This article should be read by everybody

Started Dec 12, 2013 | Discussions thread
TrojMacReady Veteran Member • Posts: 8,729
A rehash of debunked myths and misunderstandings.

K E Hoffman wrote:

This is all interesting.. but the fact is .. its all very theoretical.. at one level.. and yet the very DXO distribution graphs will bear out what I am about to write here.. and goes against the simplified idea that that there is not a pixel density to noise relationship.

Look at Generations of Cameras..

A700 vs A900 A900 has larger pixels same as about 10 MP APC and it has a better noise profile.. at the pixel level.. as it also has more resolution you have the choice of cropping or scaling the image and gaining both the better pixel noise profile AND the image scaling noise reduction.. If you crop you only get one..

A comparison that only proves sensor surface counts. The amount of photons captured to be exact. With almost 2.3 times the surface, it should come as no surprise that you will capture a lot more photons at the same f stop. Basically you changed factor Y (sensor size) to conclude something about factor X. Another variable if you want.

A77 va A700 va A99

Again.. A77 with newer sensor has less noise than A700 with double the pixels.. but it has more noise at the pixel level than A99..

Of course it does, because the pixels are smaller. But this is about efficiency per area. Same sized sensors. Why throw another variable into the mix when there's the 24MP and 36MP sensors to compare? Or the 12MP and 14MP sensor (A550/A500) for CMOS, or 10MP s 14MP for CCD (A350/A300).

Does more pixels mean more noise ALWAYS - No because there are a bunch of other Variables. like Sensor design. Camera Design., Sensor Generation. Where are the A/D converters located. [Sony likes to put them on the sensor. Canon has tended to put the off sensor.]

He talks about correlation (or complete lack of) in practise, the thing that matters to us. A point Eric Fossum often repeats, as theoretic assumptions about fully the same technology (the "what if we use that small pixel technology for larger"... etc) always fall flat on their face when there is and will be no such thing in reality. And there is no correlation worth mentioning in practise, especially once you start looking at the extremes. The cameras mentioned here are all at a similar end of the scale (all relatively large pixels).

Sony can't volate the patents Canon has on doing noise read and cancel at each pixel so they work to reduce the noise in the circuit paths.. [This of course is a broad generalization]

Actually, Sony has patents to cancel noise at the pixel level too.

If I make a sensor using the same design with the same electronics up to the A/D converter and one pixel is 50% larger the larger one will have less per pixel random noise..

Of course, but there's more pixels to gather photons in return.

The 70D is an interesting beast.. as Canon has put 40 Megapixels on the sensor.. each working separately to read focus and then as a team to read image. But this means there are in places double the supporting read circuits.. and the noise profile on the Canon 70D as soon as you get into the high ISO where the gain is being turned up before the A/D conversion it starts to get more noise faster than previous Canons tended to. (See this month's POP Photo Test of 70D)

Not supported by the RAW tests I've seen, where it shows noise levels to be pretty much identical to the 700D which does not employ the "dual pixel" technology. It would actually disprove your theory. Especially if we consider that the difference in pixel density is larger here than most of the examples you have mentioned sofar.

If I tell you there is a 16MP APC sensor and an 24MP APC sensor can you tell me which one has more noise. NO YOU CAN"T because I didn't tell you who made them and which design they were based on.. That is what he is saying in the summary.. Pixel size ONLY does not correlate 100% to noise level..

There is no correlation with any significance, see extreme examples with pixel size differences of a factor 9 or higher and similar QE and read noise.

But the trend lines on the charts and the and separation of FF to APC which also is a separation of larger sensor Pixel to smaller Sensor Pixel confirms the basic FACT. When you shrink the light bucket the % of noise leaking into the system will affect the image more PER PIXEL..

No, it proves sensor size matters, see variables mentioned in my first point. Wrong variable.

IF you scale them to the same size ALWAYS yes as TBCASS wants us to remember that act will done right will neutralize noise to a degree.. IE if I scale a 24 MP to 16 MP image size.. The noise will be very similar even in the same generation sensor.. But What if I just crop the center of the Image and scale both to 50% .. the 24 MP will have more color variation and or grain due to noise.

Not if the crop covers a similar FOV, noise will be about the same, regardless of output size.

Sensor size + Pixel Size + Technology Generation is a predictor or noise levels not only at the pixel level but collectively across the sensor. and as the gain is turned up at higher ISO the differences are magnified and will show up even in a scaled image.

False. See Sony 24MP vs 36 MP FF

Or the RX10/RX100II/RX100 sensors which are as effiecient per area as the best FF sensors currently available. A FF 147MP sensor made of 9 of those sensors, would produce results similar to a D4 sensor at the same output size.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow