... when I'm shooting Natasha Poly ...
(Tongue in cheek mode on...)
I am fed up with people criticising MF cameras and arguing nobody needs that many pixels. I had a look at fine art inkjet printers. These print on roll paper and come in width up to 60" (150cm). Using the standard 4x3 aspect ratio, they would print an image sized 60" x 80" (150cm x 200cm).
The printers have the same linear resolution than the smaller inkjets, hence they need to be fed the same ppi. Native ppi is something like 300-360 ppi.
300 ppi for 60" x 80" is 18000 x 24000 pixels, 432 million pixels. Anything less is a waste of a perfectly good printer.
Well... 432 million pixels on 24 x 36 mm implies pixels of 1.33 µm (cell-phone size...). 432 million pixels on 53.7 x 40.4 mm implies pixels of 2.24 µm. I am pretty sure that the second one gives a more usable aperture for the diffraction limit.If you put it like that then it's feasible. Samsung could do it today. I want the 400 mpix landscape version NX800E :-)
Well... 432 million pixels on 24 x 36 mm implies pixels of 1.33 µm (cell-phone size...). 432 million pixels on 53.7 x 40.4 mm implies pixels of 2.24 µm. I am pretty sure that the second one gives a more usable aperture for the diffraction limit.
The 2.24 microns would give you a 48.9 x 24.8" print on 720 PPI
The object of this thread is to discuss 60" x 80" prints (150cm x 200cm) on fine art inkjets. The printers exist, they are readily available and cost less than the average MF camera (e.g. Canon imagePROGRAF iPF9400 or Epson Stylus Pro 11880). How are we going to feed them with adequate data? There is no adequate camera on the market today.
OK: we can stitch. That is cumbersome: how many individual pictures do we need from a lowly camera like an IQ280 to feed the printer?
Well... 432 million pixels on 24 x 36 mm implies pixels of 1.33 µm (cell-phone size...). 432 million pixels on 53.7 x 40.4 mm implies pixels of 2.24 µm. I am pretty sure that the second one gives a more usable aperture for the diffraction limit.Only in that the 36x24mm format would require an f-stop of about f/2 or lower to control diffraction, while the 54x40mm could control diffraction equally at about f/3 and lower --- with each having the same pathetically shallow DOF when diffraction is equally controlled. So about 420 million of those pixels would be OOF and so wasted as far as resolution goes.
It doesn't depend on input pixels alone. It also has to do with the acuity of human vision, which is also depending on distance
Only in that the 36x24mm format would require an f-stop of about f/2 or lower to control diffraction, while the 54x40mm could control diffraction equally at about f/3 and lower --- with each having the same pathetically shallow DOF when diffraction is equally controlled. So about 420 million of those pixels would be OOF and so wasted as far as resolution goes.
The larger format would have the lens design advantage of needing less low f-stops. For example:
- an 80mm lens for 54x40mm format that is "430MP sharp" at f/3
is probably an easier lens design goal than
- a 50mm lens for 36x24mm format that is "430MP sharp" at f/2.
But do either of these lenses exist?
The iron trade-of between DOF and diffraction means that these massive pixel counts probably need focus stacking with anything but distant landscapes that have no foreground needing to be in focus.
And as Bernard will surely remind us, when you can do focus stacking, you can probably to stitching too, and the lenses have a better chance of giving adequate resolution at those low f-stops.
Hi,
What I have seen I would say 200 MP would be ideal on full frame 645.
At 6.8 microns I see a lot of aliasing that goes away at f/16. So for aliasing free rendition we would need 3.4 microns at f/8.
If we assume 54 sensor and 5/4 ratio we would have:
54 / 0.0034 * 43 / 0.0034 -> 200 MP
So, this would assume shooting at f/8.
Shooting wider apertures, with the best lenses, even more resolution would be useful.
Best regards
Erik
This can be obtained with 4*5 inch film
Nope. I want the public to be able to examine the print with a magnifier, mm per mm if needs be.
Why do you think Canon and Epson make printers capable of fine art print on 60" / 150 cm rolls?
The printers will sustain a linear resolution of at least 300 ppi on the full length of their bed. That is a feat of engineering and comes with a price tag, yet that price tag is smaller than your average MF camera. You want me to feed that miracle of technology with inferior data? Forget it.
The printers are there, I want to use them to their full potential. I want the public to watch the large picture from the distance, then move closer and closer and discover more and more details. Have you been to a gallery or museum presenting really large prints recently?
Hiroshi Sugimoto (http://en.wikipedia.org/wiki/Hiroshi_Sugimoto) believes he needs 8 * 10 inch film for his massive prints. They sure look impressive.
Only in that the 36x24mm format would require an f-stop of about f/2 or lower to control diffraction, while the 54x40mm could control diffraction equally at about f/3 and lower --- with each having the same pathetically shallow DOF when diffraction is equally controlled. So about 420 million of those pixels would be OOF and so wasted as far as resolution goes.
Others have used bigger cameras ..., contact prints are much better than enlargements ;)(https://gs1.wac.edgecastcdn.net/8019B6/data.tumblr.com/tumblr_mee72oWxPt1qz5q5oo1_1280.jpg)
I guess that the question may be a bit academic.
Not at all. The printers exist.
It has been said that Anders Gursky shoots with 50 MP MFD, and I got the impression that his prints are OK.
The object of this thread is to discuss 60" x 80" prints (150cm x 200cm) on fine art inkjets. The printers exist, they are readily available and cost less than the average MF camera (e.g. Canon imagePROGRAF iPF9400 or Epson Stylus Pro 11880). How are we going to feed them with adequate data? There is no adequate camera on the market today.
OK: we can stitch. That is cumbersome: how many individual pictures do we need from a lowly camera like an IQ280 to feed the printer?
As eronald noted, this post is satyre, but is based on truth. The printers exist. I believe that people like Andreas Gursky use them to great success. Actually, "Rhein II" is 190 cm × 360 cm (73" x 143") and is a chromogenic print (Lambda print), so it needs even more data. And there is a body of water right in the middle, which makes it extra difficult to stitch.
My opinion remains that the only thing impressive about it is the price at which it was sold. ;)
Only in that the 36x24mm format would require an f-stop of about f/2 or lower to control diffraction, while the 54x40mm could control diffraction equally at about f/3 and lower --- with each having the same pathetically shallow DOF when diffraction is equally controlled.
But before we officially revert to customary LuLa dialogue ...Your image is only 0.4MP. If a 430MP version were available, and printed large enough and viewed close enough as the OP envisions, the DOF would be vastly less: the perceived OOF effects and thus DOF (which parts of the image are detectably OOF vs which are not) depends on the apparent size of the image (size divided by viewing distance), not just features of the "file" itself. (P. S. DOF scales on lenses are an anachronistic, misleading abomination.)
I'll settle for the 400 million f/2 pathetically OOF .. sometimes the whole is greater than the sum of its parts
(http://3.bp.blogspot.com/-NQRpUtCKOj0/T8c7uxI8l0I/AAAAAAAAFH8/NQeH8Bdmcqw/s1600/Pirelli+Calendar+2008+2.jpg)
Absolutely, Bernard. I've done a 14 panel stitch that a client had printed 10' wide and it was beautiful. Stitching is just not a workflow that I'm personally interested in pursuing for the work I'm going to be doing.
100 ways to photograph a skinned cat.
I've been thinking a LOT about all of this lately, especially since I've just started making prints of my work again and thinking about exhibition. Yesterday I was at the Museum of Contemporary Photography here in Chicago and there were a couple of 40x50 prints from Richard Mosse. The images were shot on 8x10 Infrared Color Neg and they were exquisite.
I've also seen Sugimoto's big prints up close and they are beautiful too. What I love about the experience of them is that I'm drawn in close to them so that the image fills my periphery. The size creates half of the experience.
So, I've been looking at how big you can go on inkjet (one of the beauties of it being that you can print on long rolls). It also happens that I love shooting panoramic. After thinking about how big my IQ250 could go (even with stitching) I've started shooting film again. My drum scanner will do 4k dpi and the image fidelity of color neg is damn similar to the MFDBs.
