How to ensure no resampling during printing?

<p>Hi.<br>

I have large scans of 35mm film. I mean quite large, up to 10344*6888, 48 bits per pixel. This sometimes includes a number of pixels that can/should be cropped out. The point is, that the files are large. In fact they are above the effective resolving power of the scanner, and the film itself, so they can be shrunk without loss of detail, if done properly.<br>

My question is, once I prepare them for printing, I wouldn't want them to be further resampled in the photo lab. I mean, I'd like to do all the processing myself, and then just have the photo lab print them. Now the problem is I worry that lab printers may have a specific pixel count for a given size, and will either up- or downscale any photo that doesn't have the exact dimensions. So I'd like to feed them exactly the pixels they need, or at least an approximate number if they do have some tolerance. But since 1) I don't really know how they work, 2) I don't know their exact DPI, 3) print size options are often only approximate, I'm at a loss.<br>

- Does anyone know how lab printers operate in this regard?<br>

- Can I consult their DPI specs anywhere?<br>

- Can they work with 48-bit material?<br>

- Should I look for printer-specific profiles or use sRGB? Is Adobe RGB useless? CIELAB?<br>

- In case one can create chemical prints from digital material, does any of the above worries apply?</p>

<p>My first guess would be: downscale to 300dpi (2400px height for a 12x8 print), embed 24-bit sRGB, cross fingers.<br>

These are an awful lot of questions, but I'm thankful for any info. I suspect experience dictates wisdom here and I have little of the former. Thanks!</p>

<p> </p>

<p>This may be useful to you: http://www.drycreekphoto.com/icc/using_printer_profiles.htm</p>

<p>If you have questions about a specific lab's process, it would be worth your while to ask them. </p>

<p> </p>

<p>I don't think you ever need to resample a file downward in order to print it.</p>

<p>@ Mendel - I think the OP needs to downsize to make the file size manageable. 10,344 X 6,888 X 48 is a whole lot of bits to be moving around and sending to a print vendor.</p>

<p><Chas></p>

<blockquote>

<p>My question is, once I prepare them for printing, I wouldn't want them to be further resampled in the photo lab</p>

 

</blockquote>

<p>Ask to lab the PPI value for their printer and resample using this value for the print size.</p>

<p>I've read, more than once, that printer drivers and RIPs do a better job of interpolation than Photoshop. So the advice usually is to let the printer handle it by just giving it the unresampled data. That way it only gets messed with once instead of twice since the lab is going to process the file anyway.</p>

<p>If you need to reduce the file size beyond the conversion to 8bits per channel, then resampling by an even divisor (such as 2) should give you the cleanest result.</p>

<p>Images <strong>should</strong> be sampled to a fixed pixel dimension based on the output device to size because output sharpening is necessary and that’s based on size (see http://www.creativepro.com/story/feature/20357.html). <br>

There is no reason to worry about Photoshop doing this (sample down, use BiCubic Sharper). <br>

Few if any printers (drivers) will handle more than 24 bit color. Some of the newer Canon and Epson ink jets do. The benefits if any are nearly impossible to see. High bits are useful for <strong>editing</strong> to overcome data loss so you can send the best 24 bits to the printer. <br>

There is no such thing as an sRGB (or for that matter Adobe RGB (1998)) printer. These are synthetic RGB working spaces for editing images. They are not output color spaces. You need an ICC profile of the printer to get to that space (and soft proof, decide on rendering intent etc). See:<a href="http://www.adobe.com/digitalimag/pdfs/phscs2ip_colspace.pdf">The Role of working spaces in Adobe applicaitons</a><br>

You or someone scanned the images, what color space are they in (scanner RGB tagged with some profile hopefully). Convert from that to the output color space if possible. Many, many labs will not provide this essential profile or if they do, they will then tell you to send them the data in sRGB (stupid non color managed workflow). </p>

