Sony A77 low light continious autofocus

[Magnar Fjortoft]

Today I tested the A77 continious autofocus in low light. Good news! This camera is much better than the a900. The settings were center group autofocus points, auto focus priority (release priority is the other option) and manual exposure. The lens was my Sony G 70-200 mm zoom, which balances very well with the A77.

The first photograph is my cat walking along the street. I had to use the screen to get the camera as low as possible. Focal length 200 mm, exposure 1/1000 sec, f:4 at ISO 800. In spite of the dark face, the autofocus did well. Just a bit camera shake, due to a stressed photographer.

The second is the only birds I forund for my BIF test: Two croaches in bad light. About 50 % crop. Focal length 200 mm, exposure 1/1000 sec, f:4 at ISO 3200. The birds are way underexposed, but the autofocus tracked them very well. Don't expect the A900 to do the same.

The third is a really though autofocus test after sunset. Focal length 200 mm, exposure 1/1000 sec at f:4, ISO 12 800, still way underexposed. But autofocus tracked the girl and mom very well from the first to the last frame when shooting at 8 fps. Quite impressive!

[agorabasta]

OK, Magnar, your samples here are fine and dandy, but they really tell nothing of the a77 AF virtues.

First of, the cat and birds images are of very high contrast, so each and every basically functional SLR AF should have no problem with them. The cat is high-contrast wrt the background and of its black&white fur and thick bright whiskers; the birds images are against the bright sky, so AF clings to the edges.
The last two samples are much better as they really show the AF ability to function in some very low light. Still the far remote background simply cannot interfere in focus analysis in that particular sample, especially with that helmet in the frame.

In fact, your earlier samples of focus tracking were much more telling. And I really believe you that the a77 AF is a huge step forward and may be even the best in class. But the samples above can't prove that, they are simply of 'well, OK' type.

[Greg Beetham]

I still would like an actual EV value at the time of any testing and evaluation of the performance of cameras high ISO or AF accuracy etc. then you can get an idea of the actual light level that was at hand at the time, that last one could have been taken at EV5 for all I know, and if it was then the AF did a very very good job in my book, but if it was EV10 then it still did quite a good job but not as impressive as at EV5 or 6 or 7.
Greg

[Magnar Fjortoft]

Photographs doesn't tell much at their own, so there often have to be a story to explain what it is about.

For the image of the cat, the light level was about EV11, which is not that much for a black target. The autofocus did not detect the outline of the cat, but detail in the dark face.

For the image of the crows, the light level was about EV8. Not that much, but then the birds stood aout against a brighter sky.

For the last two images, light was flat and the light level was about EV7 - too low for acceptable results handheld.

[agorabasta]

Magnar, the highest spatial hf energy density in your low-res samples sits where the white hair on cat's chest is protruding over the black body hair around. At that the eyes don't even come close to that chest hair acuity.
I surely can imagine a downsampling process to play such a trick, but bet you didn't use a process like that.

So - no, your cam has focused on the chest against the surrounding black hair rather than on the mug of the cat.

[Magnar Fjortoft]

The camera was handheld, looking onto the screen, the cat was moving, the light was pretty low, there are some unsharpness due to camera shake, and focus might not be spot on, but good enough to make nice A3 sized prints. I have never had a camera before that could autofocus this good. So what to say? I am impressed with the autofocus.

[agorabasta]

Like I said, I entirely trust your judgement on AF performance. All the samples I've seen so far speak for the same very positive mark.

But you can't prove it with a high-contrast cat. For that you'd need an 18% grey cat over an 18% grey background

[Magnar Fjortoft]

Then I have to look for an 18 % grey cat! Or buy a hair bleaching cure for my cat?

[agorabasta]

Or shoot some rat running over the rocks...

But I'm afraid the AF capabilities are not there just yet. Now I'm being quite serious.

[redsim74]

My neighbour had an 18% grey cat, but it blended a little too well with the 18% grey road.

[Magnar Fjortoft]

Should not laugh, but ... LOL!

[Mark K]

I can only prove this is completely right on my own a77
I cannot focus on subjects with similar background, tested using 200G with 1.4x, of course not to mention 70-400G
The EVF failed to identify the subject as well.
I will bring my a850 to test

OK, Magnar, your samples here are fine and dandy, but they really tell nothing of the a77 AF virtues.

First of, the cat and birds images are of very high contrast, so each and every basically functional SLR AF should have no problem with them. The cat is high-contrast wrt the background and of its black&white fur and thick bright whiskers; the birds images are against the bright sky, so AF clings to the edges.
The last two samples are much better as they really show the AF ability to function in some very low light. Still the far remote background simply cannot interfere in focus analysis in that particular sample, especially with that helmet in the frame.

In fact, your earlier samples of focus tracking were much more telling. And I really believe you that the a77 AF is a huge step forward and may be even the best in class. But the samples above can't prove that, they are simply of 'well, OK' type.