brainiac wrote:
Hang on a minute, you are expecting a kind of Moore's law for cameras. There isn't one. Just because there have been occasional leaps, that doesn't mean that there is no value in incremental upgrades. The 1Ds3 and the 40D weren't targeted only at 1Ds2 and 30D owners. They are often bought by people who don't have that kind of camera yet. Should Canon hold back the technologies that went into the 1Ds3 until all 1Ds2 users are ready to upgrade? Upgrade when you feel like it, and be glad the manufacturers continue the 'evolution' of their products with regular updates....Show more →
Evolution is probably the right term, as it implies change and improvements proceeding at a snail's pace, which fits my examples!
BRAVO Brianiac!! Now, using the idea that the best equipment is only barely good enough I have continued to upgrade my capture, electronic lab, printing equipment, framing/ mounting resources and. of course, skill resources. I do want to observe that this forum has helped me bring my skills along and continues to do so with those such as Brianiac who's agenda is improvement and education-not argument.
I cannot display the result of one of my presentation 24x36" prints on your screens but suffice to say many people have never seen anything like it (if their reaction is to be believed). I also am looking forward to the near future when large, electronic wall mounted displays for "Stills" become common (and no I am not talking about a vertically mounted Plasma TV). So yes More really is More.
How about printing an 8X10 from each camera. Crop each file to 1/4 it's original size (do not change resolution) and You'll see how it will look as a 16X20.
Isn't that what really matters? Images from almost any modern DSLR will look great when viewed on screen, but what really matters is the print. If you don't print large or crop heavily then the differences become less important.
It is VERY IMPORTANT TO REMEMBER THAT WHEN YOU EXAMINE 100% CROPS FROM A 21 MEGAPIXEL CAMERA YOU ARE VIEWING THE IMAGE AT SIGNIFICANTLY HIGHER MAGNIFICATION THAN YOU ARE USED TO. This means that although the lens defects and noise are more visible in 100% crops, the print is still better than you were getting from your 12 megapixel camera. A 100% crop of a 21 megapixel file is a much higher level of scrutiny. The per pixel noise doesn't need to be as good as that in a 12 megapixel file in order for the file to equal or surpass the lesser file's noise footprint. ...Show more →
If you are comparing full frame bodies, what you say is true. However all of Canon's current APS-C crop bodies (40D, 400D, 450D and 1000D) have a higher magnification than the 5D2. If you put the same lens on a 40D and a 5D2 it will look slightly sharper in the center of the 5D2 frame.
lol
100% crop of 12mp should not be any sharper than 100% crop of a 21mp. This is the reason why reviewers do this so call "mistake" I don't want to buy a 21mp camera so that it would only be as sharp as 5D at 12mp. I want it to be just as sharp at 100% crops as well. Otherwise, why not just re sample the 5D images via photoshop!!
OP,s view doesn't make any sense
Brainiac, I take your excellent point and thank you for the info. The only issue I fail to get my head around is the "uselessness" of more pixels. Let's say we compare 2 shots from 5D and 5DII, 21Mp rather than 12 Mp, without cropping and that we print on A4 paper. The 21Mp picture is going to mean many more smaller pixels than the fewer larger ones from the 12Mp sensor, irrespective of print size. If it is meaningless, it means that either paper, printer, or both, are incapable of matching any resolution higher than 12Mp in A4 size. So we should be working hard at finding either better paper and/or better printers. Which we don't seem to be busy doing.
Conversely if more pixels make a difference, it should be visible in transition zones between volumes, objects and colours, which should be more "brutal" on the 12Mp sensor, and more delicately progressive and shaded of the 21Mp. Again, if it were meaningless, it would mean that a print off a 1DIII should be exactly the same as that off a 1DsIII, or a 5D and a 5DII unless you crop massively or print very large, not considering noise, ISO and DR.
Can you help me understand this better?
And thanks agains for the excellent thread.
Sep 19, 2008 at 12:45 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
brainiac wrote:
Well, I correctly predicted, in November last year, that the 1Ds3 would have a comparable noise performance (providing you want to make pictures, not crops) to the D3. I am going to be bold and make another prediction here: the 5D2 iso 12800 noise performance, optimally processed from raw, will look pretty much the same as that from the 1Ds3. Why?
(1) because Canon likely made some effort with the 1Ds3 sensor, so tweaks are unlikely to radically transform its performance, and
(2) sensor design and fabrication is a very expensive business in which costs are acutely sensitive to volume, so Canon would be foolish not to aim to use its latest sensor in more than one model at different price points, if it can get away with it, and
(3) Canon has stated that the 5D2 sensor is derived from the 1Ds3 design, and
(4) no detailed explanation has been offered of how the 5D2 sensor differs from that in the 1Ds3, a ...Show more →
Hi Richard,
Interesting analysis, but have you seen this description of the sensor from Rob Galbraith's web site:
"With a size of 24 x 36mm, 21.03 million image pixels and a pixel pitch of 6.4µm square, the 5D Mark II's CMOS sensor offers the identical pixel count in a sensor that's identical in size to the full-frame EOS-1Ds Mark III. The 5D Mark II's sensor also features the same light-gathering area within each pixel (called the fill ratio) and same microlens coverage over each pixel as the company's current flagship.