Long story longer: I just bought a 4x10 field camera and a set of wide glass with 8x10 coverage. All just so I can print ridiculously large. Will I ever actually print that big? Will my images even warrant the investment? Will they be better that big? Who the hell knows, but I'm gonna have fun finding out!
Now, I just wish 8x10 Portriga wasn't $14 a sheet!
CB
This thread is veering towards the [/tongue-in-cheek] mode: OFF.
But before we officially revert to customary LuLa dialogue ...
I'll settle for the 400 million f/2 pathetically OOF .. sometimes the whole is greater than the sum of its parts
(http://3.bp.blogspot.com/-NQRpUtCKOj0/T8c7uxI8l0I/AAAAAAAAFH8/NQeH8Bdmcqw/s1600/Pirelli+Calendar+2008+2.jpg)
Patrick, is that you?!
This thread is veering towards the [/tongue-in-cheek] mode: OFF.
But before we officially revert to customary LuLa dialogue ...
I'll settle for the 400 million f/2 pathetically OOF .. sometimes the whole is greater than the sum of its parts
(http://3.bp.blogspot.com/-NQRpUtCKOj0/T8c7uxI8l0I/AAAAAAAAFH8/NQeH8Bdmcqw/s1600/Pirelli+Calendar+2008+2.jpg)
I've been thinking a LOT about all of this lately, especially since I've just started making prints of my work again and thinking about exhibition. Yesterday I was at the Museum of Contemporary Photography here in Chicago and there were a couple of 40x50 prints from Richard Mosse. The images were shot on 8x10 Infrared Color Neg and they were exquisite.
I've also seen Sugimoto's big prints up close and they are beautiful too. What I love about the experience of them is that I'm drawn in close to them so that the image fills my periphery. The size creates half of the experience.
Long story longer: I just bought a 4x10 field camera and a set of wide glass with 8x10 coverage. All just so I can print ridiculously large. Will I ever actually print that big? Will my images even warrant the investment? Will they be better that big? Who the hell knows, but I'm gonna have fun finding out!
... [/still in tic mode]
This isn’t quite the end of the story though, as seen in Hans Strand’s comments where he says he is getting better results from his medium format back than he was getting from 5×4 and 8×10. Digging a little deeper, Hans was using much larger apertures that used in the tests so I did a few calculations. The following table might look really confusing at first but bear with me. What I’ve done is to provide, for each platform, a list of aperture’s used in the test where each row shows an equivalent aperture for each platform. i.e. the first row in each table is the aperture that gives the same depth of field for that platform. What follows this is the theoretical maximum enlargement based on diffraction (based on the table here) – however I’ve modified these to limit the maximum enlargement based on a couple of different factors. The first limitation is the maximum enlargement of a 35mm digital ~20Mp camera which is 12″ x 18″ (at 300dpi). The next limitation is placed on the Phase IQ180 system because it has a maximum enlargement of 26″ x 32″ (based on 300dpi). The next limitation the maximum resolution for lenses for the Mamiya 7 which is about 100 line pairs per mm. The final limitation is the resolution of LF lenses which is about 70 line pairs per mm. Each of these tables now shows the largest enlargement in mm for each platform and each f-stop for equivalent depth of fields. Fortunately you can ignore all of that maths and skip your way down to the very last table which shows the ratio of the different platforms to each other at equivalent focal lengths.
In summary, this table shows the maximum critical enlargement for each camera type at each aperture taking into account diffraction and ‘best lenses’. e.g. 35mm and Mamiya 7 are film limited at 13x but the IQ180 sensor will allow a 19x enlargement before diffraction kicks in. The last table shows the relative enlargement ratios of the camera pairs shown. e.g comparing IQ180 and 8×10 shows that at smaller apertures the advantage to 8×10 is 2.3x but this falls behind at f/90 to 0.9x – diffraction has killed 8×10’s advantage
Jerome, Epson at Photokina showed what you are looking for: A print of a very high resolution digital image on the Epson 11880 large format inkjet printer in about 160cm high and 350cm wide.
Photographer: Stefan Arand
Motive: Toledo (Spanish city)
Camera: Hasselblad H4x + IQ260 60 MP back
A large number of 60 MP images are stitched together to a total resolution of about 28.000 pixel wide x 13.000 pixel high = 360 MP.
The printed resolution of the picture is therefore about 8 dots/mm or 200 dpi.
One thing to keep in mind is that it is not possible to keep everything in focus.
Bart. Does that come with a neck strap? I think it could be great for discreet street photography
It is an impressive work and I guess Stefen Arand must have taken a lot of time to hide all the stitches in the river. I was only able to find one. We definitely need a camera which can do that in one go (I am still in tongue in cheek mode, remember? ;D ).
My guess is less than 30 minutes if he used a decent stitching software... ;)
(still tongue in cheek mode...)
I see... you believe in software.
Let us suppose that my lifetime dream would be to get a picture of the ocean. I want to see big waves and I want to have it printed at the largest size that Epson 11880 can so that I can recognise the surfers riding those waves with a magnifier. How do I do that? ;D
Hi,Still, this is all theory... one forgets the factor of Bayer pattern interpolation that messes (visible) resolution up... After all, it all depends on how high one ranks resolution in his personal ranking of ...image quality. I know most pros don't... (including those that print really big and sell expensive).
8x10" shot at f/11 should fill your needs, Airy disk diameter at f/11 is about 11 microns. You need a lens that covers 8x10" at f/11 and has reasonable MTF at 50 lp/mm. That would give you something like 18000 pixels on the short side.
With todays MFD sensors we would need 2.4 micron resolution. That is absolutely feasible as there are sensors with higher resolution, but that would need a lens capable of good MTF at around 200 lp/mm at around f/2.8.
A realistic alternative would be using say 4-8 synchronised IQ-280s and stitch.
Best regards
Erik
That is not a real problem, see the Toledo picture.Actually, the Toledo picture illustrates nicely what Erik and I are saying:
To get DoF we need to stop down and that limits resolution. Two exceptions landscape with no significant foreground and tilted plane of focus.
The iron trade-of between DOF and diffraction means that these massive pixel counts probably need focus stacking with anything but distant landscapes that have no foreground needing to be in focus.Because in that Toledo picture, nothing this side of the river is in focus even when you zoom in only part way; in fact even the cars of the other side of the river are OOF at well less than full resolution, and I wonder if anything much closer than the cathedral is in focus when viewed large enough to see the full detail in that file.
| 5.6 micron Airy diameter | 8.2 micron Airy diameter |
| (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-5.jpg) | (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-9.jpg) |
| 5.6 micron Airy diameter + FocusMagic | 8.2 micron Airy diameter + Focus Magic |
| (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-5_fm.jpg) | (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-9_fm.jpg) |
Still, this is all theory... one forgets the factor of Bayer pattern interpolation that messes (visible) resolution up... After all, it all depends on how high one ranks resolution in his personal ranking of ...image quality. I know most pros don't... (including those that print really big and sell expensive).
Why don't you buy a 432 megapixel camera?
Actually, the Toledo picture illustrates nicely what Erik and I are saying
Hi,
The question posted is not my idea…, the OP wants one pixel resolution for each resolved pixel in the print.
On the other hand the numbers I posted take Bayer interpolation into account. I would also say they are a bit conservative. In experiments I have done there is a clear degradation of image quality once Airy disc diameter gets larger than pixel size, but the figures I give are for twice the pixel size.
The samples below were shot on a 4.77 micron sensor with an 100/2.8 Minolta Macro at f/5.6 resp f/8. The second row was sharpened in FocusMagic, letting FM choosing the sharpening radius.
5.6 micron Airy diameter 8.2 micron Airy diameter (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-5.jpg) (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-9.jpg) 5.6 micron Airy diameter + FocusMagic 8.2 micron Airy diameter + Focus Magic (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-5_fm.jpg) (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-9_fm.jpg)
I would add that I don't think there is anything wrong with the Bayer matrix. It requires an OLP filter to render colour detail correctly, but any point sampled system does require OLP filtering for proper rendition of luminousity.
