Thoughts on an oversized sensor

Mal_In_Oz wrote:
amalric wrote:

BTW is you are really interested in what would be available with the present image circle, or a slightly larger one, the specialist is Riley. He did some awesome calculations. He is sometimes here, having a GH2, but mostly at 1022.

So if you do a Search there, you can get all his work. He was already discussing the issue some 2 yrs. ago.
Thanks Amalric, I did a search and stumbled on this thread:

http://forums.dpreview.com/forums/post/40758855

The key points are similar to mine, with the addition that a sensor of 18.8x18.8 would still provide all aspect ratios from 16:9 to 9:16. Personally I like the idea of even larger to provide for partial rotation to correct a wonky horizon without adjusting the tripod, or losing resolution in post. That would require the full image circle to be covered, or 21.63x21.63.

An important point that wasn't discussed was why we would want to use a smaller image circle with a larger sensor. The reason is so we can retain the tiny lenses. I like the fact that I can put a few lenses in a small kit bag and hardly notice that I am carrying them. So to keep those lenses and maximise the images we can take with them, seems like a great idea.

I don't expect Panasonic to come back from their current position with the GH3. That is, no oversized sensor. But I think it would be a killer move for Olympus if they can shoehorn a big sensor into a revised IBIS for the new Pro OMD camera. There is something cool about taking up someone else's discarded idea and making it work, that really rings with me.
We are on the same page. The Pro OM-D could attempt something similar, although IBIS might be the main problem.

My idea is also that nowadays if you change the sensor's format you don't need to provide a full blown system of lenses, just a few make do, if you keep backwards compatibility with the older format. Most people buy only an additional lens or two anyway.

Marketing might push in a similar direction since people seem hypnotized by the Megapixel count. I am not, but I would enjoy using legacy lenses closer to their original focal range.

Am.
 
I did a lot of shots recently at a dance performance and was sitting in the front row. I was very conscious of lifting my arm to do a portrait shot because it would have bothered people behind me and might even have been seen in the video.
There is a very simple solution to this problem and I always wonder why photographers put themselves in such an uncomfortable position while shooting portraits.

Just lower your right hand instead of lifting it, while holding the lens with your left hand: it works perfectly (at least for me).

Pollo
 
Pollo wrote:
I did a lot of shots recently at a dance performance and was sitting in the front row. I was very conscious of lifting my arm to do a portrait shot because it would have bothered people behind me and might even have been seen in the video.
There is a very simple solution to this problem and I always wonder why photographers put themselves in such an uncomfortable position while shooting portraits.

Just lower your right hand instead of lifting it, while holding the lens with your left hand: it works perfectly (at least for me).

Pollo
Thanks Pollo. I find that unnatural but I think that is what I did on a number of shots. It would be nicer though to be able to shoot in the same position and either select a different aspect ratio with a dial or just rely on the fact that you can recover a cropped head later in post.

The other benefit would be that you could select the aspect ration remotely via Android or IOS phone or tablet. Now that would be tricky without an oversized sensor.
 
Mal_In_Oz wrote:

I would like to hear what others think, so if anyone has an opinion on this then feel free to comment.
The current mFT sensor dimensions, 17.3mm x 13.0mm, has a diagonal of 21.6mm. So, a 21.6mm x 21.6mm sensor would cover the entire image circle, allowing the photographer to crop after the fact to any aspect ratio they like and still take advantage of all they can from the image circle.

In addition, one would not need to rotate the camera between landscape and portrait orientations, which is especially useful if using bounced flash.
 
baxters wrote:
tedolf wrote:

fabrication costs rise non-linearly with the diameter of the wafer.
I been out of the industry for decades, but believe t's still cheaper to make a pancake sized wafer with lots of big sensors than a tiny wafer with one sensor. I used to run the numbers all the time in my first job.

