[admin]

[quote name='Brightcolours' timestamp='1283435232' post='2423']

We are not talking about light reflected from the sensor, here, though.



We are talking about light entering the mirror glass at surface one. It travels through the glass, towards surface two, where most will exit it again. Part of the light, however, will reflect off surface two, upwards at 90 degrees (think of your prism experiments in science class in high school).

That light will travel through the glass till it will meet surface one again, at a 45 degree angle. There, part of the light will reflect back at an angle of 90 degrees, towards surface 2 again. And most will proceed there towards the sensor to be registered.



Since the mirror is reflecting up, the ghost image will appear down the "real" image on the photo (images get projected mirrored through the optical axis in cameras).



The reason you never noticed this with for instance an EOS RT probably has to do with the thickness of the pellicle. The RT (and 1N RS) used a very thin membrane that was actually quite delicate (and often would break just from cleaning).

It appears Sony used a much thicker mirror.



What happens with light that gets reflected back from the sensor, I do not know. I guess one thing we can be sure about is that ghost lights caused by UV "protective filters" in night photography will be less intense than on other cameras, because the light will lose 1 whole f-stop by passing through the mirror twice more, again.

[/quote]



Sorry, but your explanation is valid for all glass elements and that includes the lens elements of course.

I am slightly wondering whether the design has massive advantages regarding sensor reflections though.

Many lenses have uncoated rear elements so sensor reflections are an issue here.

[popo]

Just for some ball park numbers, according to Canon, their coatings can reduce reflections to a fraction of 1%. To put that into photographic terms, let's just round it about 8 stops down. It wont be precise or exact, but I'm just going for a ball park figure here.



For a flat piece of anti-reflection coated glass angled so it is not parallel to the sensor, the main reflection from subject reaching sensor will reflect off the inner then outer surface, so it is about 16 stops down from the original.



In Sony's case, we're dealing with a mirror surface. The direct light loses a little due to the mirror. The 1st unwanted reflection would bounce off the inner surface, then off the mirrored surface and onto the sensor. Let's say the reflected amount is about 2 stops down, so combined with the unmirrored surface that's about 10 stops down from original.



So compared to the unmirrored case, there is a fair difference of perhaps 6 stops. You can argue the figures used are wrong, and that's entirely likely. If the anti-reflection coating isn't so good, there will be more reflections and the difference is reduced. But the bottom line is still a mirror surface will reflect much more than a typical glass surface, and it will be worse.



As for sensor reflections of uncoated lens elements, I think we have a different mechanism here. I think every lens I've seen has a non-flat rear element, typically convex. The curved nature will disperse any reflections and that will help make them less noticeable.

[Brightcolours]

[quote name='admin' timestamp='1283445983' post='2425']

Sorry, but your explanation is valid for all glass elements and that includes the lens elements of course.

I am slightly wondering whether the design has massive advantages regarding sensor reflections though.

Many lenses have uncoated rear elements so sensor reflections are an issue here.

[/quote]

No, it is not true that that is valid for all glass elements. It is only true for all flat slabs of glass at an angle like here 45 degrees, no other glass element can cause that.



Sensor reflections are usually not a huge problem with rear elements. With modern lenses they are all coated (you can see that), but some have newer coatings than others. It was said that uncoated elements can cause the so called purple fringing around very "over exposed" areas. And newer coatings would counter that.

However, whether that is true I do not really know for sure. I do know that it is a bit hard to imagine the convex shape of the rear element to to actually let light bounce back the almost exact way it came from, but who knows.



The biggest problem with sensor reflections I already mentioned extensively, and that is the famous ghost lights mirrored through the optic axis, when one is shooting in dark conditions with a filter mounted (flat slab of glass).



And yes, I do imagine the Sony A55/33 to eliminate those (filter caused ghost light images) to quite a large extent, as the light passing through the mirror twice will lose about 1 f-stop of light. I however do not really see that as an advantage, as I do not want to use "protective" filters, I do not believe in their protective qualities (rather use a hood). I do believe in their image degrading qualities under certain conditions.



But again, what is happening in the glass mirror of the Sony A55V/33 only happens in that mirror, not in other glass elements. Unless you happen to have a beam splitter in your lens. What is so specific about the mirror is the flat slab of glass, and the angle.