To achieve what Canon is saying will be the highest image quality and lowest noise ever to emerge from a Canon EOS digital SLR, they introduced several refinements: the array of red, green and blue coloured filters over the sensor have been made more transmissive, which effectively bumps up the sensor's light sensitivity, plus they tweaked the way the sensor's signal (the light it has gathered during the exposure) is amplified and then read out.
Canon is showing some serious confidence in the 5D Mark II's image sensor, and in particular its noise characteristics: they've given the camera an extended ISO range of 50-25,600, which is a three stop jump from the ISO 3200 upper limit of the EOS-1Ds Mark III (and the 5D as well).
The 5D Mark II incorporates Canon's Integrated Cleaning System in front of the sensor, but with a new anti-stick fluorine coating on the frontmost filter surface that's meant to better prevent slightly moist debris from clinging on, or for it to be more readily shaken off during a cleaning cycle."
If they made the filters more transmissive, then there might be real improvement in ISO performance. I think we will have to wait for real side by side test to really know.
stockriderman wrote:
lol
100% crop of 12mp should not be any sharper than 100% crop of a 21mp. This is the reason why reviewers do this so call "mistake" I don't want to buy a 21mp camera so that it would only be as sharp as 5D at 12mp. I want it to be just as sharp at 100% crops as well. Otherwise, why not just re sample the 5D images via photoshop!!
OP,s view doesn't make any sense
You're failing to account for the fact that there is a middle ground between perfect sharpness at 100% on a 12 MP sensor and perfect sharpness at 100% on a 21 MP sensor. One of the things brainiac says is that a lens may out-resolve a 12 MP sensor but not a 21 MP sensor.
Interesting analysis, but have you seen this description of the sensor from Rob Galbraith's web site:
"With a size of 24 x 36mm, 21.03 million image pixels and a pixel pitch of 6.4µm square, the 5D Mark II's CMOS sensor offers the identical pixel count in a sensor that's identical in size to the full-frame EOS-1Ds Mark III. The 5D Mark II's sensor also features the same light-gathering area within each pixel (called the fill ratio) and same microlens coverage over each pixel as the company's current flagship.
To achieve what Canon is saying will be the highest image quality and lowest noise ever to emerge from a Canon EOS digital SLR, they introduced several refinements: the array of red, green and blue coloured filters over the sensor have been made more transmissive, which effectively bumps up the sensor's light sensitivity, plus they tweaked the way the sensor's signal (the light it has gathered during the exposure) is amplified and then read out.
Canon is showing some serious confidence in the 5D Mark II's image sensor, and in particular its noise characteristics: they've given the camera an extended ISO range of 50-25,600, which is a three stop jump from the ISO 3200 upper limit of the EOS-1Ds Mark III (and the 5D as well).
The 5D Mark II incorporates Canon's Integrated Cleaning System in front of the sensor, but with a new anti-stick fluorine coating on the frontmost filter surface that's meant to better prevent slightly moist debris from clinging on, or for it to be more readily shaken off during a cleaning cycle."
If they made the filters more transmissive, then there might be real improvement in ISO performance. I think we will have to wait for real side by side test to really know....Show more →
That's interesting. IIRC, one of the delays with implementing the new Bayer sensor pattern by Kodak, involving the use of some clear pixels, is figuring out the edges of colored regions; per-color resolution at a given total sensor resolution obviously decreases when you make some pixels clear. However, the new push for high resolutions makes that less and less of a worry. Here's hoping for clear pixels someday soon-- and maybe even a black and white sensor.
Uprezzing creates (read: fabricates, interpolates, whatever) pixels that weren't there before. A straight 21MP file from the 5D2 has actual data in all of those pixels. You can't compare the two.
It's more valid thant downresing the 21 MP file, throwing away information.
jvarszegi wrote:
It's more valid thant downresing the 21 MP file, throwing away information.
You're "throwing away" information but averaging real data to create the smaller size. With upresing you're "adding in" information that is an average of adjacent pixels, not real data. Slight difference... In the end though, neither are accurate representations.
DoubleNegative wrote:
You're "throwing away" information but averaging real data to create the smaller size. With upresing you're "adding in" information that is an average of adjacent pixels, not real data. Slight difference... In the end though, neither are accurate representations.
You're not really adding any information when you upres with an averaging method. You're not creating detail. It's like adding water to a sponge-- you can dry out the sponge, and do it all over again without changing the shape or size of the sponge.