Sensor yield drops exponentially with chip size so let's say you have to run at least 5 tiny wafers to get one sensor, while you might get 20 sensors off the big wafer. Meanwhile you still have the same number of processing steps so the big and little wafers cost about the same to make, Starting costs for the wafer will be a factor, but will it make up for the rest? Don't think so. Otherwise, sensor plants would use small wafers.
sure, but even at that you are still fighting the defect rate
up until some years ago the defect rate was 40 defects per 200mm wafer, and that determines on average the number of good sensors you get out of a wafer

Im hypothesising cheap FF has arrived due to improvements in the defect rate, but that of course is a cost saving handed on to all other formats too

theres a java based calculator to play with here:
http://tams-www.informatik.uni-hamburg.de/applets/yield/index.html



Nostalgia. Here's a tiny wafer from 1980 meeting up with the EPL1.

e4f2f360cc87438aa80d9eda185181f8.jpg



--
Riley

any similarity to persons living or dead is coincidental and unintended
support 1022 Sunday Scapes'
 
Optical1 wrote:
the viewfinder would be am interesting issue. Would you propose a circular VF as well?
with an EVF you would be limited to the component available
Im not sure where Panasonic would sit with that, but Olympus will likely only go witl OM module parts for the near future at least. I guess that means Epson, so that means cropped views for say 9x16. In the case of 9x16 (upturned 16x9 that could be pretty limiting, but it is a size seeing some favour with marketing documents like this for the ASO (Adelaide Symphony Orchestra)


GH2 1/100th sec ISO2000, Konica 40/1.8 f/2
GH2 1/100th sec ISO2000, Konica 40/1.8 f/2

--
Riley
any similarity to persons living or dead is coincidental and unintended
support 1022 Sunday Scapes'
 
Last edited:
Great Bustard wrote:
Mal_In_Oz wrote:

I would like to hear what others think, so if anyone has an opinion on this then feel free to comment.
The current mFT sensor dimensions, 17.3mm x 13.0mm, has a diagonal of 21.6mm. So, a 21.6mm x 21.6mm sensor would cover the entire image circle, allowing the photographer to crop after the fact to any aspect ratio they like and still take advantage of all they can from the image circle.

In addition, one would not need to rotate the camera between landscape and portrait orientations, which is especially useful if using bounced flash.
Thanks for your comment Great Bustard. All good points.

Have you had any experience with range-finders? Do you think it would be helpful to see around the image you are about to take? I imagine there would be a lot of range-finder effect if you could view the entire sensor image (21.6x21.6) and then shoot a image from within the image circle at 17.3x13 or 15.3x15.3.
 
Mal_In_Oz wrote:
amalric wrote:

I have proposed about the same at 43rumors. But the main interest for me is making landscapes wider - by 10-20%.
This is a biggie for me too. I have taken a lot of images at dance concerts, and many times i find I need a wider aspect like 2:1 or 3:2 to cover the stage without including heads and overhead lights.

Interestingly, when I shoot portraits I only use 3:4 or 1:1. Being able to get the most from our existing m43 lenses is what the oversized sensor is all about.
I think ideal formats vary between individuals or groups of 2 and up, but we are in an accepted environment of 4x3, 3x2, 16x9, 1x1 as they appear on GH2 multi format menu. We have 16x9 b/se it is the format for HD Video and consequently flat panel TVs or monitors as a display medium
tI wonder however why Panny abandoned the formula. Multiaspect has a lot of draw for me.
This is especially intriguing given that the GH3 is so much larger than the GH2. It's not like they didn't have the room for the bigger sensor.
I would think the GH2 project sensor hasnt been made cheaper than they could acquire from Sony which has better DR and noise performance anyway; sorta jumping at that as its just speculation as it stands. A Gh2 sensor could be made to fit this is certain, but there are a chain of cost impacts as a consequence even without including the sensor. OLED display format size, EVF format size, size of shutter etc. Non standard mirror box (for want of a better name) so they cant use mass produced parts for other cameras like G5 ...


The camera being the only use for these special parts makes it more expensive to produce due to less volume, but I must say having used multi format I quite like it.
However I think that people still not get that the strength of m4/3 is comparable resolution to FF, thanks to v. sharp lenses, light AA filters, evenness across the frame for wides, light teles.

If all of this can be preserved while gaining a marginal size/aspect advantage I am all for it.

Other solution might be a bigger jump in sensor size, with a few new lenses capable of using it, but full retro compatibility. But that's the expensive alternative :)
I suspect it would work against Olympus or Panasonic to provide another format. Sometimes more options make it harder for everyone to choose...
 
Last edited:
Mal_In_Oz wrote:
amalric wrote:

BTW is you are really interested in what would be available with the present image circle, or a slightly larger one, the specialist is Riley. He did some awesome calculations. He is sometimes here, having a GH2, but mostly at 1022.

