custom cmos better than ff? no way...

Started Jan 6, 2012 | Discussions
desinteresadamente
Regular MemberPosts: 264
Like?
custom cmos better than ff? no way...
Jan 6, 2012

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...
Thank you.

lnbolch
Senior MemberPosts: 1,862Gear list
Like?
Re: custom cmos better than ff? no way...
In reply to desinteresadamente, Jan 6, 2012

desinteresadamente wrote:

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...

No, a Sony sensor will not. Probably a Fuji sensor will, if that in my X100 is any indication.

Note that the Sony camera division and solid-state fabrication plant are run as member companies under the Sony brand. To the fab, Sony camera is just another customer. If Fuji instead used the huge IBM fab in the US, it would not make it an IBM sensor. Go to any fab with the specifications for a chip and a truckload of money, and you will get exactly what your engineers designed.

-- hide signature --
 lnbolch's gear list:lnbolch's gear list
Nikon Coolpix 8400 Fujifilm FinePix Real 3D W1 Fujifilm FinePix X100 Nikon D700 Fujifilm X-Pro1 +12 more
Reply   Reply with quote   Complain
whtchocla7e
Contributing MemberPosts: 927Gear list
Like?
Re: custom cmos better than ff? no way...
In reply to desinteresadamente, Jan 6, 2012

Just because it's 16MP, doesn't mean it's a Sony sensor.

And they never stated which FF medium it surpasses. It could mean average quality 35mm film, or the first Kodak digital 35mm sensors from the early 2000's..

There's hardly any information to just assume "no way" at this point.

Reply   Reply with quote   Complain
a l b e r t
Senior MemberPosts: 1,265Gear list
Like?
New CFA gives apparent higher resolution
In reply to whtchocla7e, Jan 7, 2012

I think Fuji customized the Sony or whatever sensor with a new color filter array, thus eliminating the AA filter. So what Fuji is saying is that even at a lower sensor resolution, it can resolve details just as great as a FF sensor. My guess is that a native 16MP sensor having the same amount of details as a 24MP FF sensor.

-- hide signature --

a l b e r t

 a l b e r t's gear list:a l b e r t's gear list
Fujifilm FinePix F31fd Fujifilm FinePix X100 Fujifilm X10 Nikon D3S Fujifilm X-E1 +7 more
Reply   Reply with quote   Complain
kuri
Regular MemberPosts: 152
Like?
Re: custom cmos better than ff? no way...
In reply to lnbolch, Jan 7, 2012

lnbolch wrote:

desinteresadamente wrote:

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...

No, a Sony sensor will not. Probably a Fuji sensor will, if that in my X100 is any indication.

I thought the Fuji X100 sensor is a Sony sensor, same as what's used in the Nikon D90...

Reply   Reply with quote   Complain
Xenofoto
Forum MemberPosts: 88
Like?
Re: custom cmos better than ff? no way...
In reply to desinteresadamente, Jan 7, 2012

desinteresadamente wrote:

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...
Thank you.

I stumbled on this, it helped me somewhat, hope it helps you as well.

After reading the article, imagine a C size sensor at 40 mega pixels,
and this is not Full Frame, with better dynamic range and low light abilities.

At its present size it would defeat any full frame sensor on the DSLR market today.
With no AA filter and no microlenses as well.

This has yet to be proven and seen!

I would wait to confirm if the organic sensor will even be inside the New Fuji mirrorless
camera, no one knows.

http://peter.vdhamer.com/2011/11/19/fuji-organic-sensor/

Reply   Reply with quote   Complain
gava
Senior MemberPosts: 1,032
Like?
Re: custom cmos better than ff? no way...
In reply to kuri, Jan 7, 2012

kuri wrote:

lnbolch wrote:

desinteresadamente wrote:

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...

No, a Sony sensor will not. Probably a Fuji sensor will, if that in my X100 is any indication.

I thought the Fuji X100 sensor is a Sony sensor, same as what's used in the Nikon D90...

A lot of people think that. A lot of people (internet pundits included) have simply assumed it to be so, without any evidence as far as I have seen.

