The Sony patent (google translated)
–
Egami (google translation here) found a new Sony patent describing a rotating semi transparent mirror tech. The reason why the mirror should rotate is because by changing the angle you will also change the kind of light that will hit the sensor. For example you can opt to transmit or not to transmit infrared light. Not only that it seems also capable to influence the RGB color separation. Really don’t know if this will ever make it into a real production camera or just lay on the desk of the Sony engineers for the years to come…
That said, what do you think of the patent?
50 Responses
I can’t understand Japanese, but seems like this is the ‘mirror flip-up mode’ we’ve been waiting for. By tilting back the transparent mirror parallel to the image sensor now the light can pass through 100%.
Just my 2 cents’ worth.
Look at the image. It actually rotates, not flips up.
You still loose 90% of light.
What they patent is basically an optional IR filter on a mirror – if it’s vertical than you get full-spectrum images (think: astrophotography), if it’s tilted you get regular photographs. SLT would work like variable hot mirror.
*30% (lol 90)
A rotating mirror could also be used for cooling the sensor.
Not sure if serious.
Of course not 🙂
Just checking – I’ve read some outlandish things around these parts. 😀
Nice joke 🙂
Awesome! Love it…! lol.
Doesn’t make sense to me. If it’s a mirror (semi translucent), and it reflechts 1/3 of a stop of light (to the AF sensor), then it means that if you rotate the mirror to vertical position, it would still reflect light, back out of the camera.
The change of angle changes which wavelenths are transmited or reflected.
Different mirror/ filters are available to achieve the same effect with differrent wavelengths…they go in an angled holder behind the mirror.
This camera is intended for use of non visible light.
These will be special purpise cameras and will NOT BE ALPHA CAMERAS.
You could have a leslie effect like in the Hammond B3.
I think Sony may have hit on a great idea with the rotating semitransparent mirror.
I posted an article not long ago on my camera history;
Olympus OM10 – died in a canoeing accident
Olympus OM40 – sold
Olympus OM2n – a keeper
Canon T90 – a keeper
Canon T90 – a keeper
Canon 300D – jury is still out on this, I said I may have it converted to infrared
Sony Alpha 77 – a keeper
There was one proviso with keeping the Alpha 77 and that was if Sony didn’t bring out a camera that could do infrared at the flick of a switch – it seemed they took notice of my post and rushed to patent this idea, (see how they value my custom)now if only they would do the same thing to bring out firmware to enable tethering on the Alpha 77.
I hope that this innovation does come to the market place, it may just be regarded as another in camera effect but I would pay another $200 dollars to incorporate such an innovation and I’m sure there would be many who appreciate the other worldly effect of infrared photography willing to pay the extra. Sony would also steal a march on the Canikon crowd. Well done for your innovation Sony, just put it into production. Please.
Sony camcorders have long had night vision mode which enables infrared at the flick of a switch.
Its great for filming your kids in total darkness.
I think night vision mode (infrared) was removed from Sony handycam long time ago. I have one that has this feature and it’s great. I don’t know if they have revive this feature for newer handycams.
Good idea. The IR info can be used in different ways; dramatic B&Ws is one of them.
This patent is clearly not for consumer cameras.
The text includes the words survielence camera and infrared.
The body illustration also looks nothing like a consumer camera.
Probable a camera for night surcielence ( security camera)
Why would a security camera need a mirror?
The patent states it can be used either for srvielence camera where by changing the angle of the miror the wavelengths reflected will change.
That would allow night survielence using multiple viewing modes without detection.
It also states it can be used for a night vision camera but would require a special camera for that purpose.
It goes on to describe the abillity to take high contrast black and white pictured.
So the inovation of this device is to allow selection of differrent light wavelength modes therefore being able to see subjects in manners not possible with visible light alone.
In both cases thes are specialized cameras and not for general purpose consumer cameras.
It is clearly at this time not intended for use in alpha cameras.
Admin
In your previous post about 54mp sensor one thing was very unclear.
Since it may be stacked are we talking about what we are used to calling 54mp ( 54mp r, and 54mp g, and 54mp b )
Or are we talking about what Foveon would call 54 mp ( 18 mp x 3 layers )?
No, that is not what we are used to calling 54 MP.
54 MP the usual way would be 27 MP green, 13.5 MP blue and 13.5 MP red.
Exactly as the 24 MP a77 is 12 MP green, 6 MP blue and 6 MP red.
Stacked 3 layer sensor would be rgb not rggb.
Only bayer sensors are rggb.
If they use the ycgg patern..2 layers instead of 3 …then your numbers are possible…it is unclear how they would call this because there is no precident for a 2 layer sensor.
I would like to see a pixel called a pxel and contain all necessary colors so we do not have to speak a differrent language every time we talk about a differrent sensor.
