fredmiranda.com

Moderated by: Fred Miranda

New fredmiranda.com Mobile Site

New Feature: SMS Notification alert

New Feature: Buy & Sell Watchlist

FM Forums | Nikon Forum | Join Upload & Sell

1 2 3

5 6 7 end

Archive 2008 · D3 v 1Ds3 at ISO 12800

gfiksel
Offline
• • • •
Upload & Sell: Off

p.4 #1 · D3 v 1Ds3 at ISO 12800

andrewd01 wrote:
Just found a great web site to explain the diffraction issue. http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm
It has a calculator so you can work out the aperture at which the sensor becomes diffraction limited (assuming perfect lens). It is clear why 11-12MP is a sweet spot for full frame cameras (1Ds, 5D, D3), these can shoot down to f11 without being diffraction limited.

The 1Ds3 can only shoot down to f8 before becoming diffraction limited. This explains why it is a great camera for studio work, but less than perfect if you like to shoot landscapes with the hyperfocal technique, like this excellent image: https://www.fredmiranda.com/forum/topic/675528

For...Show more →

You probably checked the "Set Circle of Confusion = Twice Pixel Size" option. That is for screen viewing. For most of the print sized it's not diffraction limited but CoC limited.

Edited on Sep 02, 2008 at 07:38 AM

Sep 02, 2008 at 07:38 AM

lou f
Offline
• • • • •
Upload & Sell: Off

p.4 #2 · D3 v 1Ds3 at ISO 12800

richard, there is only one way to solve this, buy a d3, buy two zeiss 100 f2, take the shots side by side at the same time, do your pp, print off A3 prints for everybody and post them off, then set up a vote at the top of the thread ;o)

Sep 02, 2008 at 07:40 AM

brainiac
Offline
[X]

p.4 #3 · D3 v 1Ds3 at ISO 12800

louis fusco wrote:
richard, there is only one way to solve this, buy a d3, buy two zeiss 100 f2, take the shots side by side at the same time, do your pp, print off A3 prints for everybody and post them off, then set up a vote at the top of the thread ;o)

Hey - why settle for A3? I'll use an Epson 9800 to send you all a pair of 44x65"!

Sep 02, 2008 at 07:47 AM

firewireguy
Offline
• • •
Upload & Sell: Off

p.4 #4 · D3 v 1Ds3 at ISO 12800

brainiac wrote:
Nearest-neighbour and bilinear only give an accurate uprezzing if you multiply the baselength by an integer. A scaling form 12 to 21 megapixels is not multiplying the baselength by an integer.

An integer is just a whole number, 21-12=9, that's an integer.

brainiac wrote:
It is not an arbitrary choice of software algorithm, it is the most accurate way to uprez that mathematics can provide. It has practically no effect on the file's detail.
It's besides the point that it's "the most accurate" method, what' you're doing is creating new data from existing data, there's no way to do that "accurately" as it's a subjective thing. So all you can say is it produces the most pleasing results under most conditions. But that doesn't mean it can't be bettered, it's not a hard limit of maths. At the end of the day it's still an arbitrary choice that has NOTHING to do with the cameras under "test".

EDIT: It would appear that the Lanczos function offers better quality interpolation than bicubic, so you could use that, but then you'd have to decide on an arbitrary number of passes to use.

brainiac wrote:
Which camera stands to suffer more by these comparisons, the one which undergoes a practically lossless uprezzing, or the one which undergoes the subtraction of 9 megapixels of detail?

The ones that loses 9MP of noise I would have thought.

Edited by firewireguy on Sep 02, 2008 at 01:11 PM GMT

Edited on Sep 02, 2008 at 08:11 AM

Sep 02, 2008 at 07:47 AM

Kit Laughlin
Offline
• • • •
Upload & Sell: Off

p.4 #5 · D3 v 1Ds3 at ISO 12800

Richard,

You use the word "scientific" as though it explains everything, or something. As someone who has a Master's degree in science, plus five year's fully funded Ph.D. research into the relationship between multiple causes in complex systems (in logic and environmental science), I can assert with some authority that claiming 'scientific' as an authority for any position is baseless. Assumptions need to be stated, methodologies explained (and weighed against competing methodologies), and so on, to even *begin* this process.

