Poll: Most wanted new features for 5D Mark IV

[Khufu]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For fixed sensor dimension, higher MP means smaller photo site size which leads to lower signal to noise.

surely that's only true when assuming all micro-engineering is equal - as I just posted... isn't the a7R the higher res shooter and better high-ISO performer out of the a7 and a7R?

 

[Lee Jay]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For fixed sensor dimension, higher MP means smaller photo site size which leads to lower signal to noise.

No, it doesn't. Not for the same sensor area.

I like to use this analogy.

Which way do you get more pizza, when you cut the 15" pizza into 8 slices or 12?

Smaller pixels tend to do better for a very easy to understand reason - larger pixels do nothing but block-average when compared to smaller pixels, and block averaging is a very primitive and poor way to reduce noise. Modern NR algorithms are far more efficient than that.

Everything is the same between these two - focal length, f-stop, ISO, shutter speed, lighting, location, processing (from raw), final image size, sensor generation, performance per unit of area, everything. Everything except pixel size. The pixels on the left are 16 times smaller than those on the right. So, which has better detail-to-noise ratio?

 

[StudentOfLight]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For fixed sensor dimension, higher MP means smaller photo site size which leads to lower signal to noise.

No, it doesn't. Not for the same sensor area.

I like to use this analogy.

Which way do you get more pizza, when you cut the 15" pizza into 8 slices or 12?

Smaller pixels tend to do better for a very easy to understand reason - larger pixels do nothing but block-average when compared to smaller pixels, and block averaging is a very primitive and poor way to reduce noise. Modern NR algorithms are far more efficient than that.

Everything is the same between these two - focal length, f-stop, ISO, shutter speed, lighting, location, processing (from raw), final image size, sensor generation, performance per unit of area, everything. Everything except pixel size. The pixels on the left are 16 times smaller than those on the right. So, which has better detail-to-noise ratio?

Which cameras are you comparing in these pictures?

 

[tcmatthews]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For a specific technology I think they do at least a little.

For example you can compare Sony 7, 7R and 7S.

Scale A7R to A7 and the difference really isn't there. Maybe the A7S, with somewhat different tech too keep in mind, would still do a little bit better.

Isn't the a7R meant to be both a higher resolution image producer and a marginally better High-ISO performer than the a7? If so, it kinda messes up the resolution vs low light argument, doesn't it?...

A7r is meant to be the higher resolution image producer the A7 has marginally better High-ISO performance. In reality the A7r has slightly more noise but retains more detail because it has no AA filter.

Your whole smaller pixel produce better ISO performance completely ignores reality.

 

[Lee Jay]

Your whole smaller pixel produce better ISO performance completely ignores reality.

I posted some reality above, and the smaller pixels just crushed the larger pixels.

 

[LetTheRightLensIn]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For fixed sensor dimension, higher MP means smaller photo site size which leads to lower signal to noise.

No, it doesn't. Not for the same sensor area.

I like to use this analogy.

Which way do you get more pizza, when you cut the 15" pizza into 8 slices or 12?

Smaller pixels tend to do better for a very easy to understand reason - larger pixels do nothing but block-average when compared to smaller pixels, and block averaging is a very primitive and poor way to reduce noise. Modern NR algorithms are far more efficient than that.

Everything is the same between these two - focal length, f-stop, ISO, shutter speed, lighting, location, processing (from raw), final image size, sensor generation, performance per unit of area, everything. Everything except pixel size. The pixels on the left are 16 times smaller than those on the right. So, which has better detail-to-noise ratio?

Which cameras are you comparing in these pictures?

I think those came from a P&S vs DSLR test from a few years back. With P&S on the left and with equal areas o the sensor compared on each and the DSLR upscaled to the same pixel dimensions as the P&S shot.

 

[RGF]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For fixed sensor dimension, higher MP means smaller photo site size which leads to lower signal to noise.

No, it doesn't. Not for the same sensor area.

I like to use this analogy.

Which way do you get more pizza, when you cut the 15" pizza into 8 slices or 12?

A photo site that is larger will have more signal. For equivalent electronic noise, which is doubtful since the weaker signal will need more, statistic very clearly tell us that the S/N in the larger site will be lower. The noise from the signal (number of photons reaching the site) is governed by the Poisson distribution and the variance is equal to the signal. Noise is typically reports as the std dev which is the square root of the variance. For if the signal is 10,000 the variance is 10,00 and the std dev is 100. S/N is 10,000 / 100 = 100. If the area is 1/4 the size the signal will be ~ 2,500 with a variance of 2,500 and a std dev of 50. S/N is 2,500 / 50 = 50.

Which S/N would you rather have, 100 or 50?

 

[Lee Jay]

Interesting that two competing features are chosen, both by about 50% of the respondents. These are higher megapixels and the other is better low light noise.

Those don't compete.

For fixed sensor dimension, higher MP means smaller photo site size which leads to lower signal to noise.

No, it doesn't. Not for the same sensor area.

I like to use this analogy.

Which way do you get more pizza, when you cut the 15" pizza into 8 slices or 12?

A photo site that is larger will have more signal. For equivalent electronic noise, which is doubtful since the weaker signal will need more, statistic very clearly tell us that the S/N in the larger site will be lower. The noise from the signal (number of photons reaching the site) is governed by the Poisson distribution and the variance is equal to the signal. Noise is typically reports as the std dev which is the square root of the variance. For if the signal is 10,000 the variance is 10,00 and the std dev is 100. S/N is 10,000 / 100 = 100. If the area is 1/4 the size the signal will be ~ 2,500 with a variance of 2,500 and a std dev of 50. S/N is 2,500 / 50 = 50.

Which S/N would you rather have, 100 or 50?

That's not the proper comparison.

This is.

Which would you rather have, one pixel with a SnR or 100 or four pixels each with an SnR of 50?

 

[StudentOfLight]

...

A photo site that is larger will have more signal. For equivalent electronic noise, which is doubtful since the weaker signal will need more, statistic very clearly tell us that the S/N in the larger site will be lower. The noise from the signal (number of photons reaching the site) is governed by the Poisson distribution and the variance is equal to the signal. Noise is typically reports as the std dev which is the square root of the variance. For if the signal is 10,000 the variance is 10,00 and the std dev is 100. S/N is 10,000 / 100 = 100. If the area is 1/4 the size the signal will be ~ 2,500 with a variance of 2,500 and a std dev of 50. S/N is 2,500 / 50 = 50.

Which S/N would you rather have, 100 or 50?

That's not the proper comparison.

This is.

Which would you rather have, one pixel with a SnR or 100 or four pixels each with an SnR of 50?

Which would you rather have, one pixel with SnR of 2, or four pixels with an SnR of 1?

 

[Lee Jay]

Which would you rather have, one pixel with a SnR or 100 or four pixels each with an SnR of 50?

Which would you rather have, one pixel with SnR of 2, or four pixels with an SnR of 1?

Neither. However, either one will do.

The ideal situation is to have each and every photon's position and wavelength recorded separately. That's equivalent to an infinite number of pixels. That will produce the most information possible and thus the best image possible from that exposure (scene, FOV, illuminance, shutter period, etc.).