Best regards
Erik
If we had the design files for trees, flowers, mountains etc we wouldn't need to photograph, we could just generate.
Edmund
All this know-how on maximising achievable resolution is of course (as with all knowledge) good to know... My concern has much more to do with the O/P... I still don't understand why printing on 300dpi is a necessity, why is it that we need more pixels and how achieving maximum resolution is related with improvements in ones photography...? My impression (and I think most of those that print big and sell their pictures) is that decreasing pixel size may increase resolution to some extend, but OTOH, they affect picture qualities that are much more important than resolution...
What is enough resolution is of course both personal and subjective, but since resolution has been improved a lot in the (say 7 - after 6 micron backs first arrived) later years, even from what 8x10 film can give, but photographs haven't improve at all (Its the opposite that happens actually), I don't see how seeking for more resolution (by inevitably decreasing pixel size which will affect negatively other pixel qualities) will lead into improving photographs....
In fact, the cases I miss more resolution in prints I see published, are extremely rare... it's rather the opposite... there is a good number of todays photographs that one finds that resolution is more than it should be! In fact, in some cases depending on the subject, resolution is so much that distracts ones attention to the subject itself... which of course isn't a good thing to happen.
They are not readily available. Which is the whole point of this thread: internet forum analysts argue that resolutions beyond 20 (argued around 2009), 50 (argued around 2011) or 80 mpix (argued today) are not needed, yet printers capable of taking advantages of over 400 mpix are readily available and cost less than the average MF back (the Epson 11880 discussed earlier costs 7000$ at B&H).I am not sure how you interpret the availability of "400 mpix" printers to say anything about the utility of 400 mpix cameras? There can be many reasons why "400 MP" printers are available, not all of them pointing to 400MP prints looking very different from 200MP prints for the majority of cases.
The same internet forum analysts argue that the properties of human vision are such that large prints can only be seen from far away,Not sure what you are saying here, but I am pretty sure that no one have ever stated that it is impossible to approach a large print?
internet forum analysts ...It seems to me that you pick your words in order to ridicule those that disagree with you. While this may make for an interesting argument, it is less likely that anyone of us will learn anything from the discussion.
I am not sure how you interpret the availability of "400 mpix" printers to say anything about the utility of 400 mpix cameras? There can be many reasons why "400 MP" printers are available, not all of them pointing to 400MP prints looking very different from 200MP prints for the majority of cases.
Focusing on a given spec (e.g. "I only ever print at 300 dpi, no matter what") seems less productive than a more pragmatic "let's see what works".
Not sure what you are saying here, but I am pretty sure that no one have ever stated that it is impossible to approach a large print?
It seems to me that you pick your words in order to ridicule those that disagree with you. While this may make for an interesting argument, it is less likely that anyone of us will learn anything from the discussion.
-h
I am very suspicious with ultra small pixels and I never had resolution in more than the 5th place on my criteria when buying... In fact, in most cases I do prefer my Nikon's 16mp prints than my other Nikon's 36mp prints, not to mention my 22 MFDB prints that are superior than both?
Theodoros: what makes the 16mp prints better than the 36,p and the 22 MFDB better than the opther two?
| (http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/feather_a.png) | (http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/feather_na_small.png) |
Theodoros: what makes the 16mp prints better than the 36,p and the 22 MFDB better than the opther two?Better DR, more contrasty mids, better balance between LLs and HLs, more accurate colour.... I find that with small pixels, if one is after keeping the captured DR, he ends up with a dull look and thus he is forced to process in a way that affects DR negatively.
Better DR, more contrasty mids, better balance between LLs and HLs, more accurate colour.... I find that with small pixels, if one is after keeping the captured DR, he ends up with a dull look and thus he is forced to process in a way that affects DR negatively.
Hi,Its noise per sensor area that is lower with larger sensors Erik... at least that's what Physics suggest. Back to the artefacts matter though... I see that you only mean moire when you refer to artefacts, isn't it? We all have to understand that if Bayer pattern was absent, there would have been no moire problem... Detail would also improve considerably since what the human eye understands as resolution is the colour difference between pixels and thus interpolated colour mistakes affects resolution too and DR would improve considerably...
I would say it may be the other way around. The small pixel cameras have high DR, due to low read noise. Making use of that DR make the image flat, as neither screen or print has similar brightness range.
Best regards
Erik
Its noise per sensor area that is lower with larger sensors Erik... at least that's what Physics suggest. Back to the artefacts matter though... I see that you only mean moire when you refer to artefacts, isn't it? We all have to understand that if Bayer pattern was absent, there would have been no moire problem... Detail would also improve considerably since what the human eye understands as resolution is the colour difference between pixels and thus interpolated colour mistakes affects resolution too and DR would improve considerably...
IMO, the problem is more with the technology used than with pixel density. I believe that soon there will be a technological revolution and what we now use, in a few years will be considered as ancient equipment... From that POV, one has to see on how he may improve with the short period he'll be using the current technology, to do that he has to judge what is more important to him as far as the benefits and the handicaps of using smaller pixels is concerned, which of course is highly subjective and personal...
Only thing I'm saying, is that there is a high number of photographers that consider resolution low in their list of priorities and that what is higher in the list for them is better served with larger pixels. Take a view camera user for instance....
Hi,
I would suggest that you need to read up a little bit on basic signal processing as your understanding of the issues involved seems to be incomplete. But I agree on the resolution not being a major issue.
Two direct problems with your interpretations I would like to point out.
One is that moiré has no relation to bayer pattern, even monochrome implementations have moiré, but not colour aliasing. You can just check the wikipedia page: http://en.wikipedia.org/wiki/Moiré_pattern
Noise levels are not really related to pixel size. Basically there is shot noise that depends on statistical properties of light and more dependent on sensor size than pixel size and readout noise which depends on readout circuitry.
At the present stage all full frame small pixel sensors are coming from Sony and use on chip converters. These how low readout noise, thereby delivering cleaner and deeper shadows than older designs with external ADCs. The reasons that the on chip converters work so well is probably in part that are many of them working in parallell, so each converters only processess say 5000 pixels/frame while a traditional external ADC may need to process 4 million pixels/frame (assuming four readout channels).
It seems that the on chip converters have something like 1/3 - 1/4 of the readout noise levels of external converters.
Now, readout noise is not dependent on pixel size, so larger pixels would have somewhat better dynamic range, but as said the small pixel cameras all have superior readout circuitry that more than compensates for the pixel size.
This sample is from a P45+:
(http://echophoto.dnsalias.net/ekr/Articles/MFDJourneyEOY/Noise/20131117-CF044323.jpg)
And this one is from a Sony Alpha 99 SLT
(http://echophoto.dnsalias.net/ekr/Articles/MFDJourneyEOY/Noise/20131117-_DSC3262.jpg)
The second image is brighter than the first one. I guess that it has to do with the zoom lens on the Sony having more lens flare. Exposure had been very similar.
Best regards
Erik
I hate to argue on theory... Larger pixels DO HAVE less read noise per area they cover and this is theory... (that's why the "LL champion cameras" have larger pixels). On the images that you posted, the second one isn't "brighter", its considerably duller (luck of contrast) and the HLs have been "blown" (which can't be an effect of flare - flare does decrease contrast but it doesn't blow HLs) while in the first one whites do hold some surface information... I also suspect, that the first image is not processed using C1 (which would have increased DR considerably), but rather an inadequate processing method that has been designed for general processing that has the masses of "photographers" in mind... Finally... moire on a B&W sensor is as rare as rain in the moon and in a true colour multishot is simply non existent... Also... the cases that moire is not the result of colour alising (which you correctly say is absent if BP is absent) are rare to the extent that they may be considered negligible.
One more thing that is questionable is the phrase "the small pixel cameras all have superior readout circuitry that more than compensates for the pixel size." that you used... is that theory? ...Is comparing (years back of) different processing technology a criterion that one may use to prove his case? ...its not to me (and the most sensible thinking).
