full frame size vs. 4/3 - enlargement limits?

<p>Question for 4/3 users - what's your biggest print before you start seeing resolution or quality degradation? I spent last weekend doing comparison shoots with a Canon full frame vs. Olympus 4/3 at comparable focal lengths using pro lenses. I was pleased that the original image quality was comparable between the two, but fell off quickly once I enlarged the 4/3 beyond a certain point while holding up for the full frame. These comparisons were viewed entirely within Lightroom 2.</p>

<p>The Canon rep I talked with said one limitation of the 4/3 sensor is you'l reach a point where enlargements will suffer, while the full frame will continue enlarging more. On a first pass examination I see what he's talking about.</p>

<p>I'm planning the same comparison with Nikon full frame. However, what's the forum's experience with this? Not screen viewing but actual printing? What size limitation are you seeing?</p>

<p>Thanks.</p>

<p>So far I've printed up to 36 by 48 inches with really good quality. But I've also post processed first for best results.</p>

<p>2 cents worth. You can debate between the quality of each pixel between a D3X, 1DS MK3 or an E30. With digital enlargement, say if you believe for a moment all pixels carry equal quality (image info) and the lens in use exceed the resolution limit of the pixels, it is still all about the number of mega pixels. A 12.3 mega pixels E30 will max out 100% sooner (area wise) then a 25 Mega pixel D3X. But then, you can alway say you don't print more then A4 size (~ 8 by 12), can't really see more then 300DPI and really hate to crop. For me, 12MP are plenty enough, most of the time. The question is how happy one get with just most of the time :-) If a lot, f1.4 lens won't sell.</p>

<p>I think that talking to a Canon rep about the merits of 4/3 is maybe not the best way to get impartial advice. But yes, full frame sensors carry more pixels and there comes a point where you can enlarge a bit more. If that point falls outside your normal print size range, it's totally irrelevant. As Tommy mentions, a lot of us can do everything we need with a 4/3 image.<br>

I don't routinely make prints bigger than A4, and I have no problem with 4/3. Also, I don't know where the acceptable size limit lies. I can do my photography without concerning myself with these issues.</p>

<p>Well, everyones' standards vary so it is hard to put a fixed limit. It also in my experience is somewhat image dependent. I find landscape shots with normal to wideangle lenses to stress crop sensors more than full frame. (Unlike some I dont find all that much difference between aps-c and 4/3) Also printing is more forgiving than viewing 1:1 on a computer screen. But I think most would agree that at 8x10/A4 there is little advantage for full frame and that probably extends to 11x14/A3 which is what I print my best at.</p>

<p>But 11x14 is probably approaching the limit for crop sensors if you are looking for critical quality in a print viewed at close range. I am sure Rich's 36x48 prints look great when viewed at a reasonable distance for that size but if you stick your nose up close you will see the difference.</p>

<p>You can go ahead and compare the Nikon FF as well but you wont find much difference from the Canon version. Its just like in the film days a bigger piece of film gave you more to work with, the same is true for silicon sensors. </p>

<p>First you need to define the actual printing size rather than just talking about enlargements. Unless you are doing contact printing then everything is an enlargement... Talk to the Hasselblad rep. They will explain the limits of full frame 35mm sensors and how you will reach a point at which enlargements will suffer. Then talk to the Better Light rep and they'll explain how 4x5 scanning backs out do tiny medium format sensors like the Hasselblad. Then talk to the gigapixel people who stitch hundreds of images together and they'll explain how even 8x10 is a tiny format.</p>

<p>There's no such thing as a "contact print" from a digital camera. ;-)</p>

<p>I had to laugh. John Isaacs in a 2006 interview on "The Candid Frame" talked about the beautiful 20x24 inch prints he was making with his 5Mpixel Olympus E-1, and how they were better than his older work. And if you've ever seen his work in exhibition, you'd agree that they are stunning, gorgeous prints. If he can do that with a 2003 generation, 5Mpixel E-1, that's good enough for me.</p>

