Does size matter?

  • Thread starter Thread starter Deleted member 8670
  • Start date Start date
D

Deleted member 8670

Guest
I use Nikon gear, specifically a 5 year old D300 and a bunch of lenses. My camera is getting long in the tooth, and I've been waiting for something similar with the latest sensor technology, to buy as a replacement.

I was very interested in the D800, as it does a very respectable 15.5MP in DX crop mode. It has excellent noise performance too. But the price?! I like the DX format, as I often shoot nature photography with a 300mm lens, and DX gives me that extra reach. I'm not sure I want to spend all that money and only use half of the viewfinder/sensor. So I kept waiting for a "D400".

I want a "D400", but many people think one is unlikely to materialise. And although the D7100 specs seem good, it looks slightly disappointing in ergonomics, build and buffer size. I await the test results with interest. But here comes the question...

Does the sensor size (FX/DX) have a direct effect on the noise performance of a sensor? My best guess is No, why should it? But the figures seem to disagree. If you look at the data on the DXOMark website you get (in chronological order of release):

D300 (12MP): 679 ISO
D7000 (16MP): 1167 ISO
D800 (36MP): 2853 ISO
D3200 (24MP): 1131 ISO
D600 (24MP): 2980 ISO
D5200 (24MP): 1284 ISO

The best the current DX models seem to provide is a 1-stop advantage over the D300, whereas I get 2 stops with FX. Surely, the sensor size is irrelevent?! Wouldn't noise performance be more affected by pixel size?

When the D800 was released, it had roughly the same pixel size as the D7000, but twice the noise performance. I hoped that future DX cameras would have similar performance, but so far, it seems not, although they have decreased the pixel size by 33% too.

I suppose we will find out when DXO tests the D7100. But, in the meantime, can anyone with a technical brain think of why the figures look the way they do?
 
Last edited by a moderator:
Yammer said:
Does the sensor size (FX/DX) have a direct effect on the noise performance of a sensor?
Click to expand...
Yes.

Think about it for a minute... Take a photo with the appropriate lens to frame properly.... When you print, you print to a specific size. The larger the sensor, the less you have to magnify the resultant image. The less you have to magnify, the less you will see the noise.

Simples ;)
 
Yammer said:
Does the sensor size (FX/DX) have a direct effect on the noise performance of a sensor? My best guess is No, why should it?
Click to expand...

You're right, why should it? The figures you quote are distorted by a number of factors - pixel size/density, age/technology of sensor, technology/efficiency of A-D converter and amplifiers, etc.

Wouldn't noise performance be more affected by pixel size?
Click to expand...

Absolutely, if everything else is equal then pixel size is the major contributor to noise performance. As the absolute amount of noise seen in the output from each pixel is relatively the same irrespective of physical size the ratio of that noise to the complete output from the pixel becomes smaller as the pixel gets bigger.
 
Bristolian said:
Absolutely, if everything else is equal then pixel size is the major contributor to noise performance.
Click to expand...
Nope.... For a well exposed image (i.e. where signal is totally dominant over noise), it is sensor size that matters most (given the same sensor technology).
 
Bristolian said:
You're right, why should it? The figures you quote are distorted by a number of factors - pixel size/density, age/technology of sensor, technology/efficiency of A-D converter and amplifiers, etc.
Click to expand...

Also what manufacturers would like to limit their cameras to.
 
If the sensor technology is the same and the shot is taken from the same camera to subject distance with the same lens, the same ISO, the same aperture setting and the same shutter speed all an APS-C image is is a crop of the centre portion of the full frame image and it'll be identical to that centre portion.

That's my theory... but in the real world you'll be lucky to get two sensors which are the same tec and other things will alter too, you wont take exactly the same shot from the same distance with the same settings etc. so it all gets messy.

For example when shooting with FF and a 50mm lens at f8 you may end up with ISO 400 and your noise will equal "X." Take the same shot with a 35mm lens at f5.6 (that's close enough) and you may be using ISO (I give up... my head hurts, a lower one anyway...) and your noise will equal "Y." You've then got your image output size to think about and how that and the viewing distance will affect your perception of noise.

My own theory is that any difference in noise between the best APS-C and FF cameras (and MFT for that matter) will only be apparent at the highest ISO settings (if the sensors are within spitting distance of each) and you'll need a sharp knife to tell them apart at all but the largest of print sizes, the closest of viewing distances or when doing extreme pixel peeping :D
 
arad85 said:
Yes.

