A77 Video PROBLEM!! huge disappointment

ET2 wrote:

SouthElginDad wrote:

Again, I don't get this. If the autofocus sensor needs lots of light, it ought to work better with apertures larger than 3.5. Can someone explain this?

The problem is that in most daylight conditions, the PDAF sensor would not be able to focus if the lens is open wider than F3.5 (let's say F1.4, or F1.8). That's too much light for the sensitive AF sensor.

In stills mode, all video-capable Alpha DSLR's/SLT's happily autofocus with the lenses attached at max open aperture. So how can there be "too much light" for the PDAF sensor?

Also, why are there extra sensitive AF sensors in the top end Alphas that only get active when lenses with 2.8 or even brighter opening are used?

My guess for the 3.5 in video AF is that this allows the AF algorythm in video mode to smoothly approach an area of focus achieved due to the somewhat wider DOF compared to wide open. AF with an 85/1.4 or 135/1.8 for example wide open has the potential to completely miss focus and start hunting - with that totally ruining the clip - or that focus transitions would appear as too sudden and yerky, also taking away from the viewing experience. That could be ...

... why it's fixed at F3.5. Less user errors, less problems of people complaining AF is crappy, etc.

Focus transitions in good movies are mostly done manually anyway (focus pull...).

I can see the OP getting (a bit too) exited as this limit is not exactly proactively advertised by Sony. I tend to also perceive the forum members jumping at him as overreacting. YMMV on that aspect of the discussion here.

Finally, shooting an 85/1.4 in video wide open allows all optical issues (LOCA bokeh shift often perceived as "purple fringing", lesser contrast etc.) to challenge the resulting clip. At 3.5, all versions of the 85/1.4 (be they the various Minolta incarnations or the CZ) look a lot better across the frame.

ET2 wrote:

AK Productions wrote:

Well my friend, that was a stupid reply, the manual said nothing about this "technical limitation" and what a poor explanation of why the camera's firmware/software will not allow diaphragm blades of a lens to fully open while a phase detection sensor is active.

Grow up, dude.

Strange way to start a post, considering that you don't have a technological leg to stand on...

It's a "technical limitation" with PDAF AF system. This applies to all PDAF AF cameras. The lens has to be F3.5 or wide open for slower lenses for the PDAF AF sensor to work.

Except that it does not apply to "all PDAF AF cameras".

Pretty much all PDAF cameras (Nikon, Canon, Sony, etc. SLRS) set their lenses wide open when using their PDAF systems. The AF sensor has offset microlenses, cylindrical lenses, aperture masks, or barrier bars (lots of ways to solve this problem) that aim the sensors at two points about 10 degrees apart on a lens's exit pupil (that's the angle subtended by f5.6) and they can do this just fine on a wide open lens.

High end cameras often have a second set of AF sensors that cover a 20 degree angle (subtended by f2.8) that increases AF speed and accuracy with faster lenses.

The only "technical limitation" that applies here is Sony's belief that their focus motor and AF sensor system cannot respond accurately enough to achieve acceptable focus for any aperture with a DOF shallower than f3.5. f1.4, for example, would require about 2.5x the accuracy that f3.5 required. The deeper DOF at f3.5 hides the focus errors.

That's how it works, dude.

SouthElginDad wrote:

Can someone explain why this is the limitation?

No, no one can explain that, because it is not "the limitation".

I thought the autofocus needed a lot of light in order to work (hence no smaller apertures than 5.6). Why is it also a problem with too much light? I don't get it.

Most people don't, because it's tricky. It's all about angles, not really the amount of light.

A modern PDAF system can deal with a huge variation in light, generally over 20 stops. They're usually specced from something like EV-1 (sunlight outdoors at noon) to EV19 (Candles).

The AF sensors in a camera are directional. They're aimed at two spots on the "exit pupil" of the lens. That's the lens's aperture, seen from the camera side. If you aim the sensor "beams" along a 10 degree V, you cover the part of the exit pupil that corresponds to f5.6.

• 1/(2 tan(10deg/2))

If you have a lens faster than f5.6, it doesn't matter, because the parts of the exit pupil that correspond to apertures larger than f5.6 fall outside the 10 degree V of the AF sensor.

Now, it's pretty obvious that a wider beam spread will give you a "better" AF signal, more "distance error" reading than you get from a narrower beam. So, why did they pick 10 degrees, instead of something friendlier to faster lenses, like 20 degrees (f2.8). Basically, PDAF systems (late 80s Nikon, Canon, and Minolta) started to catch on just about the same time zooms started to replace primes as the popular "kit" lenses. When I was learning, photographers started out with a 50mm f1.8 or 1.4, and added on a 35mm f1.4, 2.0 or 2.8, and a 135mm f2.8. The 105mm f2.5 was also popular, as was the 24mm f2.8. Nothing really slower than f2.8.