<p>ANY printing shop you go to will interpolate data for you. If 8MP shooters are making wall-sized prints, it's because of interpolation. The data is not actually there. Considering that the ACTUAL image resolution of a 10MP camera is just enough to fill 8x10 at 300dpi (print industry standard magazine quality resolution) everything above 8x10 on any sub-20MP camera is an enlargement with a loss in quality which is masked with interpolation software, which is really the powerful part of the whole process, not the camera at the beginning.... despite what your camera magazines may be telling you. A "modest" poster-sized 3 foot print is a massive enlargement, and it's very likely that so much of the image is interpolated data that it wouldn't matter if you shot it with a point-n-shoot camera through a hotel door peep-hole, because the software is working overtime to make the print even possible. But you can't sell new DSLR's or photography magazines telling people that can you?</p>

<p>The important thing is to do what you can to avoid noise... that means shooting low ISO with lots of light and avoiding deep shadows.</p>

<p>Patrick,</p>

<p>In a photograph, unless being credited for documentary or investigative purposes, "what you see is what is there."<br /> Whether the <em>ACTUAL</em> resolution, pixel for pixel, is "real" or not is irrelevant.</p>

<p>BTW, I do my own printing in sizings up to 13x19 inch image area. I don't normally do any scaling or interpolation. A 5Mpixel 3:4 proportion image prints quite nicely at 150ppi to ~13x17 inch image area. The wrong kind of image, or one badly treated, can show pixelation and other artifacting at which point upscaling to 7.5 or 10 Mpixel is useful to clean up the print. A G1's 12Mpixel output produces a very clean 240ppi at 13x17 inch image area without any scaling at all, and minimal sharpening too.</p>

<p>The 300ppi print 'standard' you name is significant to the offset-press world of printing devices, and for scanned film due to issues with grain. Digital capture images, printed with inkjet printers, can print to the same quality with 40-60% of that pixel resolution</p>

<p>Actually, I learned 300dpi in my very short stint as a Computer Graphics major before moving on to the high-profile world of Fine Art. ;) We were printing to photo-quality CMYK digital printers from 100% digital PS paintings. Most inkjets I find blur the image enough that 300 or 150 looks the same... and many are bad enough that 100 looks the same as well. I wasn't suggesting you CAN'T print 13x17... obviously you can print at any resolution you want... I was just giving a point of reference. Typically, the most important factor in usable print size isn't capture resolution or interpolation or print resolution, it's viewing distance. And that's the killer. Gallery prints need to be able to be viewed up close and personal. 8MP won't do 16x20's without some major interpolation work no matter how many people have tried to pull it off... oh, and I've seen lot's of them... people attempting to charge $300 for a rastery print made from their Canon Rebel... who had either invested too much in bad prints to back out or were so in love with themselves they couldn't see it. </p>

<p>But the big picture here is that 12MP is 12MP. It doesn't matter if the plastic box says Olympus or Canon or Nikon or Sony. It doesn't matter if it's an SLR, a ZLR, a PNS or an EVIL. The print doesn't care if it was a zoom lens or a fixed lens or a hotel door peephole. The Megapixel resolution tells you how many pixels there are in an image. 4000 pixels x 3000 pixels = 12,000,000 pixels = 12 megapixels. The major difference in sensor size is how well the individual sensor sites deal with issues such as noise and light diffraction at small apertures. Smaller sensors are more prone to noise and because the individual pixel sites are smaller, they pick up diffraction rings at lower apertures. So for the highest resolution images out of a 4/3's size 12MP camera you need to be sure to avoid dark shadows or other situations and encourage noise and be sure to shoot at apertures between f/1.4 (the fastest lens available for 4/3's) and f/8. Supposedly the EP1 features better noise reduction right in the box, which allows it to handle all way up 6400 ISO... although the high end ISO's are almost always just marketing hype I think.</p>