<p>@ Charles: I see now.</p>

<p>Learn to WORK WITH the service you are using to make the print. If you are using a high quality service, they will tell you exactly what they need in a file to make the best print. Depending upon the output device, the requirements will vary. An Epson printer needs one thing, a Canon or HP another, and a Ligthjet or Lambda something else. If you are trying to do this on the cheap using a low cost printing service you won't see much of a difference if you just resample (as Andrew Rodney has suggested) to a multiple of the device's native format. </p>

<p>As an example, the native format for an Epson printer is 720 dpi so you could resample to 360 ppi. If it is a Canon or HP printer the native format is 600dpi so you could resample to 300 ppi. If it is a Lightjet or Lamda, you'll have to contact the service because the machines work at different dpi settings and you'll have to know what they're using.</p>

<p>However, you have to understand that if you're sending the data to an inkjet printer, at a multiple of the native format, the printer is still doing some interpolation to get it to 720 dpi, and then applying dithering - so the printer driver is never really out of the equation unless the printer is being sent data by a RIP, and then the RIP is doing the interpolation to the device's native format.</p>

<p>Thank you all, and thanks Don and Andrew for the links. And Steve for the advice. I haven't yet settled on a lab, but there's one or two in my area that maybe will turn out to be good. It's something I hadn't thought of, but if indeed the lab is good, the people will be pleased to give you the info you need. <strong>And</strong> complementing that with what I can learn here, I think I may be able to get exactly what I want. Conversely, if the people in the lab are unable or unwilling to talk to you, then the odds are that they aren't the best bet for a quality service.</p>

<p><Printers rant on:></p>

<p>Unless you know the target printer's hardware and software and the person running it and how their settings are set and software and hardware versions; there is *no answer to a question like this.*<br /> <br /> Here I am a printer</p>

<p>*****follow Steves words: "Learn to WORK WITH the service you are using to make the print."</p>

<p><br /> Here our scan back from 1997 makes a 50 megapixel image. A common 24x36 color map scanned 400 dpi with our 36" RGB scanner is 9600 x 14440 pixels; our scanner 10 years ago did this. Your image is large; but really nothing new. It was new say about in 1996 when our best computer held 1 gig of ram and had a 200Mhz cpu on a server board.<br /> <br /> Even us printers who have worked with digital for over 20 years do not know what a new printer or rip does inside.<br /> <br /> The long term trend is to have a doofus filter inside the black box. Thus if you send a 9600 x 14440 pixel image to make a 4x6" print the doofus filter chops down you useless pixels. The 9600 pixels cannot be mapped on the 4 inch width; a lessor amount is used.</p>

<p><br /> In a lay sense if is like going to Home Depot and have them cut up a 4x8 foot piece of plywood into 6 pieces to build a dog house; but you give them dimensions of 24.000 inches by 32.000 inches. Their cuts may be only good to 1/10 inch; they ignore your 1/1000 inch wishing. If you make a big deal about the 1/1000 inch value; you get talked about later; we had one that was confused.</p>

<p><br /> In a way today everybody is an expert; when the long term trend is the average input is worse. <br /> A typical input like " 10344*6888, 48 bits per pixel 48 bits per pixel" comes across as oddball; ie warning. It is odd because printers (hardware/softwre) do not usually 48 bits. It is like bringing in something on a jazz drive; or some other fool thing to get like missing fonts; or some password. ie we have to get your file to be normal number of bits; something that printers use.</p>

<p><br /> If the 10344*6888 pixels is for a small print; this is less odd. With some rips the useless pixels are dumped and the internal file is a practical one that works. If you give your accountant your income data to 1/1000 of a cent; he will think you have a screw loose; since he will round all your data to cents; or even dollars on the returns.<br /> Every printer that has accepted a digital input for the last 20 years has limits. As the public's usage of digital has grown; the average input is stupider.</p>

<p>Today a lot of a printers time is with "talking about printing". The printer drivers and printers have added doofus filters that make absurd inputs be practical ones. The software on some settups is with sane/rational input; the filter is not used. If you breach the limit; the filter can hack it down and actually make a poor copy.</p>

