How important is sensor pixel density really?

Neil S.

TPF Noob!
Joined
Jul 16, 2010
Messages
1,128
Reaction score
21
Location
Japan
Can others edit my Photos
Photos NOT OK to edit
I am starting this thread because there seems to be some pretty strong opinions about how important sensor pixel density really is.

Please state your opinions on this, and provide examples, etc.

This is especially relevant now in my opinion because of the move toward higher and higher MP counts, while still keeping the same size sensors.

The question has to be asked if more MP is really better or not without first increasing sensor dimensions.

Is 20+ MP for an APS-C too much for the dimensions of the sensor for example?

I for one have a very open mind on this, and hope that if we put our heads together we can provide some good info for the community on this one.

I would like to request that we keep this strictly about sensor density, and not about specific camera brands or models (unless it directly relates to the topic).

Here is some info on how sensor size and MP count relates to sensor density:

Sensor length (mm) x width (mm) = square mm (sensor area)

Next you need to divide the Megapixels by the sensor area, and this gives you the pixel density per square mm

Example: Canon EOS 5D mk II (21.1MP) 36mm x 24mm FF sensor

36mm x 24mm = 864 square mm

21,100,000 (pixels) ÷ 864 (sensor area) = 24,421 pixels per square mm (rounded down)

* For comparison the 7D has a much higher sensor density of 53,157, and APS-C sensors in general have a way higher density compared to FF. Also many sensors used in point and shoot and cell phone cameras have a lot higher density than even APS-C.

- Neil
 
I think it's fairly important, for the camera designers anyway. In practical terms, it doesn't matter to me while I'm shooting.

As I understand it, the lower the pixel density, the easier it is to create a nice clean image. That's why large sensor cameras are better at high ISO (less noise)...and the reason that P&S digicams are so bad at high ISO (more noise).

We also need to consider the cameras processor and it's programing. 6 years ago, 17 MP on an APS-C sensor would have seemed crazy...the noise would have been outrageous. But improvements in both sensor technology and processing technology, have allowed it to happen, with an acceptable level of noise.

I mentioned sensor technology, which I admittedly don't know a lot about...but I think that some of the improvements have been the 'lenses' that are the 'front/top' of the sensor and how they gather the light.
 
We also need to consider the cameras processor and it's programing. 6 years ago, 17 MP on an APS-C sensor would have seemed crazy...the noise would have been outrageous. But improvements in both sensor technology and processing technology, have allowed it to happen, with an acceptable level of noise.

Good point Mike.

I had not thought about the processing.

It seems to me that as time goes on this will improve greatly.

All they really need to do in my opinion for the short term is develop a way for the camera to apply noise reduction itself as good as you can get in post processing.

Who knows how long that may take though...
 
All they really need to do in my opinion for the short term is develop a way for the camera to apply noise reduction itself as good as you can get in post processing.
Actually, I believe that a big part of the in-camera processing is noise reduction...or maybe noise suppression.

As far as making it 'as good as post processing'...that's tricky. Keep in mind that most images will need a different level of noise processing...and much of that will depend on the particulars of the shot and the preference of the person doing it. In most/all cases, you loose detail when you reduce noise. So the camera companies especially will be careful about loosing too much sharpness.

A bit off topic, but sort of related. Most sensors have an anti-aliasing filter on them. This is meant to prevent Moire patterns that will tend to show up when shooting something with a repeating pattern. I think it has to do with the grid pattern of the sensor. This reduces sharpness...but the camera companies feel that this is an acceptable trade off....well most of them. Some don't have anit-aliasing filters and can thus produce excellent sharpness. There are some Sigma cameras and Leica as well.

Aliasing and Moire patterns
 
A bit off topic, but sort of related. Most sensors have an anti-aliasing filter on them. This is meant to prevent Moire patterns that will tend to show up when shooting something with a repeating pattern. I think it has to do with the grid pattern of the sensor. This reduces sharpness...but the camera companies feel that this is an acceptable trade off....well most of them. Some don't have anit-aliasing filters and can thus produce excellent sharpness. There are some Sigma cameras and Leica as well.

Aliasing and Moire patterns

This is interesting...

I hadnt heard of this before today.
 
Derrel,

Would love to get your thoughts on this...
 
