Reading mode: Light Dark

What does a pixel look like?

Go to last post
B

Bjorn Rorslett

Leading Member
Messages
912
Reaction score
0
Location
Oslo, NO
We all talk about them and always want them to be more plentiful. The pixels, of course. All this without anyone ever having seen a single pixel in reality. Well, while we're waiting for Nikon's next announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification with a D1H. The CCD is from a D1, and I had to remove the anti-aliasing and hot-mirror filters in front of the CCD in order to get a clear picture of the pixels. They form the well-know Bayer-pattern with 50% green, 25% red, and 25% blue units. There are 2 catchlights on each pixel (in case you wonder) from the fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most interesting feature is that each pixel (photosite) actually is made up of 4 sub-pixels, each with its own micro-lens on top of it. So, it is easy to see how D1X got a doubled pixel count compared to D1. More interesitingly, since the CCD already is laid out with 4 times the actual pixel number of D1, it would only take some additional electronics to have the camera to read out 4 times the data. This implies a 11 Mpix CCD would be feasible even with today's line of cameras. Is this the news we're going to encounter tomorrow (18th February)?
 
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
 
As always, very interesting stuff from you. I wonder if you are tipping your hand a little since you seem to have an "in" with Nikon. What's up?

Sam
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
 
Samuel A. Anderson said:
Sam
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
Click to expand...
Your guess is no better than mine. I have absolutely no clue as to what Nikon may announce. My contacts have been extremely tight-lipped about this, either because they themselves don't know anything, or do but are not permitted to disclose anything.
 
Very interesting Bjorn. I heard the rumor before that the D1H pixels are 4 photosites masked as one and the D1x two photsites are masked as one. Now I see it is true. This Bayer patters is quite strange. Does not the pixels have to be arranged as G-R-G-B. In your picture all the plue pixels are on one row and all the red pixels are on the other row. There was a rumor on Nikonians.com mentioning that Nikon will offer an upgrade to the D1x to change it to 10.6 MP. However that will make an untypical looking Bayer mask with two blue/ red pixels under each other. I guess Canon expected the D2 to be 10.6 MP and wanted to cut off this expected development and so added 1 MP to that number.
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
 
I said "This Bayer patters is quite
HAE said:
strange. Does not the pixels have to be arranged as G-R-G-B. In
your picture all the plue pixels are on one row and all the red
pixels are on the other row. "
Click to expand...
Sorry I see this is how they should be arranged.
HAE said:
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
Click to expand...
 
That's a cool image, and you might be right about what it means. However it might also be the case that there are simply 4 microlenses per photosite. This should allow the lenslet array to be closer to the actual imaging surface, which would (I think) decrease the angular sensitivity of the sensor. Just a thought.

Of course, I would hope your explanation is the correct one. That way it should be possible to get to a 24.4MP full frame sensor in the forseeable future.

Brian
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
--
J. Brian Caldwell
http://www.caldwellphotographic.com
 
I'm curious about your removing the AA/IR filter. Was it difficult, and did you do it to create a dedicated IR camera?

Brian
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
--
J. Brian Caldwell
http://www.caldwellphotographic.com
 
I posted on this to nikonians
http://www.nikonians.org/dcforum/DCForumID71/1060.html#1

but it may be of iterest here too: The reson for 2x2 CCD wells as compared to the camera resolution is probably to comply with the Nyquist sampling criterion for avoiding aliasing in the final picture. See the Nikonians post for more details, but the main point is that in order to avoid aliasing you must sample a signal (image) at twice the signal's bandwidth. This means that to get a 1000 x 1000 pixel image without aliasing, we must:
  • blur the image so it contains no features smaller than 1/1000 the size of the image sensor
  • then sample it with 2000 x 2000 resolution
  • and finally, low-pass filter it down to 1000 x 1000.
There is no additional information in the 3 million measurements that we discard and all you would get from a direct readout of the 11 million D1 sensors would be the equivalent of using Photoshop to increase the image size to 11 million pixels with a simple interpolation algorithm. This limit is imposed by the anti-aliasing filter.

I do not know for a fact that I am right in this interpretation, but it is consistent with Nyquist's theorem which is of fundamental importance for digital signal processing.
Bjorn Rorslett said:
since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
 
brian said:
Brian
Bjorn Rorslett said:
We all talk about them and always want them to be more plentiful.
The pixels, of course. All this without anyone ever having seen a
single pixel in reality. Well, while we're waiting for Nikon's next
announcement, here is the image (a small crop from the original),

http://www.naturfotograf.com/images/B030217207.jpg

(178K file)

This is taken with 19 mm f/2.8 Macro-Nikkor at 25X magnification
with a D1H. The CCD is from a D1, and I had to remove the
anti-aliasing and hot-mirror filters in front of the CCD in order
to get a clear picture of the pixels. They form the well-know
Bayer-pattern with 50% green, 25% red, and 25% blue units. There
are 2 catchlights on each pixel (in case you wonder) from the
fibre-optical lights employed for the shot.

