Pixel density, please explian.

[Great Bustard]

But what he actually wrote was

'Anyway, moving right along, pixel density is an irrelevant measure in terms of IQ (image quality).'

Quite clearly that statement is wrong, which is why I was initally confused by his explaination. He understands it, and I understand it, but his explaination was so poor that it confused me and made no sense.

Indeed. I should have written, "'Anyway, moving right along, pixel density, by itself , is an irrelevant measure in terms of IQ (image quality).

 

[Lee Jay]

GB - You are right, of course, but I'd just like to add that pixel density is relevant to image quality when you are focal length or magnification limited and your goal is primarily detail extraction. In those cases, more PD is better. Yes, I realize you already know this.

Indeed -- and that is an important clarification to make. Thanks!

I just realized this case isn't all that special. When you are focal length or magnification limited, essentially your sensor size is fixed by the focal length and subject diatance or by the magnification (because you are going to crop to a fixed size) which leaves the other parameters as before.

--
Lee Jay
(see profile for equipment)

 

[Cedarhill]

There is one issue that I haven't seen mentioned yet that appears to me to have as much influence on image quality as pixel count and sensor size. What I am talking about is sensor technology. New generations sensors produce better images than older generation sensors even though they may have less area. For example, the new Olympus micro 4/3 cameras outperform older generation cameras like the Canon XTi (according to DPR at least), even though the sensor is half the area. You can not reasonably make any conclusions about sensor performance without taking new generation technology into account.

 

[Lee Jay]

As do read noise, dark current, etc.

Not sure how much recent progress is "real" versus processing.

--
Lee Jay
(see profile for equipment)

 

[Cedarhill]

The reason I bring this up is that I was playing with the new comparison tool provided by DPR and compared Canon E-PL1 images with the Nikon D3000. It appears to me that the Olympus performed better than the Nikon by any measure you want to apply, even though the sensor is half sized. Both JPEG and RAW files were compared and I didn't notice that detail was being unduly obliterated in the process. This contradicts traditional thinking in regards to image quality and sensor size. The most reasonable explanation I can come up with is superior sensor technology, though superior image processing could also be involved.

 

[Great Bustard]

The reason I bring this up is that I was playing with the new comparison tool provided by DPR and compared Canon E-PL1 images with the Nikon D3000. It appears to me that the Olympus performed better than the Nikon by any measure you want to apply, even though the sensor is half sized.

I'm assuming that you're talking about these two images:

http://a.img-dpreview.com/reviews/NikonD3000/samples/comparedto/raw/d3000-ISO100-ACR.jpg

http://a.img-dpreview.com/reviews/OlympusEPL1/samples/comparedto/EPL1_ISO200-ACR.JPG

The first thing to note is that the EPL-1 sensor has more pixels, which gives it an advantage. The second thing to note is that the Nikon pic is less sharpened than the EPL-1 pic. If you sharpen the Nikon pic until it has the same noise as the EPL-1 pic, you may come to a different conclusion.

That said, the D3000 sensor measures 23.6 mm x 15.8mm (diagonal = 28.4mm, area = 373 mm²) vs 17.3mm x 13mm (diagonal = 21.6mm, area = 225 mm²) for the Olympus E-PL1. So, if we go by diagonal measure, the D3000 sensor is 31% larger, and if we go by area, it is 66% larger.

As a side, since DPR's lens test are in lp/ph (linear pairs per picture height), then if comparing the two cameras, we would not go by the sensor diagonal, but by the sensor heights (effectively cropping the D3000 image to 4:3), which would make the D3000 sensor 22% larger.

Both JPEG and RAW files were compared and I didn't notice that detail was being unduly obliterated in the process. This contradicts traditional thinking in regards to image quality and sensor size. The most reasonable explanation I can come up with is superior sensor technology, though superior image processing could also be involved.

When comparing jpgs, it's not only the jpg engine, but the jpg settings that make a huge difference. When comparing RAWs, the RAW converter being used, as well as the settings, can also make a huge difference, and there has been a lot of debate on that point in these forums, with many arguing that using the same RAW converter and settings for different cameras is not "fair", and others arguing that not using the same RAW converter and settings is not "fair".

Of course, as Lee Jay noted above, factors such as QE (quantum efficiency), read noise, etc., all matter. How much they matter depends greatly on the particular scene in question. The differences between an "ideal RAW conversion" and a default in-camera jpg, however, can be staggering.

So, for sure, it is not enough to know merely the sensor size and number of pixels -- there are many other factors involved. What can be said with certainty , however, is that pixel density, as a single measure, tells us nothing about the IQ of a system.

 

[Cedarhill]

This is the tool I used for the comparisons I did. Pretty useful.

http://www.dpreview.com/learn/?/Guides/dpreview_comparison_widget_01.htm

 

[CeeDave]

GB - You are right, of course, but I'd just like to add that pixel density is relevant to image quality when you are focal length or magnification limited and your goal is primarily detail extraction. In those cases, more PD is better. Yes, I realize you already know this.
--
Lee Jay
(see profile for equipment)

Doesn't pixel density also affect critical f/stop. As pixel density gets large, the f-number at which you start losing resolution to diffraction gets smaller.
--
Chris in Red Stick

 

[Chris59]

GB - You are right, of course, but I'd just like to add that pixel density is relevant to image quality when you are focal length or magnification limited and your goal is primarily detail extraction. In those cases, more PD is better. Yes, I realize you already know this.
--
Lee Jay
(see profile for equipment)

Doesn't pixel density also affect critical f/stop. As pixel density gets large, the f-number at which you start losing resolution to diffraction gets smaller.
--
Chris in Red Stick

No. The higher the pixel density, the higher the resolution and the larger the aperture at which you start to see the effects of diffraction. However, this is not a fault, and at no point does a high resolution (high pixel density) sensor actually produce worse images at any aperture, regardless of diffraction.

Of course as you increase sensor resolution you do get the "law" of diminishing returns kicking in. The amount of useful detail you can extract from a lens gets smaller and smaller as you increase the number of pixels on a sensor, but at no time does it actually become less.

The term "diffraction limited" when applied to a lens is a good thing and a sensor which can record this "limitation" ought also to be considered a good sensor.

 

[Rriley]

Sensor size divided number of pixels gives you density.
Number of pixels multiplied by density equals sensor size.
Sensor size divided by density equals number of pixels.

How can they be separated so that two matter and one is irrevevant? From my understanding of basic mathematic formula what you are saying makes no sense.

Maybe what he means is that as a sole parameter, it tells us absolutely nothing about the image potential. With just the pixel size, you have no idea how much resolution the sensor can have, or how much light it can collect. The number of MP gives you potential image resolution, alone, and the sensor size tells you how much light can be collected, alone.

and to add, commonly used pixel spacing wont tell us how large or small a pixel is, and where nMOS have inherently less surface wiring than CMOS, most surely that allows an nMOS to have larger pixels than it otherwise would.

--
Riley

any similarity to persons living or dead is coincidental and unintended