If there is a differrence in quality compared to different sensors then they can claim EQUIVALENT TOO.
I realize Sigma has set a president with foveon..but it is a bad president..other manufacturers should ignore that precident and call a pixel a pixel therefore leaving Sigma as the only manufacturer making these blatently dishonest/misleading/confusing ( take your pick) claims.
Bayer call a pixel a pixel.
A unit of rggb is called one pixel and when you look at the resolution of the resulting file it shows that. A 24mp bayer produces 6000x 4000 pixel file.
A. 46 mp. Fovion sensor is a blatent lie as it only produces a file with 15 megapixels.
Thats why the fovion lies are the worst kind because people like yourself actually believe it.
Look at the specs..compare the file resolutions.
This is why I am adamently against what sigma claims for thier resolution…you. yourself are proof of thier lies…look at the file resolution and you will see that Sigma has played you for a fool and you too will be justifiably outraged.
No. A pixel on a sensor with a Bayer CFA is not a full pixel with red, green and blue information. Each pixel contains EITHER red, green or blue information. The two other color channels for each pixel are guessed in later processing.
As I wrote, the 24 MP a77 only has 12M green pixels, 6M red pixels and 6M blue pixels. That is not 24M true pixels but something more comparable to 6-8M true pixels.
So to use your own words (with the spelling corrected), the pixel count of a Bayer sensor is a blatant lie, .
Foveon sensors started out as true pixels, but in marketing they could not compete with the Bayer lie. So Foveon started lying too, and now Bayer and Foveon pixel counts are comparable.
Allen
One of us is mistaken about how bayer pixels are counted.
Can you possibly provide a link to document what you are saying?
I would like to find out if I am correct or mistaken before I respond to your comment.
IF I find that I was wrong I will say so but I would like to check the facts first..thanx.
Alen
Please provide a link to the statement you made about bayer pixel count.
I would like to check it before I respond.
I just did some refferrence reading about bayer pixels and demosaicing.
Apparently demasaicing creates an interpolation of a 4 pixel array from 4 differrent locations.
Therefore you are correct..bayer sensors simply have a more clever lie than foveon therefore I instantly realized the foveon lie but for a long time I falsely believed the bayer lie.
Scaling up a foveon inage at 200% would therefore produce an interpolated image similar to a much higher resolution bayer image.
Thank you for pointing this out to me..you have corrected an error that I have falsely believed for a long time.
I now look forward to stacked sensors even more.
Bayer sensors don’t “lie” about megapixel count. A 24 MP sensor has 24 million active photosites and produces an output image with 24 million pixels. So in terms of linear resolution a 24 MP sensor is a true 24 MP sensor. However, in terms of colour resolution it is not 24 MP, since 2/3 of the colour information has to be interpolated for each pixel.
It has 24 million active monochrome photosites, yes.
A true 24 MP sensor would require 72 MP monochrome photosites to form 24 M true pixels.
Prof’s correct, Bayer output MP is exactly as reported MP. For example, if the claimed MP is 6mp, the picture does have 3000×2000 pixels. The catch is 3/4 of those are guessed(calculated based on surrounding pixels). I don’t know much about Foveon, but based on my preliminary reading, they are claiming 3x more resolution then what it actually is.
An interesting note
The 2 layer yyccgggg. Two layor patent that sony has will as a 54 mp sensor provide 1/3 more resolution than the 3 layer stacked sensor with only minimal interpolation of the checherboard pattern y and c.
It seems this is the most promising desighn for high resolution output.
Well, honestly this looks like half way back to a traditional reflex mirror – is this really the sort of news A-Mount users are frequently crying for here?
So where does this fit in the Z Shit camera?
sorry, *SHIFT camera
Clealy sar readers cant read….
This is NOT FOR ALPHA
It is for special purpose cameras ONLY.
That fact is CLEARLY STATED IN THE PATENT!!!!
All these comments are written as if this patent were in some way relevent for consumer cameras..it is not.
Yes it may be for surveillance but if Sony saw there was a market for such a sensor on an Alpha body would they say, “oh, we would love to put it on an Alpha body but it is only for surveillance” no, they would put it on a consumer camera and steal a march on Canikon. It may take a number of years for this to occur but all of the camera manufactures are looking for a unique selling point to attract customers to their brand
Mate, this is SonyAlphaRumors here and since E-Mount does not have a mirror it’s A-Mount news. Of course I’m unable to READ the patent, because egami is only a blog writing about the patent and unfortunately I’m also unable to understand the Japanese blog. But the Google translation does not say what you’re claiming is stated in the patent: “It’s fit to α mount”, Google translates to me plus: “This patent application seems to be assumed also night vision camera or the like surveillance camera.” Usually “also” means another option, but since you’re claiming this is an exclusion, I’m sorry, here I’m obviously also unable to read.