I still don't see why we need higher resolution sensors.... Is the proposal that the makers should increase per area read noise so that one may be able to print at 300ppi but with considerably more noise on the print? Wouldn't a (same technology) sensor with considerably smaller pixels have much worst charasterics as far as read noise is concerned? Isn't "per pixel noise" more difficult to control the more the size of the pixel is compressed? Shouldn't we prefer that the makers solve the issues of artefacts that are present on todays pixel size before they proceed into making smaller pixels?
Hi,This is irrelevant to my post... No one says that noise is affected by pixel size in print... Printing (as well as sampling up or down in order to print at a certain size) is a method that also requires extensive skills which have to be adopted differently to the equipment used (an Epson requires different sampling than a Canon for instance). My post has to do with smaller pixels and the increased per image area noise that they inevitably have if compared with (same technology) larger sensors.
Noise (shot noise) is not really affected by pixel size in print. Is the image is upsized the upsizing algorithm will distribute each image pixel over more printed pixels, if the image is downsized the downsizing algorithm will merge more pixels into a printed pixel. Upsizing and downsizing algorithm have of course their own issues.
You don't get aliasing "every time the lens out resolves the sensor" as you say..., you need a repeated pattern at the right frequency for moire to exist, or cross talking between neighbouring pixels with different BP colour filtering for colour aliasing..
Aliasing artifacts like moiré and staircase effects are caused by large pixels. You get them any time the lens out-resolves the sensor. There are essentially three ways around this:
- Making the pixels smaller
- Stopping down so lens resolution is reduced
- Adding an OLP filter that reduces lens resolution
In practice, OLP filters are not strong enough to eliminate aliasing, it seems that their purpose is to reduce colour artefacts.
True colour sensors (either MS or Scanning backs) never present any aliasing although the pixels are huge and the lens out resolves the sensor.
Take a look at the Betterlight Super 6K rendition of the ISO 12233 target rosette here: http://blog.kasson.com/?p=4802I use an Imacon 528c at 16x and its totally moire free, I thought it would be the same with scanning backs but obviously forgot that sampling there is tri-linear... :)
Look at the horizontal lines around the 7. (Nyquist is about 5.5) They diverge as you go from left to right when the target is converging. That's aliasing. It occurs because the lens is resolving at high MTF beyond the Nyquist frequency of the sensor. So it does happen. On the other hand, ISO 12233 is designed to find things like that, and you'd be hard pressed to ever discover the effect in a natural image scene. I have seen moire in BL images with window screens in them.
Note the admirable lack of false color; that's the beauty of the BL back, and MS backs as well, as you point out.
So this is merely a quibble with your point.
Jim
I use an Imacon 528c at 16x and its totally moire free, I thought it would be the same with scanning backs but obviously forgot that sampling there is tri-linear... :)
Take a look at the Betterlight Super 6K rendition of the ISO 12233 target rosette here: http://blog.kasson.com/?p=4802
Look at the horizontal lines around the 7. (Nyquist is about 5.5) They diverge as you go from left to right when the target is converging. That's aliasing. It occurs because the lens is resolving at high MTF beyond the Nyquist frequency of the sensor. So it does happen. On the other hand, ISO 12233 is designed to find things like that, and you'd be hard pressed to ever discover the effect in a natural image scene. I have seen moire in BL images with window screens in them.
Note the admirable lack of false color; that's the beauty of the BL back, and MS backs as well, as you point out.
So this is merely a quibble with your point.
Jim
Hi Jim,Wouldn't smaller pixels increase cross-talking between them and thus increase colour aliasing Erik? (additionally to noise per area increase and the DR reduction)... Moire doesn't seem to be a problem that occurs often with pixels of 6microns, (I don't see the owners of those sensors complaining at all for moire), it seems to me that 6microns is good enough for a modern sensor with no AA filter, I am curious to see how things will develop with CMOS sensors for MF from here on... I expect that there will be no more CCD sensors here after and I am really curious to see a comparison between Leica's new 37.5mp CMOS sensor (6 microns) with respect to Sony's 50mp one.
Thanks for contributing to the thread.
I would say that colour aliasing is a major problem that is caused by the Bayer pattern while monochrome moiré and aliasing are lesser obvious artefacts. On the other hand I feel the monochrome aliasing also creates false detail, that may be seen as texture. That is part of the reason I use the feather shots as a sample, it illustrates the issue well.
I guess that we will live with the Bayer pattern for a long time. The other approaches are not that practical. Right now those are multishot, scanning backs and Foveon type sensors. Foveons have problems with noise and colour interpretation, as far as I understand. Multishot and scanning backs can have problems with things that move.
My understanding is that the best way of getting rid of aliasing (colour or not) is to reduce the pixel size. Increasing the fill factor would also help, I am pretty sure. Making the pixels smaller reduces DR at the pixel level, but I would say that the loss is probably acceptable with say 3 micron pixels and present levels of readout noise.
Best regards
Erik
I would say that colour aliasing is a major problem that is caused by the Bayer pattern while monochrome moiré and aliasing are lesser obvious artifacts. On the other hand I feel the monochrome aliasing also creates false detail, that may be seen as texture.
My understanding is that the best way of getting rid of aliasing (colour or not) is to reduce the pixel size. Increasing the fill factor would also help, I am pretty sure. Making the pixels smaller reduces DR at the pixel level, but I would say that the loss is probably acceptable with say 3 micron pixels and present levels of readout noise.
With respect to MS cameras, why not bring back color wheels? You'd have a mono camera that would be versatile, and a color one that would be limited. You'd be able to use any filter set you wanted, for any color rendering desired. There could be a cottage industry developing filter sets. Photographers could dye their own glass, just like they used to mix up their own developer.
Jim
If I , say, have a P1 Monochrom back or Leica Monochrom and want to shoot a painting through the filters, is there a Software to automatically mix the files? And have a reliable color management?
I imagine that any camera profiling software should work, providing the filter spectra aren't too weird.Thank you for the explanation.
Jim
Jim,Hi Yevgeny,
Very interesting idea. Getting back to XIXth century.
If I , say, have a P1 Monochrom back or Leica Monochrom and want to shoot a painting through the filters, is there a Software to automatically mix the files? And have a reliable color management?
Yevgeny
Hi Theodoros,There are certainly a lot of advantages (some disadvantages too) in using film depending on the case... Certainly the major disadvantage of digital is the fact that captures are not in colour but an interpolated result and the associated equipment (Bayer pattern etc) that is involved... One that shoots MS like you, (especially one that uses such a capable back as the one you have) surely has understand the issues... Unfortunately we'll have to wait until single shot captures are also in true colour before we can get rid of artefacts... One more thing to mention is Sinar's work on profiles (which you stated above), it seems that they are ahead in working on the matter, they seem to care about accurate colour reproduction as their major market and they target people or organisations that concentrate in painting reproduction... Their profiling methods look far more advanced than "competition"...
Thanks. I am not tech savvy at all and struggle with the tech language. I do not understand a lot of things on digital even in my own language. I will have to google for Astro photography.
Last year I had a problem with a shooting of a collage on fine silk with a lot of small 3-dimentional elements and newspaper clippings. It was mounted on a wall of a barn in Provence and I had no MS back with me.
I tried my Leica s2 on it, with a shift lens first, than made a number of captures with the idea to stitch them. I looked at the files at the hotel and saw a lot of artifacts. I went to a local photographer and hired him to shoot the collage on his old 5 by 7 film camera, scanned the film on our Leaf flatbed and there were only minor color and falloff corrections needed.
Just a story to share.
What I meant is that indeed parts of the Toledo picture are OOF. This is not a problem and did not preclude that picture to be chosen for photokina.We are not really disagreeing there, since I (like Erik) just stated the limitation to scenes where OOF foreground is not a problem. Though frankly ,"Toledo" is a poor example: the OOF "foreground" is about 80% of the image, and as soon as I enlarge anywhere close to "400MP", it is very disappointing to hunt almost in vain for something truly sharp. I suspect that its selection was I more about demonstrating a technological possibility.
Also, about Epson's making these printers proving that 400MP is needed; I can see another reason:
- cameras exist giving files of about 10,328 pixels on the long edge, and somewhat more is probably coming soon.
