Real-world samples added to Sony SLT-A77 II First Impressions

Indeed the AF system sounds to be quite a beast.
But it seems that the IQ from the samples, is rather... noisy. The image at ISO2500 looks pretty bad. And surprisingly also does the images at ISO100.
Now, if what you need is a powerful AF, and can't afford 1DX, D4s, it's ok to make some compromise... however, for most indoor sports (or weddings, if you like) ISO 3200-6400 is rather usual, and they became rather standard since ... maybe 5DII.

I never said that the A77 II wasn't a good camera. I absolutely think it's a good camera, but for the way that I use my cameras, there's nothing about it that could compel me to switch from what I already own or add it to my existing kit for some specialized purpose that couldn't be handled by my kit. I shoot in small, cramped venues so much so that my EF-S 10-22 Wide angle zoom lens is hardly ever removed from my 70D body. That body/lens combo has proven to be perfect for my needs, so why would I switch? Do I need subject tracking across the frame? No. What about 12fps capture? No. Video? Not really, but the 70D does have it. Everything else that I use my cameras for is handled by my 6D plus assorted lenses, so why switch? Someone with different needs would likely make different choices. This is the point I'm trying to make. No more, no less.

Donnie:

Quite, fast, fast with live view, good video, good buffer.

Now of course it may not be the camera for you, but ..

badi: Which ISO 100 image is noisy?

If you're talking about the shadows after a large shadow push, then do keep in mind that you're trading off some noise here for more visible dynamic range. And even then, the resulting noise is less the camera's fault and almost entirely due to noise in the light itself (shot noise).

You'll do slightly better by going to non-SLT APS-C, but only about a half a stop better. You'll do approx. a whole 1.5 stops better going to full-frame non-SLT with good sensor performance (read: good shadow recovery ability).

Yes, you are probably right, that should be the case, but in almost all images, for ex the one with the lotus leaves: "28-ISO100-DSC00631", even in the normal areas (not under the leaves), in 100% there is not "clean detail" but rather a pattern of noise.

Of course the image is good enough for pretty everything, even printing on a huge wall :) and i am not necessary a "pixel peeper". But I think it is something in the late generations sensors that loose a little in the base iso to gain at the high iso levels. And for those who really enjoy APSC reasonable good light type of shooting, it's not that cool. For me ISO100-1600 is much more important than pushing to astronomical ISOs and loosing ISO100 (as many have).

The lotus shot was a good +2 EV exposure push, plus more shadow boosting - b/c of the large difference between brightness at the upper left & lower right of the frame. Naturally there'll be more visible noise b/c of this.

Why do you feel late generation sensors sacrifice base ISO performance for higher ISO levels? I'm curious if you have any specific examples?

Some cameras even start from ISO 200up. For example the FujiX (I have and love the X-E1, and i think the 2nd generation is similar), but i remember i saw others as well when trying to compare images at low iso in the comparison tool here.
Yes, i am more than happy with my X and i will not change it too soon (maybe when the X-E 3,4,5 will shoot reasonable 4K video), but this limitation in itself is a good proof of my opinion.

Rishi, I don't see why lenses will be an issue with A-mount. You can pick up Sony 70-200/2.8 SSM II and use a similar lens from another mount. If you desire to keep FOV identical, you can always select a specific FL range to work with (for example, 70-135mm on a77 will compare to 105-200mm on a FF).

@ Rishi Sanyal

" I'd ideally like to compare FF SLT vs. 1DX, using the same FOV lenses. Or at least 35mm equiv. FOV on both. Which is impossible right now..."

You do have the option to test the FF A99 with SLT, albeit without the newer AF in the A77M2, but the A99 also features the dual PDAF with AF-D mode, whch the A77M2 does not.

As far as lenses... Umm... maybe you need to revisit the Sony lens lineup... certainly not impossible to match lenses - unless you HAVE to be using your 1.2 L glass lol. You just need to make the effort to get the equivalent lenses. You didn't say which lens you plan to use on the 1Dx... which may lend some validity to your comment.

I see the source of confusion. I elaborated on what I wanted to test down in comments further below, but not here.

I want to compare AF tracking across the frame for wider angle shallow DOF shots. B/c recomposing causes a larger change in focal plane w/ wider angle lenses.

I typically shoot 24/1.4, 35/1.4, and 85/1.2 primes at weddings, w/ emphasis on 24/35. This requires pinpoint focus accuracy. I find my 5DIII good when it comes to shifting the AF point around b/c the joystick is pretty easy/quick to use. But I'd prefer even faster AF point selection using the center point & recomposing, having the camera automatically track my subject. This also aids in tracking moving subjects (running brides) across the frame, etc.

So, again: I'm looking for performance of AF tracking across the frame as well as along the Z (depth) axis.