<p>Thus if one sends a 16x20" image at a practical 300 ppi to an ACME printer; it might sail right through. If you send a 16x20" image at 1800 ppi; the doofus filter might hack it down to 240 ppi or less; and the saved file in the RIP is less than the 300 ppi.</p>

<p>In the photo.net world; you want a simple model; there is only one 50mm lens; only one beer; only fishing lure; only one rip and printer.<br /> <br /> In reality it is super complex; and there can be software bugs too.</p>

<p>Some print set ups when the bloaded file is hacked down to size; the software weanies did not use the printers current calibration; they used the prior one. Thus bloaded file that gets hacked in the RIP can in weird cases make a poorer print due to software issues; your bloaded file got sent to the dunce area for downsizing and they software is applying the wrong color calibrations. I have seen this twice; on different machines.<br /> <br /> I have seen where if I send a TIFF and JPEG to the same printer/rip; one gets messes with at XXX ppi and the other at YYY ppi.<br /> <br /> Like fixing a car if you take up the car repair guys time all the time being an expert; he will charge you more. The same goes in printing. If one guys input requires a mess of stuff; micromanaging; zillion goofy questions; you charge him more.</p>

<p><br /> What we really want is a decent input. Folks who bring in bloaded inputs tend to to have bloaded egos and want a black and white 1 bit model to a complex question. Thus makes dealing with certain customers is difficult; they no nothing about the ins and outs of using the stuff; but want to dicate how to use it. the long term trend is worse every day. At some point there is no use scanning that average 35mm sunset or cat image at a zillion dpi! :) <br /> Look at plumbing versus printing; you tell the plumber what you want; in printing you tell the plumber to level hot water pipe to a nano-radian!:)<br /> <br />****Tattoo Steves advice on your arm :</p>

<p>*****"Learn to WORK WITH the service you are using to make the print. If you are using a high quality service, they will tell you exactly what they need in a file to make the best print.".</p>

<p>The BEST customers of ours tell us what they want; the WORST dictate how to do it with tools they really know only enough to be dangerous<br /> <br /> Any input that is "In fact they are above the effective resolving power of the scanner, and the film itself," is a giant red flag.</p>

<p><br /> ***What you want is the best print and NOT tie our hands with goofy requirements that reduce the final prints quality. have too many pixels is NOT the best thing aways. Resampling upward or downward can often make a better print</p>

<p><br /> <printers rant off></p>

<p>Saying:<br /> "How to ensure no resampling during printing" is like saying:<br>

<br /> (1) you only want a plumber to use blue wrenches; or a<br /> (2) wedding photographer has to use iso 100 film; or a<br /> (3) author/writer has to use a computer or a<br /> (4) math major has to use a pen only<br>

<br /> You just box in the persons control of their task at hand. It means you think you know more than they do; the things they do everyday. It starts off the customer/client relationship in a bad way.</p>

 

<p>Hi Kelly,<br>

<intrigued><br>

So now that you've basically restated my problem - my files will need processing, and I'm unsure offhand what is the best input to give the photo lab -, do you have any particular advice, given your experience in the field?<br>

(I'm also wondering how could I get information from the photo lab people if they abhor questions as you say.)<br>

(But you also seem to say they don't know either what their equipment does, though I guess they can always look at the output and form an opinion.)<br>

You mention using 9600*14400 files. My impression was that printing equipment wouldn't accept files that large. Do you mean to say that a Lightjet, for instance, will be glad with the 10344*6888? Because in that case there would indeed be little point in downsampling them.<br>

And I'm really sorry that those files are above the effective resolving power of my scanner and film, not to mention the cameras that took most of the shots, but unfortunately that's the way it goes. Not all of us scan Nikon-shot Velvia on an Imacon.<br>

</intrigued></p>

 