<p>"I don't normally do any scaling or interpolation."</p>

<p>You may not be doing any scaling or interpolation but are you sure that your printer driver is not?</p>

<p><strong><em>Peter:</em></strong> if I set a size and density, and what comes out on the paper is that size, I can only deduce that the density I set is what the print driver used. I can't do much else anyway. ;-)</p>

<p><strong><em>Patrick:</em></strong> Perceived quality isn't measured resolution. </p>

<p>amateur photographer in the UK did an objective test of sensor size recently G1, D3 and D300. They concluded that the G1 produced the best results at low ISO in good light - yes better than the D3 but that under a variety of conditions the D3 was best, D300 next and G1 last - they also discussed DOF. For me the ability to get a shallow DOF from a full frame sensor is a big advantage but others have different opinion. I can tell you that up to 11x17 it is hard to see a difference between the G1 and the EOS 5DII when I use similar quality lenses. Cropping the image or going larger gives an obvious advantage to the full frame Canon. I love my G1 - even the EVF but lenses are it's weakness. The standard lens is slow and not very good - thus you can never really use a narrow DOF (as the lens is slow) and the image quality suffers. If I put a top quality Canon FD lens on the G1 (the reason I bought it) then as I say the results are for practical purposes as good as the 5DII up to 11x17. With Genuine fractals or similar you may even be able to get bigger prints. It will be interesting to see how the EP1 performs as the Amateur Photographer tests consistantly peg the G1 above the Olympus 4/3 bodies for resolution, noise and image quality. They just showed the GH1 as being slightly better than the G1. While you can always criticise tests Amateur Photographer is probably the best general photography magazine published today (I pay to have it sent across the Atlantic) and has over 100 years of history.</p>

<p>Two major factors should be mentioned:<br /> <br /> 1- that 4/3 lenses are generally <em>sharper</em> than full frame lenses, due to the fact that the FT lenses have a smaller area to cover!<br /> <br /> 2-the NEW factor is the elimination of the mirror with micro-FT will lead to still <em>further</em> sharpness enhancements for wide angle lenses in particular due to symmetrical lens design used already by rangefinder cameras. <br /> <br /> Thus for architectural and landscape photographers who don't need low noise high ISO and narrow depths of field from fast lenses, the FT system might be the best solution even now. I'm looking forward to seeing how well the Panasonic 7-14 micro zoom does when it's finally released, but I'm with Philip in wanting fast lenses on larger sensors. I used to have the OM Zuiko 100F2.0 and 50F1.2 lenses (along with 21F3.5 and 28F2.0).</p>

<blockquote>

<p><strong><em>Peter:</em> </strong> if I set a size and density, and what comes out on the paper is that size, I can only deduce that the density I set is what the print driver used. I can't do much else anyway. ;-)</p>

</blockquote>

<p>Godfrey, my original statement was a bit rhetorical. Print drivers have a native resolution that they operate at. If you send it data that is less than the native resolution then the driver uses interpolation to make up the difference. So even if you are not interpolating you data your print driver is.</p>

<p>Oh, is that what you mean.</p>

<p>Well, if I'm being critical, I set the output ppi to 180, 240, 300, or 360 and print whatever size that ends up being. (Epson drivers are natively modulo 60 in output resolution, and the difference between 240 and 360 ppi is just barely visible with a magnifying glass.)</p>

<p>Let's debunk "resolution".<br>

Firstly a Bayer sensor digital camera interpolates resolution upwards. A typical subarray is 4 pixels with 2 green, 1 red and one blue. (Actually the photosites have filters over them ).<br>

G R<br>

B G<br>

A 12MP camera has 6 million green photosites and 3 million red and blue each. What if the lens projects a red image the shape and size of two photosites that happens to fall on the two diagonal green photosites? It does not register . It would register if the sensor was Foveon type, but these have 4-5 MP real resolution, the rest is also interpolated.What we get from digital cameras is already interpolated image. The quality of the demosaicing algorythm (in Bayer cameras), whether by the in-camera processing into jpeg or offline in a raw developer, decides how close is the image to what the lens projected onto the sensor.<br>