- owners of such printers might well want sometimes to print such files with 10,328 pixels across the width of the roll, so needing about that many pixels across the roll.
- owners might also want at other times to make far larger panoramic prints, with the short edge of the image oriented across the roll.
And voila: 10,328 on the short edge is needed for the printer, even if not for any print.
Who was it that said "if you can't make it good, make it big!" ?
You cannot possibly be serious: nobody in their right mind would buy a 60" (150cm) wide printer to print images which would fit in a 40" (120cm) printer sideways, since the second printer is half the price and a lot less floor space.If some one (or some printing shop) wants to do both huge prints (60" on the short edge) and less huge ones (a mere 60" on the long side), it makes sense to get one printer that can handle both tasks, rather than having to pay for and make space for two huge printers.
If some one (or some printing shop) wants to do both huge prints (60" on the short edge) and less huge ones (a mere 60" on the long side), it makes sense to get one printer that can handle both tasks, rather than having to pay for and make space for two huge printers.
If some one (or some printing shop) wants to do both huge prints (60" on the short edge) and less huge ones (a mere 60" on the long side), it makes sense to get one printer that can handle both tasks, rather than having to pay for and make space for two huge printers.
Right you are. Not to mention that you'll spend more on ink for a printer than you'll ever spend on the printer itself. In a commercial environment, you'll spend much, much more. So the added cost of a 60" printer over a 44" one is a drop in the bucket, and a wise investment if you've got the room, even if you'll put 44" rolls in it most of the time.
You are aware than one can use the smaller paper rolls in the larger printers, right?
Additionally: you are aware that for this kind of printers, ink costs are watched by the customers so that ink is considerably cheaper per liter than for your typical desktop printer?
Apparently, one of us is missing something because I still fail to understand the advantage of buying a 60" printer for double the price if one is always going to print on 44" rolls, so please you help me out.No one said always printing on 44" rolls:
Apparently, one of us is missing something because I still fail to understand the advantage of buying a 60" printer for double the price if one is always going to print on 44" rolls, so please you help me out.
No one said always printing on 44" rolls:
- I referred to users with a mix of some jobs needing the 60" roll printer and other jobs than can be done on either fit or a smaller printer, and using the single bigger printer to avoid needing a second slightly less big printer- Jim said 'even if you'll put 44" rolls in it most of the time.'
Yet another debate where by overlooking the qualifications in a statement, or changing the wording, an opponent's argument is morphed into a more easily refuted straw man.
BJL and Jim's argument is that the 60" printer is mainly for someone who needs 44" prints in "300 dpi" resolution and 60" prints some of the time, presumably in less than "300 dpi" resolution.
Am I the one who changes the arguments here? I doubt it.You inserted the word "always" in your words "when one is always going to print on 44" rolls" in "Reply #101" in this thread, which clearly contradicts what either of us had said.
My contention is not that the 60" printer is always purchased with the idea that it will infrequently be used with 60" paper. My contention is that some people will buy the 60" printer with the idea that it will infrequently be used with 60" paper.
I don't think you mean "dpi" in this context. Usually, dpi stands for dots per inch, and is a measure of the closest spacing of the halftoned image sent to the print heads. Perhaps you mean "ppi", which stands for pixels per inch, and is a measure of the spacing of the 8 or more bits per color plane image that is sent to the printer driver for halftone processing. 300 ppi is a Canon standard, and 360 ppi is an Epson one.
Jim
My contention is not that the 60" printer is always purchased with the idea that it will infrequently be used with 60" paper. My contention is that some people will buy the 60" printer with the idea that it will infrequently be used with 60" paper.The 300 vs. 360ppi argument is very interesting discussion... I use Epson 9900 although I sometimes need larger than 44" just because Epson never made an 11900... but I did choose Epson than Canon, simply because of the 360 standard and the much better compatibility with 72 multiples and all of the modern monitors.
I don't think you mean "dpi" in this context. Usually, dpi stands for dots per inch, and is a measure of the closest spacing of the halftoned image sent to the print heads. Perhaps you mean "ppi", which stands for pixels per inch, and is a measure of the spacing of the 8 or more bits per color plane image that is sent to the printer driver for halftone processing. 300 ppi is a Canon standard, and 360 ppi is an Epson one.
Jim
Hi,The printers print at much higher dpi than the ppi input they receive..., on an Epson, the input is sampled up to 360ppi and then it is printed at either 720 or 1440 or 2880dpi without more sampling, it simply uses 2 or 4 or 8 dpi for each one of the 360ppi that are input... On a Canon, the input is sampled up to 300ppi and then there are 300dpi multiples that are used (by choice of the user) from the printer to print...
As we have seen in the discussion, it is very hard to utilize 60" wide paper at 360 PPI fully, with high transferred modulation at the 1/360" pixel level. So quite obviously, the prints normally don't fully utilize the resolution of the printer.
The most reasonable way of achieving extreme resolutions are probably either 8x10" film using very good lenses at medium apertures and scanned at high resolution or stitching multiple images.
In both cases DoF will be a problem.
The reason that wide printers have high resolution is probably that the technology is there and also that most prints will be smaller, and need higher resolution.
As a small remark, when I print larger than A2, my printing service recommends 200 ppi on Durst Lambda, while they recommend 400 PPI for A3 size or smaller.
Best regards
Erik
The printers print at much higher dpi than the ppi input they receive..., on an Epson, the input is sampled up to 360ppi and then it is printed at either 720 or 1440 or 2880dpi without more sampling, it simply uses 2 or 4 or 8 dpi for each one of the 360ppi that are input... On a Canon, the input is sampled up to 300ppi and then there are 300dpi multiples that are used (by choice of the user) from the printer to print...
Actually the above is one of the reasons I never understood why we need more camera pixels than much better image quality (and absence of artefacts during capturing)... A perfect capture, will print superbly at 72ppi (on an Epson - 75 on a Canon), while a bad capture will print poorly on anything... Thus, 24mp is enough to print perfectly well on a 60" Epson using the full 60" for the smaller image side... Now, sampling or not (up to 300ppi for a Canon or 360ppi for an Epson) to print is another extremely interesting conversation...
EDIT: One thing is for sure... with todays offered sensor resolution, having better pixels should be much more preferable than having more of them...
Hi,Obviously Erik you are talking having Epson in mind... Of course the limits one puts for what he considers as minimum ppi is personal, but OTOH, nobody can argue that a picture is created and printed to be viewed as a whole... I never thought of looking close on parts of a print to judge the quality of a picture... I do that sometimes only for study purposes, but never include it in my criterion to judge a print... I wouldn't print scanned film (on an Epson) at less than 144ppi, that's for sure! But speaking about "edge contrast", that cannot be achieved if sensors out resolve the lenses, can it? At least not without "heavy" sharpening... Another thing to consider is (of course) the subject... I don't see why one should insist on 180, or 144, or 120, or even 90ppi on a distant landscape... I see why one may insist on 180 or 240 or 360 ...or even 720(!) with some rare small pieces of art though... It all depends on the subject and the quality of the capture IMO... Printing skills is another extremely important part of the process...
I would rather put the limit at 180PPI for an excellent print, that can be viewed reasonably close. The OP suggests that 360 PPI is needed to fully utilise the printing process, and he previously made the point that he has extremely good vision. Better vision than 20/20 is not uncommon.
On the other hand I would possible argue 20/20 vision, which is about 1' of arc, applies to high contrast edges but vision is dominated by lower frequency detail, so having good edge contrast at say 90 PPI may be more important than resolving detail at 360 PPI.
Best regards
Erik
Obviously Erik you are talking having Epson in mind... Of course the limits one puts for what he considers as minimum ppi is personal, but OTOH, nobody can argue that a picture is created and printed to be viewed as a whole... I never thought of looking close on parts of a print to judge the quality of a picture... I do that sometimes only for study purposes, but never include it in my criterion to judge a print... I wouldn't print scanned film (on an Epson) at less than 144ppi, that's for sure! But speaking about "edge contrast", that cannot be achieved if sensors out resolve the lenses, can it? At least not without "heavy" sharpening... Another thing to consider is (of course) the subject... I don't see why one should insist on 180, or 144, or 120, or even 90ppi on a distant landscape... I see why one may insist on 180 or 240 or 360 ...or even 720(!) with some rare small pieces of art though... It all depends on the subject and the quality of the capture IMO... Printing skills is another extremely important part of the process...