So if you do a Search there, you can get all his work. He was already discussing the issue some 2 yrs. ago.
Thanks Amalric, I did a search and stumbled on this thread:

http://forums.dpreview.com/forums/post/40758855


The key points are similar to mine, with the addition that a sensor of 18.8x18.8 would still provide all aspect ratios from 16:9 to 9:16. Personally I like the idea of even larger to provide for partial rotation to correct a wonky horizon without adjusting the tripod, or losing resolution in post. That would require the full image circle to be covered, or 21.63x21.63.
according to the original patent specification 43rds would have been allowed to be within a range of sizes up to a 25mm diagonal. Such a sensor would measure 20x15mm for an area of 300 sqmm, not to far below that of Canons APSC variously 327 sq mm or so.

theres a sideways impact of that where in 4x3 it would pick up most of that 0.52 stops deficiency it has to Canons sensor size. Or looked at another way, rather than an e/v of 1.94, it would be 1.53. So better noise performance by approx half stop, and a half stop impact on DoF


Im pretty sure that 43rds glass could cope with that increase, indeed some of it can operate near FF sizes, but mFT Im unsure how far mFT can be pushed, I would think less than 43rds



An important point that wasn't discussed was why we would want to use a smaller image circle with a larger sensor. The reason is so we can retain the tiny lenses. I like the fact that I can put a few lenses in a small kit bag and hardly notice that I am carrying them. So to keep those lenses and maximise the images we can take with them, seems like a great idea.
in the sizes of sensor size increase we can effect the impacts on lens size are likely immaterial
I don't expect Panasonic to come back from their current position with the GH3. That is, no oversized sensor. But I think it would be a killer move for Olympus if they can shoehorn a big sensor into a revised IBIS for the new Pro OMD camera. There is something cool about taking up someone else's discarded idea and making it work, that really rings with me.
This has been speculated for some time, and I think Olympus labs in the US gave it a name, Thom Hogan referred to it as Super four thirds
 
Mal_In_Oz wrote:
Great Bustard wrote:
Mal_In_Oz wrote:

I would like to hear what others think, so if anyone has an opinion on this then feel free to comment.
The current mFT sensor dimensions, 17.3mm x 13.0mm, has a diagonal of 21.6mm. So, a 21.6mm x 21.6mm sensor would cover the entire image circle, allowing the photographer to crop after the fact to any aspect ratio they like and still take advantage of all they can from the image circle.

In addition, one would not need to rotate the camera between landscape and portrait orientations, which is especially useful if using bounced flash.
Thanks for your comment Great Bustard. All good points.

Have you had any experience with range-finders? Do you think it would be helpful to see around the image you are about to take? I imagine there would be a lot of range-finder effect if you could view the entire sensor image (21.6x21.6) and then shoot a image from within the image circle at 17.3x13 or 15.3x15.3.
yes this is a famous quip by Leica RF users, well I use them as theyre the only organised company still making RF cameras. In order for that to work out for an EVF camera you would have to forgive some available FL. That isnt going to pan out so nicely with UWA lovers, but I note that at 21.6 sq most of the additional is going to be at the top of the frame where the lens hoods impact most

this is Zuiko 7-14 on a FF 5D, the curly bits are from the fixed lens hood

43lens7-14-7mm_internet_zpsfc19d2dc.jpg


at 14mm it looks like this

43lenses7-14_internet_zpse327a7fb.jpg
















--
Riley

any similarity to persons living or dead is coincidental and unintended
support 1022 Sunday Scapes'
 
baxters wrote:
tedolf wrote:

fabrication costs rise non-linearly with the diameter of the wafer.
I been out of the industry for decades, but believe t's still cheaper to make a pancake sized wafer with lots of big sensors than a tiny wafer with one sensor. I used to run the numbers all the time in my first job.

Sensor yield drops exponentially with chip size so let's say you have to run at least 5 tiny wafers to get one sensor, while you might get 20 sensors off the big wafer. Meanwhile you still have the same number of processing steps so the big and little wafers cost about the same to make,


This is where we disagree. Large waffers (they are very large now) require large epitaxal reactors and they are expensive. If you use a small diameter silicon bar, you can probably get an old, small diameter reactor for free! Once the photolithography is done, all the sunk costs are covered.


Starting costs for the wafer will be a factor, but will it make up for the rest? Don't think so. Otherwise, sensor plants would use small wafers.

Nostalgia. Here's a tiny wafer from 1980 meeting up with the EPL1.

e4f2f360cc87438aa80d9eda185181f8.jpg

Tedolph
 
Mal_In_Oz wrote:
Great Bustard wrote:
Mal_In_Oz wrote:

I would like to hear what others think, so if anyone has an opinion on this then feel free to comment.
The current mFT sensor dimensions, 17.3mm x 13.0mm, has a diagonal of 21.6mm. So, a 21.6mm x 21.6mm sensor would cover the entire image circle, allowing the photographer to crop after the fact to any aspect ratio they like and still take advantage of all they can from the image circle.