Rudolpho wrote:

You have not even attempted to argue, discuss, or dialogue about WHY you think noise-per-image is the correct metric over noise-per-pixel or noise-per-inch. You have simply stated it, and repeated it, and added that you're right but some people don't understand it. That line of argument has zero... repeat, zero... validity.

That seems like a reasonable proposition; at least there is something to argue about. Personally, I think your reasoning, and your test, is fundamentally flawed. Nothing personal; just my opinion.

and your quote:

"Any effectively generated theory capable of expressing elementary arithmetic cannot be both consistent and complete. In particular, for any consistent, effectively generated formal theory that proves certain basic arithmetic truths, there is an arithmetical statement that is true, but not provable in the theory."

is a pr�cis of G�del's Incompleteness Theorem; interesting, but not relevant here, IMHO. And, crucially, it is a *theorem*: neither true nor false, but an assertion that admits of testing. The validity of this theorem lies in its structure (thinking otherwise is a basic error in logic); its truth or falsity may be tested; the testing lies *outside* the domain of the theorem itself.

Very happy to discuss, from first principles. That's why we are here. Cheers, KL

Edited by Kit Laughlin on Sep 03, 2008 at 01:07 AM GMT

Edited by Kit Laughlin on Sep 03, 2008 at 01:12 AM GMT

Edited on Sep 02, 2008 at 08:12 AM

Sep 02, 2008 at 08:03 AM

nikt
Offline
• • • • •
Upload & Sell: Off

p.4 #6 · D3 v 1Ds3 at ISO 12800

Ok, I'll give you that; you did say 'maybe' the D3 is the prince.

Tell me though Rich, the title "1DS MK III king of high ISO"...... I can read it slowly if you like. Maybe I didn't understand it.

Sep 02, 2008 at 08:07 AM

brainiac
Offline
[X]

p.4 #7 · D3 v 1Ds3 at ISO 12800

nikt wrote:
Tell me though Rich, the title "1DS MK III king of high ISO"...... I can read it slowly if you like. Maybe I didn't understand it.

Apologies - I left out the question mark, although the text of the first post makes it explicit that I don't know the answer to that question. Thanks for spotting it - I have put in the question mark now, although I don't think it's the only sensationalist thread title on FM.

Sep 02, 2008 at 08:19 AM

sjms
Offline
• • • • • • •
Upload & Sell: On

p.4 #8 · D3 v 1Ds3 at ISO 12800

oh no far from it. but with that question mark it changes the entire aspect and attitude of the thread.

Sep 02, 2008 at 08:26 AM

brainiac
Offline
[X]

p.4 #9 · D3 v 1Ds3 at ISO 12800

Kit Laughlin wrote:
You use the word "scientific" as though it explains everything, or something. As someone who has a Master's degree in science, plus five year's fully funded Ph.D. research into the relationship between multiple causes in complex systems (in logic and environmental science), I can assert with some authority that claiming 'scientific' as an authority for any position is baseless. Assumptions need to be stated, methodologies explained (and weighed against competing methodologies), and so on, to even *begin* this process.

Rudolpho wrote:

That seems like a reasonable proposition; at least there is something to argue about. Personally, I think your reasoning,...Show more →

I used the word scientific to denote lack of bias in a comparison. I do not mean lack of bias in the observer, but lack of bias in the test itself. An unscientific test of the speed of two 100 yard runners would be to give one a 30 yard head start, and declare the first one over the finish line to be the faster. That is precisely what is happening when you compare per pixel noise in 100% crops from a 21 megapixel image and a 12 megapixel image, and therefore the comparison is unscientific. It is biased towards the lower megapixel file, because you are examining per pixel noise from a less magnified (i.e. physically smaller) image. When you come to printing the image from both cameras at any given size, the pixels from the camera with more pixels will be smaller than those from the other. For instance, for a 20x30 cm print, each 1Ds3 pixel fills 0.00285 square millimetres, whereas each D700 pixel fills 0.00497 square millimetres. If you compare pixels from the two cameras in 100% crops you are failing to take into account that for any given print, the pixels from the 12 megapixel crop would be considerably bigger. The more you enlarge noise, the more disruptive it appears to be, so to view crops at the same enlargement (i.e. 100%) you are handicapping the file which needs less enlargement.

