Various sensor sizes

[retrosi]

Just been reading some much older threads & noticed a comment 'is about half the size'.
So I did a couple of calculations.
I know that between the 1.6 crop Canons & 1.5 crop Nikons, the sensor sizes vary a small amount between model to model.
I'm not referring between Nikon & Canon but each makes models themselves.

Therefore aside from the the oddities of say the 350d having 22.2x14.8 & the 7d having 22.3x14.9 & the old 10d having 22.7x15.1
averaging those three sensors gives a total mm area sq of 335.02mm sq
Averaging the d3200 @ 23.2x15.4, the d3100 @ 23.1x15.4 & d5200 @ 23.5*15.6 gives an average of 359.87mm sq

An oddity I did find is that the Nikon full frame sensors, well aren't? They are either 35.9x24mm or 36*23.9mm, weird...
All the Canons full frame sensor are 36x24.

Anyhoo... asuming full frame for all is 36x24 then that area sq = 864mm sq
the now defunct APSH of the Canon 1D sq area = 518.94mm sq
And averaging the two APSC of Canon & Nikon = 347.445mm sq

So the APSH is 60% of FF
the APSC is 40% of FF
& the APSC is 67% of the APSH

Given these massive reductions in areas sq the pixel sizes are that much smaller also, we know this.
So when you take a smaller pixel which is in itself a window to let the light through, this now massively reduced in size window, the pixel, because with each new camera the sensor is not increased, does what to the image quality?

Are we being tricked by the old ways when the sensor count of 8 million(ish) was reached back in 2005-7 & noise became such an issue there was a lull in forwarding the pixel count of cameras .
During this lull they sorted things out with 'a' software in camera fix to reduce the noise.
Things got better with the actual sensors & we moved on but how much of the improvements are down to the in camera corrections.
Even with RAW the camera has to record this info which in a way is an interpretation of what it is receiving, thus as much as it is commonly accepted that a RAW file is a digital negative,
I ask, how much 'processing' is actually going on during the 'recording' of that data?

A scenario I use is that of a normal window, lets make that window 3 foot by 2 foot.
Now put some lead in it to make it pretty as in the older style & split it into say ten smaller sections.
Those lines of lead have now reduced the actual area of glass that can actually let light in.

Now I'm not clever enough to know this next part but here goes...
Does each smaller pane now see the whole image or just the part of the area as seen by the previously single window.
I'm going to assume just the small part because if your eyes are the sensor & you are a distance back as is the sensor from the front of the lens, with out going right up to the individual smaller section, you see the whole image but through smaller cut up sections.

So surely less light coming through the now smaller panes / pixels has to be / give a lower quality of light for the sensor to record / process.

My point to all this is as the makers are cramming in more pixels into the same sensors surely the quality of light squeezing through can't be as good.
Again, a single unobstructed window / sensor has more light than the same size with lots of smaller panes / pixels?

Have we reached a point where the makers are fooling us into these massive pixel counting sensors do give higher dynamic range, or is it some very clever in camera processing?
Even with RAW files? Canon-CR2, Nikon-NEF. Different files must have different processing going on?
As I mentioned these have to be recorded in some fashion otherwise there wouldn't be the difference in RAW file formats.

This isn't meant to be a thread over whose sensors are better, so please don't start one of those.
This is a general thought about when has a sensor too many pixels & have they already & have we been deluded already for some time now?

Is the medium format camera & its larger sensors the way to go before those manufactures go out of business?

I noticed the older Pentax 645D with 40 million pixels can be bought for under £2.5k it has a 44x33mm sensor
But their new 645Z has the same size sensor & now 53 million pixels.

There are of course other medium format cameras with larger sensors & higher pixel counts, but I'm thinking they have a long way to go before they hit trouble.

 

[Carl Hall]

Interesting post, I have never thought about it but it does make sense that more pixels means less light to each pixel. I'm interested to hear what people say.

Or, you could just use medium format film... roughly 3800 square mm for a 6x7 and not a single one of those pesky pixels in sight

Seriously though, I try to explain this to people all the time when they come out with "my camera phone has 20 megapixels, that means it's better". Yeah, more pixels on a tiny sensor. Nice image quality. Can't argue with them though because they're usually complete walnuts that don't listen.

 

[woof woof]

It's not only the size of the sensor and the number and size of pixies that matter as the electronics layout, design and processing also have an effect and explain why newer smaller format cameras with higher mp counts perform better than they would have some years ago.

 

[ancient_mariner]

Sensor development has moved rapidly, so that a 20mp sensor in APSC will give a decent, low-ish noise image and dynamic range has also increased.

In order to cope with smaller pixels that receive less light it is necessary to increase the sensitivity of the detecting medium, and to do that may increase noise levels. Development will have been done, not just to increase sensitivity, but also to reduce noise, whether through physical improvements to materials or through software that can lower baseline noise. This is to be expected, and it's not a case of anyone fooling anyone, so much as a natural outcome of intensive R&D in a race to increase or maintain market share. No-one is fooled by the old 'more megapix=better image' now.

A good example of how this has been turned on it's head is the Sony A7R - nominally full frame, but just 12mp and with super-high sensitivity/low noise.

 

[BillN_33]

Just been reading some much older threads & noticed a comment 'is about half the size'.
So I did a couple of calculations.
I know that between the 1.6 crop Canons & 1.5 crop Nikons, the sensor sizes vary a small amount between model to model.
I'm not referring between Nikon & Canon but each makes models themselves.

Therefore aside from the the oddities of say the 350d having 22.2x14.8 & the 7d having 22.3x14.9 & the old 10d having 22.7x15.1
averaging those three sensors gives a total mm area sq of 335.02mm sq
Averaging the d3200 @ 23.2x15.4, the d3100 @ 23.1x15.4 & d5200 @ 23.5*15.6 gives an average of 359.87mm sq

An oddity I did find is that the Nikon full frame sensors, well aren't? They are either 35.9x24mm or 36*23.9mm, weird...
All the Canons full frame sensor are 36x24.

Anyhoo... asuming full frame for all is 36x24 then that area sq = 864mm sq
the now defunct APSH of the Canon 1D sq area = 518.94mm sq
And averaging the two APSC of Canon & Nikon = 347.445mm sq

So the APSH is 60% of FF
the APSC is 40% of FF
& the APSC is 67% of the APSH

Given these massive reductions in areas sq the pixel sizes are that much smaller also, we know this.
So when you take a smaller pixel which is in itself a window to let the light through, this now massively reduced in size window, the pixel, because with each new camera the sensor is not increased, does what to the image quality?

Are we being tricked by the old ways when the sensor count of 8 million(ish) was reached back in 2005-7 & noise became such an issue there was a lull in forwarding the pixel count of cameras .
During this lull they sorted things out with 'a' software in camera fix to reduce the noise.
Things got better with the actual sensors & we moved on but how much of the improvements are down to the in camera corrections.
Even with RAW the camera has to record this info which in a way is an interpretation of what it is receiving, thus as much as it is commonly accepted that a RAW file is a digital negative,
I ask, how much 'processing' is actually going on during the 'recording' of that data?

A scenario I use is that of a normal window, lets make that window 3 foot by 2 foot.
Now put some lead in it to make it pretty as in the older style & split it into say ten smaller sections.
Those lines of lead have now reduced the actual area of glass that can actually let light in.

Now I'm not clever enough to know this next part but here goes...
Does each smaller pane now see the whole image or just the part of the area as seen by the previously single window.
I'm going to assume just the small part because if your eyes are the sensor & you are a distance back as is the sensor from the front of the lens, with out going right up to the individual smaller section, you see the whole image but through smaller cut up sections.

So surely less light coming through the now smaller panes / pixels has to be / give a lower quality of light for the sensor to record / process.

My point to all this is as the makers are cramming in more pixels into the same sensors surely the quality of light squeezing through can't be as good.
Again, a single unobstructed window / sensor has more light than the same size with lots of smaller panes / pixels?

Have we reached a point where the makers are fooling us into these massive pixel counting sensors do give higher dynamic range, or is it some very clever in camera processing?
Even with RAW files? Canon-CR2, Nikon-NEF. Different files must have different processing going on?
As I mentioned these have to be recorded in some fashion otherwise there wouldn't be the difference in RAW file formats.

This isn't meant to be a thread over whose sensors are better, so please don't start one of those.
This is a general thought about when has a sensor too many pixels & have they already & have we been deluded already for some time now?