In addition, one would not need to rotate the camera between landscape and portrait orientations, which is especially useful if using bounced flash.
Thanks for your comment Great Bustard. All good points.
Glad you approve!
Have you had any experience with range-finders?
I'm afraid I don't.
Do you think it would be helpful to see around the image you are about to take?
Personally, I don't think so, but perhaps that's 'cause I've never used one.
I imagine there would be a lot of range-finder effect if you could view the entire sensor image (21.6x21.6) and then shoot a image from within the image circle at 17.3x13 or 15.3x15.3.
Indeed. However, it's also a simple matter to have the LCD / EVF crop to a preset aspect ratio, if the photographer so desires, whilst at the same time capturing the whole of the image circle, allowing for "crop adjustment" after the fact.

The thing is, such a sensor would have a diagonal only slightly larger than an APS-C sensor, which means it would cost about the same to produce, so I cannot think of any economic reason not to do it.
 
Last edited:
Rriley wrote:
Mal_In_Oz wrote:
Great Bustard wrote:
Mal_In_Oz wrote:

I would like to hear what others think, so if anyone has an opinion on this then feel free to comment.
The current mFT sensor dimensions, 17.3mm x 13.0mm, has a diagonal of 21.6mm. So, a 21.6mm x 21.6mm sensor would cover the entire image circle, allowing the photographer to crop after the fact to any aspect ratio they like and still take advantage of all they can from the image circle.

In addition, one would not need to rotate the camera between landscape and portrait orientations, which is especially useful if using bounced flash.
Thanks for your comment Great Bustard. All good points.

Have you had any experience with range-finders? Do you think it would be helpful to see around the image you are about to take? I imagine there would be a lot of range-finder effect if you could view the entire sensor image (21.6x21.6) and then shoot a image from within the image circle at 17.3x13 or 15.3x15.3.
yes this is a famous quip by Leica RF users, well I use them as theyre the only organised company still making RF cameras. In order for that to work out for an EVF camera you would have to forgive some available FL. That isnt going to pan out so nicely with UWA lovers, but I note that at 21.6 sq most of the additional is going to be at the top of the frame where the lens hoods impact most
Thanks for joining in the conversation Riley. I accept your example of a hood that will limit how much extra view you get, but in many circumstances we are still going to get extra viewing area outside the captured image. While shooting we can use this extra view to 'see around corners' so to speak, but we could also choose to use the extra view in post if we save the whole sensor image in RAW. We take for granted the ability to crop smaller when needed, sacrificing some image quality along the way. But wouldn't it be nice to crop larger also with possibly improved image quality?

I did this with a couple of test images taken with the 12-60 on a NEX5N. There was a lot of vignetting in the corners that i had to avoid due to the width of the Sony sensor, but it still provided a larger image in both area and pixels than what I could achieve on the EP3 at the time, especially in 1:1.
 
Optical1 wrote:
Mal_In_Oz wrote:

3. It would be possible to view a larger image than the captured image. This would provide a range-finder style of view that could help in framing. The view would be more circular (or perhaps the sensor could be even larger than the image circle), but it would be possible to show the full image with the captured image shown with markings in the viewfinder. A quick turn of the dial and the marked image could be rotated or aspect ratio changed to make a better composition.
the viewfinder would be am interesting issue. Would you propose a circular VF as well?
I think worst case is that the sensor is round and the maximum view possible would be round also, like looking through a telescope.

Ideally, the sensor would be square and so could provide a regular square view with a digital marking outlining where the captured image will be taken from. The view could be switched between rangefinder view (full image from the sensor), or simply show what will be captured as we already have now.
 
Rriley wrote:
Optical1 wrote:
the viewfinder would be am interesting issue. Would you propose a circular VF as well?
with an EVF you would be limited to the component available
Im not sure where Panasonic would sit with that, but Olympus will likely only go witl OM module parts for the near future at least. I guess that means Epson, so that means cropped views for say 9x16. In the case of 9x16 (upturned 16x9 that could be pretty limiting, but it is a size seeing some favour with marketing documents like this for the ASO (Adelaide Symphony Orchestra)
In the short term you are probably right, the hardware is just not there to support some of these ideas. I can only hope that manufacturers have the capacity to develop what they need to allow these ideas to become reality.