Click on the link in bold that says egani( google english translation)
Imeadiatly below the piture and caption.
Then you can read the actual patent rather than my interpretation of it.
It’s mostly the writer’s interpretation and thought on possible use.
Maybe the mirror can be gold so it can reflect light onto the subject through the lens and be a useful portrait camera
just give us flip up mirror..
I’m not in a position to read the translation, so I have a question. Is it possible to have a infrared mirror that transmits 100% visible light and reflects only infrared to the AF sensor, as what *may* be shown in “State 1”? To me it seems feasible to have a mirror that transmits everything in visible wavelengths. At least that is what the “State 1” graph looks like to me. If that is the case, why couldn’t this be used on Alpha cameras? If so, it’s actually fairly clever.
Hmmm…. Wait a second, don’t camera manufacturers use infrared filters on their sensors, currently? If so, this is a clever way of using the reflected infrared.
Depending on how wide of a spectrum Sony sensors can detect, this may be great for multi-spectral imaging. Now I think about it, multilayer sensors could be used to detect light as well as more energetic spectrum at the buried layers.
All very interesting…. and there is a huge market for this right now.
If you relied on infrared light for focus, nearly all shots would be front focused. This kind of explains it: http://photo.stackexchange.com/questions/23221/how-to-use-ir-index-on-canon-ef-24-105mm-f-4-l-usm
Another way to explain it, if nearly all lenses have chromatic abberation, meaning different colors focus at different points, what happens when you go past what is visible (i.e infrared is more “red” than visible red; ultraviolet is more “violet” than what we can see).
The only lenses I know of that focus without these issues are mirror lenses.
@ZytaZiouZ
Thank you for taking the time to patiently explain that to me and to include a reference, too. You were very gracious to take that approach. Too many times I’ve read significantly less gracious approaches, here.
So, that was a good reminder that different wavelengths behave differently as they pass through the optics. (That is one reason why I appreciate DxO software, because they actually take that into account when the calibrate their lens modules to correct for “lens softness”. That is, they have different math for the red channel, the blue channel, and the green channel.)
Anyway, here is my question. Since we can model the different optical behavior of infrared v. visible, couldn’t the engineers use a infrared-to-visible conversion factor when calculating focus? Does my question make sense?
Every lens is different, so while it’s possible to make an educated guess, without having the camera look up a conversion table for every lens, it would never be accurate.
Also there is a large range of infrared (700-1000ish or 700-1100 depending on how sensitive the sensor is). That is about as much variance as the visible light spectrum, and 720 will surely focus differently than 950, and without some sort of infrared Bayer pattern to differentiate between them, there is no knowing which the camera is seeing.
In my humble opinion, on sensor pdaf and cdaf are the way to go, but then again I’m a NEX shooter lol.
That kind of mirror is called hot mirror. I don’t know what’s the transmission rate, but probably close to 99%. One of the possible problem is variance of focus offset(IR vs visible) from lens to lens. But this can be solved by the use of lens profile. The bigger problem is some lenses may block IR altogether.
Forget the mirror – you are riding on a dead horse. Develop better PDAF on sensor and improve CDAF. That’s the way to go.
it looks like that the sony-made on-matrix PDAF is eating about 0.5 stop of light.
at lease it could be seen in the DXO mark measurements for A7 vs A7r, A7 vs Nikon D610, A7 vs Nikon D800, nex 5t vs nex 5n
so for the sony case it is does not matter will there be a SLT or on-matrix PDAF. you will loose 0.5 stop in both cases.
which one to choose?
well, for the compact cameras, like nex or A7 there is no place for the translucent mirror. so on-matrix PDAF is the only remaining choice.
but SLT PDAF works much better. so for the bigger cameras, like A57 or future full-frame A-mount cameras SLT PDAF is preferable choice.
But sony, please not this nonesence like you did for A99 any more. Why did you put both on-matrix PDAF and SLT pdaf to the same camera. Thus losing 0.5 stop at translucent mirro (which is ok for this sort of camera) and yet loosing additional 0.5 stop at the on-matrix pdaf. That is probable explanation of the low sensitivity of the A99 compare to the competition.
I need 1/3 stop = 20-30% light when shooting ballet at F2.0, 1/250 sec and ISO 6400.
http://www.jackdevant.com/
SONY, JUST GIVE US FLIP UP MIRROR MODE IN YOUR NEXT SLT ALPHA, this mode will be 99% ill use. With the same a7 pdaf sensor it should be enough af speed for me. I don’t need extra slt mirror for sports shoot, I don’t shoot sports, and I have only one ssm tele.