<p>So what you're saying is that if they need a file that's 5k pixels wide, and I can give them exactly that, since I have a 10k pixels original, I should instead give them a 4k or 6k pixels one, and let them resample it since it's their job and they resent the customer interfering with it and asking questions. Doesn't sound intuitive, but...</p>

<p>Again, ideally you size the image to what you want on output, based on a PPI (DPI for output) the printer will use because output sharpening is based on this <strong>exact</strong> size (and output device) and sending them more than necessary is useless and excessive in terms of file handling. <br>

Proper output sharpening is going to produce a far more helpful effect on your print than sending them more than 24 bits by a large amount. Sending them data sized and in the appropriate color space will as well. </p>

<p><br />"I have large scans of 35mm film. I mean quite large, up to 10344*6888, 48 bits per pixel."</p>

<p>Antonio, how were these scans produced?</p>

 

<blockquote>

<p>"I have large scans of 35mm film. I mean quite large, up to 10344*6888, 48 bits per pixel."<br>

Antonio, how were these scans produced?</p>

</blockquote>

<p>Hi, Don. With Silverfast and this: http://www.filmscanner.info/en/ReflectaRPS7200.html, which seems to be the same as this: http://www.scanace.com/en/product/pf7250pro3.php, which looks like an mkII of the Kodak RFS 3600.<br>

It doesn't do anywhere near 7200dpi, of course. With luck, it goes almost to 4000 when in scanning at 7200. Scanning at 3600 yields a little less. The dynamic range is imo quite reasonable if one uses multi-exposure, and one reason to use 48 bits. ICE works ok, again imo. Scanning time is actually small.<br>

Of course, it's hard to justify 8x the size of 3600/24 just for a little more detail, but size is the thing that gets less important year after year. Conversely, very little if anything of a good 3600/24 scan is bloat.<br>

Negative strips are not mounted but just fed through a motor. It can scan whole strips in batch mode, in fact whole rolls if you have them uncut. No wet mounting, of course, and only good for 135 film. You can scan 126 film, but the batch mode doesn't seem to work, and it will be cropped - needlessly, since the sensor reads from top to bottom, so it could perfectly read the whole 35mm if they had just thought of that. As it stands, it can read almost but not exactly 36.5*24.3mm.</p>

<p>"Of course, it's hard to justify 8x the size of 3600/24 just for a little more detail"</p>

<p>Is there a little more detail? My guess is the scanner has added pixels which contain no additional information -- 'upsampled', 'padded' the image. </p>

 

<p>Annoyingly, few of the commercial printing services will tell you <em>exactly</em> what pixel dimensions to send for a specific print size. The only one I know of that does is Winkflash; see <a href="http://www.winkflash.com/content/pro.asp#2">http://www.winkflash.com/content/pro.asp#2</a>. For example, if you send them a 2456 x 3070 pixel file and request an 8 x 10 inch print, presumably they, their software, and their hardware do absolutely no resampling or scaling of any kind.</p>

<p>I'm sure that Mpix and Shutterfly and whomever could tell you the same thing. I've explicitly asked them to. But as far as I know, they have not.</p>

<p>Now be aware that all of these printers print a hair beyond the borders of the paper, to avoid possible white edges, so not all of the pixels you send will be printed, regardless of how many you send. Also, the sizes and resolutions are nominal. I suspect that Mpix's nominally 250 ppi printers are really 254 ppi (10 ppmm), and Shutterfly's nominally 300 ppi printers are really 304.8 ppi (12 ppmm). Also, that 8 x 10 inch might really be, say, from a 200 or 250 mm-wide roll, and really be, say, 7.87 x 9.84 inches (200 x 250 mm). Between resolution, paper-size, and full-bleed / over-print issues, it's pretty hard for you to know unless they tell you explicitly.</p>

 

<p>Antonio;<br>

I suggested after Steve that you deal with the printer you are going to have print your stuff.</p>

