which lens to buy?

[agorabasta]

There's no self-correcting process. It is easy enough to confirm this.

Yes, that seems very easy to do. I'd propose shooting an incandescent bulb and an old fluorescent tube separated by an opaque wall so that the light sources don't mix in two halves of the image. It would be better to use spot metering at some point on the fluorescent source itself and then fix it with AEL. Then fire a series of very short exposures about 1/1000s or shorter. If there's any gain/scaling really happening, the image of incandescent bulb would also show some brightness modulation between the shots.

I have just tried shooting an entrance into a room that's lit from the outside with incandescent bulbs and is lit inside with old fluorescent tubes. So far, I can say I see some modulation where it should not be present without the post-shot gain/scaling correction. But some further checks are necessary, since the modulation is only about 7%, and I'm not sure it's not caused by reflections.
Then I tried shooting two bulbs, one halogen incandescent and another a modern fluorescent, so that the image of the bulbs was not overexposed, but without a separating wall between them and got max 20% modulation in the captured brightness of the halogen incandescent.

So far, it seems that if the cameras are doing any soft image brightness corrections, they do so at the analogue gain stage since the raw levels in the shadows are not shifted.

Btw, I used an a700 at 5fps and then an a55 at its lower continuous speed.

[David Kilpatrick]

Well ok, reluctantly the post adjusting is out until proven otherwise, but what about exposure compensating DURING the exposure, by boosting the gain a tad or something like that.
Btw I still have my X-700 handbook (in pristine condition) and I’m fairly sure that camera can adjust the exposure during the exposure as long as the MD lock button is engaged (MD lens needed)….I’ll have to dig it out and check sometime.
Greg

The X-700 has no metering during exposure except for flash. Although Minolta invented off-the-film metering and sold the idea to Olympus, they never went down this route themselves because they were bright enough to realise different films have different colour/density emulsions and therefore produce exposure errors.

Nothing happens during digital exposure - nothing at all. The sensor simply sits there receiving photons for a set period, decided in advance. Even during video, the wells are emptied (read out) 50 or 60 times a second so it's like 50 or 60 still frames, there is never a continuous process happening. Everything happens after exposure, when the image is read out - collected off the sensor. That's when gain occurs. The only significant factors during both stages may be the temperature of the sensor, or some fractional shift in light values during the exposure itself.

I've now done some tests on Firmware v1.04 and conclude that if using the electronic first curtain has any effect, it's not significant, and also that v1.04 does not produce any improvement in image quality which I can detect in raw files.

David

[agorabasta]

Everything happens after exposure, when the image is read out - collected off the sensor. That's when gain occurs.

Sure! But that's exactly the problem! If the iris overshot, the processor would be trying to scale to match the expected brightness by commanding more analogue gain at the input stages of ADC.
Then likewise, if even the AEL is locked, it may be locked not only in the shutter speed and aperture, but also in the expected brightness at the measured area - and that could employ gain readjustment if that flaky mechanical aperture fails to work as expected.

There's no really hard evidence to deny the reality of gain adjustments. Moreover, the gain adjustments are all too regular when using fast lenses on digital, as a nominal A-stop may differ from the actual T-stop by more than a full EV easily. So the 'cheating' mechanism is always there, all they have to do is use it to their needs, that's it!

[Greg Beetham]

I just went back and had a look at David’s 5 Euro price label photo again and the ISO640 shot looks pushed too my not so wonderful pp ability, it may have other things wrong with it from other unknown causes but too my sensibilities it looks like it was initially underexposed by an unknown amount and then pushed to the point of over exposure.
If that is not the case then what is?
Greg

[David Kilpatrick]

Simply that ISO 640 is far worse than ISO 500. I've made further tests and can confirm that from ISO 400 up, the noise levels increase very rapidly to 800. I would not have believed that such a small step could make such a difference but if you remember, the A700 and A900 are the same - ISO 160 is far better than ISO 200, and actually ISO 320 is one of the optimum settings on the A900.

If the step seems odd, remember that there is an SLT mirror in the way!

That means that what is called ISO 500 is actually ISO 640-800, but what is called ISO 640 is really faster than 800 - so maybe it's not unexpected that the more obtrusive noise kicks in around that step.