Think about it for a minute... Take a photo with the appropriate lens to frame properly.... When you print, you print to a specific size. The larger the sensor, the less you have to magnify the resultant image. The less you have to magnify, the less you will see the noise.

Simples ;)
Click to expand...

I have thought about it - a lot!

I can't follow your logic though. Surely the D600 and the D3200 both have 24MP sensors and need the same enlargement?
 
Bristolian said:
You're right, why should it? The figures you quote are distorted by a number of factors - pixel size/density, age/technology of sensor, technology/efficiency of A-D converter and amplifiers, etc.
Click to expand...

Yes, which I why I listed them in chronological order. My logic was that technology improves with time, and that the technology that gives the D800 & D600 great S/N performance should also be available in the D3200 and D5200 cameras.

However, as you point out, the pixel size isn't the same. A D800 DX crop is 15.5MP, and the new DX cameras are 24MP--that's an extra 50% pixels in the frame, or, to put it another way, the pixels have been made smaller, at about 65% of the size. Simplistically, we might expect something like 65% of the light grasp, and maybe a performance of about 1800 ISO.

Also, logic dictates, that if pixel size is so important, why are the D600 and D800 results so similar?! Surely the D600's pixels are much bigger. I'm confused. This is why I asked the question, and am grasping to understand why sensor size seems to dictate the performance, irrespective of pixel size.

arad85 has stated this, but I can't follow his logic, and I want to understand why for myself, not take someone's word for it. Sorry to be a pain.
 
I just revisited the DXOMark website, and had a play with another of the stats tools. These charts are quite revealing. They clearly show that the FX models are well above the DX, even allowing for the advancing technology with time.

DX
DX.jpg


FX
FX.jpg
 
A bit of googling, and I'm starting to work things out:
http://simon.tindemans.eu/essays/sensorsizes
http://www.clarkvision.com/articles/does.pixel.size.matter/
http://photography-on-the.net/forum/showthread.php?t=706255

I have been taken in by the myth that small pixels equals noisy pictures. A small pixel IS a noisy pixel, but more pixels reduces noise. The type of noise is determined by the light grasp of the sensor, and larger sensors generally have larger pixels as well as high pixel counts. Even a camera like the D800, which has smaller pixels than a D600, has effectively reduced noise due to the increased number of pixels averaging things out.

The design of photocells over the years, seems to have been the biggest single influence on performance.

I think I can probably also guess that running a D800 in crop mode will reduce its noise performance, as you are effectively reducing the noise reduction of large pixel numbers. So a D800 in crop mode is probably no better than a D7000.

I need a lie down.
 
Last edited by a moderator:
Of course the simple way around it is to sell the 300mm and purchase a 400mm + the D800E, you then have a similar effective focal length or use a 1.7 TC on the 300mm
 
I have fretted over many a laboratory test data sheet, and my epiphany was to realise that good light and good eyes matter far more than any discrete value I could quote you. Relax in the knowledge that the camera is just a tiny piece of the process. Your D300 is still an excellent camera, and a good shot with a D300 is way better than a mediocre shot with a D800 or D4.
 
Yammer said:
I can't follow your logic though. Surely the D600 and the D3200 both have 24MP sensors and need the same enlargement?
Click to expand...
Nope... The D600 is FX, D3200 DX which means to get the same sized print (6x4, 18x12 etc) you have to magnify the image that is created on the sensor more. Rather than thinking of pixels, think of sensor area...
 
In the most fundamental terms, ISO and noise performance is simply about the amount of light and photon capture. FX has more than twice the surface area of DX (crop factor squared) so all other things being equal, noise performance has the potential to be at least one stop better.

Of course all other things are rarely equal and get complicated by things like pixel density, sensor development and processing engines etc etc, but FX always has that big head start.
 
danbroad said:
Your D300 is still an excellent camera, and a good shot with a D300 is way better than a mediocre shot with a D800 or D4.
Click to expand...

See the point you are making but with the same person behind the camera the shot would be the same good or mediocre shot with any camera they use which is why some people put too much bias on the the camera part of the equation.

But I agree, if you take an outstanding shot that you took using a D300, how much more outstanding would it have been on a D800 and what is that difference worth and what does it matter?
 
danbroad said:
I have fretted over many a laboratory test data sheet, and my epiphany was to realise that good light and good eyes matter far more than any discrete value I could quote you. Relax in the knowledge that the camera is just a tiny piece of the process. Your D300 is still an excellent camera, and a good shot with a D300 is way better than a mediocre shot with a D800 or D4.
Click to expand...

ernesto said:
See the point you are making but with the same person behind the camera the shot would be the same good or mediocre shot with any camera they use which is why some people put too much bias on the the camera part of the equation.