<p>What you are asking is for others opinions on the size file to send to an unknown printer; in an unknown town; using an unknown coarseness of paper; with an unknown type of image; using a file yourself said is bloaded; ie has useless pixels. </p>

<p>Re " In fact they are above the effective resolving power of the scanner, and the film itself"</p>

<p>The really is no black and white anwers to questions like this; your image might just be a 10344*6888 that really holds only 1034 by 688 pixels; since you shot has some blur. A local printer can tell you this; we cannot; none of us have a crystal balls. Thus you might actually be sending in a file that is 100 times too big; ie has 100 times the pixels.</p>

<p>Typically customers who get all wrapped up in pixels worries and micromanaging a service bureau have teh worst inputs; bloaded and have little soul.</p>

<p>There is really not rational way to say your 35mm original has 10344*6888 pixels worth of info. I have scanned slides since 1989 and *NEVER* seen one have that amount of info; that is a 7200 dpi scan. Only a B&W Microfilm piece of film gets near this; with a great lens at F8 with the rig on a granite block.</p>

<p>Useless pixels do not matter.</p>

<p>Printers want customers to get out of their heads all the gobble gook of making bloaded files with useless info,</p>

<p>One can make a better giant poster out of a iPhones cellphones great image; than yours if your 10344*6888 image is really just poor and thus has only 400 by 300 pixels worth of info.</p>

<p>What matters is the usefull pixels; not the made up bloaded ones</p>

<p>THERE ARE A MESS OF UNKNOWNS:</p>

<p>(1)a bloaded digital file of unknown bload; probably at least 5 if a tack sharp best case original; maybe by 20 if an average one; maybe by 100 if a fair or slightly blured one.<br>

(2) unknown printer<br>

(3) unknown printer settings; high quality; average; draft<br>

(4)unknown RIP or driver settings<br>

(5)unknown paper type. Super coarse canva might only support 60 to 100 ppi; fine glossy 400 to 700 dpi.<br>

(6)unknown viewing distance<br>

(7)unknown viewing light level<br>

(8)unknown purpose of the final print<br>

(9)unknown budget<br>

(10)unknown color balances and printer calibrations<br>

(11)unknown printer ink if inkjet; dye or pigmented<br>

(12)unknown life of final output<br>

(13)unknown after unknown</p>

<p>****You are basically asking a public forum how to bring raw food to an unknown master cook to cook a GREAT unknown meal; with an unknown kitchen; with unknown tools.</p>

<p> Thus again here is STEVES comment:<br>

"Learn to WORK WITH the service you are using to make the print. If you are using a high quality service, they will tell you exactly what they need in a file to make the best print. Depending upon the output device, the requirements will vary."</p>

<p>If you are scared; then ask them; they want to make you happy. If you want to tie the cooks hands; you might get a plate full of poor food. If you bring in a bloaded 10344*6888 pixel file to a printer and it really only looks like a VGA image; he should tell you that it does not have much usefull info. It is really quite common; many folks somehow thinking bloaded files help.</p>

<p>When you ask another cook what type of stuff he needs; or printer it often is different. There are many thousands of printers in use and kitchens too.</p>

<p>In the old days with films; one did not spend an hour talking about enlarging lenses; staples; how to adjust for focus; how a safelight works with each print customer. Today the major expense in prinring is often NOT the inkjet ink or paper; it is the customers all wrapped up in a tar ball over pixels. Today everybody is an expert; even on another shops settup they have never used. Thus when customers image for a 24x36" print really only is about 50 ppi in details; many folks want it printed at 600 dpi. As Truman said you cannot polish a turd.</p>

<p>Find a local printer; focus on providing unbloaded file, Try to focus more on a image with impact instead of pixels. There is a direct correllation better poor images and pixel worry warts. Folks who worry the most about pixels create the worst inputs; all the brains power is on what mostly doent matter.</p>

<p> Try a local printer; talk to them; get samples made; that is what matters.</p>

