the white pop-up backdrop.


Only 3 shots were taken for each focus mode - I wanted to see how many "keepers" I get from each set of three photos. I don't shoot a lot of exposures, so I am not interested in a hit rate 800 shots out of 1000, but rather 9 out of ten, or in this case, 3 out of three.

The shootout: I started with the Rokinon to set the baseline. It's an f1.4 lens, and I'll be shooting all the lenses wide open today. We all know that lenses are not their sharpest either wide open or closed down, however we are not shooting a resolution chart, we're shooting shallow DOF portraits. Manually focusing with the Rokinons, I used the Peaking Level / Peaking Color feature set to yellow, on fine, as well as the camera's focus magnifier feature. I set my usual "back button focus" as the magnify button and click twice to zoom all the way in on an eye. I also use an LCD loupe on the rear LCD as it gives me a clear view, with an additional fatigue reducing camera anchor point on my eye, and its additional magnifying of the screen lets me see the focus and peaking at pixel level.

The LCD Loupe  is a magnifier attached to the rear LCD screen that lets you perform critical manual focusing in conjunction with the camera's Focus Magnifier function.  Is it more accurate than the camera's Autofocus?
I missed focus by a millimeter out on the beach. Due to the midday sun as backlight, the model's face was washed out and made manual focusing difficult. This is a situation I face a lot, and a big reason I want to switch to a reliable AF lens system. In this lighting situation, I could see the focus peaking on the model's eyelashes, but not her irises, so I went with that, assuming it would be within tolerance for an acceptable shot. I was correct in that the shots are more than useable, and that the irises can be sharpened up, but they are not perfect. Technically these are "missed critical" for the iris, but they absolutely hit critical for what I focused on - the eyelashes, so take that as you want. The shots are useable, and that's my bottom line requirement.

In the studio, I didn't have the backlight situation I had outside, and the shots are 3 for 3 on the iris. I can't ask for more.  (Click on any image to see it larger.)

3 manually-focused shots - top row is outside; bottom is in the studio.
GMaster Lens
Let's now go from the cheapest lens to the most expensive - the G Master. It's a big lens and it feels great, and I love it. I am happy to say that Eye-AF hit all three shots critically on the iris. I had my model keep her eyes "on plane", so there is no forward or back eye, and they are both in focus, but the Eye-AF would jump back and forth from eyes, while tracking. This didn't cause any problems in the final shots, but when I have decided which eye I want to focus on, and then the system jumps around, it can be distracting and disconcerting.

3 shots each: AF-C with Eye AF, AF-C with Face Detection, and AF-S on the eye.
In the studio, Eye-AF didn't do quite as well. The first is unacceptably off, but the second two are great.


I didn't know what to expect from face tracking. We know that when say we want the eyes in focus, we want the irises. But if we choose to focus on the "face", what does that mean for the eyes, especially with mm-level focus depth? We'll, in face tracking, the GM quickly, accurately, and consistently tracked the face, and two of the shots hit critical focus on her eyelashes, missing her iris.

In-studio, Face tracking did very well, with the first and third with good focus, but the second is just a bit too off to be worth trying to save, unless you need the model's expression and you wanted to work the image to save it. I'd call it a firm miss.

Finally, AF-S. This is a traditional focusing method, used for decades, where you use a single focus point, AF on the feature you want in focus - in this case her eyes - and then recompose the shot. Conventional wisdom says that the recomposing will shift the axis enough that you end of with an out of focus image, but I used an off-center spot, and the model's eyes are just off-center, so there was barely any recomposing. Interesting results here. The first shot hit the tips of the eyelashes, leaving the irises unusably defocused. The second shot also hit the eyelashes, but the irises are savable, and the third shot missed entirely.

In-studio, AF-S again lets us down, only the first is good, and the third misses entirely.

The good news with the GMaster lens is that Eye AF feature (the feature we really want to use as portrait shooters) performed the best of the three modes tested.

Batis
Next down on the list is the Batis. The Batis is the first of two f1.8 lenses today, but twice the price of the second. All three Eye-AF shots using the Batis hit the eyelashes, and not the iris, but only the first and third are recoverable - the second's irises were too defocused.


In-studio, all three Eye tracking hit the iris.


In Face tracking, we have the worst results. The first shot hits the iris perfectly, but the second two miss entirely - the second hits the forehead and the third hits the cheeks, leaving only one useable and one barely recoverable photo.

In-studio, Face tracking lets us down again. The first two are unusable, with various facial planes focused on, and the third is useable with a slightly defocused iris.

AF-S results are no better. The first shot is completely out of focus, the second misses the face entirely, catching maybe the tip of her chin, but the third hits the iris nicely.

In-studio, AF-S is a disappointment, with the first just short of useable, the second way out of focus, and the third is not critical on the iris, but is useable.

The "Budget 85"
Now for what the photographers with a bottom line are rooting for. Will it give us useable images while keeping us from spending the cash on the GM or Batis?


On the beach, in Eye-AF, the budget 85 hits critical iris in two out of three shots. The second, however, is unuseably out of focus. In-studio, Eye-AF hits the iris twice, and the first hits the eyelashes, but is more than useable. In-studio, critical iris is hit all three shots.


Face tracking on the beach isn't horrible. The first is saveable, the second is just barely saveable, but the third nails the iris. In-studio, Face tracking doesn't do very well, with only one good image.

Finally, AF-S on the beach hit the eyelashes for the first two, the first being unsaveable, but the third nails the iris. In-studio, AF-S nails all three shots perfectly.

I will also mention that there is no noticeable difference in operation of any of these AF lenses - none appear to be any slower, faster, or more accurate in locking on and tracking either the face or eyes.

Before we get to a quick comparison in image sharpness between the lenses, here is a summary of the focus hit rate for each lens:

Rokinon 85/1.4
Location: 3/3
Studio: 3/3
        Total: 6/6 Usable Shots

While the Rokinon gets a 6/6, only the studio shots are sharp on the iris, but all are more than useable. This is due to me focusing on the eyelashes as my false color spot. Definitely something to think about if you are manually focusing - you need to be able to clearly see the area you are choosing to focus on.

G Master 85/1.4
Location
EyeAF: 3/3
Face: 2/3
AF-S: 1/3

Examining the in-focus shots more closely, the location shots had 4/6 shots perfectly focused on the iris.

Studio
EyeAF: 2/3
Face: 2/3
AF-S: 1/3

Examining the in-focus shots more closely, the studio shots only had one image perfectly in focus on the iris, but the other four were very close.
Total: 11/18 Usable Shots

Zeiss Batis 85/1.8
Location
EyeAF: 2/3
Face: 2/3
AF-S: 1/3

3/5 location shots were perfectly in focus. When the Batis nails focus, it nails it.
Studio
EyeAF: 3/3
Face: 1/3
AF-S: 1/3

Again, 3/5 with perfectly focused irises.
Total: 10/18 Usable Shots

FE 85/1.8
Location
EyeAF: 2/3
Face: 3/3
AF-S: 2/3

Only 2/7 have perfectly focused irises, but this is 7/9 useable shots. Next to the Rokinon, it's our best keeper rate so far.

Studio
EyeAF: 3/3
Face: 1/3
AF-S: 3/3

And here is our best critical focus hit rate - 5/7 useable shots were perfect on the iris.
Total: 14/18 Usable Shots

Regarding focus, no AF lens was perfect, but neither was I with my manual focus. But of all the lenses, the Batis missed more shots than any other. Way more. I am really pleased to discover that the budget FE gave the least amount of misses, along with the highest critical focus count, I think a 5/6 for EyeAF is good enough for me to trust this lens in EyeAF for headshots. Of course the GM and Batis also gave 5/6 in EyeAF, but at two and three times the cost as well as more missed shots in other focusing modes.