But in the 80s films got faster, and zooms got better, and we saw the 35-70mm f3.5-4.5 replace the 50mm f1.8 as the standard "kit" lens, and the 70-210mm f4-5.6 replace the trusty 135mm f2.8.

So, the camera makers did two things. They changed the screens on manual focus cameras to work better with slow kit zooms ("bright screens", microprism collars, and narrower spread angles on the prisms that make the "split image" focusing aids work) and they introduced AF, with sensors aimed at 10 degrees to accommodate the popular 70-210mm.

An f8 lens only lets in 1 stop less light than an f5.6 lens. So, if it were just a matter of the amount of light, you'd expect the f8 lens to work well outdoors, because what's 1 stop when you've got 70,000 lux of daylight. But it doesn't work with such a sensor, because you've only got a 7 degree exit pupil.

• 2*arctan(1/f8/2)

So the f5.6 beams are aimed outside the actual exit pupil. It might appear to "work" in really, really bright light, because there's lens flare outside the exit pupil, and the AF sensors are sensitive enough to try to focus on the flare, but doing this gives you focus errors.

Same thing with the idea of too much light. The AF sensors create an "effective aperture" of their own. The V is spread about 10 degrees, but the two sensors patterns themselves are only about 3 degrees. That's like two f22 "masks" on the f5.6 "ring" of the exit pupil (I won't repeat the equation). So, it doesn't matter if the lens is f5.6 or f1.2, the AF sensor masks it down to f22, as far as the amount of light it can receive. Faster lenses don't "focus better", they tax the AF system more.

Because faster lenses have shallower DOF, they demand more accurate focusing to look like the image plane is in focus. If you focus on a face, and shoot at f8, you might have 15 inches of DOF, so a 3 inch focus error is "hidden" in that deep DOF. But at f1.4, with 3 inches of DOF, a 3 inch focus error could take the eyes out of the sharp zone, and bring the ears into it.

That's the "technical limitation" that Sony has to deal with on their SLT,

• A given amount of time (1/24, 130, even 1/60 sec)

• A given lens communication protocol (Minolta, from the 1980s, where no one thought of using this stuff for video)

• A given motor and gear configuration (again, Minolta, from the 1980s)

Given X amount of time, and Y ability to move the lens, you can figure out exactly how accurately you can focus in the available time. Apparently, Sony did the math, and decided that they could focus well enough to look OK with f3.5 DOF, but not with f1.4 DOF, so, they gave it an artificial limit.

Not simple, but not beyond comprehension, either.

We're in the "horseless carriage" days. Many years ago, when gas, diesel, and steam motors just started to be practical, carriage companies started fitting their carriages with them, making "horseless carriages". Eventually, they gave way to people who made "cars", vehicles "purpose built" to be powered transportation, instead of adapter to it.

SLT is one of the horseless carriages. The car, in this case, is EVIL. The pieces are just starting to come together. Oly and Samsung pioneered the format, and started reaping the advantages, in compact, high performance lenses. Sony followed suit with NEX, a pretty refined system which really pushed the limits on the EVF. Fuji and Nikon built hybrid sensors that could do PDAF without diverting light to a second AF sensor. All those things will get better, and the SLT and SLR will eventually die off.

Wizfaq AF SLT

Ralf Bliesener wrote:

ET2 wrote:

SouthElginDad wrote:

Again, I don't get this. If the autofocus sensor needs lots of light, it ought to work better with apertures larger than 3.5. Can someone explain this?

The problem is that in most daylight conditions, the PDAF sensor would not be able to focus if the lens is open wider than F3.5 (let's say F1.4, or F1.8). That's too much light for the sensitive AF sensor.

In stills mode, all video-capable Alpha DSLR's/SLT's happily autofocus with the lenses attached at max open aperture. So how can there be "too much light" for the PDAF sensor?

There can't be. As I explained elsewhere in this thread, AF sensors effectively "mask" the lens's exit pupil down to a smaller opening, with 3 degrees, (f22) being the most common.

http://forums.dpreview.com/forums/read.asp?forum=1037&message=39997161
http://forums.dpreview.com/forums/read.asp?forum=1037&message=39997053

Also, why are there extra sensitive AF sensors in the top end Alphas that only get active when lenses with 2.8 or even brighter opening are used?