It is definitely not the SAME sensor; it has been customised at the micro-lens level at the very least. It may or may not be based on the older Sony sensor, it may or may not be manufactured in the Sony plant.

-- hide signature --

Fuji X100 Fanboy #1
X100 blog -> http://peri.org.uk/wp/?tag=blog

Reply   Reply with quote   Complain
desinteresadamente
Regular MemberPosts: 264
Like?
Re: custom cmos better than ff? no way...
In reply to Xenofoto, Jan 7, 2012

I think it makes no sense to claim better than ff and refer to a dinosaur ff film, even at ff 2003 benchmarchs. If the sensor is not organic, only a different cfa will not make it surpass ff output of, to say, 1ds mark 11. I think
REgards and thank you for the insights

Reply   Reply with quote   Complain
Jon Stern
Contributing MemberPosts: 923
Like?
Re: custom cmos better than ff? no way...
In reply to lnbolch, Jan 7, 2012

IBM's sensor fab is not a match for Sony. Trust me I know.

Reply   Reply with quote   Complain
lnbolch
Senior MemberPosts: 1,862Gear list
Like?
Re: custom cmos better than ff? no way...
In reply to kuri, Jan 7, 2012

kuri wrote:

lnbolch wrote:

desinteresadamente wrote:

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...

No, a Sony sensor will not. Probably a Fuji sensor will, if that in my X100 is any indication.

I thought the Fuji X100 sensor is a Sony sensor, same as what's used in the Nikon D90...

Does the D90 have variable photosite spacing to match a 35mm lens, allowing the light to enter at the optimum angle both in the centre and the corners? And every point in between? What on earth gave you the idea that this would be the D90 sensor. It would produce pure crap with long lenses. It would also produce intense vignetting with ultra-short lenses. It is designed to work with only the 35mm equivalent—23mm.

The combination of the superb lens with a matching sensor is why owners call it their best f/2.0 35mm equivalent prime that comes with an attached camera as a free bonus.

-- hide signature --
 lnbolch's gear list:lnbolch's gear list
Nikon Coolpix 8400 Fujifilm FinePix Real 3D W1 Fujifilm FinePix X100 Nikon D700 Fujifilm X-Pro1 +12 more
Reply   Reply with quote   Complain
kuri
Regular MemberPosts: 152
Like?
Re: custom cmos better than ff? no way...
In reply to lnbolch, Jan 7, 2012

Seems that this is a touchy subject?

I'm not here proclaiming it is 100% indeed a Sony sensor, but just I had thought it was commonly understood to be so. Even on Japanese sites, the common understanding is that the basic sensor is a Sony (same as D90 etc), but Fuji customized it for the X100.

lnbolch wrote:

kuri wrote:

lnbolch wrote:

desinteresadamente wrote:

How on earth if the sensor is not newly developed, will the sony sensor surpass full frame?
No way, if someone could explain me...

No, a Sony sensor will not. Probably a Fuji sensor will, if that in my X100 is any indication.

I thought the Fuji X100 sensor is a Sony sensor, same as what's used in the Nikon D90...

Does the D90 have variable photosite spacing to match a 35mm lens, allowing the light to enter at the optimum angle both in the centre and the corners? And every point in between? What on earth gave you the idea that this would be the D90 sensor. It would produce pure crap with long lenses. It would also produce intense vignetting with ultra-short lenses. It is designed to work with only the 35mm equivalent—23mm.

The combination of the superb lens with a matching sensor is why owners call it their best f/2.0 35mm equivalent prime that comes with an attached camera as a free bonus.

-- hide signature --
Reply   Reply with quote   Complain
Bernie Ess
Veteran MemberPosts: 6,899Gear list
Like?
Re: custom cmos better than ff? no way...
In reply to desinteresadamente, Jan 7, 2012

desinteresadamente wrote:

If the sensor is not organic, only a different cfa will not make it surpass ff output of, to say, 1ds mark 11.

It think it is pretty sure it is not the organic layer sensor described in the dpreview artikel a while ago.