My big lesson here is that Face Tracking is not the mode to use for close-range, super shallow DOF portraits. This makes sense, but may not be entirely intuitive for many users. I will have to do a later test on full body portraits using face tracking, where the distance to subject will be more forgiving for keeping the whole face in focus. Finally, AF-S Flexible spot did horribly on the most expensive lenses, but performed as expected with the budget FE.

On focus alone, and the percentage of "keepers" from a session with these lenses, the Rokinon and the FE85 are the two clear winners. That's only on focus. Now let's look at other attributes, starting with the sharpest image delivered from each lens.



I have chosen to judge the SOOC JPGs with picture settings set to "neutral". This sets the sharpness, color, saturation, and other settings to zero. While RAW can be processed in many ways, I find the Neutral JPG settings to be a very conservative process - giving decent color, contrast, and sharpness without overdoing any attribute.

Now, in this sharpness comparison, I see the Budget 85mm f/1.8 as the clear winner for both the location and the studio tests.  Next is the Batis, and then at the bottom is the GM followed very closely by the Rokinon. The Rokinon and GM are so close that it's really only the higher contrast of the GM that sets it apart. After processing, I can't imagine the GM would have much advantage for this particular sensor. And that brings us to other attributes.

The Rokinon gives a flatter, less contrasty image. It's also not as sharp as the others, especially the FE or Batis. The FE gives what I would call a "natural" color recording, while the GM and Batis are richer and more saturated. This gives a better SOOC image, but I've been shooting for 15+ years with flat picture profiles in order to record the most image data possible, so it's part of my long established workflow to bring the color and contrast out of the image. That said, in the era of instantly uploading client images to Instagram, while still on the shoot, it's nice to have a slightly more punchy image SOOC. It not something I care to spend 4-6x the money on, though.

So, finally, which lens was used or the final image selection that I delivered to the model? After scrolling though, the same way I do for every shoot, as I select the images to finalize based on hair, expression, and other non-optical things, the image I chose happened to be from the Batis.

Here is a 100% crop from the final processed image.


Processed in Adobe Camera RAW, and edited in Photoshop, I lightly corrected the skin, evened the skin tones, brought the eye color up to match reality, and increased the color of the background sky and sand. I never over-work skin, nor over-sharpen eyes and eyelashes. SOOC, the sharpness is gorgeous and the color is great, but a lot goes into selecting an image - model expression, flyaway hair, framing... etc. So, the Batis can give a great image that I am more than happy to deliver - but I won't subject my work to its 56% hit rate, especially when I can get a FE for half the price with a 78% hit rate, or even a Rokinon for half the price of the FE with 100% hit rate. No need to bother with the 61% hit rate of the $1600 GMaster.

Conclusion

So, for my money, the FE85 is my new go-to 85mm, with the Rokinon is staying in my bag for the times that I need/want the extra shallow DOF from the f1.4. I got rid of my Batis and GMaster lenses; happy to never use them again. It's a nice, solid lens, and if you get good results from it, then great, but I am confident that I have proven here that it is not worth the extra money.

Let me remind you that all of these lenses can produce sharp images; the purpose of this test was to determine autofocus consistency and "hit rate".  In theory all should have had a 100% hit rate since the E-mount uses a closed-loop focusing system, meaning the AF points are on the sensor and therefore there's no separate optical path that can be misaligned (as can happen with any DSLR), so these differences in hit rates were a little annoying.  Nevertheless, there they are.

