Pixel Density?

[Chuck A]

Since this is a small sensor forum I thought this might interest the readers.

I don't know if anybody has been following the newest trend over at Dpreview. They have started following and charting Pixel Density (PD) as a measure of the quality of a sensor. You can get the background here: http://www.dpreview.com/news/0807/08070301pixeldensity.asp and there are some interesting threads here: http://forums.dpreview.com/forums/readflat.asp?forum=1012&thread=28498478 & here: http://forums.dpreview.com/forums/readflat.asp?forum=1013&thread=28502048

Basically the gist is that the smaller the number the better.
Fuji F100 25MP/cm²
Fuji F40 18MP/cm²
Fuji F30/31 14MP/cm²

DSLRs seem to run between 1.4 and 3.3MP/cm².

The Ricohs are :
Ricoh GR Digital 21 MP/cm²
Ricoh GX8 21 MP/cm²
Ricoh R4 24 MP/cm²
Ricoh GR Digital II 25 MP/cm²
Ricoh GX100 25 MP/cm²
Ricoh R8 35 MP/cm²

The Canon is:
Canon PowerShot G9 28 MP/cm²

Look at some of the older smaller mp models:
The 5MP Nikon Coolpix 5000 8MP/cm²
The 3MP Nikon Coolpix 990, 995 8MP/cm²

The 2MP Olympus C-2100 UZI 6 MP/cm²
The 1.3MP Olympus E-100 RS 4 MP/cm²

It seems to me that there is much more involved here. First you have to compare models with similar mp counts to judge image quality at a particular print size. I am not sure that pixel density is the whole story. Not being an engineer I am sure that my thinking may be flawed but I thought that the size of the pixels really mattered. Just having less of them per cubic cm doesn't help if they are small and poor quality. Also, camera processing, lens sharpness and contrast and many other variables come into effect here.

For example, just looking at the test crops at Dpreview for these cameras the Nikon CP5000 has a PD of 8MP/cm² while the Fuji F31 has 14MP/cm². They have roughly the same MP and by the PD the Nikon should be a better camera. Look at the noise crops and it is clear that the F31 has much lower noise characteristics than the Nikon.

This is just one example but I think that as you went through and actually compared real life output and correlated it with PD it may not be all that cut and dried. Look at the small MP cameras and the PD numbers are 4 to 6MP/cm². But my main interest is in the quality of a certain print size. Even at 25MP/cm² the 10mp GX100 would certainly make a much better 11x14in print than the 1.3mp Olympus E-100 RS with its' 4MP/cm².

Dpreview is including pixel density numbers with all of their camera reviews. They say that it gives a clearer picture of image quality than just megapixels. While I agree with that it certainly is not the end all here. I would hate to see another flawed way of evaluating sensors by numbers to replace another flawed system.

Am I wrong here. Anybody interested in this who can maybe shed some light as to whether PD is a good number to have? Perhaps someone like Sean Reid could weigh in and help us out.

 

[jonoslack]

Hi Chuck
I think it's interesting stuff, and I agree that it isn't everything . . Mind you, it goes some of the way to explain why the E100RS produced such fine files, despite the very low pixel count.

I think it also goes to show how much scope there is for higher pixel count in dSLR cameras - one wouldn't want to go up to densities of 20 or 30, but the highest density at the moment is in the 10mp Olympus 4/3 and the 14mp APSc cameras (also around 4). These may not be the kings of low noise, but they do very well up to 1600 ISO.

It also shows what scope there is for camera makers to bring out some excellent high quality compacts using slightly larger sensors (2/3 seems to me to be something of a sweet spot).

 

[Terry]

I agree that it isn't the only story and the quality of the pixels is important. For instance if progress has been made in sensor technology (which I believe it has) you can effectively pack more pixels on a 2/3 sensor today than what you could a few years ago. The question is what is optimal and how do you get consumers to understand.

On a number of todays cameras the sensors are so jammed packed that the real problem becomes noise reduction and smearing.

I do think there may be something new at Photokina and hopefully all of the manufacturers read all of the buzz on the DP1. Ultimately that camera has its issues with the user interface and IMHO the Foveon sensor's quirkiness but it did generate lots of good discussion and proved that there is a market for this kind of camera.

 

[Lili]

I agree too that density is not the whole story, the improvements in sensor technology keep the relationship between pixel density and image quality from being perfectly linear.
Improved sensors, noise reduction algorithyms, et al all contribute to the whole.
In some cases, at least for JPEGS, the OEM's have gotten so paranoid about noise that image detail suffers horribly.

 

[Chuck A]

Thanks for the replies. We will have to keep watch on this. For the serious small sensor cameras that have RAW there is some extra leeway. The smearing and loss of detail in JPG files makes most other small sensors unusable IMHO. While the GX100 has a PD of 25 (which seems rather high) its' JPGs are pretty usable because of a lower level of noise reduction. They are a bit fragile but they can take some PP before they suffer.

I agree that education on this subject is a good thing. If more people are educated and request better sensors then maybe the manufacturers will comply. So maybe if the sensor makers keep the PD as low as they can and also include good quality pixels we will be out ahead. Numbers like PD can help to educate.

 

[Bob]

Keep in mind that pixel density although an important factor, needs to be paired with complementary lens resolution. In some of the small sensor cameras, I have a distinct impression that the sensor is better than the lens. OTOH, it is a lot easier to make a high quality small lens than a big one.
-bob

 

[Lili]

Bob,
I have to second that on the smaller lens being possibly easier to correct.
The zooms on both my Fujis;
S5200 20mp/sq cm
S6000 14 MP/sq cm
are really quite good for the money.
The S6000 especially is happy combination of the sensor from the F30/31fd with a wide range zoom/RAW/and decent image controls.
As to NR and JPEGS; give the GRD I's jpeg engine any day.
If one could combine that with the RAW buffer on the GRD II...
Sigh.....