The Canon EOS RT has a very thin membrane of plastic. It has been known for decades that to avoid ghost images the pellicle membrane needs to be very thin, not just in SLRs. But obviously, the Sony glass mirror is less fragile.



To visualize what happens because of the mirror thickness:





Theoretically it would be possible for more than one ghost light to appear.

[admin]

[quote name='Brightcolours' timestamp='1283456315' post='2430']

No, it is not true that that is valid for all glass elements. It is only true for all flat slabs of glass at an angle like here 45 degrees, no other glass element can cause that.



Sensor reflections are usually not a huge problem with rear elements. With modern lenses they are all coated (you can see that), but some have newer coatings than others. It was said that uncoated elements can cause the so called purple fringing around very "over exposed" areas. And newer coatings would counter that.

However, whether that is true I do not really know for sure. I do know that it is a bit hard to imagine the convex shape of the rear element to to actually let light bounce back the almost exact way it came from, but who knows.



The biggest problem with sensor reflections I already mentioned extensively, and that is the famous ghost lights mirrored through the optic axis, when one is shooting in dark conditions with a filter mounted (flat slab of glass).



And yes, I do imagine the Sony A55/33 to eliminate those (filter caused ghost light images) to quite a large extent, as the light passing through the mirror twice will lose about 1 f-stop of light. I however do not really see that as an advantage, as I do not want to use "protective" filters, I do not believe in their protective qualities (rather use a hood). I do believe in their image degrading qualities under certain conditions.



But again, what is happening in the glass mirror of the Sony A55V/33 only happens in that mirror, not in other glass elements. Unless you happen to have a beam splitter in your lens. What is so specific about the mirror is the flat slab of glass, and the angle.



The Canon EOS RT has a very thin membrane of plastic. It has been known for decades that to avoid ghost images the pellicle membrane needs to be very thin, not just in SLRs. But obviously, the Sony glass mirror is less fragile.



To visualize what happens because of the mirror thickness:





Theoretically it would be possible for more than one ghost light to appear.

[/quote]





Yes, I surely understand what you mean but this effect is also present with all glass elements - just not with a 45 degree reflection. There're reflections all over the place in there. Just imaging a near center ray which is (in part) bounced back by near-180 degrees and then reflected again. This would cause the same kind of issue you described here.

[Brightcolours]

[quote name='admin' timestamp='1283493195' post='2443']

Yes, I surely understand what you mean but this effect is also present with all glass elements - just not with a 45 degree reflection. There're reflections all over the place in there. Just imaging a near center ray which is (in part) bounced back by near-180 degrees and then reflected again. This would cause the same kind of issue you described here.

[/quote]

No, it can't. Like Popo also already pointed out, a rear element is not a flat slab of glass. Anything that would reflect internally would be spread all over the place, since it is convex. It would never leave a ghost image. Further, the angles within the rear element are much less than 45 degrees, so reflections to the same extent are much less likely.



Also, rear elements have very rarely a reflective coating applied to one of the surfaces <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/wink.gif' class='bbc_emoticon' alt='' />.



I have never seen any lens cause ghost lights (except when they have "protective filters", see an earlier post which illustrates the ghost lights mirrored through the optical axis). Lenses with lesser coatings (this is not about the rear element, but all elements) are less contrasty, not producing ghost images.



What occurs in the A55V/33 is really quite exclusive for those cameras. Even other cameras with beam splitters solved this problem: The Minolta E-10/E-20 DSLRs used prisms instead of a flat slab of glass

http:[img]//a.img-dpreview.com/reviews/olympuse10/images/e10cutaway.gif[/img]

and Canon used a very thin plastic reflective membrane in for instance the EOS RT.



Actually, strictly speaking Sony does not use a pellicle (which stands for membrane), but a mirror instead, a "plate beam splitter".

The 2nd surface reflections are a well known problem in the beam splitter world (used in telescopes, optical measurements with lasers, a.o.).