Yes, lens. Another limiting factor in resolution. There is no point talking about resolution in terms of megapixels if the lens does not deliver. Assuming (and that's a big IF) that we have the best possible lens (usually a prime) perfectly focused on the chosen plane of focus at a non diffraction limiting aperture AND the camera is rock solid steady, the actual resolution of obtained images will be always lower than the megapixel count.<br>

Based on DPreview tests we find the following effective resolutions of cameras with their best glass (primes, G1 used zuiko) in LPH-lines per picture height. In brackets print resolution<br /> for 10inch x 15 inch picture (E3 and G1 10x13 inch)<br /> <br /> In the 10-12Mpix range:<br /> Canon 5d 2000 lph (dpi 200)<br /> Nikon d3/700 2200 lph (dpi 220)<br /> Olympus E3 1800 lph (dpi 180)<br /> Panasonic G1 2300 lph (dpi 230)<br /> Leica M8 2400 lph (dpi 240)<br /> 20+ Mpix FF:<br /> Canon 5Dm2 2700 lph (dpi 270)<br /> Canon 1Dm3 2700 lph (dpi 270)<br /> Sony Alfa 900 2700 lph (dpi 270)<br /> Nikon D3x 2600 lph (dpi 260)<br /> <br /> If we choose to print a 10 inch x15 inch picture, which is effectively A3 size with "art" framing, the picture height resolutions in dpi will be the above figures divided by 10.<br /> 180-200 dpi of 5D and E3 slightly below standard. 220-240 dpi of D3/700, G1 and M8 at acceptable standard. All 20+ Mpix cameras at 260-270 dpi.<br /> <br /> The talk of 20+ Mpix cameras being able to print way bigger pictures than 10-12 mpix cams doesn't take into account the fact that the megapixels do not translate proportionally to increased detail even with the best prime lenses. They show only 15% more detail than eg. G1 , the rest of resolution is below Nyquist=blur. That given perfect conditions: accurate focus, no shake.<br /> What's interesting the size of the frame in the 10-12 MPix range does not correspond to resolution: G1 and M8 sensors are 25% and 56% of 36x24mm respectively, yet offer better resolution than FF 5D and D3/700.<br>

The conclusion is that only 15% of the full frame 20MP+ picture is real detail increase over the G1 picture. Uprezzing the latter in a good software allows to print the same image size as those from D3x, A900, 5D2/1D3, without much discernible difference in real detail. Other picture quality elements (noise, dynamic range, color) require of course a separate analysis.</p>

 

<p>Looking at the image on a computer screen is not the same as making a print of the image. A good print requires an entire chain of interlinked processes that must be carefully handled in order to arrive at the highest quality print possible.</p>

<p>The thread contains a lot of good information, but your question is not one that can be easily addressed with a single, quantified answer - mainly because only you know what is acceptable in a final print for your personal aesthetics.</p>

<p>As has been pointed out, if you are sending a printer anything other than its native format, then the printer driver is interpolating the image. When the printer driver is doing modest up-sampling or down-sampling, there is little, if any, degredation to the image. However, when the printer driver is asked to interpolate past a certain point you will see stair steps in the edges of diagonals, circles, etc. as the printer driver is rather primitive in how it upsizes the image.</p>

<p>For example, I use an M8, and if I make a print at an 18x29-inch image size, the printer driver would be seeing a file that was 137 ppi. I print with an Epson 9800 and the native resolution of the 9800 is 720 dpi. So, in this case the printer driver would have to interpolate the 137 ppi to the 720 dpi required by the printer. Printer drivers are best at moving the printing head, regulating the ink droplet size, and moving the paper - and not image interpolation. While the printer driver will generate the needed information, it is not very sophisticated in how it interpolates. </p>