Lower density always leads to larger pixels. Larger pixels can capture more photons. Larger pixels are more-capable at capturing large amounts of light. Lower pixel density makes it easier to get good to excellent optical performance out of "merely mortal" lens designs. When a physically large sensor, like a 24x36mm sensor or a Full Frame sensor, is paired with an average lens, the larger,low-density sensor with average lens can optically out-perform a much smaller APS-C camera that has been paired with an excellent lens. Analogy: a journeyman heavyweight MMA fighter can beat a top-level lightweight MMA fighter....bigger works better.

When the pixel density leads to exceptionally small pixels,like say, those that are just over 4 microns (we all know what camera I mean), dynamic range suffers, and color richness suffers, and the higher-ISO settings are not as good at collecting light as when the pixels are nearly twice as large...

There is a balancing act that the camera designers need to wage in terms of pixel size and density, and what the marketing arm of the company wants to try and lure consumers in with. The thing is, on-chip noise reduction can be done at the pixel level with CMOS sensors, so CCD sensors are basically now relegated to lower-level,cheaper bodies and medium format backs. On-chip NR has made it possible to make small, dense sensors with "acceptable" noise at the ISO settings most people use, most of the time. However, in each generation, higher density always leads to more noise at higher ISO settings, and the trend is to apply more and more NR to counteract that; if a company wants to try and make an ultra-high ISO camera, they need to go with bigger pixels; case in point being the Nikon D3s versus the Canon 1D Mark IV...the D3s blows the Canon away at the higher ISO settings, because the pixels are both lower in density AND larger, but fewer in number. Nikon's D3s is a vastly better high-ISO camera at the top settings than the 24 megapixel D3x is,even though both are from the same sensor technology generation.

We have no choice as consumers, except to buy what the camera makers offer for sale. If you want a super-duper, low-light, HIGH-ISO camera, buy the largest sensored camera you can get in your brand. The camera makers need to sell new cameras every generation. Who remembers the days when a 32 inch color TV was "a big TV set"???
 
This is really good stuff here Derrel...

I have never seen this explained anywhere near this well.

TY for your input.

I agree that a larger less dense sensor is better overall, for obvious reasons.

This is why at the moment I got my eye on the 5D mk II's replacement.

- Neil
 
I want to address the factor of resolution here as well.

All other things being equal more resolution is better than less resolution in my opinion.

That being said; the question as I see it is "more resolution at what cost?"
 
DxOMark - Compare sensors

Exactly: more resolution, at what cost? Higher MegaPixel count cuts down on High-ISO performance, color depth, and total dynamic range, with all other things being held equal. But, things are usually not "equal". Frankly, I think more could be done by going to larger sensors at this time; the APS-C 1.5x and 1.6x sensors are good these days, they really are, but the above comparison of the D90, T2i,and 7D shows that using a larger sensor (D90 at 370 sq. mm in area) versus the 1.6x (329 sq. mm) size, and keeping the MP count lower, actually can improve dynamic range, High-ISO performance, and color depth and richness. I think it would be great if Canon's APS-H sensor, like that in the 1D-series bodies, would become the new standard for prosumer bodies....it has a 548 sq. mm area...FF is 864 sq. mm.

I HOPE, strongly, that Canon moves the APS-H sensor into its next high-performance $1700 body line.

The question is "at what cost" do we want or need more resolution? I think there is ample resolution with the 12 MP to 13 MP Nikon D3,D700,and Canon 5D Mark 1 (12.8 MP). For me. For most people. I would rather have killer high-ISO performance, like the D3s has, rather than ultra-high resolving ability...I would rather be able to get a killer-clean, low-noise capture that is 12 to 13 MP in size, and then be able to up-scale it later in Photoshop, than I would have a camera like a D3x or SOny A900 that produces sharp 40x60 inch prints and chews thru CF cards like a fat guy thru Doritos...

I would really though, rather have wider dynamic range, than more resolving power, at the current point in time. The Fuji S3 and S5 Pro bodies had that extra, smaller set of pixels that responded ONLY to really bright values, which allowed those cameras to capture incredible highlight detail and deep,dark values as well with the bigger pixels. For me, wider dynamic range would be more welcome than higher resolving power; we can now up-scale images amazingly well with Genuine Fractals,sSpline, etc.etc.

It's like gas mileage...when gas was 39 cents a gallon, 10 MPG from a Chevy 396 engine was okay...when gas hit $1.00 per gallon, Honda cars became a desired commodity...we can have 45 MPG cars...but they cannot tow a boat or trailer...
 
Well said.

I think that part of the issue, is that the marketing departments of these companies, are steering the ship, so to speak. I do think that the engineers at Canon & Nikon could give us sensor technology, along the lines of what Derrell is talking about...but the marketing department wants more Megapixels, because that is an easy stat to sell. 90% of people's technical knowledge of digital cameras, starts and ends with how many megapixels it has.
 