Some links to illustrate further:

http://www.datasheetarchive.com/datasheet-pdf/05/DSA0084689.html

[url="http://www.thorlabs.de/NewGroupPage9.cfm?ObjectGroup_ID=915"]Plate beam splitter (Sony A55)[/url]

[url="http://www.thorlabs.de/NewGroupPage9.cfm?ObjectGroup_ID=898"]Pellicle beam splitter (EOS RT)[/url]

"Eliminates Ghosting"

[url="http://www.thorlabs.de/NewGroupPage9.cfm?ObjectGroup_ID=754"]Prism beam splitter (Olympus E-10)[/url]

"Since there is only one reflecting surface, this design inherently avoids ghost images."

[wim]

[quote name='Brightcolours' timestamp='1283456315' post='2430']

No, it is not true that that is valid for all glass elements. It is only true for all flat slabs of glass at an angle like here 45 degrees, no other glass element can cause that.



Sensor reflections are usually not a huge problem with rear elements. With modern lenses they are all coated (you can see that), but some have newer coatings than others. It was said that uncoated elements can cause the so called purple fringing around very "over exposed" areas. And newer coatings would counter that.

However, whether that is true I do not really know for sure. I do know that it is a bit hard to imagine the convex shape of the rear element to to actually let light bounce back the almost exact way it came from, but who knows.



The biggest problem with sensor reflections I already mentioned extensively, and that is the famous ghost lights mirrored through the optic axis, when one is shooting in dark conditions with a filter mounted (flat slab of glass).



And yes, I do imagine the Sony A55/33 to eliminate those (filter caused ghost light images) to quite a large extent, as the light passing through the mirror twice will lose about 1 f-stop of light. I however do not really see that as an advantage, as I do not want to use "protective" filters, I do not believe in their protective qualities (rather use a hood). I do believe in their image degrading qualities under certain conditions.



But again, what is happening in the glass mirror of the Sony A55V/33 only happens in that mirror, not in other glass elements. Unless you happen to have a beam splitter in your lens. What is so specific about the mirror is the flat slab of glass, and the angle.



The Canon EOS RT has a very thin membrane of plastic. It has been known for decades that to avoid ghost images the pellicle membrane needs to be very thin, not just in SLRs. But obviously, the Sony glass mirror is less fragile.



To visualize what happens because of the mirror thickness:





Theoretically it would be possible for more than one ghost light to appear.

[/quote]

Your drawing is actually incorrect. The light should be refracted, twice (once at each mirror surface, see for example [url="http://www.thorlabs.de/NewGroupPage9.cfm?ObjectGroup_ID=915"]http://www.thorlabs....ectGroup_ID=915[/url]), therefore only causing at worst two images slightly offset to each other, as the beams before and after the mirror have to be parallel. Also, the scenario as you draw it here with more than one reflection is not possible, unless they had chosen the refractive index of the reflective mirror glass deliberately so to create those reflections (IOW, a very, very high refractive index). And good coatings will reduce these reflections by 99.8 % or more.



I therefore conclude, that final proof for the cause of the reflections in those images, unless scientifically proven to be from the mirror, may well be something different. They could well be reflections of the sensor, re-reflected by the AA-filter, or even a similar thing as the black dot phenomenon, IOW, a processing problem when reading the raw data. Or something completely different for that matter, but very likely not any reflections caused by the mirror, unless they really made a terrible engineering error.



Kind regards, Wim

[Brightcolours]

[quote name='wim' timestamp='1283544989' post='2473']

Your drawing is actually incorrect. The light should be refracted, twice (once at each mirror surface, see for example [url="http://www.thorlabs.de/NewGroupPage9.cfm?ObjectGroup_ID=915"]http://www.thorlabs....ectGroup_ID=915[/url]), therefore only causing at worst two images slightly offset to each other, as the beams before and after the mirror have to be parallel. Also, the scenario as you draw it here with more than one reflection is not possible, unless they had chosen the refractive index of the reflective mirror glass deliberately so to create those reflections (IOW, a very, very high refractive index). And good coatings will reduce these reflections by 99.8 % or more.



I therefore conclude, that final proof for the cause of the reflections in those images, unless scientifically proven to be from the mirror, may well be something different. They could well be reflections of the sensor, re-reflected by the AA-filter, or even a similar thing as the black dot phenomenon, IOW, a processing problem when reading the raw data. Or something completely different for that matter, but very likely not any reflections caused by the mirror, unless they really made a terrible engineering error.



Kind regards, Wim

[/quote]

My drawing is very correct. No idea why you think otherwise. The 3rd (2nd ghost image) I only put in to show that in theory that can happen (but obviously, you won't see it as the light intensity would be very low by then), just disregard that 3rd image.



The diagram from thorlabs shows something different. They show the problem of getting ghost measurements from the part of light they are splitting off the main beam, in the A55's case the light that would be going to the AF sensor.



They have NO diagram that shows what happens to the image itself. But MY diagram does show that. Just a fact of 45 degrees slabs of glass.



Anyway, whether or not you want to believe it or not, the cause of the ghost lights is evident, and it is in the use of that mirror. Not something I dreamt up, Wim.



And also the reason why others either used a pellicle instead of a glass mirror, or a prism assembly.

And yes, even the Thor diagram does show how the 2 reflective surfaces are problematic.



That I did not draw in unknown diffraction should be forgiven, as it is not important to make what is going on clear.



But anyway, if we take an unknown refraction index into account, things might look like this:

[popo]

I think that makes sense.



Looking at it another way, any light ray hitting the mirror that is not perpendicular to the mirror will be refracted to some other non-perpendicular angle. It doesn't matter what that is exactly, but internal reflections will occur and could provide additional light paths offset from the main image. The level of that would be significantly reduced, but this is similar in principle to front filter caused flare.



I think we can all agree there is a mechanism for a front filter to cause unwanted images by reflection from the sensor. That is with glass, even anti-reflection coated glass. In the case of a semi-silvered mirror, one surface is much more reflective, and has potential to make things much worse.



For an angled mirror, as a rough 1st thought, I don't think sensor reflection to the mirror could cause additional images, as the light would be going in the wrong directions after refraction/reflection and wouldn't return to the sensor again. So for now, I think reflections within the mirror is the most likely cause.



I'd have to revise my optical physics equations to put numbers to that... and its a bit late in the night to start now.

[Guest]

I think Joseph Wisniewski at DPReview gives a pretty good explanation why back reflection from lenses is nowhere as bad as ghosting from Sony's partial mirror. Allow me to quote from:

[url="http://forums.dpreview.com/forums/read.asp?forum=1000&message=36225053"]http://forums.dprevi...essage=36225053[/url]



"Lens surfaces are generally

• curved - which means that any rays that are converging on the sensor (and therefore, capable of generating a sharply defined image) will have their point of convergence changed by reflecting from a curved surface. The curved surface will either increase convergence and make the ghost image focus in front of the sensor and then diverge again before it hits the sensor, or decrease convergence and make the ghost image not converge before it hits the sensor.
• reasonably perpendicular to the optic path - so anything that does converge will do so pretty well aligned with the original image.

The pellicle mirror reflections are from a

• flat surface - that doesn't really affect converging rays: so the ghosts converge as sharply focused images.
• at 45 degrees to the optic path - to displace the ghost by the thickness fo the mirror (less some effects due to refraction).

This is why Canon used a much thinner mirror in their Pellix and EOS RT cameras, and why Oly used solid cube beam splitters in the E-10 and E-20"



This agrees with what Brightcolors has been trying to convey here.



Sigh, such a pity that partial mirrors in DSLRs have suddenly dropped off my radar screen.

[popo]

I just played about with some calculations. Of course there are a lot of unknown variables.



If we assume the mirror is 45 degrees from vertical with the silvered surface on the outside, and it has a refractive index of 1.5, then an unwanted reflected image would be formed at a spacings multiples of about 1.07x the thickness of the mirror.



For the mirrored surface reflecting at 30% and the glass surface at 1%, each additional reflected image is about 7.2 stops down from the previous. If you assume the coating is even better at 0.1% reflection then the figure becomes 10.5 stops. With such huge losses, the reflection of the reflection is going to be so far down it is unlikely to be significant short of pointing the camera at a very bright point source.



So looking back at the sample image earlier, or more precisely the original of it, the spacing of the secondary image is about 12 pixels offset from the original, working out to be about 0.06mm. That's pretty small! Even if they used a high refractive index material and a more vertical angle the mirror would still have to be pretty thin, a fraction of a mm.



I can refine the above if anyone has more precise values for any of the variables.