Sony’s Translucent Mirror May Reduce Detail in Photos by up to 5%

[kombizz0]

Apparently one of the main downsides to having a translucent mirror is that the light hitting the sensor passes through an additional layer (the translucent mirror), which reduces the amount of light and the image quality.

Ray Lemieux attempted to quantify how much the mirror affects the resulting image quality by removing the mirror on his A55 and comparing the results. He confirmed that about 1/2 stop of light is lost, and estimates that 5% of the detail in each shot is lost due to the mirror.
For more reading, please click here.
http://thesybersite.com/sony/a55/index.htm#slt_mirror_image_quality
--
I was born and brought up in Iran, a beautiful country full of history.
http://www.kombizz.com/

 

[Steve West]

Something looks odd in the first car scene. The result with the mirror looks like there is general OOF. He must have manually focussed without the mirror, and I wonder if he manually tweeked the focus with the mirror or just let the AF unit do the work (if the latter, then he might have to fine adjust the fine focus with that lens).

I certainly agree that there will be degradation with the mirror installed and that it will increase it's effect with faster beam convergence (more spherical aberration at faster f-stops), but something else looks wrong with that first comparison.
--
Steve W
weather photos: http://home.comcast.net/~scwest/atmo/

 

[mdesouki]

It's not a secret or anything, Sony says this on its website, there is no need to remove the mirror and run the tests. 30% of the incoming light is lost and used for focusing, so you get only 70% of the light, since every f stop is half (or double) the amount of light as the stop after (or before), so half of a stop is 25% of the light, so we're loosing a little more than half a stop with the mirror. But that is a price that you should be willing to pay for the faster, full time, always on, AF, otherwise you shouldn't buy an SLT. Also, I don't get where he got the 5% loss of IQ. Loosing light should (in theory) have nothing to do with loosing detail. I don't get the link. Taking pictures in very strong outdoor sunlight, loosing half a stop due to the mirror requires you to only adjust the shutter speed using the same ISO, why and how would the IQ drop? I have an A55, works just fine, if you want to know how not having the mirror in the way would perform better, read about the A580, which is exactly the same but an SLR.

 

[ET2]

Maybe Sony uses that 5% blur as part of AA filter compensation.

The right comparison is A55 vs A580, both use the same sensor.

There is is no difference in IQ at lower ISO. Go look at IR images.

In fact, A55 scores slightly better than A580 in resolution test.

 

[ET2]

In fact, A55 scores slightly better than A580 in resolution test.

And A55 resolution numbers at IR are better than D7000 (and far better) than K-5 too.

Go look it up yourself.

 

[Neeyon]

It's not a secret or anything, Sony says this on its website, there is no need to remove the mirror and run the tests. 30% of the incoming light is lost and used for focusing, so you get only 70% of the light, since every f stop is half (or double) the amount of light as the stop after (or before), so half of a stop is 25% of the light, so we're loosing a little more than half a stop with the mirror. But that is a price that you should be willing to pay for the faster, full time, always on, AF, otherwise you shouldn't buy an SLT. Also, I don't get where he got the 5% loss of IQ. Loosing light should (in theory) have nothing to do with loosing detail. I don't get the link. Taking pictures in very strong outdoor sunlight, loosing half a stop due to the mirror requires you to only adjust the shutter speed using the same ISO, why and how would the IQ drop? I have an A55, works just fine, if you want to know how not having the mirror in the way would perform better, read about the A580, which is exactly the same but an SLR.

Any time you put something in the way of the beam of light there is the potential for some detail to be lost due to refraction. The mirror is denser than air, and so the path of the light passing through the mirror is affected, however slightly that may be. So even the 70% of the light which hits the sensor would be subtly affected which could - in principle - affect detail.

This is all very theoretical though...I think the amount of effect is so tiny that it it is very hard to notice unless you're doing some strenuous pixel peeping. And there's no doubt that the SLT mirror, like a good UV filter, would have been designed specifically to minimize this interference anyway. The reality is also that light passes through multiple lens elements before it gets anywhere near the sensor so it has already been modified anyway, even in a traditional SLR. Oh, and don't forget that each little sensor node has something resembling a mini lens element over the top of it too, so that's another barrier the light has to pass through which could also potentially affect IQ. So while in principle yes, the mirror does affect IQ to some extent, I think sometimes some SLR purists forget that even in a traditional SLR there are number barriers which the beam of light has to pass through already. So the SLT mirror is simple "another" barrier rather than "the" barrier.

My personal opinion is that the IQ of the A580 has a minor edge over the A55, especially at higher ISOs, but I had to look long and hard at a lot of pictures to come to that conclusion so I don't think it's a difference most people would notice in everyday use.

 

[ET2]

The "test" doesn't consider that that 5% blur is used as part of AA filter compensation. So you can put a weaker AA filter.

Compare A55 resolution (at base ISO) with other cameras with the same sensor.

A55 resolution score is better than D7000
It's better than A580
And far better than K-5

You can find these numbers in IR review for these cameras.

 

[headofdestiny]

IR has long been known to have focus and lighting inconsistencies, but that is a topic on its own. Either way, the bottom line is, take off the mirror and get more resolution and light, regardless of how it compares with other cameras.

 

[ET2]

IR has long been known to have focus and lighting inconsistencies, but that is a topic on its own.

IR is not the only site with these figures. Photozone had similar result, especially regarding K-5 resolution which was worse than A33.