Why is per image noise more important than per pixel noise? Because Cameras are for making images, not for recording and excising fixed numbers of pixels. Per pixel noise is quantifiably 4/7 as important in a 21 megapixel file than in a 12 megapixel file. That is, a D3 pixel needs to suffer from 4/7 times the noise of a 1Ds3 pixel in order for the picture itself to look as noise-free.

That's why comparison of 100% crops from two cameras always handicap the higher resolving camera, precisely in proportion to the ratio of megapixels between the two cameras.

Edited by brainiac on Sep 02, 2008 at 03:21 PM GMT

Edited on Sep 02, 2008 at 10:21 AM

Sep 02, 2008 at 08:51 AM

brainiac
Offline
[X]

p.4 #10 · D3 v 1Ds3 at ISO 12800

sjms wrote:
oh no far from it. but with that question mark it changes the entire aspect and attitude of the thread.

Not really, bearing in mind that the first post explicitly states that the answer is unknown. How many threads entire aspect and attitude is defined by the title? Almost none, as far as I can see.

Sep 02, 2008 at 08:53 AM

sjms
Offline
• • • • • • •
Upload & Sell: On

p.4 #11 · D3 v 1Ds3 at ISO 12800

yours did it originally. it came out as a statement not a question. thus setting a completely different un deniable tone. someone with your education should know that after writting so many papers to get your degree.

in the long run it doesn't matter to me because i have reduced my work load gotten a higher in focus rate using the nikon then any canon i have owned to this day. i have needed considerably less agressive post NR at high iso values. you keep atacking the issue you have created from one (your) side and of course the owners of that side are going to believe and side and see what you see. after 5 years with the canon product i'm afraid i no longer do. at the moment i have given up on the massive pixel count for my snapshots. maybe next time. i like shooting pictures and easily getting them. i'm back to doing that now. i have higher productivity when i'm happier.

Edited on Sep 02, 2008 at 09:21 AM

Sep 02, 2008 at 09:20 AM

andrewd01
Offline
• • •
Upload & Sell: Off

p.4 #12 · D3 v 1Ds3 at ISO 12800

gfiksel wrote:
You probably checked the "Set Circle of Confusion = Twice Pixel Size" option. That is for screen viewing. For most of the print sized it's not diffraction limited but CoC limited.

Yes, thanks for clarifying, so for most prints at reasonable viewing distance, the eye cannot resolve the detail in the high resolution image (therefore no value in high MP). You need to view the print at an unusually close distance or on screen at 100% to be diffraction limited. In either case I see no strong case for high MP except for studio cameras that need to make images for billboards.

If per-pixel noise was the same between high and low res cameras you could argue a case for high res cameras to allow more cropping, but we know that per-pixel noise is a function of pixel size (in a given vintage of sensor).

cheers

Sep 02, 2008 at 02:43 PM

Kit Laughlin
Offline
• • • •
Upload & Sell: Off

p.4 #13 · D3 v 1Ds3 at ISO 12800

Richard,

That is precisely what is happening when you compare per pixel noise in 100% crops from a 21 megapixel image and a 12 megapixel image, and therefore the comparison is unscientific.

Sorry, no. To assert this is to 'beg the question': a descriptor of one of the many fallacies in logic. This one means that you are taking for granted the very matter that the experiment is designed to reveal understanding of. Further, to label this comparison as unscientific simply does not explain, or add any new information to, the original assertion. The conjunctive term "therefore" is entailment relation, as well: it says that the relation is a necessary one, and the only one. This is not the case here. There are many possibilities and many positions, some of which have been canvassed here. Many are just opinions�again, no problem, provided they are labelled so.