Heaven Art Gallery

Real quick...last month I briefly mentioned the Heaven Art Gallery which will display anyone's work electronically on giant screens.  Your cost to have your work in rotation is just $10/month per image - pretty cheap for bragging rights.  Well, they've now officially had their grand opening and the place is abuzz.  A new photo contest is in the works with the theme "Earth - Water - Fire - Air" and you can enter your favorite images here.  Tell them Gary sent you. :-)

Next time

I'm leaving for the UK in a few days armed with a new Sony A9, which is optimized for fast-moving subjects and I'm going to be using it for landscapes and stock shots for a few weeks.  Yes, the camera will be underutilized, but it's important that I use a new camera day in and day out if I'm going to be writing a book on it.  (Still not sure I'm going to write one, since there have been very few requests for such a book so far.  Send me an email and let me know of your interest; if the requests get high enough I'll definitely take it on.)

Until then,
Yours Truly, Gary Friedman

Celebrating the end of a long day of shooting.


16 comments:

  1. Thanks for this detailed comparison. Did you have the latest firmware on the G-master? Firmware version 2 was posted on May 25 and my experience is that focusing is much more assured with this firmware upgrade.

    ReplyDelete
    Replies
    1. This test was done on April 8th, so no.

      Delete
    2. as I read it, it was the quality of the photo for the value of the lens. That makes the FE the better value.

      Delete
    3. Not just value as per results per dollar, but the FE beat the GM in this test in every possible way.

      Delete
  2. Thank you. Interesting'.
    A9, yes, please, sir!

    ReplyDelete
  3. I am interested in the A9 book as well!

    ReplyDelete
  4. Very interesting discussion on focusing for portraits. Of course, the AF lenses can also be focused manually (to get the best results), and that would take advantage of the (presumably) higher lens quality of the expensive lenses. Also, I wonder if using the DMF function wouldn't work better here - the AF would get you "close" and then you could manually focus the rest of the way. I find DMF to work very well when I am photographing my orchid flowers - AF works pretty well but I can fine-tune the focus manually if needed. (Orchid flowers are three-dimensional and it is a battle with shallow depth of field.

    ReplyDelete
    Replies
    1. Hi David, I replied to this from my phone last week, but I guess there was an error.

      I don't use DMF because it's a mode that doesn't make sense to me. I did this whole 85mm shootout in order to find an AF lens to replace my MF Rokinon. Using an AF focus mode that then had me manually focus to refine it isn;t a mode I would want to use. I want an AF lens that hits critically hits the focus point, otherwise I am happy (happyish) to just fully manually focus.

      Thanks for reading!

      Delete
  5. The details of your lens analyses are amazing -- kudos for such in-depth comparisons.
    Many thanks for mentioning Heaven Art Gallery -- we are receiving fantastic artwork from all over the world.
    Have a safe trip to the UK!

    ReplyDelete
  6. I think that the title of this article is a bit misleading. The fault of the missed focus is not on the mirrorless system per se, but it depends on it's implementation, mnamely, on the eye sensor technology. On the contrary, in a DSLR the focusing system is per se prone to errors, due to use-induced inevitable misalignment of the mirror.

    ReplyDelete
    Replies
    1. A closed-loop feedback system should always nail it. Always. If the software gets it wrong then it would affect all lenses equally. If a lens is off then the closed-loop feedback system will compensate for it by telling the lens to move more until it's right.

      Delete
  7. Hi Gary
    I'm very interested and really hope you write an a9 handbook.
    Regards
    /Michael

    ReplyDelete
  8. A little off topic but I found a manual focus lens from Venus Optics - Laowa 105mm designed for Sony A-mount with supposed excellent sharpness. The lens has a apodization (APD) element which the manufacture says "functions as a radially graduated neutral density filter, and as such limits the light transmission to a T-stop value of 3.2, while permitting the depth of field and selective focus qualities of a true f/2 design". Does anyone know what this means and does anyone know of Venus Optics? Thanks
    GEGJr

    ReplyDelete
  9. You have a great blog and I enjoyed your informative posts. Thank you so much for sharing.

    ReplyDelete
  10. When will your book for the Sony a9 be available? I am anxiously waiting.
    Thank you
    sharonyoung@yahoo.com

    ReplyDelete
    Replies
    1. Working on it now... I'll add you to the notification list.

      Delete