<p>If you contrast the printer driver to raster image processor (RIP) software, or a program like Qimage - there is no contest, the RIP or Qimage will make a far superior print. And either will also make a better print than interpolation software like Genuine Fractals, PhotoZoom Pro, Imagener, Photoshop, etc. as the RIP software or Qimage size the final file to the printer's native resolution and do not generate the artifacts often seen with interpolation software.</p>

<p>However, even using a RIP or Qimage for printing requires specific image handling in noise reduction and sharpening to ensure the final print will be as good as possible - and the image handling is image-dependent - meaning, once again, an exact, single answer to your question is not possible.</p>

<p>As to your last question, "What size limitation are you seeing?" - with my M8, I think the limit for my aesthetics is the 18x29 inch image size. At that size, you can get 6-inches from the image and still see details without artifacts, or the "pixel putty" look of an over-interpolated image. I also use a 6x7 film camera and the M8 prints can be displayed along side the same size (vertical dimension = 18 inches) print from a 6x7 transparency, without the M8 image looking inferior. </p>

<p>At 20x32, the image is starting to look a bit soft with fine details getting blurred; at 24x36 I find the image noticably soft and textures and fine details degraded - unless, of course, you subscribe to the "viewing distance" answer - meaning a billboard at 8 dpi looks great from a 1/4 mile away. However, the viewing distance response doesn't work for me as I like to approach the photograph and see more and more detail until I'm about 6-inches from the print surface.</p>

<p>What you can do with a 4/3 camera can only be determined by testing several different types of images in final print sizes with the appropriate image workflow - and only you can evaluate the final prints and decide the limit based upon your aesthetics - the Canon rep can't do that for you.</p>

<p>I also shoot M8 (and G1) and the Leica picture despite its pitiful pixel count (10MP) is excellent at large format printing. I actually read somewhere an interview with a renowned gallery art printer (Luminous Landscape? not sure) who compared M8 picture to that of medium format, so he confirms in a sense the above post info.<br>

Again, it's not how many pixels, but their quality and workflow. Eg for portraits I find the RD1+Hexanon 60mm/1.2 combo unbeatable. Better than g1, better even than M8. And it's only 6 MP. Maybe my preference is for less detail with this kind of photography?</p>

<p>Thanks to everyone who took the time to respond. This has been a very informative thread that will help me as I continue the comparisons. Some of the more technical information is beyond me right now but I am learning and hope to catch up soon.</p>

<p>To clarify one point, I discussed this with the Canon rep because he's shot several different systems in his career from 35mm to large format film, worked for Kodak while early sensors were being developed, and is more involved with Canon's education program than sales. He shot Olympus film in the OM days and respects the company's commitment to their digital vision. His information was based on what he's seen in his comparisons and those of others.</p>

<p>I realize the resolution vs. quality argument rages on the web, and that ultimately it's my decision what system delivers the image I'm trying to produce. I'm starting to get really serious about my photography as a business. This comparison represents research for myself on whether the system I use might create limitations for me in the future, specifically print size and "acceptable" sharpness. Right now I'm sticking with what I'm using to continue learning perspective, composition, exposure and lighting. As Tommy says, I want to get the most out of every pixel I've got before even thinking of my system as a barrier to success.</p>

<p>You should also remember to consider that your Canon rep is a Canon rep. It is in his interest to sell you a full-frame Canon pro camera and if you give him any hints that you are curious about one and have the money to spare, he will undoubtedly steer you in that direction. You should also be aware that a very very large proportion of the worst myths I've ever seen regarding photography posted to this website have originated from the lips of salesmen. It's one thing to respect a man's opinions and another thing to believe everything he says. The basic point I mentioned above and others have talked about is that resolution is resolution. Any camera over 8MP will meet most people's needs who are printing below 11"x14". The question you should be asking yourself and others is if this camera (or any other) has acceptable noise for your planned usage. Noise (not resolution) is the deal breaker with any digital camera system. Cameras with larger sensors make less noise, although in-camera noise reduction software is specifically aimed at closing that gap.</p>

<p>deleted</p>