They don't really get any extra brightness. They still use a pair of 3 degree sensors, but instead of aiming them on a 10 degree V, they aim them down a 20 degree V, which doubles the amount of "error signal" (the "phase" in PDAF). But because those sensors are at 20 degrees, they'd "straddle" any smaller aperture. So, when you have 20 degree sensors, you also include conventional 10 degree ones.

Those other posts also explain the historical reasons for the choice of 10 degrees as the V angle.

My guess for the 3.5 in video AF is that this allows the AF algorythm in video mode to smoothly approach an area of focus achieved due to the somewhat wider DOF compared to wide open.

Your guess is 100% spot on. So, kudos for...

• guessing better than anyone else here.

• labeling your guess as a "guess" instead of stating it as fact, and even putting it in bold face, the way some of the other guessers did.

• being polite about it, instead of doing things like telling people to "grow up".

AF with an 85/1.4 or 135/1.8 for example wide open has the potential to completely miss focus and start hunting - with that totally ruining the clip - or that focus transitions would appear as too sudden and yerky, also taking away from the viewing experience. That could be ...

Indeed, it is. You get the same experiences from hand-pulling shallow DOF focus. There are scenes that allow it, and scenes that don't. You can't, in real time, analyze scenes and say "this one might work OK at f1.4, but we can't do better than f5.6 on that one". So, the camera maker picks a "safe" aperture that shouldn't cause trouble with their particular AF system.

Eventually, the AF systems will have better speed vs. accuracy curves. If f3.5 is the "safe limit" for a particular level of technology, you need to get up to 2.5x the speed for acceptable accuracy at f1.4.

... why it's fixed at F3.5. Less user errors, less problems of people complaining AF is crappy, etc.

Focus transitions in good movies are mostly done manually anyway (focus pull...).

Yup.

I can see the OP getting (a bit too) exited as this limit is not exactly proactively advertised by Sony. I tend to also perceive the forum members jumping at him as overreacting. YMMV on that aspect of the discussion here.

Get that sanity, reason, and logic out of this forum, right now!

Finally, shooting an 85/1.4 in video wide open allows all optical issues (LOCA bokeh shift often perceived as "purple fringing", lesser contrast etc.) to challenge the resulting clip. At 3.5, all versions of the 85/1.4 (be they the various Minolta incarnations or the CZ) look a lot better across the frame.

And stop using "experience" to back up your arguments, too!

Joseph S Wisniewski wrote:

Wizfaq AF SLT

Have you considered making these user articles? I think you can edit those later on if you ever wanted to add to them or whatever...

Makinations wrote:

Joseph S Wisniewski wrote:

Wizfaq AF SLT

Have you considered making these user articles? I think you can edit those later on if you ever wanted to add to them or whatever...

I should look into doing that.

Thanks.

AK Productions wrote:

how nice is it going to be when I throw this thing on my "soon to be complete custom made 15' Aluminum boom I am having built" set my lighting, set my A77 up, flip that bad boy on Continues focus and get some killer video shots how sick?? yeah thats what I thought......

You'll do your tests, but I wouldn't expect a PDAF module with a limited set of AF points to do too well in that situation. Hunting expected...

ET2 wrote:

AK Productions wrote:

Joseph S Wisniewski wrote:

Thank you very much for helping me better understand this, I cant believe how many wanna-bees are on this feed/Site just looking to boost their ego, and have no experience or idea what their even talking about, blows my mind. So thanks Again for clearing this up.

Don't take this guy too seriously. Have a look at Nex-7 rumor thread on News forum where he declared the posted leaked images so obviously faked that even DPR staff should be ashamed of themselves for not deleting the thread.

He is almost always blowing hotair with these long technical ramblings that he loves to post.

I've seen quite the opposite. Granted, Mr. Wisniewski erred with respect to the NEX and its lenses, but none of us walks on water. In most cases, Mr. Wisniewski is technically accurate and has a lot of experience with sensors that he's willing to share. So I would sooner not take you seriously. Do you differ with Joseph's analysis of the A77's focusing limitations? If so, what explanation do you offer?
--
Leonard Migliore

He's usually right. 1 counter example doesn't remove all credibility.

ET2 wrote:

AK Productions wrote:

Joseph S Wisniewski wrote:

Thank you very much for helping me better understand this, I cant believe how many wanna-bees are on this feed/Site just looking to boost their ego, and have no experience or idea what their even talking about, blows my mind. So thanks Again for clearing this up.

Don't take this guy too seriously. Have a look at Nex-7 rumor thread on News forum where he declared the posted leaked images so obviously faked that even DPR staff should be ashamed of themselves for not deleting the thread.

He is almost always blowing hotair with these long technical ramblings that he loves to post.