Do you see the next plateau in high resolution imaging having sensors with very fine pitch, or with significantly larger sensors than we have now?]
Eric, I'm thinking finer pitch. My reasons:
We can get lots of pixels that way more cheaply, by riding on the coat tails of CMOS processes developed for other, higher-volume purposes.
We are nowhere near using all the available performance of the best lenses, at least, if my simulation results are right.
There's a lot of small pixel work being done for cellphone cameras.
Much bigger sensors means much bigger cameras and lenses, with cost and portablility issues.
Of course, I've been wrong before...
Jim
Hi,Now you are coming in my (and others) position Erik... why one would look at a (say) 2m^2 (more than 12 sq.ft) print from a 50cm distance? ...can he see the picture from there? ...a picture is printed to be viewed as whole Erik, it's photography we are talking about, not testing one's vision ability. OTOH, in some rare product photography, or some pieces of art reproduction where the recipient often orders the print to study the item because he doesn't have direct access to the original, even higher than 360ppi may be required.
I used the 180 PPI figure for two reasons:
- It is often mentioned in litterature as the resolution needed for a very good print, like in the Bruce Fraser/Jeff Schewe book on sharpening. I am pretty sure that Norman Koren also mentioned that figure.
- It corresponds to the resolution of the human eye at 50 cm viewing distance.
Best regards
Erik
Moving away from printers for the moment….A common sense point... which leads (again) into the superiority of MS capturing conversation! ;) ...love it! :-*
There's pixels and then there are quality pixels. Just have lots of pixels doesn't mean very much to me.
Do you see the next plateau in high resolution imaging having sensors with very fine pitch, or with significantly larger sensors than we have now?]
Discus
A common sense point... which leads (again) into the superiority of MS capturing conversation! ;) ...love it! :-*
Hi,
This paper is a good read: http://www-isl.stanford.edu/~abbas/group/papers_and_pub/pixelsize.pdf
The impression I have is that the papers says that 4-5 microns is optimal with 0.18 micron technology, but I also feel that the pixel model in the paper is a bit dated. Modern pixel designs use shared transistors and utilise the pixel area better than older designs.
Best regards
Erik
| 6.8 microns | 3.8 microns |
| (http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/feather_a_large.png) | (http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/feather_na.png) |
Actually the conclusion suggests a 6.5μm pixel as perfect, here is the conclusion only...
CONCLUSION
We proposed a methodology using a camera simulator, synthetic CSF scenes, and S-CIELAB for selecting the optimal pixel size for an image sensor given process technology parameters, imaging optics parameters, and imaging constraints. We applied the methodology to photodiode APS implemented in CMOS technologies down to 0.18μ and demonstrated the tradeoff between DR and SNR on one hand and spatial resolution and MTF on the other hand due to the selection of pixel size. Using mean ∆E as an image quality metric, we found that indeed an optimal pixel size exists, which represents the optimal tradeoff. For a 0.35μ process we found that a pixel size of around 6.5μm with fill factor 30% under certain imaging optics, illumination range, and integration time constraints achieves the lowest mean ∆E. We found that the optimal pixel size scales with technology, albeit at slower rate than the technology.
Given the last phrase (on technology dependance), but also considering the photon's reception angle for the edges/corners of a huge MF sensor, I wouldn't trust anything less than 6μm...
Yes,You should take in mind the last phrase Erik... "We found that the optimal pixel size scales with technology, albeit at slower rate than the technology." 4-5 microns is huge difference (about 55-70% reduction) than the 6.5microns conclusion... It's no where near a sensible difference.
With 0.35 micron technology, with 0.18 microns it would be 4-5 microns. Please keep in mind that the article is based on a couple of years old pixel designs.
Technology sort of evolves over time…
Best regards
Erik
You should take in mind the last phrase Erik... "We found that the optimal pixel size scales with technology, albeit at slower rate than the technology." 4-5 microns is huge difference (about 55-70% reduction) than the 6.5microns conclusion... It's no where near a sensible difference.
See post above.Erik, it's YOU that posted the article (Great article, thanks for posting!) ...and now you disagree with the conclusion of the article you posted!
Best regards
Erik
As we have seen in the discussion, it is very hard to utilize 60" wide paper at 360 PPI fully, with high transferred modulation at the 1/360" pixel level. So quite obviously, the prints normally don't fully utilize the resolution of the printer.
But still... is it a "perfect" capture?
I'm not sure what you mean. I suppose, from one perspective, nothing ever is. From another, it's pretty great, with about 160 source pixels being averaged into each pixel in the working copy of the final edit. In the orthogonal direction to the time one, I could have used some more resolution, since I needed f/45 on the 120mm AM-ED Nikkor for the DOF, and the photodiode array is maybe 3 inches high. There was some CA that I had to process out.Sounds good... but since this was obviously a still capture, using a Betterlight for so long as 16 captures require can be an issue... I just wonder, I don't say you did anything wrong by any means... :) My quote was on my opinion on the printing process... One may ignore the "capture" part... it's only there to underline my opinion that printing big requires a "perfect" capture.
Jim
Sounds good... but since this was obviously a still capture, using a Betterlight for so long as 16 captures require can be an issue...
Well, it was a 14-hour series of exposures, and I did have to write some Matlab code to find and deal with instances of subject motion as well as doing one-dimensional downresing and median filtering, so in that sense the capture wasn't perfect.I will stay with your comment: "That's not to say that the final result was bad in any way; I was delighted."... A creation belongs to the creator! Well done. ;) What do you think on my printing proposals?
Jim
What do you think on my printing proposals?
Sending the image at 90 ppi to the printer driver wouldn't be my first choice. The Epson driver resamples the image it is fed to either 360 or 720 ppi (depending on the state of the "fine detail" check box, using nearest neighbor. With a 90 ppi image, that means that every pixel in the input image becomes a 4x4 square of equal-value pixels in the image the halftoner works from. The canvas will probably smooth that out enough so that you won't notice it, but I prefer not to take chances, and upsample using a more sophisticated algorithm such as Lanczos 3 or the proprietary algorithms in QImage.OK... But, when downsampling to 90 or up-sampling to 120 (in first stage), there is still a choice of algorithm involved... it's just kept to the minimum possible influence... If one aims to affect the capture as little as possible it seems a logical approach, ...doesn't it?
Your other suggestion was to do the upsampling in two stages. That has proven to be a successful regime for some, and, if I remember right, was the basis for the Genuine Fractals plug-in many years ago. My own personal experience is that doing the resolution changing in one shot is at least as good, and I'd rather spen my time making sure I'm using an algorithm suited to the image than trying various staged possibilities. But to each his own, and I'm sure that, done with care, the staged approach can yield successful results. In any event, I think there is more leverage in the choice of algorithm than in the choice of ratio, but that's just a personal opinion.
We have some experts on this board, Bart van der Wolf and Nicolas Robidoux, who have forgotten more than I'll ever know about resampling.
Jim
OK... But, when downsampling to 90 or up-sampling to 120 (in first stage), there is still a choice of algorithm involved... it's just kept to the minimum possible influence... If one aims to affect the capture as little as possible it seems a logical approach, ...doesn't it?
What does quality pixel even mean. It's a dimensionless entity with a RGB value as it's only property. I'll take more of them before I pay for a bigger sensor. The cost of a chip is closely related to size. With improved technology more pixels in a given size basically comes for free. If I can keep dynamic range I'll take more. When we are diffraction limited wide open we have enough.
But speaking of DR. What is the limit of DR that could be obtained in a single capture?
Question:
Do you judge how a print will look like on screen by viewing at 50% (I don't know if this is true)?