<p>Thanks again.<br>

To Don: yes, the detail is there (provided the film has got it, of course), I've measured it, and I can see notice it. The reason scanners don't ever achieve their nominal resolution hasn't necessarily to do with interpolation but is that the effective resolution is the product of several factors, only one of which is the nominal resolution. Optics, sensor layout, focus, motor precision, all of these influence the final result.</p>

<p>The sensors are in a linear array that moves along the film (let's say from top to bottom):<br>

- horizontal resolution is obtained in each line by reading the value captured by all the cells and resampling that to the desired output resolution. If one uses the same resolution as there are cells, then hopefully no resampling is done, but good luck finding the specs for the number of cells in the array. Anyway, this may not matter as much as it seems, because regardless of the number of cells, they will overlap in their reading of the original, and that will ultimately determine the sharpness of the scan.<br>

- vertical resolution is given by the motor steps, but here again the overlap of the cells' readings will determine it. If you ask for 7200dpi, then samples are taken at each 1/7200 inch (that's why scanning takes longer, but only 2x the time of a 3600dpi scan, not 4x). But since the readings overlap - not with those of the adjoining cells, in this dimension, but with those made by the same cell in the stops before and after, you never get as many true dpi as there are motor steps (notice in this case it's not a matter of interpolation, though the end result is similar).</p>

<p>So the only way imo to determine whether a given DPI setting is worth it is 1) to measure it with resolution targets <strong>and </strong>2) to look at the pictures and decide whether you like them or not. In this case, the 7200 scans do have a bit more detail than the 3600 ones (and the 48 bits do have their use), and do look better. The real issue is whether the admittedly small improvement is worth it. The only drawbacks are scan time, which isn't very important and file wieldiness, which will improve with time.</p>

 

<p>Hi Kelly,<br>

Thanks for all that effective info (each lab person is different - each equipment is different - different situations require different approaches). I don't quite understand what the remaining 96% (mostly about p*x*ls) have to do with what I asked, but certainly that's just me.<br>

Again, you seem to imply that lab equipment can eat large files. In that case I'd be glad to leave all resampling to the lab. But can it really? From what I read here (thanks again to all of you with your links and info) and elsewhere, most machine have certain limits.</p>

<p>Ask your lab what they require. If your scanner will give you 4000ppi of real resolution from 35mm film then send them 4000ppi scans or smaller if you are making small prints like an 8x12inch print. If you want to make a 40inch wide print give them the full 4000ppi scan and let then do the rest. They will either let the rip resize it or they will use software to resize it. There is no point in sending a 7200ppi scan if it is bloated.</p>

<p>Hi Dave,</p>

<blockquote>

<p>Now be aware that all of these printers print a hair beyond the borders of the paper, to avoid possible white edges, so not all of the pixels you send will be printed, regardless of how many you send.</p>

</blockquote>

<p>I've heard that some people introduce padding or margins in their image files to account for this. What's you own opinion?</p>

<blockquote>

<p>Also, the sizes and resolutions are nominal. I suspect that Mpix's nominally 250 ppi printers are really 254 ppi (10 ppmm), and Shutterfly's nominally 300 ppi printers are really 304.8 ppi (12 ppmm). Also, that 8 x 10 inch might really be, say, from a 200 or 250 mm-wide roll, and really be, say, 7.87 x 9.84 inches (200 x 250 mm). Between resolution, paper-size, and full-bleed / over-print issues, it's pretty hard for you to know unless they tell you explicitly.</p>

</blockquote>

<p>Yes, I've come to that conclusion. I guess that one must just accept such imprecisions with the change of medium.</p>

<p>António what exactly are you trying to get printed as it seems you are making this more complicated than it needs to be. I assume you have negs that you have scanned and you want to make some prints. When you print say an 8x12 from a 35mm neg you will usually lose something from around the edge but unless you had a camera with a 100% view finder you did not see it when you shot the photo in the first place. One easy way would be to give your neg to the lab and let them do the whole job.</p>