Go Back   Steve's Digicams Forums > Digicam Help > Newbie Help

Reply
 
Thread Tools Search this Thread
Old Jan 29, 2007, 11:29 PM   #21
Senior Member
 
Corpsy's Avatar
 
Join Date: Dec 2005
Posts: 879
Default

Eric, what I'm asking you is where do you get the information making the comparisons that you made? I have not seen any sites where any of the cameras you mentioned could be directly compared and all you've provided were your own opinions.

I never suggested that you made any comparisons to a Pentax camera. I brought those up as an example of different cameras using the exact same sensor getting far different results in noise and detail only due to in-camera processing.

The point that I made was that, "on different sensors built using the same technology, the one that has the most surface area per pixel will produce the least noise. " When you provide a couple examples of cameras you think dispute the point I made, and say that, "These examples directly *violate* what you're saying", I was just wondering if you could provide any evidence to support your own arguments as I have for my own.

Here's another example to support my own argument: http://www.dpreview.com/reviews/canoneos350d/page19.asp

There are charts at the bottom comparing the noise levels between the 6mp Digital Rebel and the 8mp Digital Rebel XT. They both have the same size sensor, but the newer more advanced XT has higher luminance and higher chroma noise levels.
Corpsy is offline   Reply With Quote
Old Jan 30, 2007, 9:49 AM   #22
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

I will happily post pictures from my own experimentation with descriptions of how I did it. I've taken thousands of pictures with the 20D and 1D MkII N. Unfortunately I sold my 10D to buy the 20D back when it came out. So I'll have to look through old pictures to see...

Wait a sec... I don't think I need to go that far. Look at the dark noise between these two pictures:
http://www.steves-digicams.com/2003_...s/IMG_9968.JPG
http://www.steves-digicams.com/2004_...s/img_9941.jpg

both are taken at 400 ISO. Since I didn't take them, I can't comment on how they were done - the comments say nothing about post-processing. But it does say this for the 20D's pictures:
We used a custom SET 1 parameter with Sharpening at +1 and used a
Canon EF 28-105mm F3.5/4.5 USM lens and ISO 100 unless noted.

Note that those are original pictures, so they are quite large. By the comments I'd say they were taken as JPGs, which probably means software incamera noise reduction was done.


I feel the noise is slightly worse to the same on the 10D (certainly not better) than in the 20D. And the chip in the 20D is higher resolution but the same size as the 10D. As to the 1D MkII N.... I'll shoot some examples that demonstrate it and post them (or I'll try, but I'm fairly sure I can do it!) I'm working right now so I'll have to do it later.

The point I'm trying to make is not that your point is inherently wrong. I'm not saying "all cameras with higher resolution will have less noise." I agree that that *can* be the case. What I'm saying is that it doesn't *HAVE* to be the case. Sure, there are examples where this is true, I don't dispute it (i.e. you don't have to post examples to demonstrate it - I accept it as fact.) But just because a sensor has more photosite packed into the same area doesn't mean it *MUST* have more noise. It means that it will *if* they leave everything else the same. It just means that the designers of the camera will have to work to prevent it.


Why can't, for example, sensor design engineers, find a way to better isolate the photosites so that there is less signal bleed between them? This would reduce the noise. Now, that require space - so eventually you'll pack so many that it won't work... until they come up with a better way to isolate the photosites from each other. Technology marches on. What was once difficult becomes easy. Noise will get worse as they pack more photosites on a sensor, until they find a way to reduce the effects of them being packed together. It only seems logical to me.

Eric
eric s is offline   Reply With Quote
Old Jan 30, 2007, 3:59 PM   #23
Senior Member
 
Join Date: Aug 2006
Posts: 822
Default

A couple points.

1. With point and shoot cameras, with 1/2.5" and 1/1.8" sensors, we've already reached the point where more pixels makes no difference at all, even in the largest possible prints. This is because they have already reached the point where they are diffraction limited by around f2.8, the brightest their lenses get. There are some estimates involved here, and it can be debated at exactly what point this occurs, but at the very least we're close to that point now. Shooting these cameras at a typical apperture of f5.6, it's almost physically impossible for them to sharper with 10 MP than the same camera would be with 8MP.