Each pixel is 11.9 x 11.9 µm (1 µm = 1/1000 mm) and the most
interesting feature is that each pixel (photosite) actually is made
up of 4 sub-pixels, each with its own micro-lens on top of it. So,
it is easy to see how D1X got a doubled pixel count compared to D1.
More interesitingly, since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
--
J. Brian Caldwell
http://www.caldwellphotographic.com
Click to expand...
I had been given a wrecked CCD to play with, so in principle could do whatever I wanted. The AA/IR filter is held in position by a metal frame fixed with 4 set screws and is not very difficult to remove.

Initially I shot directly through the filter onto the CCD, the ordinary (or rather direct opposite) way we take our pictures. However, the AA filter makes imaging details of the pixels (photosites) extremly difficult, which after all goes to show the filter works as it should ("anti-aliasing"). For the curious, here is a similar shot at 25X of the same CCD, before the AA filter is removed. The greenish tinge is from the hot-mirror (IR) filter which is bound onto the AA filter. The crop represents the border of the CCD;



; (204 K file)

The darkened edge presumably is by design of the AA filter. I took several shots with different lighting of the CCD and all showed a similar view. The central part of the CCD was uniform in colour and contrast.
 
Jonathan1 said:
  • blur the image so it contains no features smaller than 1/1000 the
size of the image sensor
  • then sample it with 2000 x 2000 resolution
  • and finally, low-pass filter it down to 1000 x 1000.
There is no additional information in the 3 million measurements
that we discard and all you would get from a direct readout of the
11 million D1 sensors would be the equivalent of using Photoshop to
increase the image size to 11 million pixels with a simple
interpolation algorithm. This limit is imposed by the anti-aliasing
filter.

I do not know for a fact that I am right in this interpretation,
but it is consistent with Nyquist's theorem which is of fundamental
importance for digital signal processing.
Bjorn Rorslett said:
since the CCD already is laid out with 4 times
the actual pixel number of D1, it would only take some additional
electronics to have the camera to read out 4 times the data. This
implies a 11 Mpix CCD would be feasible even with today's line of
cameras. Is this the news we're going to encounter tomorrow (18th
February)?
Click to expand...
Click to expand...
You ignore the fact that the CCD is overlaid with an anti-aliasing (AA) filter which functions to remove the spatial frequencies above the Nyquist frequency. See my picture of the AA filter.

Furthermore, the D1X already have utilised 2 of the 4 photosites within each quadrat to give an effectively doubled horizontal resolution. The logical step would be (in my opinion, not necessarily Nikon's) to employ both photosites in the vertical direction also. doing so would produce about 11 MPix from the current size of their CCDs.

My point was to show how the pixels are laid out, and I was quite surprised by my observations. Feel free to have any interpretation of the findings as you like.
 
I suspect it is just a fabrication simplification. There are probably 11M individual photosites but not 11M electrically addresable signals. The interconnection layer most likely parallels 4 sites per pixel for the D1H and 2 for the D1X. Nikon (Sony) probably felt that fabrication improvements would allow acceptable sensitivity and yield for 11M pixels in the future while still using a majority of the masking of the present chips.

Ron
 
I often wondered if the Sony CCD in the D100 comes from the same wafer as the consumer CCD's. They just change the filters, group the pixels and get a larger more sensitive chip. It would also gain them a higher yield since a defective photosite may not kill a whole "pixel" and they wouldn't have to scrap the chip.
Sean
Ron Schroeder said:
I suspect it is just a fabrication simplification. There are
probably 11M individual photosites but not 11M electrically
addresable signals. The interconnection layer most likely
parallels 4 sites per pixel for the D1H and 2 for the D1X. Nikon
(Sony) probably felt that fabrication improvements would allow
acceptable sensitivity and yield for 11M pixels in the future while
still using a majority of the masking of the present chips.

Ron
Click to expand...
 
JoseMaza said:
I have read something similar here:
http://www.lonestardigital.com/techcorner.htm
Click to expand...
The funny thing about it is that when I posted this article and a thread about same in this forum a year ago, a very well known expert posted a reply to the thread AND a personal e-mail to me that said it was pretty much a ridiculous idea, that he had personally looked at Nikon's patent and knew all about it, and that (additionally) my theory was technologically unfeasible.

--
John Cowley
[email protected]
http://www.lonestardigital.com
 
Actually ...those are photosites or photodiodes rather than pixels.

Pixels are more to do with software side of things

Jose
 
sorry for another crazy bump, BUT does this mean that the 10mpx d80 and d200 could be using the exact same sensor as the 2.7 mpx d1h?

if so, I need to go find me a d1h to buy for around $300
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top