Its noise per sensor area that is lower with larger sensors Erik... at least that's what Physics suggest.This can be approached by introducing arguments and references to relevant theory, or introducing real-world side-by-side tests that confirm or fail to confirm your view. Simply claiming "because... physics" is not very fruitful.
Back to the artefacts matter though... I see that you only mean moire when you refer to artefacts, isn't it? We all have to understand that if Bayer pattern was absent, there would have been no moire problem... Detail would also improve considerably since what the human eye understands as resolution is the colour difference between pixels and thus interpolated colour mistakes affects resolution too and DR would improve considerably...There is aliasing in any sampling system as long as the input signal has a bandwidth >= fs/2. Man-made sampling systems might include some kind of "anti-aliasing" prefiltering (explicit or implicit), but like any man-made filters they won't have infinite attenuation in the stop-band. What might be realistic (for some systems/cases) is for this aliasing to be sufficiently attenuated to levels where they are no longer a big problem. There is no need for the signal to be periodic, by the way. A single impulse (think single-sensel-size star against a black sky) contains a wide range of frequencies and may cause aliasing.
16 stitched Betterlight captures
This can be approached by introducing arguments and references to relevant theory, or introducing real-world side-by-side tests that confirm or fail to confirm your view. Simply claiming "because... physics" is not very fruitful.
There is aliasing in any sampling system as long as the input signal has a bandwidth >= fs/2. Man-made sampling systems might include some kind of "anti-aliasing" prefiltering (explicit or implicit), but like any man-made filters they won't have infinite attenuation in the stop-band. What might be realistic (for some systems/cases) is for this aliasing to be sufficiently attenuated to levels where they are no longer a big problem. There is no need for the signal to be periodic, by the way. A single impulse (think single-sensel-size star against a black sky) contains a wide range of frequencies and may cause aliasing.
With the CFA used in "Bayer" sensors, the 3 sampling frequencies are simply shifted downwards. If you prefilter at this new minimum fs/2, I believe that there will be no Bayer artifacts (and Debayering will be a very simple linear operation). One reason that we don't do such pre-filtering is that the sensel density is simply not high enough. The resulting (luminance) resolution would simply be too low, and the artifacts from the CFA pre/post processing are often surprisingly small anyways. The current level of prefiltering and postprocessing may be a compromise that works well for current sensel densities and print sizes/viewing distances.
Hi,
This paper is a good read: http://www-isl.stanford.edu/~abbas/group/papers_and_pub/pixelsize.pdf
The impression I have is that the papers says that 4-5 microns is optimal with 0.18 micron technology, but I also feel that the pixel model in the paper is a bit dated. Modern pixel designs use shared transistors and utilise the pixel area better than older designs.
Best regards
Erik
Ok, enough of the tech talk. ;D"Use your magnify lens"... "Makers are stupid"! "P1 is right for being the only one that has moved (a little) below 6microns"... "We know better"... "We know everything !" ..."You need prints to be captured at 300dpi"! ..."Move closer"! ..."Less ppi is because there is a conspiracy against the public"! ..."Gursky is overrated because he doesn't use enough ppis"... "I close examined that expensive print and it was rubbish"... "I watch prints to judge them at less than 50cm no matter the size of them"... I want, I want, I want... I know, I know, I know...
Anyone want to see how much resolution is enough? or required?
Go to a Rodney Lough JR Gallery (las Vegas or San Fran (the Wharf)) and look at some of the VERY large prints on display. Get close. Yes, real close, no, 50cm is not close enough. Get your nose to almost touch the glass. See the detail? Yes, you can count the blades of grass on a wide angle shot. Count every leaf. Those prints do not leave you wanting for more resolution. Only color prints I have seen that large that do that. Almost every print is just stunningly well made and look amazing even under close scrutiny.
The Peter Lik prints are almost as good but not quite. He gets a touch sloppy on some although most are awesome also.
Again, I am talking color prints.
Google what kind of gear they use and techniques.
Rodney Lough JR used a 8x10" view camera for much of his work and presently uses an RL3d with an IQ180 back. His maximum print size appears to be 45" (114cm) on the shorter side. The prints are probably about 180-200 dpi.Actually It's 44" Jerome..., but them being at 180-200ppi, wasn't your O/P about someone needing 300ppi?
Rodney Lough JR used a 8x10" view camera for much of his work and presently uses an RL3d with an IQ180 back. His maximum print size appears to be 45" (114cm) on the shorter side. The prints are probably about 180-200 dpi.
It is also consistent with the observation in (my) above article that high contrasts texture and line patterns need higher resolution to avoid aliasing and pixelation.
So, I guess we are in a race between good enough and more is even better.
Best regards
Erik
Interesting, that's something I have been considering trying to. How did you move the camera/back?
What that spherical stitching or did you use shift? If you used shift, what lens did you use with such a coverage? One of the mega schneider monsters covering 11x14?
Eric, I'm thinking finer pitch. My reasons:
We can get lots of pixels that way more cheaply, by riding on the coat tails of CMOS processes developed for other, higher-volume purposes.
We are nowhere near using all the available performance of the best lenses, at least, if my simulation results are right.
There's a lot of small pixel work being done for cellphone cameras.
Much bigger sensors means much bigger cameras and lenses, with cost and portablility issues.
Of course, I've been wrong before...
Jim
That would be 185 PPI, pretty consistent with the suggestion that 180 PPI is needed for a very good print. Pretty much consistent with this article of my (http://echophoto.dnsalias.net/ekr/index.php/photoarticles/82-why-i-cannot-see-a-difference-in-a2-size-prints).
I see your logic, and probably this is the way the tech will go since devices such as cell phones will push for smaller pixels and higher quality small optics.
It's an interesting dilemma because that route will be one which pushes the glass to be shorter, faster, smaller and higher cost for the reasons of diffraction etc. Going to a smaller pixel will up the cost of lenses, and also affect range of DOF the photographer can use on both sides - Shallow DOF will not be as easy since format is smaller and faster glass not so easy to come by, and deep DOF not possible because diffraction limits are found sooner when stopping down because of the smaller pixels.
Technology on small pixels is still developing but the limits of physics are there - smaller pixels will ultimately have less dynamic range than larger pixels - all else being held equal.
If FF is "small", we're already there with the Oti (is that the plural of Otus?), the Sigma ART series, the Coastal Optical 60mm macro lens, and some others. It's interesting as the lens quality for FF lenses increases, the lenses themselves get much larger, until they are almost as big as some MF lenses. That's not always true: the Coastal 60 and the Leica 50 f/2 APO are counterexamples.
Going to a smaller pixel will up the cost of lenses, and also affect range of DOF the photographer can use on both sides - Shallow DOF will not be as easy since format is smaller and faster glass not so easy to come by, and deep DOF not possible because diffraction limits are found sooner when stopping down because of the smaller pixels. Technology on small pixels is still developing but the limits of physics are there - smaller pixels will ultimately have less dynamic range than larger pixels - all else being held equal.
Actually, my experience of large prints has been that ~200dpi are fine, but the quality visually deteriorates below 200 dpi. How much depends on the subject. Still: apparent sharness definitely improves between 200 dpi and 300 dpi.There is no print than can be made as low as 200dpi or even 300dpi Jerome... Not on a Canon or an Epson printer anyway... Perhaps you mean ppi and just confused it with printer's dpi.... is that what you mean?
300 dpi for 60" x 80" is 18000 x 24000 pixels, 200 dpi for 60" x 80" is 12000 x 16000 pixels.
We can discuss whether 200 dpi or 300 dpi is needed all day long, the fact remain that the 60" printers demand resolution that cannot be attained by present cameras for best output. Which was always the point of this thread.
This thread started tongue in cheek about the realisation that available inkjet printers are better than any camera on the market today and, therefore, that the idea that digital MF cameras have too high a resolution is absurd. It then evolved about out of my control about bizarre theories about printing, optimal pixel size, moire and lenses, etc... But this was never my intention.
My intention was always about large prints and large printers. I have printed somewhat big and reached the limits of my cameras of the time (24 mpix 24 x 36). I have been to galleries and museums and have experienced large prints, both analogue and digital. I know perfectly well what can be done, because I have tested it myself. And I know just as well that few photographers can produce large prints which appear sharp when looked up close, because I have been to galleries and museums and watched prints up close.