Is the medium format camera & its larger sensors the way to go before those manufactures go out of business?

I noticed the older Pentax 645D with 40 million pixels can be bought for under £2.5k it has a 44x33mm sensor
But their new 645Z has the same size sensor & now 53 million pixels.

There are of course other medium format cameras with larger sensors & higher pixel counts, but I'm thinking they have a long way to go before they hit trouble.

You need to start shooting Birds and join the Bird Section

 

[retrosi]

Hello Bill, can I politely ask why?

 

[Swanseajack]

As Alan (woof woof) states, I would say the type of sensor and processor has a lot to do with it, along with the removal of the AA filter. As an example the Fuji cameras I have owned have produced some excellent images at APSC size and to my eyes and computer monitor will give FF a run for there money, especially in JPEG SOOC.

 

[BillN_33]

Hello Bill, can I politely ask why?

Simon, no offence ….. more tongue in cheek

it's just that we can tend to get into detail on many "bird" type questions

 

[retrosi]

Oh! thanks for clearing that up, I certainly knew there was no offense meant, but your remark caused my thought process to kick in & not having been to the 'bird area!
I was wondering where you were going in your thought pattern, now it's clear, cheers.
Perhaps I'll go take a 'read' and see if I can conjure up some deep questions with the detail required to invoke the necessary replies...

 

[woof woof]

As Alan (woof woof) states, I would say the type of sensor and processor has a lot to do with it, along with the removal of the AA filter. As an example the Fuji cameras I have owned have produced some excellent images at APSC size and to my eyes and computer monitor will give FF a run for there money, especially in JPEG SOOC.

I used to work in computing / electronics and most people wouldn't believe the effect that the shape, size, length and material a circuit is made up of has. Two electronically the same circuits (the same electronic design and identical components) but with different circuit path routing, layout, thickness of material etc could have very different noise and signal characteristics. That sort of thing has to have an effect when dealing with relatively low analogue and digital signals levels. I've never worked on camera systems but I'd imagine that the effect of changes to design even with the same components within a digital camera image pipeline would be significant. And then you've got microlenses and the like...

 

[Swanseajack]

I used to work in computing / electronics and most people wouldn't believe the effect that the shape, size, length and material a circuit is made up of has. Two electronically the same circuits (the same electronic design and identical components) but with different circuit path routing, layout, thickness of material etc could have very different noise and signal characteristics. That sort of thing has to have an effect when dealing with relatively low analogue and digital signals levels. I've never worked on camera systems but I'd imagine that the effect of changes to design even with the same components within a digital camera image pipeline would be significant. And then you've got microlenses and the like...

Only thing I have electronic experience with is an EBCU (Electron Beam Curing Unit, used to change the molecular structure of polymers) machine which we would strip and re-assemble the window in a clean room environment. Complex and over my head at times.

 

[arad85]

If the image is well exposed (and by that I mean the signal to noise ratio is high) then all that matters is sensor size for image quality, not pixels.

Why? Well, a well exposed image has way more signal to noise, so noise is minimal compared to the overall image. Given that, when you print out a picture, you are taking a sensor that is Xmm x Ymm and blowing it up to A inches x B inches. No matter how big or small the pixels are (assuming they are both big enough to capture more signal to light and small enough to not be seen), 1mm x 1mm of the sensor will cover a certain area of the picture. The smaller pixels will be smaller on the paper, and individually contribute less to the overall picture. So whilst you may look at each individual pixel at 100% on your screen and perceive it as of lower quality for signal to noise as its bigger brethren, it covers a smaller area of print, so is less noticeable.

Take the "same" image with a bigger sensor (i.e. change lenses so the field of view is the same) and now you have a bigger sensor producing the same image in print. Now each 1mm x 1mm of the sensor produces a correspondingly smaller part of the actual print, so the print is perceived to be of better quality. It will also appear sharper (assuming the lenses are similar quality) as you are magnifying the imperfections less.

All this goes away when you crop though as by cropping, you are effectively changing the sensor size.

If you want to search out the effect of sensor and electronic noise on images, look for the guys doing astrophotography. Not only are some of them extremely talented, they have also analysed noise from sensors and how it is produced quite theoretically as noise is their biggest bugbear (very little light, long exposure photography and even then an underexposed image).