But I agree, if you take an outstanding shot that you took using a D300, how much more outstanding would it have been on a D800 and what is that difference worth and what does it matter?
Click to expand...

No, sorry, I don't buy into that philosophy. Of course, the photographer makes the picture, but the camera provides the limitations, and when you're trying to shoot a moving subject 80ft away with 1/400s, or shoot handheld inside a public building with twitchy security at 1600+ ISO the noise spoils the picture. It's nothing to do with light, or creativity, or how luscious my hair is, it's just physics.

Ever since the D7000 was released in 2010 with a 1-stop advantage, I have looked forward to the time when I can shoot at much higher ISO handheld. I don't like to carry a tripod around with me, unless I'm planning on a long-exposure shot.

I already shoot raw, and expose to the right. My handholding technique is pretty good. But the D300 struggles in low light, and always did. It's still a great camera though, and yes it still takes great photos. We're still not driving Ford Cortinas!
 
HoppyUK said:
In the most fundamental terms, ISO and noise performance is simply about the amount of light and photon capture. FX has more than twice the surface area of DX (crop factor squared) so all other things being equal, noise performance has the potential to be at least one stop better.

Of course all other things are rarely equal and get complicated by things like pixel density, sensor development and processing engines etc etc, but FX always has that big head start.
Click to expand...

The devil is in the detail.

What always confused me was the fact that the D800 and the D7000 have the same pixel density, and yet the D800 noise performance is much better. I put this down to advancing technology. But now it seems, quantity matters as much as quality, even though, for years, many people have been saying that too many MP is bad.

What DXOMark doesn't say about the D800 is what its Sports ISO rating is in DX Crop mode, and I suspect it will be lower than its FX rating. This is the concept I'm having trouble getting around. It's counter-intuitive to me: Surely, if the pixels are the same size, and the lightfall is the same, it doesn't matter how many pixels take the image. The cropped FX sensor is the same as a DX sensor, so it should produce the same results.

Perhaps it does. Perhaps its the way DXO expresses noise in relation to total image size that makes the difference. I read in one of the links above that bigger sensors don't have less noise, they just make the noise smaller! :lol:
 
Yammer said:
No, sorry, I don't buy into that philosophy.
Click to expand...

Hardly a philosophy. If the difference is worth it and justified (as it is to you for low light and hand held) then fine. The point was aimed at those that change to a newer model because it is a newer model rather than based on any real need.
 
Yammer said:
The devil is in the detail.

What always confused me was the fact that the D800 and the D7000 have the same pixel density, and yet the D800 noise performance is much better. I put this down to advancing technology. But now it seems, quantity matters as much as quality, even though, for years, many people have been saying that too many MP is bad.

What DXOMark doesn't say about the D800 is what its Sports ISO rating is in DX Crop mode, and I suspect it will be lower than its FX rating. This is the concept I'm having trouble getting around. It's counter-intuitive to me: Surely, if the pixels are the same size, and the lightfall is the same, it doesn't matter how many pixels take the image. The cropped FX sensor is the same as a DX sensor, so it should produce the same results.

Perhaps it does. Perhaps its the way DXO expresses noise in relation to total image size that makes the difference. I read in one of the links above that bigger sensors don't have less noise, they just make the noise smaller! :lol:
Click to expand...

Don't get hung up on pixels and density, and trying to make empirical sense of DXO on anything too specific will drive you mad. It's a useful indicator of some general trends though.

Pixels etc is (mainly) just technology, and very important technology, but that changes. The basic fact that FX gathers more than twice as much light/photons as DX is (one of) the fundamental advantages of larger sensors.

It used to be 'truth' that smaller pixels and higher pixel density resulted in more noise, but that's technology. Things are changing, with more efficient photon capture like capless microlenses and better processing engines - the latter being much more significant IMHO than it's usually given credit for.

If it helps, Nikon told me at the launch of the D800 that in crop mode it's performance was very similar to the D7000, but that the D800 actually had a very fractional advantage because the technology was newer.
 
HoppyUK said:
If it helps, Nikon told me at the launch of the D800 that in crop mode it's performance was very similar to the D7000, but that the D800 actually had a very fractional advantage because the technology was newer.
Click to expand...

It does help, thanks. At least, it restores a little sanity. I'm not so much "hung up" as someone who likes to properly understand the way things work. That's why I'm an engineer by trade.
 
You must log in or register to reply here.
Back
Top