Some don't have anit-aliasing filters and can thus produce excellent sharpness. There are some Sigma cameras and Leica as well.

This is interesting...

I hadnt heard of this before today.
Because of it's low cost, most of today's dSLR cameras have a Bayer type image sensor.
The image sensor does not record colors directly. Instead there is a Bayer Array filter in front of the image sensor:
BayerArray.png
this much of a Bayer Array would cover just 4 pixels. Two of the 4 squares are green because human eyes are most sensitive to green light. An anti-aliasing (AA)filter would be in front of the Bayer Array.

With this type of image sensor a software algorithm must be used to de-moisaic the image and interpolate the color actually represented by each 4 pixel array based on the voltage developed at each of red, 2 green, and blue pixels.

Sigma, and some other cameras, use a Foveon X3 sensor. The Foveon sensor has layered pixel sensors that separate light via the inherent wavelength-dependent absorption property of silicon, such that every location senses all three color channels, red, green, and blue.

Much less common is the 3CCD image sensor using three discrete image sensors, with the color separation done by a dichroic prism. Considered the best quality, they are pretty expensive so they don't get used in consumer cameras.

Noise reduction takes a lot of processing time, witness what happens to fps rate if high ISO or long exposure NR is used.

As Derrel mentioned, pixel density affects pixel size and pixel size affects how much signal (number of photons) a single pixel can gather. If a pixel is small, the amount of signal it can gather may not be enough to offset the noise the pixel and downstream processes also generates with every image.
 
DxOMark - Compare sensors

Exactly: more resolution, at what cost? Higher MegaPixel count cuts down on High-ISO performance, color depth, and total dynamic range, with all other things being held equal. But, things are usually not "equal".

More good stuff here...

TY again.

I had never thought of the dynamic range part of it really. I had seen it mentioned before when we were discussing a certain camera lol, but didnt think of it in relation to sensor density.

Just from this one thread so far my knowledge and understanding of FF vs. APS-C image quality has increased a lot.

I was already planning on getting a FF body in the next few years, and this just confirms the importance of that desire.

My 7D is great for me right now though, and it was a huge upgrade from my 30D in just about every way.

It also allows me to keep shooting the 17-55 too, which is kind of in its own class because there is no other Canon 2.8 wide zoom with IS.

I still got my fingers crossed on a 24-70 2.8L IS coming out soon lol.

That + the 5D mk II's replacement + my 70-200 mk II, would be an amazing combo to shoot with imo.

Speaking of the 5D mk III (just a codename lol), how many MP do you guys predict it will be? I am going to say 28 even.

- Neil

Edit: I realize that I am getting off topic with the last part. Unelss anyone else has anything to add on sensor density though, I think Derrel has pretty much nailed the answer here...
 
Last edited:
Well said.

I think that part of the issue, is that the marketing departments of these companies, are steering the ship, so to speak. I do think that the engineers at Canon & Nikon could give us sensor technology, along the lines of what Derrell is talking about...but the marketing department wants more Megapixels, because that is an easy stat to sell. 90% of people's technical knowledge of digital cameras, starts and ends with how many megapixels it has.

You are right, and I know its true.

I know this because when you look at a P&S in the store they often have a giant sticker on them with (only) the MP count advertised in huge numbers.

This brings us to another question.

Will the drive for more resolution and higher MP counts cause the camera makers to change to larger lens mounts that will allow the use of a larger (than FF) sensor?

There is no right or wrong answer to this, because its all based purely on opinion and speculation.

I would just like to get peoples thoughts on this...

- Neil
 
No. Firstly there are already larger sensors and companies fill that niche quite nicely. I don't think Nikon has anything to gain other than headache of trying to compete with yet another camera manufacturer by creating a medium format camera.

Secondly there would be two things that impact the change of the format. 1) either a new technology is created and companies work together en mass to take advantage with an interchangable format (4/3rds did this). or 2) the lens catalogue is backwards compatible (aps and micro4/3rds did this).

Creating good lenses is hard and very expensive. So if Nikon / Canon chose to create a larger sensor they'd need to redesign a camera and create a whole lot of new lenses at the same time (a kind of a resource drain that can potentially bankrupt companies), or develop a new camera for an existing format competing directly with new competitors in a field where they have zero reputation for only a TINY subset of professional customers.


My speculative opinion is that it would be madness.
 

Most reactions

New Topics

Back
Top