1ds MkII has 16MP, but it has a vibrating mirror and an AA filter. The Fuji hasn't either one, also the Fuji sensor (Sony base I guess) is roughly 7-8 years old, whereas the Sony is only 1-2 years old.

I imagine that it has a resolution not far from a Canon 21MP.

Bernie

-- hide signature --

'All the technique in the world doesn’t compensate for the inability to notice.' (Elliot Erwitt)

Reply   Reply with quote   Complain
Bernie Ess
Veteran MemberPosts: 6,899Gear list
Like?
Re: New CFA gives apparent higher resolution
In reply to a l b e r t, Jan 7, 2012

a l b e r t wrote:

My guess is that a native 16MP sensor having the same amount of details as a 24MP FF sensor.

That would be 1,5x as many pixels? As it still has a Bayer mask, I cannot imagine how that could happen.

Bernie
--

'All the technique in the world doesn’t compensate for the inability to notice.' (Elliot Erwitt)

Reply   Reply with quote   Complain
HAE
HAE
Senior MemberPosts: 1,028
Like?
The only way I can think of
In reply to desinteresadamente, Jan 7, 2012

to make sense of Fuji's previous claim of equal/superior noise and resolution to 36*24 sensors (say the existing offerings of 21 MP and 24 MP FF sensors) in light of the new information about an APS-C 16MP sensor with 6 color sampling is that they might have managed to create a 2 layer sensor (i.e 2 colors are sampled per photosite). That sensor will be quasi-Bayer / quasi-Foveon. It probably has equal or better resolution that the current FF sensors since 1 color only will be interpolated by photosite rather than 2 for the standard Bayer configuration. The color array will also be designed to pass more light, which should help with noise performance (assuming they got other contributors to noise at the current level of the industry). So instead of the typical Bayer filters that pass predominantly the red or blue wavelengths, you would have a filter that passes red or blue and part or all of the green wavelengths and you would have a another filter that passes green and part of red and blue in place of the Bayer filter that passes predominantly Green.

I do not think that adding more colors to the Bayer arrangement will improve resolution and noise performance the way Fuji claims. It is also a simple idea and if it had an advantage somebody would have done it by now. I also remember that Sony long ago used a fourth filter color in some of their compacts for some time and later abandoned that idea.

In all cases, it is also a few days till we know for sure.

Reply   Reply with quote   Complain
maple
Senior MemberPosts: 2,567
Like?
Re: custom cmos better than ff? no way...
In reply to desinteresadamente, Jan 7, 2012

Here's what intrigues me from the French magazine Responsephoto:

“Fuji took some ideas from the classica film technology to create the new sensor which will beat the image quality of all current full frame sensors.”

To me, this is the most exciting and crucial bit about the whole Fuji X Pro1 leak. It effectively reaffirms Fuji’s earlier claim that their new sensor will exceed FF both in resolution and noise property.

But such prospects are really hard to imagine, if the sensor is indeed based on a conventional sensor, as how most tend to interpret the leaked info, with only a smart tweak of the color filter, and the removal of the usual AA filter. We know a regular APS-C sensor resolves about 70% of its nominal resolution. So the removal of AA filter can’t possibly give a sensor more than 40% (1/0.7 – 1 = 0.43) gain in resolution. As such a 16mp APS-C sensor without AA filter shall not exceed the resolution of a 23mp regular CMOS sensor of the same size, let alone a FF. What’s more, the new filter is rumored to have 6 colors instead of 4 with the regular Bayer pattern. I suspect that 2 of them are transparent, thus the name X-Trans. That means the sensor has 1/3 less primary color information. I wonder whether that may or may not result in lower resolution in the final reconstruction of full color pixels

High ISO performance might be a different story. Sony’s 16mp is very close to the current generation of (already fairly old) FF sensor in terms of noise. If 1/3 of the filter sites are indeed transparent, it could translate into meaningful improvement in low light performance.