See DPReview numbers too.

There was another site with A580 and A55 resolution numbers. A55 scored better.

 

[ET2]

Either way, the bottom line is, take off the mirror and get more resolution and light, regardless of how it compares with other cameras.

That's true.

it does affect high ISO performance by half-stop. This is well known and not disputed. However, the difference is not visible to human eye until ISO 6400

I was especially commenting about the "blur" claim. Sony has definitely used this fact to put a weaker AA filter. There is no resolution difference at lower ISO compared to A580 that probably has stronger AA filter.

A55 resolution score is also better than D7000 and K-5.

 

[SteveGJ]

Even assuming this 5% is true (and was hardly a rigorously derived number), just how many photographers out there actually realise what is necessary to actually see this tiny difference? Firstly you can forget hand-held shots. It's virtually impossible to hold a camera that securely at even high shutter speeds to eliminate camera shake from the image. Forget any action shots - 5% difference in detail is going to make no difference; composition, timing, position and so on are vastly more important.

Then you have better use the best lenses you have, in the sweet spot of the aperture range. Avoid anything above f5.6 or difraction losses will lose you more than that 5%. Then you'll have to output your images to a large size, or crop them closely and inspect closely to perceive a difference that might be there.

Frankly, if you are into exhibition-quality, very large prints of landscapes, architectural images, fine art and other still images, then an APS-C camera is not the right tool. Better save up for that MF camera.

nb. would somebody care to provide a rigorous definition of what the loss of 5% in detail content actually is? That is something mathematically based. I'm just working on a subjective impression of the crops from those test photos.

 

[ThinCat]

It's well known that the translucent mirror loses light and detail. However, Sony obviously thinks that the benefits of having a faster, more accurate focus mechanism, faster frame rate, simpler design and better video outweigh these drawbacks. There are also obvious drawbacks to having a moving mirror - moving at exactly the time that you take the photograph. As the resolution of these sensors gets progressively better the camera shake becomes less and less acceptable.

There is no perfect solution yet.

 

[Mike Fewster]

Why is there image degradation?

-- If incoming light is reduced 5%, why doesn't a 5% decrease in shutter speed compensate just like every other exposure increment. 5% is as close to negligible as you can get in exposure. I don't think mechanical shutters are even consistent in speed to a 5% tolerance.
Mike Fewster
Adelaide Australia

 

[ser dpreview]

Why is there image degradation?

-- If incoming light is reduced 5%, why doesn't a 5% decrease in shutter speed compensate just like every other exposure increment. 5% is as close to negligible as you can get in exposure. I don't think mechanical shutters are even consistent in speed to a 5% tolerance.
Mike Fewster
Adelaide Australia

have evidence of this?

 

[underxposed59]

What percent of detail is lost by mirror slap in a traditional DSLR?

 

[iso rivolta]

The "test" doesn't consider that that 5% blur is used as part of AA filter compensation.

Do you have any documentation to support your repeated claim? An AA filter is made from birefringent optical materials and designed specifically for each sensor, it's not a random contrast loss inducing element, like the pellicle is.

The pellicle is creating something similar to lens flare. A nice and expected solution from Sony's part could have been to allow the swing out from the optical axis of the the semi-transparent mirror or at least a movement that would put it at a right angle with the optical axis when desired.

So you can put a weaker AA filter.

A weaker AA filter can be used by the manufacturer for limiting the loss of contrast induced under certain conditions (not always and not uniformly) by the pellicle, but that would mean anyway more remaining aliasing artifacts. A loss is a loss.

 

[DutchMM]

I had some landscape photos from my A55 blown up to 60cm by 40 - about 24 * 16 inches - for framing, and they are as detailed as one could ask. Are they exhibition quality? If not, I blame the photographer (myself) rather than the camera!

I would happily try a 60X90, but my wife would not allow me to take up that much wall space with a single photo ROFL.

--
Paypal? Just say "No!"

 

[mike_2008]

Hi iso, I wonder if you'd like to share some of your photographs taken with a sony camera here. Given your enormous interest in sony cameras I guess you must be a sony user, right?

PS. If you think you're saying anything new, think again. We went through all these discussions 10,000 times when the A55 was released. I suggest you look at the (excellent) real-world IQ of the released SLT cameras, rather than try to conclude anything when using such hand-waving terms about the tradeoffs in imaging with a removable lens camera.
--

The fact that a-mount lenses are still worth any money at all is thanks to the success of SLT.

 

[Ron Poelman]

my dumb guess is that another layer to pass through
is introducing all the expected garbage from refraction/reflection effects,
and that's where the quality loss starts.
How you derive that to be 5% is anyone's guess.
Shoot a grid and measure the fuzzy edges ?

Steve's point about horses for courses is the right one to make;
when is enough, enough ?
--

 

[tbcass]

So that means that at 24mp the actual resolution of the A77 would be equivalent to 22.8 mp and the A55 would be 15.2. I guess that means something to the IQ purists but in practical use it has been demonstrated many times that it requires a 50% increase in mp count to make a significant IQ difference. The fact is 5% is not visible at all. There are many other factors that would swamp the 5% difference. According to DPR the A55 has the same resolution in tests as the 18mp Canon cameras. To all those claiming to see a significant difference due to the mirror, get glasses!!!
--
Tom

Look at the picture, not the pixels

http://www.flickr.com/photos/25301400@N00/