We all have opinions; the ancient Greeks set the agenda many years ago, when they said the purpose of philosophy was to distinguish between 'doxa' (opinion) and 'episteme' (truth). This is stil the case. The point about opinions is that everyone has one; we want to know why we should value one over another. I am interested in your opinion here, but feel that in no way have you demonstrated that yours is to be values because of QPR. I hope I am clear in saying this.

You may have the basis of an assertion that might be tested ("a comparison of XYZ will reveal that the smaller resolution camera's 100% crops will be sharper", for example, and, personally, I find this interesting), but to assert, as you have been doing, that to compare in this fashion is unscientific is the very point that is yet to be demonstrated, from what you have written.

I am not commenting about the possible results of your tests�only on your way of trying to dismiss those who do not see the situation the way you do, and the use of the term "unscientific" to dismiss that same group.

Not a problem, but if you are appealing to a 'higher authority' (in this case science, whatever that might mean) to support your position, it's good to get it right.

Edited on Sep 02, 2008 at 05:21 PM

Sep 02, 2008 at 05:09 PM

Rodolfo Paiz
Offline
• • • • • •
Upload & Sell: On

p.4 #14 · D3 v 1Ds3 at ISO 12800

brainiac wrote:
I will also show crops of the 21 megapixel file downrezzed to 12. That certainly does have an effect on the file's detail, namely it trashes 9 megapixels of it. Which camera stands to suffer more by these comparisons, the one which undergoes a practically lossless uprezzing, or the one which undergoes the subtraction of 9 megapixels of detail?

If ever you had a hope of showing unbiased objectivity, this little tidbit destroys it utterly. Claim that you want to test your views and beliefs, and I'll accept that. But claim that your view (per-pixel noise is a false deity) is the one true voice of science, and that everybody is else is missing the point, and you get called on the carpet and asked to prove your point. You simply cannot write a rhetorical question comparing "practically lossless uprezzing" with "subtraction of 9MP of detail" and expect anyone to take you seriously! If you lean your "scientific" self any further over to one side of the argument, you'll fall over.

I'm going to bow out of this discussion, for two reasons:

1. I have now understood the testing methodology you propose to use, and I believe to be a biased method which will skew the results in favor of the Canon. In every actual image I've ever worked with, reduction in size reduced or eliminated noise, or made its appearance more tolerable and pleasing to the eye. Magnifying the image or printing larger at lower resolution made noise worse, and more noticeable.
So for any given print size and resolution (say, 20"x30" @ 300dpi) the 1DsIII will require less magnification to reach the required pixel dimensions and will thus be favored in the test. Similarly, reducing the 21MP file to 12MP will eliminate a great deal of whatever noise was present, again favoring the Canon.

2. I have also now understood the key assumption on which you wish to base your test (per-image noise is all that matters, and so on), and I believe that assumption to be false. I believe that comparing actual pixels is a far more valid method of comparison (e.g. print at 300 dpi without any image resizing, then compare print quality) than what you propose.

In the end, I do hope you carry out those tests. Despite what appear to me as glaring flaws in the assumptions and methodologies you plan to use, I'd still be interested in seeing the tests and maybe being able to learn something from them. But until and unless those tests are done and I can see the results, I have lost interest in continuing this discussion. Honestly but bluntly, you strike me as an intelligent, highly-educated individual who is bent on proving that he's right. And as far as I am concerned, you are welcome to do so... but you do have to prove that you are right... quoting G�del and verbal sleight-of-hand will get you nowhere. I shall await the proof, and see what conclusions the tests allow me to draw.

Good luck, and Godspeed.