It probably goes without saying that an oversized sensor is something Sony could do more easily than anyone else. And probably has the most to gain from it. If they did this trick with the APSC sensor and their NEX range, they would certainly get some gains in captured image size.
 
Last edited:
One of the problems with the articulated display is that it doesn't tilt in all directions, so it's not as useful when holding the camera sideways to take a portrait. Not a problem with an oversized sensor.

Dan
 
The idea of an oversize sensor, either circular or dodecagonal, where the vertical 3:4/9:16 crop has the same number of pixels as the horizontal 4:3/16:9 crop and where the square 1:1 crop is also at maximum for the 4/3 image circle, seems a great idea to me. The rear screen and viewfinder may need to be square to make best use of the new crops but this should not hinder adoption of such a sensor as current rectangular viewing screens show cropped or rotated images anyway.
 
OM User wrote:

The idea of an oversize sensor, either circular or dodecagonal, where the vertical 3:4/9:16 crop has the same number of pixels as the horizontal 4:3/16:9 crop and where the square 1:1 crop is also at maximum for the 4/3 image circle, seems a great idea to me. The rear screen and viewfinder may need to be square to make best use of the new crops but this should not hinder adoption of such a sensor as current rectangular viewing screens show cropped or rotated images anyway.
There are rumors about an extraordinary viewfinder on a yet-to-be-released OMD. Perhaps they are implementing an oversized sensor with an oversized viewfinder.

A taller body would be needed to show the square screen (I am a fan of the smaller bodies though so perhaps they could do away with it altogether), but taller and narrower screen might provide better ergonomics.

Another option would be just one small high definition screen that takes a loupe with a focus lens to provide the biggest built in viewfinder in history. Just fold back the loupe to reveal a small but detailed screen for the times it is needed.
 
Its nigh on 20mm wide and has 5000 odd pixels in 16:9 ratio! This is what the GH3 should have too, sadly it doesnt, ahhhhh.
Mal_In_Oz wrote:
OM User wrote:

The idea of an oversize sensor, either circular or dodecagonal, where the vertical 3:4/9:16 crop has the same number of pixels as the horizontal 4:3/16:9 crop and where the square 1:1 crop is also at maximum for the 4/3 image circle, seems a great idea to me. The rear screen and viewfinder may need to be square to make best use of the new crops but this should not hinder adoption of such a sensor as current rectangular viewing screens show cropped or rotated images anyway.
There are rumors about an extraordinary viewfinder on a yet-to-be-released OMD. Perhaps they are implementing an oversized sensor with an oversized viewfinder.

A taller body would be needed to show the square screen (I am a fan of the smaller bodies though so perhaps they could do away with it altogether), but taller and narrower screen might provide better ergonomics.

Another option would be just one small high definition screen that takes a loupe with a focus lens to provide the biggest built in viewfinder in history. Just fold back the loupe to reveal a small but detailed screen for the times it is needed.
 
Adventsam wrote:

Its nigh on 20mm wide and has 5000 odd pixels in 16:9 ratio! This is what the GH3 should have too, sadly it doesnt, ahhhhh.
Mal_In_Oz wrote:
OM User wrote:

The idea of an oversize sensor, either circular or dodecagonal, where the vertical 3:4/9:16 crop has the same number of pixels as the horizontal 4:3/16:9 crop and where the square 1:1 crop is also at maximum for the 4/3 image circle, seems a great idea to me. The rear screen and viewfinder may need to be square to make best use of the new crops but this should not hinder adoption of such a sensor as current rectangular viewing screens show cropped or rotated images anyway.
There are rumors about an extraordinary viewfinder on a yet-to-be-released OMD. Perhaps they are implementing an oversized sensor with an oversized viewfinder.

A taller body would be needed to show the square screen (I am a fan of the smaller bodies though so perhaps they could do away with it altogether), but taller and narrower screen might provide better ergonomics.

Another option would be just one small high definition screen that takes a loupe with a focus lens to provide the biggest built in viewfinder in history. Just fold back the loupe to reveal a small but detailed screen for the times it is needed.
I feel your frustration. The Panasonic got a lot of supporters for using an oversized sensor in the GH2, so it was a bit strange that they chose not to continue this feature, or even to make it larger so that it covered the full image circle and portrait aspect ratios. Is it too much to hope for that Olympus will take on the idea in a super OMD?
 

Keyboard shortcuts

Back
Top