2. There is a tradeoff between pixel density and noise at high ISO, although it's not anything nearly as dramatic as the 1:1 ratio suggested by Corpsy. Moreover, there is an inverse tradoff with resolution, which does bring up the one major exception to my point #1 above. Packing more photosites onto the same size sensor does increase "noise", which becomes visable at higher ISO (for DSLRs, particularly ISO 800, 1600, 3200). But packing more pixels into the same size sensor can also increase resolution. This is seperate and apart from point one above. This is because more sensors in a smaller area is neede to reduce moire, or aliasing. Therefore, this may allow the use of less agressive anti-alias filters. That means sharper photos. So if the camera manufacturer uses less agressive anti-aliasing in conjuction with greater pixel density, you will get sharper photos, regardless of megapixels or sensor size. This will be true right up until the point where pixel density is sufficient that anti-aliasing filters are not needed.

4. Whether you want better quality at ISO 1600 or higher reslution at the expense of higher noise, really depends on what you are shooting. If you are a sports photographer, or shooting alot of low light without flash, you want the higher ISO ability rather than the ultimate in sharpness. A landscape photographer might opt instead for more resolution, and never use anything over ISO 800.

5. Anything digital offering with the Leica brand is apt to put the emphasis a bit more on resolution. You can see this even looking at the Panasonic point and shoots with Leica branded lenses. They don't measure up to some others in noise at higher ISO (some even showing too much at ISO 200). But compare the review test shots in good daylight, the landscape type shots. They're killing Canon, Sony, Fuji, and everyone else in sharp jpeg detail out of camera. A good part is obviously those lenses. I'm not really sure if lss agressive anti-alias filtering also plays a role in that case, but it might.

6. For DSLRs, there still is a use for more MP for very large prints and cropping. But very large means very large. A typical high quality magazine photo is printed at 180 dots per inch, and that's as much detail as most people's eyes can actually resolve. A 6MP image can deliver that level of reolution at greater than 15"x10". See this for info on print sizes and megapixels:

http://www.cambridgeincolour.com/tut...mera-pixel.htm


kenbalbari is offline   Reply With Quote
Old Jan 30, 2007, 5:26 PM   #24
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

Ken,
I generally agree with all you said, just one sorta question/point.

You mention "typical high quality magazine photo is printed at 180 dots per inch". My definition of high quality magazine photo is something you'd see in the big glossy magazines. Like Elle, Lenswork, or Architectural Digest - the Big Glossy magazines. And I expect those have gotta want more than 180 dpi for their pictures.

But it's really a definition of what is "a high quality magazine." I do generally agree that many magazines don't use really fancy printing techniques (just too darn expensive) so you can get away with pictures from 6-8MP cameras.

Eric
eric s is offline   Reply With Quote
Old Jan 30, 2007, 6:36 PM   #25
Senior Member
 
Corpsy's Avatar
 
Join Date: Dec 2005
Posts: 879
Default

Eric, thanks for posting the examples. Because these were likely JPGs direct from the camera is difficult to ascertain an accurate measure of noise levels due to differences in in-camera processing, and usually the differences are more noticeable at ISO settings higher than 400.

In this case, it would appear the noise levels are comparable between the two. They don't look the same, but I can't really say if one looks any better than the other (except that the 20D appears to have better color accuracy) However, the 20D photo was a 2 second exposure and the 10D was a 6 second exposure, and since noise accumulates over time I'd say the 10D performed quite admirably to have such similar noise levels.

For magazine printing, megapixel helps (shooting with the Canon 5D was very nice for magazines and posters), but pixel accuracy is just as important. Most magazines that I've submitted ad images to won't accept an image less than 300dpi though, so enlarging is often necessary.