Edited on Sep 02, 2008 at 07:06 PM

Sep 02, 2008 at 06:59 PM

gfiksel
Offline
• • • •
Upload & Sell: Off

p.4 #15 · D3 v 1Ds3 at ISO 12800

Rodolfo Paiz wrote:
So for any given print size and resolution (say, 20"x30" @ 300dpi) the 1DsIII will require less magnification to reach the required pixel dimensions and will thus be favored in the test.

I don't get it. Isn't it what we are supposed to do - look at the final print? Suppose, the test has established that for ... any given print size... the print from one camera has less visible noise, will you still be arguing that the other camera has less noise

I'm actually appalled by the animosity this offer has been received. Richard is offering to make a test using the same subject under the same condition. Thus, two cameras are needed at the same location. The prime most thing in any test is the quality of primary data. That's what Richard is offering to provide. Everybody else can then analyze it to death by any way they want it. I would still be interested to know the results. I don't have any allegiance, I'm just interested in the results. Why is such an animocity. Is everybody so insecure?

Sep 02, 2008 at 07:54 PM

surf monkey
Offline
• • • •
Upload & Sell: Off

p.4 #16 · D3 v 1Ds3 at ISO 12800

Sorry if this is rehash, ignore it if you have addressed this.

Rodolfo wrote:
(e.g. print at 300 dpi without any image resizing, then compare print quality)
This makes some sense to me, printing at 300DPI.

But, unless I missed something, shouldn't you resize the images to your print size at the correct DPI for your printer. Resizing to either a 12MP or 21MP file first doesn't seem to make sense to me.

I would pick a couple "popular" print sizes (determine that as you will) and take the "unprocessed" files from the two cameras, resize them for each print size and away we go.

Sep 02, 2008 at 08:23 PM

brainiac
Offline
[X]

p.4 #17 · D3 v 1Ds3 at ISO 12800

Rodolfo Paiz wrote:
So for any given print size and resolution (say, 20"x30" @ 300dpi) the 1DsIII will require less magnification to reach the required pixel dimensions and will thus be favored in the test.

The error is in comparing 100% crops without resizing to the same magnification. Comparing prints at a given size is not an error. When 100% crops are compared without resizing, the 1Ds3 file is being examined at a higher magnification, for per pixel quality. That favours the less magnified image, i.e. the D700.

Edited on Sep 03, 2008 at 06:04 AM

Sep 03, 2008 at 06:03 AM

firewireguy
Offline
• • •
Upload & Sell: Off

p.4 #18 · D3 v 1Ds3 at ISO 12800

Perhaps you're going to be better off putting your suggestions to the people at DPR that do the reviews and get their opinion. Perhaps they'll change their tests.

Sep 03, 2008 at 06:31 AM

brainiac
Offline
[X]

p.4 #19 · D3 v 1Ds3 at ISO 12800

firewireguy wrote:
Perhaps you're going to be better off putting your suggestions to the people at DPR that do the reviews and get their opinion. Perhaps they'll change their tests.

I emailed them about this a few months ago and had no response at all.

Sep 03, 2008 at 06:39 AM

panos.v
Offline
• • • •
Upload & Sell: Off

p.4 #20 · D3 v 1Ds3 at ISO 12800

I find this thread just amazing. A guy asks whether someone with a D3 is available in London so that they can go to the pub and try some photo geekery and we now have 4 pages of comments from people outside London, actually oustide this whole damn country bar 1 poster, who attack the methodologies, motives, intelligence or anything else of the OP.

Do you guys really try to turn this into a trolling post? I didn't see anything offensive or trollish in the original question and I didn't see any point in any of your replys when only a simple "yes, I can do next Saturday at 6pm" would be relevant. I've seen brainiac's other threads but that doesn't mean he might no be entitled to asking a D3 user for a comparison.

Who cares how it's done. It's for his own pleasure and test. And sounds like much better than the standard brick wall (TM) shot.

Edited on Sep 03, 2008 at 06:44 AM

Sep 03, 2008 at 06:44 AM

1 2 3 4 5 6 7 end
FM Forums \| Nikon Forum \| Join Upload & Sell 1 2 3 4 5 6 7 end