There is no print than can be made as low as 200dpi or even 300dpi Jerome... Not on a Canon or an Epson printer anyway... Perhaps you mean ppi and just confused it with printer's dpi.... is that what you mean?
But all else is not being held equal. Thinking about smaller feature size, different layout, copper, microlenses and so on.
Diffraction is not a limit on aperture. When we are diffraction limited wide open you are free to worry about what the image need for the intended viewing size. 1:1 pixel view is going to be soft anyway.
Hi,What one judges as priority to him... is obviously personal... PHOTOgraphy though, is not personal... it's by definition a print that involves lighting difference and the treating of lighting difference... by definition! ..That's what a PHOTOgraph is!
I would say that DR is more about sensor area than pixel size, equivalent technology presumed. But, I don´t really think that DR is decisive for image quality, I would say that less than 0.1% of my 70000 images are affected by DR.
Best regards
Erik
Bernard, it's a little complicated, and all is not what it seems to be. I can explain.
The idea behind this series, which I call Timescapes, is one dimension of the image is distance, like a normal photograph, but the other dimension is time, quite unlike a normal photograph, but like the images from finish-line cameras at racetracks. I use a Betterlight Super 6K back on a Linhof Master Technica, and occasionally an Ebony field camera, which provides a delicious combination of old=tech and high-tech.
There is a rotating platform available for the Betterlight back that allows it to be used to make panoramas. When the camera software is in panorama mode, the sensor goes to the center of its travel and stays there, expecting the rotating platform to provide the scanning. In this mode, the back can make images of up to 6000x64000 pixels with no interpolation.
So how do I get the camera to make the Timescapes pictures? I lie to the software. I tell it that the camera is on a rotating platform, but it's firmly attached to the head of a normal tripod. Therefore, any changes visible in the picture that results are the result of subject motion. To get multiple exposures for stiching, I use the camera software's built-in intervalometer.
Now, about this image.
I have some succulents that my wife propagated from a cutting Don Worth (yes, that Don Worth; AA's assistant and, with Jack Welpott, creator of one of the great photography education programs) gave to me. The succulent was created by Don — it’s a hybrid of Echeveria Shaviana and Echeveria Subrigida, and it’s called Echeveria Afterglow.
Slit scans need some motion to rise above banality, and succulents are not known for their athleticism. I figured, if I can’t get the plant to move, I can at least get the light to change. In fact, since I’m taking the photographs in direct sunlight, I can’t get the light not to change. So I made a series of images with long exposures. When the slit was vertical, I had time run from right to left, and with the slit horizontal, time goes from top to bottom.
You can see more of the series here (http://www.kasson.com/galleries/timescapes.php).
Jim
So these are not really photographs, they are waitographs :)
So these are not really photographs, they are waitographs :)
Edmund
Hi,
The question posted is not my idea…, the OP wants one pixel resolution for each resolved pixel in the print.
On the other hand the numbers I posted take Bayer interpolation into account. I would also say they are a bit conservative. In experiments I have done there is a clear degradation of image quality once Airy disc diameter gets larger than pixel size, but the figures I give are for twice the pixel size.
The samples below were shot on a 4.77 micron sensor with an 100/2.8 Minolta Macro at f/5.6 resp f/8. The second row was sharpened in FocusMagic, letting FM choosing the sharpening radius.
5.6 micron Airy diameter 8.2 micron Airy diameter (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-5.jpg) (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-9.jpg) 5.6 micron Airy diameter + FocusMagic 8.2 micron Airy diameter + Focus Magic (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-5_fm.jpg) (http://echophoto.dnsalias.net/ekr/images/DoF2/A55_100Macro_small1-9_fm.jpg)
I would add that I don't think there is anything wrong with the Bayer matrix. It requires an OLP filter to render colour detail correctly, but any point sampled system does require OLP filtering for proper rendition of luminousity.
Best regards
Erik
6.8 microns 3.8 microns (http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/feather_a_large.png) (http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/feather_na.png)
Two images, same camera position and focal length. Which one do you prefer?
Bernard, it's a little complicated, and all is not what it seems to be. I can explain.
Thanks Jim, fascinating usage of the Betterlight!
Mike Collette calls this "intentional misuse".
..."Gursky is overrated because he doesn't use enough ppis"...
I keep hoping one day you'll mix it up and surprise us with a $2 bill
I don't know if this could be of relevance in this discussion.
Hi,
There are a few reasons for using a one Dollar bill.
One such reason is that it is widely available so any can take one up, shoot it adequately and reproduce my results. That is also the reason not to use a two Dollar bill, they are quite rare.
The other reason for using a bank note is that they have a lot of fine and hard to reproduce detail, in order to hinder counterfeiting. So they are a good test for proper reproduction.
The reason for using the feather is that I wanted to see if aliasing occurs on natural objects and if that aliasing can be mistaken for real detail.
The discussion here is about high resolution needed for large prints. So I feel the issues demonstrated here are relevant. The Dollar bill is used to demonstrate the effect of diffraction. The feather shows that smaller pixels reproduce fine regular structures better than large pixels.
None of this may be of interest to you if you are still shooting film, but may be of more interest if you shoot digital. Film doesn't have a regular structure like a digital sensor, so it does not alias for instance. Film is affected by diffraction, but large enlargement is needed to make it obvious, except in macro photography. But I have been very familiar with diffraction effects since the late seventies. So it is not a "digital era invention".
If you scan film, aliasing may still be of some interest to you. There is something called grain aliasing (http://www.photoscientia.co.uk/Grain.htm), scanning at low resolution can create large amounts of extra grain. So that is a reason to scan at high PPI.
Best regards
Erik
I was teasing you Erik...
Hi,
It still depends on the viewing distance.
Human visual acuity limits visible resolution to approx. 1 arc minute, which translates to 286.48 PPI at a 1 foot viewing distance. The required output PPI is then determined by dividing the 286.48 PPI by the distance in feet. So viewing your 27'' Retina display from a distance of something like 16-18 inches should approximately match average visual resolution.
Of course, if you use Jerome's loupe, you'll just see larger LED pixels, not more resolution.
Cheers,
Bart
>Within reason, that all sorts itself out: http://www.josephjamesphotography.com/equivalence/
Looks interesting, thanks for the link.
>But your original question was about sensor getting bigger than they are, and they're already a bit under 56mm in the long dimension. I think they could get >a little bigger -- I'd sure like to see a 56mmx56mm sensor, but I don't expect to see 4x5 inch and 8x10inch sensors as standard photographic tools in my admittedly >limited prospective lifetime.
Yes I was thinking 6x6 square myself but also 4x5 though admittedly not much of a chance to bad I can't just program it into the "matter compiler".
I'm still of the opinion that the images from smaller cameras look flatter - as in lacking the perception of depth. Even the DSLR's, but then I haven't shot with an Otus or a costal optics lens. I'm going to have a hard time believing that it all sorts out until I see it at myself. There's a test waiting… same subject, same framing and DOF, but shot with every size format we can through at it.
What happens when the pixel pitch goes like 10x finer - say fine enough to resolve diffraction effects and lens effects? Can any special math be applied in that case to improve IQ?
>Good luck with Rollei!
Thanks!
>Jim
Hi Bart,
Correct me if I am wrong, but the one minute of arc figure applies to high contrast detail.
Maximum contrast sensivity is at much lower frequencies. So I feel that for normal subjects high contrast (MTF) at say 4 arcs of minute would dominate over low contrast at 1 arc of minute.
(see also here (http://webvision.med.utah.edu/book/part-viii-gabac-receptors/visual-acuity/))[/center]
In addition, the Vernier resolution of our eyes is much higher than that for parallel features.
A properly output sharpened image can somewhat take advantage of all that, by by adding a boost of the contrast for higher spatial frequencies. A PS plugin such as Topaz Labs Detail can do that very effectively. That of course does assume normal viewing conditions, not people with magnifying glasses walking up to the output to do close inspection.
Cheers,
Bart