And I'm not interested in f/2.8 or higher equivalent lenses. Hence, there are no lenses in the A-mount lineup that allow me to do this for 24/35mm FOV.

A99 has very little AF point spread, negating the entire point of testing AF tracking across the frame after recomposing.

A 24/1.4 lens on APS-C would at least give me 35mm f/2.0 equiv. Which is shallow enough for my work. But no such lens exists in A-mount. I'm hoping Sigma releases an Art 24/1.4 soon.

I think the bigger point might be that I'm interested in a different use of advanced AF than what sports/wildlife photographers are looking for. Is that fair?

Maybe try it with the Sigma 18-35mm f/1.8 or Zeiss 24mm f/2.0 It's not the same, but it's closer.

I would rather see AF tests on F2.8 70-200 lenses than other weird 1.2 lenses which will is be so lens dependent and disputed. Every camera maker has 70-200 F2.8 lens, and that is the lens that must be used for AF tests

@Rishi
"I want to compare AF tracking across the frame for wider angle shallow DOF shots. B/c recomposing causes a larger change in focal plane w/ wider angle lenses."

a bit OT, but ...
It would be really interesting if lens reviews could tell more , or show, the focus plane(?) of tested lenses. That would tell how good or bad a lens is for focus-and-recompose style of shooting.

As for AF testing, you have Kempton nearby!!
That should give you photo ops a few times a week, right? And fairly repeateble one, if not stricty scientific.

Sorry Rishi, are you saying you can't compare focus tracking until you get a 24mm ~f1 lens to compare to a 35mm f1.4 on the 1DX? Because I don't think that is really going to happen very soon.

Maybe a more appropriate 'test' would be to test the same two actual FL lenses, preferably from the same manufacturer (say Sigma 35mm f1.4), at the same distance. While framing won't be the same most else will be.

Otherwise, using something like the Sigma 18-35mm f1.8 should be 'fairly' close to a 24mm f1.4 on the same camera (if a f1.4 is what you are waiting for?). The difference in DoF between a 24mm f1.4 and f1.8 is rather minor to be honest. Besides when viewing at 1:1 (to pixel peep the tracking effectiveness) you are essentially reducing the circle of confusion to the point of being rather tiny, which is a punishing test for any camera, even at f1.8.

Yes, waiting for a 24/1.4. Problem with using the same focal length on APS-C & full-frame is that you recompose less on the crop sensor - therefore, there's typically less of a need for the off-center AF point to re-focus at all. AFAIK, focus has to fall outside some pre-defined tolerance to re-engage the PDAF measurement. If it doesn't, then the AF system doesn't re-engage & you get the same results as simply focusing & recomposing.

OTOH, if we had a test that actually shows the ability of the camera to track the initial subject laterally (x-y axes) across the frame, then this discussion would be moot. We don't, though, and so I want to use # of in-focus images as a proxy for this.

All that said, we may still try to get a sense of tracking & AF hit-rate using a Sony 16-50 f/2.8, or that Sigma f/1.8 lens you mention. Problem is getting a hold of the latter, and wether or not Sigma have reverse-engineered AF protocols properly...

Well lens rentals has the 18-35mm f/1.8 in stock and this forum member in the US posted a little while ago using it on his A77ii: http://www.dpreview.com/forums/post/54141956

...

While I'll save more commentary for a review, I can say that the camera often does well in good light, with slightly better performance when the camera is oriented in landscape (not portrait) orientation. Yes, that last bit is very weird & I can't explain it. In lower light, subject recognition struggles, and the selected AF point tends to jump around quite a bit.

Another thing I noticed is that the camera appears to be eager to move the lens' focus element rather readily. Therefore, I found AF to be rather 'jumpy', with focus falling on a pretty far-off plane every now and then in this AF-C mode. I tried adjusting modes in accordance with Sony's descriptions of the settings (be more/less persistent, etc.), but was not able to satisfactorily avoid this behavior from time to time.

As I said before, controlled tests of this nature are very difficult to develop, so I'd encourage you to take my experience with a grain of salt.

Thanks Rishi, I didnt realize the fact that it was darker when you took the Empress picture. BTW we travel to Victoria many times a year its beautiful and a good place to take pictures particularly at night.

There's definitely some very interesting technology & potential here. What remains to be seen, though, is if the communication between the image sensor & AF sensor leads to better, more accurate subject tracking than Canon & Nikon's RGB sensors. While SLT technology makes sense, it has to be backed with fine-tuned algorithms for AF tracking.

Where SLT could be particularly exciting is in tracking objects even during the 'blackout' of the exposure - something traditional DSLRs can't do.

I do the same thing, but that seems a bit strange to do with sample images- I imagine many people take them at face value at first glancing, thinking what they see is what they will get...