I have now changed my everyday auto ISO setting to 100-400, not 100-1600 as I used on the A55. What's really annoying is that you can not select 1/3rd stop steps for the auto ISO range.

David

[Greg Beetham]

Well ok above ISO500 you are pushing your luck, but is it consistent; does the camera do the same thing ALL the time? And what does f11 look like at low ISO? no problems?
Greg

[David Kilpatrick]

f11 is always softer than wider apertures on the 24 megapixel sensor. Diffraction may begin in theory at anything smaller than f/5.6, but you don't really see this at f/8, f/9, even f/10. By f/11 it is visible and I would now hesitate to use anything smaller than f/11. It can be fixed to some extent by careful use of Sharpness, Detail in ACR.

What I find surprising is how much more visible this is than with the A55. I guess that in normal photography, apertures like f/8 to f/11 are common. On the A55 you don't see a clear difference between these, on the A77 you probably will.

David

[stevecim]

david, does the scene you're shooting have a impact has well? i.e is it more of a problem with landscapes?

[alphaomega]

Simply that ISO 640 is far worse than ISO 500.

David, what happens to the A77 noise comparison with A55 using the same lens if the 24Mp image is reduced to 16Mp? Would that equalise the noise at ISO640?
I find that discussions such as this one re-inforces my wish to acquire the NEX-7 rather than a SLT with 24Mp as noise is less likely to be an issue with the NEX.

[alphaomega]

I actually went straight to SAR after submitting the above post and found this DxO comparison between A77, NEX-7 and A900 http://www.sonyalpharumors.com/sony-nex ... all-score/
The NEX-7 is the winner here.
At Sports (low light ISO) the NEX-7 scores 1016 ISO) and A77 801 ISO. I know that some people pay little attention to DxO but at least it kind of confirms David's findings. The translucent mirror must have a deleterious effect at higher ISOs when a sensor treshold has been surpassed. It would appear that the jump to a 24Mp SLT is a jump too many if noise at higher ISOs in an important issue.

[agorabasta]

Strictly speaking, a 24Mp sensor should be sharper and also provide more image detail than a 16Mp one at any lens aperture. That's quite obvious in comparisons between Nex7&5N. But the sharpening radius must be 1.5 times larger for the 24Mp sensor whenever the resolution gets optically limited, as it surely is e.g. at f/11. The 1.5 multiplier comes as sqrt(1.5) due to pixel pitch times about as much to equalise the effect of LPF on a 16Mp sensor where it has to be more optically limiting.
If a 24Mp system output doesn't appear sharper and more detailed than a 16Mp one, there must be a problem with the design/defects of the sensor toppings.

[Greg Beetham]

David I’m sort of worried about Macro now, what happens with say f16 on a good macro lens, does one get an image that we are used to getting i.e. fairly reasonable considering, or pretty average? Do we really have to think about stacking at a max of f8 as the only means of getting a good macro photo now? And if you take bellows and micro lenses, macro lenses extenders and tubes into consideration where you can end up with f48 before you even start stopping down, what’s going to happen there?
I guess there is one good reason why Sony hasn’t been exactly keen on giving us mirror up capability with the SLT’s, they’d have to have two complete sets of in camera lens parameters, one with the mirror down and one with it up; as there would be a different ISO and lens T curve for each lens right across the range so that the camera could position the quadrant relay in the right position to accomplish the correct amount of light coming through the diaphragm and regulate the matching sensor gain accordingly.
Greg

[agorabasta]

I guess there is one good reason why Sony hasn’t been exactly keen on giving us mirror up capability with the SLT’s, they’d have to have two complete sets of in camera lens parameters, ...

They did exactly that for the 5N+LA-EA2, and promised to do the same for all earlier Nexes. Was no problem, obviously...

[Greg Beetham]

If a 24Mp system output doesn't appear sharper and more detailed than a 16Mp one, there must be a problem with the design/defects of the sensor toppings.

I'd be inclined to add another proviso, one that covers the capability of the lens in having a sufficient LPPM/MTF for the 24MP APS-C sensor as well. This baby has opened a can of worms I think, with the size of the sensels it has in conjunction with the lenses we have.
Greg

[Greg Beetham]

So the LA-EA2 has mirror up capability then agorabasta?
Greg