Downresing actually discards detail that you can't get back; it's a destructive transformation.
I don't see how simple interpolation is somehow inaccurate. The image looks fuzzier blown up, just as it would in the film days.
There may not be much difference on SMALL prints (up to 8x10), but there is a big difference on large prints (16x20 and up). The larger the print, the more visible the difference.
stockriderman wrote:
100% crop of 12mp should not be any sharper than 100% crop of a 21mp. This is the reason why reviewers do this so call "mistake" I don't want to buy a 21mp camera so that it would only be as sharp as 5D at 12mp. I want it to be just as sharp at 100% crops as well. Otherwise, why not just re sample the 5D images via photoshop!!
OP,s view doesn't make any sense
Well, if your goal is to compare 100% crops, then yes, you are right. Brainiac's view doesnt make sense.
But for most people, the goal is prints. With any image, the goal is a certain sized print - be it 4x5, 16x20, 4'x6', billboard or whatever. When you compare 2 cameras, what you are trying to figure out is how well the prints from the 2 cameras would match up (at that chosen print size).
In such a case, it makes more sense to compare the 2 cameras with one image up-rezzed to the other (not down-rezzed - that destroys information and also leads to blindness and hairy palms, I think). This gives you a better idea of how the 2 cameras stack up.
It is a bloody shame that camera reviews today seem to be driven by the 100% crowd, most of whom dont appear to actually make large prints or actually use the camera in real-world conditions. It affects those few remaining people who actually seem to care about images and prints, b/c it forces manufacturers to re-allocate their priorities to appease the pixel-peeping vocal minority/majority.
I've already ranted about this in another thread, but I read these posts from people who need 100% weathersealing and wonder if they have actually ever shot in the rain - what they do with water droplets on the front element and viewfinder window, for example. And for those people who apparently shoot fast-developing action and at the same time want AF points on the edge of the image frame (!!!). My best-selling action shot to date was taken with an Elan IIE (3 AF points, 3fps).
Anyway, I digress. Coming back to the subject - I hope it makes sense now why 100% is not the best arbiter of IQ.
DoubleNegative wrote:
Actually, they did... Kodak is coming out with Ektar 100, due in October or November. Promises to be VERY good (better than Ektachrome E100G) and replaces the Ultra Color UC100/400 films.
Fuji I believe is also fully comitted to film, mentioning something recently.
My point is that before digital matured, it was understood that to get a major improvement in print quality you moved to a larger format. There were periodic improvements in film emulsions, but they were generally more subtle than some of the jumps in sensor technology that seem to show up about every two years. We can now get print quality out of a 24x36 mm ("FF") sensor that would have required a large format scanning back about 10 years ago (no hope of action stopping location exposures).
Regarding all the fuss over whether to compare cameras by uprezing the smaller file to match the larger - remember when the 4 MP 1D had been around for a couple of years and people would say "why bother with that when you can get 6 or 8 MP for less money?" Users of the 1D would reply that its images responded well to interpolation, and made clean large prints. (They still do, within reason). No one was complaining how unfair it was to upres the 1D to match a 20D or whatever. I see nothing spurious about upresing 5D or D3 images either. Make the prints if you have the chance and don't mind the small expense.
Also, I wouldn't form any negative ideas about 5DII sharpness until I'd seen well exposed raw files using great lenses and careful technique.
That was well written and discussed. BUT, are you suggesting that for some lenses we may get a better resolved image if we were to use sRAW v RAW with the 5DmkII?
Thank you so much for putting this much useful information. I too have a 1Ds MarkIII and distinctive remember all the forums talk about lack of sharpness without taking into consideration the 21MB image size.
Buying this new 20+ MB bodies is like buying a JET. If you think you are done expending the big money because you got the JET (Body) , you are wrong You just starting to spend money - You are now obligate to buy the best glass so you can justify the body. Point in case My 16-35L did decent on my 5D camera, but no-way I would use it on my 1Ds Mark III. I have several friends that had to get rid of many lenses that just don't cut it when used with the 1Ds Mark III, and that is attributed to the fact that many of the optics flaws on the current glass offerings are better noticed with this new superMB bodies.
I recently looked at side-by-side portraiture prints from a phase one and a D3 and, as one in the group rightly commented: "...the D3 print shows a perfect picture of hair. The Phase One print shows hair..."
Having said that, one of the D3 prints was 24"X36" and it held up pretty well on that size.
You just starting to spend money - You are now obligate to buy the best glass so you can justify the body. Point in case My 16-35L did decent on my 5D camera, but no-way I would use it on my 1Ds Mark III. I have several friends that had to get rid of many lenses that just don't cut it when used with the 1Ds Mark III, and that is attributed to the fact that many of the optics flaws on the current glass offerings are better noticed with this new superMB bodies.