I brought my 8mp Panasonic FZ30 into work once to compare directly to our 6mp Canon Digital Rebel. I shot both at their lowest ISO setting on a tripod under studio lighting. While the FZ30 photos looked sharper and held a bit more detail, the Rebel was much more accurate and had a far better signal to noise ratio. This meant that if enlarged, the Rebel's photos still looked very smooth and could be greatly enhanced with a moderate amount of sharpening. The FZ30 photos however could not be sharpened much at all because all the noise that wasn't very noticeable at first was brought to the surface by the enlarging and sharpening and made for some grungy looking images.
Corpsy is offline   Reply With Quote
Old Jan 30, 2007, 8:03 PM   #26
Senior Member
 
Join Date: Aug 2006
Posts: 822
Default

Quote:
You mention "typical high quality magazine photo is printed at 180 dots per inch".
Well, I'm not sure where I got that from, but it does appear to be wrong. I think I read something somewhere that claimed photos in National Geographic were really only 180 dpi, but I can't find it, and everything I can find right now suggests 300 dpi as the standard for highest quality print. So that's probably more like it.

A better point is probably that whatever the reolution is in theory, you'd best check the results of the actual camera you're looking at. There are just too many other factors, from the lens, to the low pass filter, to the processing, that have as much an impact as megapixels (or more).

But, but the 300 dpi standard, that would mean that a 7.2 MP camera was good up to an 8"x10" print.
kenbalbari is offline   Reply With Quote
Old Jan 30, 2007, 11:41 PM   #27
Junior Member
 
Graphico's Avatar
 
Join Date: Jan 2007
Posts: 13
Default

Thanks for your help, I think I've decided on this camera.
http://www.dcviews.com/press/Hasselblad-503CWD.htm
Graphico is offline   Reply With Quote
Old Jan 31, 2007, 4:27 PM   #28
Senior Member
 
JDar's Avatar
 
Join Date: Jan 2007
Posts: 956
Default

That's a great choice. May be a little heavy, but it's probably adequate.<g>
JDar is offline   Reply With Quote
Old Jan 31, 2007, 6:48 PM   #29
Senior Member
 
Join Date: Aug 2006
Posts: 822
Default

Quick answer:
The camera is 16.6 MP, a 36.7mm x 36.7mm sensor, 4080 x 4080 pixels.

Using the 300 PPI rule, you'd be good to about 13.5" x 13.5" [4080/300=13.6]

Reality is the large pixel size will begin to limit reolution a bit sooner, at about 10" x 10".

You have a pixel pitch or about 9 microns (micrometers).

[36.7*36.7=1346 sqr mm
1346/16.6 MP = 81.14 sqr micrometers/pixel
sqrt(81.4)=9.02 micons pixel size]

At that size, the sensor will be limited to about 56 line pairs per mm resolution [1/(9*2/1000)=55.56]. The low pass filter will likely limit reolution to this point to prevent moire.

The human eye can resolve up to about 8 lp/mm. Therefore you an magnify up to about 56/8= 7 X and still have maximum visible resolution. Your sensor measures 1.44 inches (36.7mm/25.4). So you can enlarge up to about 10" x 10" (1.44*7) with no perceptible loss of resolution.

In reality it will still look sharp at around 16"x16" --at about 5 lp/mm-- but you would begin to notice a difference at this point from resolution achievable with medium format film using the highest quality film and lenses. At typical viewing distances for such large prints, however, you could go even quite a bit larger and still have acceptable quality.

The large pixel pitch also means very low noise. So, even if reolution is a bit less than you could get with a larger film format, the result will still be very smooth and clean. It won't get grainy as quickly as film does when enlarged.

So as long as you know the limits--you're still not going to match larger format film--you're good to go.

http://luminous-landscape.com/tutorials/sharpness.shtml

kenbalbari is offline   Reply With Quote
Old Jan 31, 2007, 8:38 PM   #30
Junior Member
 
Graphico's Avatar
 
Join Date: Jan 2007
Posts: 13
Default

After reading this, http://kenrockwell.com/tech/mpmyth.htm and this http://pogue.blogs.nytimes.com/2006/...ogues-posts-2/ I've changed my mind.
Graphico is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 7:59 AM.