Just speculation for fun to make the waiting easier to endure.
--
Maple

Reply   Reply with quote   Complain
MiTaka
Regular MemberPosts: 353
Like?
Re: custom cmos better than ff? no way...
In reply to maple, Jan 7, 2012

X-trans will be RGB. It is just the way different colors are arranged. RGB pixels are arranged in 6x6 random pixel sets. This randomness allows to skip the need of AA filter. There is R, G and B pixel in every vertical and horizontal pixel series. As a result resolution will be enhanced, and colour accuracy also (no false demosaic colors).
Also Fuji state that the sensor is benchmark for all aps-c sensors.

-- hide signature --

Dimitar Ivanov

Reply   Reply with quote   Complain
Anadrol
Regular MemberPosts: 346
Like?
Re: New CFA gives apparent higher resolution
In reply to Bernie Ess, Jan 7, 2012

The X-Trans sensor doesn't use Bayer, but random patterns of 6 photosites.

Reply   Reply with quote   Complain
Anadrol
Regular MemberPosts: 346
Like?
Re: The only way I can think of
In reply to HAE, Jan 7, 2012

Extremely interesting post, thank you !

Reply   Reply with quote   Complain
knickerhawk
Veteran MemberPosts: 3,115
Like?
Re: custom cmos better than ff? no way...
In reply to MiTaka, Jan 7, 2012

MiTaka wrote:

X-trans will be RGB. It is just the way different colors are arranged. RGB pixels are arranged in 6x6 random pixel sets. This randomness allows to skip the need of AA filter. There is R, G and B pixel in every vertical and horizontal pixel series. As a result resolution will be enhanced, and colour accuracy also (no false demosaic colors).

Hmmm...First, in response to Maple, I don't understand how two transparent + R + G + B = six colors. At most, it's four (if you think of transparent as a color). In response to your speculation, how do you reconcile your statement that the array is a simple RGB pattern as being the same as "six colors?" Either it's very poorly worded by the article (certainly possible) or your guess is incorrect. Moreover, the idea of only one green pixel per column seems to be counter to all the development that led to the classic Bayer pattern that doubles the number of green pixels relative to red and blue because of the human eye's greater sensitivity to green. I'm also not following you when you refer to 6x6 random pixel sets. Are you asserting that these 6x6 grids are physical and randomness is defined only within each block or are you saying that the demosaicing process will use the 6x6 grid surrounding a pixel to define it's color?

Also Fuji state that the sensor is benchmark for all aps-c sensors.

-- hide signature --
Reply   Reply with quote   Complain
MiTaka
Regular MemberPosts: 353
Like?
Re: custom cmos better than ff? no way...
In reply to knickerhawk, Jan 7, 2012

knickerhawk wrote:

MiTaka wrote:

X-trans will be RGB. It is just the way different colors are arranged. RGB pixels are arranged in 6x6 random pixel sets. This randomness allows to skip the need of AA filter. There is R, G and B pixel in every vertical and horizontal pixel series. As a result resolution will be enhanced, and colour accuracy also (no false demosaic colors).

Hmmm...First, in response to Maple, I don't understand how two transparent + R + G + B = six colors. At most, it's four (if you think of transparent as a color). In response to your speculation, how do you reconcile your statement that the array is a simple RGB pattern as being the same as "six colors?" Either it's very poorly worded by the article (certainly possible) or your guess is incorrect. Moreover, the idea of only one green pixel per column seems to be counter to all the development that led to the classic Bayer pattern that doubles the number of green pixels relative to red and blue because of the human eye's greater sensitivity to green. I'm also not following you when you refer to 6x6 random pixel sets. Are you asserting that these 6x6 grids are physical and randomness is defined only within each block or are you saying that the demosaicing process will use the 6x6 grid surrounding a pixel to define it's color?

Also Fuji state that the sensor is benchmark for all aps-c sensors.

-- hide signature --

You will have a 6*6 pixel array with 18 green 9 red and 9 blue. They will be arranged randomly. Current sensors are arranged in fixed 2*2 grid.
The picture is coming on monday
--
Dimitar Ivanov

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads