How many DPI do I need?

[dreed]

I'm getting ready to venture off to the local printer to get some prints made and before I do, I started to think about how big the print will be vs DPI and pixels.

What made me pause was this.

The bigger a print is, the less sense it makes to be up close to it because you stop being able to take in the entire image properly and if you're not as close to the print then the DPI print resolution doesn't need to be as high (unless you're going to "pixel peep" a print with a magnifying glass, etc.)

A good example of what I mean here is HDTVs. The bigger a TV is, the further back you sit (generally speaking.) A 50" HDTV is ~42DPI - even lower than your computer monitor. If you were to double that for UHD, you're at about 80DPI - or to be generous, 100DPI. That's a third of what we print at - 300DPI - but people are not saying that their 50" isn't sharp, rather that they can't see what the difference is unless their nose is against the screen (well not exactly but I hope you get what I mean.) Incidently, the 4K TV is only 8MP... well, maybe if you thought about the TV resolution in camera sensor terms, a 4K TV would be 24MP (each TV pixel is a RGB and each triplet is only counted once, whereas in your camera, every pixel is counted so the RGGB is four pixels, not one.) But the point remains, if you display a 2MP image on a 1080p TV on a 60" screen (with sub-100DPI output), people will still sit/stand back and go "wow".

So printing out a card or A4 sized document at 300dpi makes sense because the way we interact with it is fairly intimate. But if you print out to A2 or A1, the relationship with the image isn't nearly as intimate. This would suggest (to me) that the output doesn't need to be as fine (150DPI would be perfectly ok?)

Unfortunately I don't have my own printer to experiment on, so I'm reaching out to the community and asking for some guidance.

Besides the two essays below, are there any worthwhile resources to read?

http://www.luminous-landscape.com/tutorials/understanding-series/und_resolution.shtml
http://www.luminous-landscape.com/tutorials/sharpness.shtml

Something that I'm also curious about is if I wanted to print an image at 200DPI, is it better to first scale up the image in Photoshop (increase the size in each axis by 50%) so that a 300DPI image gets sent to the printers (and their printer driver) or will whatever print software/drivers do that for them or ...?

The important point here is that it won't be me doing the printing, rather I'll be giving the output to someone else to print so I need to get the output right the first time or it becomes a rather costly endeavour.

[Mark D Segal]

You don't say how large you want your prints to be, so it is hard to give you specific guidance on the floor of PPI (NOT dpi) to send to the printer.

There is tons of information on this that a comprehensive Internet search will bring up; however, if you are looking for specific hands-on advice from people experienced with making prints, my general advice (baded on what I've done with a string of Epson professional printers over the years) would be as follows:

For anything printed at least up to 13*19 inches, printing upward of 240 PPI is generally a safe place to be.
For much larger prints, I would be reluctant to print below 180 PPI; visible degradation of image quality really starts to break-out below this threshold.
It is preferable to do any resampling in your photo editing application (Lightroom, Photoshop, etc.) and give the printer a print-ready file in this respect.
It would be good to discuss with the printer the machines they are using and their experience with resolution outcomes on their equipment.
It would be good to soft-proof you files for tonality and colour rendition using a paper/printer profile that best matches what your print service provider will be using - this is a general purpose precaution to avoid disappointment when you get the results back.

[Manoli]

What made me pause was this.

The bigger a print is, the less sense it makes to be up close to it because you stop being able to take in the entire image properly and if you're not as close to the print then the DPI print resolution doesn't need to be as high (unless you're going to "pixel peep" a print with a magnifying glass, etc.)

The Right Resolution, Jeff Schewe

[dreed]

My goal is A0 or A1 sized prints - A2 at the smallest. Why so big? Because once it is mounted on a wall above a couch, table, desk, etc, the minimum viewing distance will be at least 50cm to 1m, so A3/A4 are a bit small. And thus the question and tradeoff of *PI vs output print quality.

As for web searching, if it doesn't show up on Google's first page then it doesn't exist

This page has a table that presents "Viewing distance vs Minimimum PPI vs Max Print length" which is kind of where I was heading:

http://www.photoreview.com.au/tips/outputting/how-large-should-you-print

[Paul2660]

If you are having them printed on an Epson, I would strongly recommend, 360dpi for larger 18 x 24 or larger and 720 for smaller.  If you print from CC, then the CC/driver combination will do an uprez using Nearest Neighboor, which is the worst possible, and you have no ability to sharpen during the printing.  Epson needs 360 or 720.  LR allows you to set the dpi and add sharpening the printing which is most beneficial since uprezing is being done at the LR level, then being sent to the driver.  This is much better than letting CC do it.  

If you are printing to an Canon it's 300dpi.  But same rules apply.

These rules apply to inkjets, not sure what your printer is using, a RIP, subdye etc. as for those 2 rules are different.

But letting the driver/CC do the uprez is just not the best solution.

You can also uprez the file to the correct dpi before you send it, and there are plenty of tools out there there for that.  CC 2014 with it's latest algorithims does a good job as does Photozoom pro.  I never have and still don't see any advantage to Genuine Fractals (or whatever it's now called by Topaz unless you are making a billboard).  

Since I started 100% of printing from LR, I find my results on larger prints just look better to my eye.  Just set the dpi to 360 or 720 (for Epson) and let LR do the rest.  Again not sure what your printer company will use, but they should also be very very well versed in this if not I would not be using them.   The only things I uprez outside of LR are canvas prints which by nature have more issues than paper prints, if they are to be gallery wrapped.  If they are to be substrate mounted, I go straight to LR.

Paul

[dreed]

The Right Resolution, Jeff Schewe

As usual, Jeff excels. Thanks for that link.

[Paul2660]

Or pick up "The Digital Print"  by Jeff.

Paul

[Bart_van_der_Wolf]

I'm getting ready to venture off to the local printer to get some prints made and before I do, I started to think about how big the print will be vs DPI and pixels.

What made me pause was this.

The bigger a print is, the less sense it makes to be up close to it because you stop being able to take in the entire image properly and if you're not as close to the print then the DPI print resolution doesn't need to be as high (unless you're going to "pixel peep" a print with a magnifying glass, etc.)

Correct. You can use my free on-line tool, section 1 is all you need, to judge the required PPI for a given viewing distance, and it will also tell you how the resulting perspective will look when the image was taken with a certain focal length. That will allow to better plan a shot if you have not yet shot it.

Something that I'm also curious about is if I wanted to print an image at 200DPI, is it better to first scale up the image in Photoshop (increase the size in each axis by 50%) so that a 300DPI image gets sent to the printers (and their printer driver) or will whatever print software/drivers do that for them or ...?

No, it's better to upsample than to let the printer do it. There are two reasons for that. The first is that you can use much better resampling algorithms (although +50%) is perhaps not enough to make a big difference, and the second reason is that you can apply output sharpening at the printer's native resolution, and get a much better quality than doing it before the printer upsamples while using a mediocre algorithm.

Cheers,
Bart

[simplify]

PPI or DPI as some people wrongly say it - doesn't matter in my opinion.  I can print a 100ppi file (uprezed to 200ppi) from a really sharp lens/camera combination and it could be sharper than a 300ppi file from a camera with a cheap lens.

The proof is in the pudding.  Ask your shop to make some tests strips for you at the largest size they think it can go, or better yet a few different sizes.  Then pin them up on the wall and view from the distance the finals will be viewed from.

Making thousands of prints a week for 10 years has taught me that everybody has a different view of print quality.  What may work for some, may be unacceptable for others.

[Schewe]

PPI or DPI as some people wrongly say it - doesn't matter in my opinion.

Yeah, ya know, I can't let this stand. The PPI you print at absolutely makes a difference...as indicated in my article The Right Rez (maybe you should read it), if you send any PPI to an inkjet print, your image will end up being resampled using (at best) Bilinear (and more likely Nearest Neighbor) which as very, very poor resampling algorithms. You are far better off resampling to the printer's reported DPI requirement either 360 or 7020 PPI for Epson or 300/600 for HP & Canon. If you don't do this, your print output is compromised...upsampling and sharpening at the native resolution of a printer is demonstrably better. Now, if one doesn't care about getting the optimal result, sure, fine, go ahead and send any old PPI. Again, read the article, the proof of what I'm saying is there.

[PeterAit]

Yeah, ya know, I can't let this stand. The PPI you print at absolutely makes a difference...as indicated in my article The Right Rez (maybe you should read it), if you send any PPI to an inkjet print, your image will end up being resampled using (at best) Bilinear (and more likely Nearest Neighbor) which as very, very poor resampling algorithms. You are far better off resampling to the printer's reported DPI requirement either 360 or 7020 PPI for Epson or 300/600 for HP & Canon. If you don't do this, your print output is compromised...upsampling and sharpening at the native resolution of a printer is demonstrably better. Now, if one doesn't care about getting the optimal result, sure, fine, go ahead and send any old PPI. Again, read the article, the proof of what I'm saying is there.

There is unfortunately a lot of confusion between "different" and "better." I read Jeff's article and indeed he demonstrated a clear difference between printing with this PPI and that PPI. But, the differences were shown by extreme blowups of print details, and really, who gives a bat's bunion (other than the pixel-peepers). The differences are real, in the technical sense, but do they make a print better in the aesthetic sense? After all, we are supposed to be photographic artists and not techo-nerds trying to squeeze every tiny bit of sharpness out of our photos (well, some of you may be, but not me). A dull photo that is super-sharp is still a dull photo, and a beautiful photo that is not super-sharp is still a beautiful photo. Let's face it, sharpness is easy, and anyone with half a brain who puts some money and effort into getting sharp photos will be successful. But, beautiful and meaningful photos? That's something else, and can be a life's work. Don't get me wrong, I have nothing against sharpness per se, but - as my sig line says - it should be a means to an end and not the end itself.
 

[Mark D Segal]

Peter,

Jeff is right. Optimum technical printing quality is best obtained using the native resolution of printer and getting there in the photo editing software, not the printer driver. Test after test confirms this.

You are right, insofar as depending on the image and the paper used, and the size printed, one may have a hard time when looking at a photograph on paper perceiving a visual quality difference within a range of 240 to 360 or more.

It is valid to make the technical representation Jeff made, and the practical one you made. They are different subjects with different purposes.

But the bottom line: why sacrifice technical quality if you don't need to? Yes, come a certain print size there may be trade-offs between resampling to achieve optimum resolution, vs simply lowering the resolution with relatively less resampling. But that territory needs hands-on case by case assessment. I think what we're trying to do here is provide practical general guidance to the OP. Ultimately he will need to make decisions and sometimes there are trade-offs, other times not.

[Schewe]

Don't get me wrong, I have nothing against sharpness per se, but - as my sig line says - it should be a means to an end and not the end itself.

So, is it too hard for you to resample to 360/720 (for Epson, 300/600 for others). If you are printing from Lightroom, all you need to do is click a box and enter the resolution you want...LR does the resampling and sharpening on the fly. Seems a small price to pay for getting an optimal output from your printer...now, as to what you want to print, that's an entirely different subject. But to fail to take all the steps (that are pretty easy) to make your printed image look as good as it can, well, sorry, that's a mindset I don't agree with.

[Paul2660]

To me, there is a difference in the use of LR and letting it upsample/downsample to 360dpi, over say using CC/ or CS6 to print from/with the driver.   At first I thought it was just on large prints 23 x 33 and larger, but actually you see it in small prints also, 11 x 14 etc.  Crisper looking details for sure.  But also, in noise.  One issue I have always had with CS3 or CS6 printing is that noise in an image can be made to look worse, as if the driver is not sure what it is and makes it look much worse.  Example, is an image at 240 dpi, printed in CS6 to the Epson driver, vs same file printed from LR at 360.  I can see a difference. 

The only area I would like to see LR improve on is alternative for the the uprez engine, as you don't get any choices currently. 

On the flip side, I still will work up an image in either Photozoom pro, or CC 2014 to 300 dpi, and then let LR take that to 360 on the print.  More often than not this is for canvas prints, where I am having to create special edges for a gallery wrap. 

The knowledge that the native dpi of Epson is 360/720 to me is key to a good result, and printing straight from CC or CS 3,4,5,6, you don't have that control as the driver is making the uprez. 

Paul

[Dan Wells]

I thought the big printers were 300 or 360 dpi devices when printing photos - I wasn't aware the doubled resolution existed (with the possible exception of text and vector graphics) above 13" printers. I believe the smallest drop sizes are found on desktop printers, some of which have resolutions as high as 2400x2400 or even higher for text. I believe that there is at least one desktop printer which claims to print photos at 1440x1440 dpi (and has a claimed text resolution of something like 5760x1440)? This raises the question of whether the mechanism is actually capable of that precision? Sure, the software lets you address that many dots, but can the hardware put them down in the proper relationship to each other? I'd guess a 7900 is actually much more precise than a desktop Epson with a higher claimed resolution (I'd be far more confident in a 200 lb chassis holding everything in proper alignment than a 14 lb piece of plastic!). The desktop printer with the smaller droplet size can squeeze more dots in, but the 7900 gets them in the right place? Anyone confirm or deny this? Also, I'd think HP's page-width inkjet printers (none yet with >4 colors) might be unusually precise - they only have to deal with one axis of motion...

[Mark D Segal]

Dan,

Your post seems to suggest you are commingling considerations of printer dots with PPI output resolution. I assume you know they are not the same thing. You can't infer much from the dots per inch statistics in any event. These machines have very sophisticated dithering and droplet placement algorithms that defy any simple inferences from the these specification numbers. These factors would also determine accuracy of dot placement. And I don't think the heft of the machne has much to do with it. As long as the print head can travel properly and the paper is held properly it should place the dots just fine whatever it's made of. Perhaps more robust construction and stronger materials will be conducive to longer machine life for equal usage...........another issue.

[Paul2660]

I thought the big printers were 300 or 360 dpi devices when printing photos - I wasn't aware the doubled resolution existed (with the possible exception of text and vector graphics) above 13" printers. I believe the smallest drop sizes are found on desktop printers, some of which have resolutions as high as 2400x2400 or even higher for text. I believe that there is at least one desktop printer which claims to print photos at 1440x1440 dpi (and has a claimed text resolution of something like 5760x1440)? This raises the question of whether the mechanism is actually capable of that precision? Sure, the software lets you address that many dots, but can the hardware put them down in the proper relationship to each other? I'd guess a 7900 is actually much more precise than a desktop Epson with a higher claimed resolution (I'd be far more confident in a 200 lb chassis holding everything in proper alignment than a 14 lb piece of plastic!). The desktop printer with the smaller droplet size can squeeze more dots in, but the 7900 gets them in the right place? Anyone confirm or deny this? Also, I'd think HP's page-width inkjet printers (none yet with >4 colors) might be unusually precise - they only have to deal with one axis of motion...

Most times I am using 360dpi, but on smaller prints, I will push the dpi to 720, 8x10 11 x14 or 12 x 18, much larger and I stay at 360.  With the 79/9900, at the max 2880/1440 I feel there is a slight difference, enough that I use it.

Paul

[simplify]

Schewe, Your a god, I am a follower.

I am not sending any old PPI, I always uprez to at least 200 ppi for my canon ipf printers if the native ppi is under 200 then for larger files 300ppi and 600ppi are sent to the printer.  What I was talking about in my response was the native ppi before uprezing for print.  Also in the Canon IPF plugin you can determine rendering intent for uprezing.

That being said I have done a ton of testing printing at native PPI below 200 then uprezing to 300 and 600 with different softwares and in the final print there is very little if any difference compared to the canon plugin uprezing.  Epson really needs a plugin, it saves a ton of time for shops that produce a lot of prints.

My main point is make a test print or strip at size and determine with your own eyes what you think will work.  People get caught up on the numbers when its really about the final print. Proof is in the pudding

I see it everyday, people supply native 300 ppi files shot with crappy lenses, then others supply files that were shot with cameras with no AA filter and Zeiss lenses that are 100ppi at the size they want to print.   The later would be much sharper after uprezing than the former.

[Bart_van_der_Wolf]

I am not sending any old PPI, I always uprez to at least 200 ppi for my canon ipf printers if the native ppi is under 200 then for larger files 300ppi and 600ppi are sent to the printer.

Hi,

Why stop at 200 PPI? All it takes is to change the upsampling scale. Doesn't cost money, and gives better quality (assuming proper resampling and output sharpening).

That being said I have done a ton of testing printing at native PPI below 200 then uprezing to 300 and 600 with different softwares and in the final print there is very little if any difference compared to the canon plugin uprezing.  Epson really needs a plugin, it saves a ton of time for shops that produce a lot of prints.

Strange, I have also done such testing (as several others in this thead have also done), and usually the quality improved by using proper tools and technique. Sure, paying attention to achieve the best output quality is not always necessary for large volume printing operations, but why deliberately reduce quality? That would not be an operation who I would use/recommend for taking care of my output.

My main point is make a test print or strip at size and determine with your own eyes what you think will work.  People get caught up on the numbers when its really about the final print. Proof is in the pudding

Again, why reduce the taste of the pudding, instead of even improving it?

I see it everyday, people supply native 300 ppi files shot with crappy lenses, then others supply files that were shot with cameras with no AA filter and Zeiss lenses that are 100ppi at the size they want to print.   The later would be much sharper after uprezing than the former.

Aim for low quality, and low quality is what you'll get. Comparing poor technique with other poor technique doesn't make sense. The thread is about good or even best practice.

Output sharpening after upsampling can make a significant visible difference to the quality of the output. To get the most of it, it helps to use proper technique all the way from capture to print. It won't save images of mediocre artistic quality, but it won't hurt them either.

Cheers,
Bart

[JRSmit]

Peter, it is not just sharpness, it is overall resolution, also of fine details, nuances in color, smoothness, etc. If the image you print is dull, then the print should also give this dull feeling. It is not however a justification to let go the quality level of the print.
personally i print at highest resolution , unless image is really low in pixels per inch.
A good image to start, combined with high quality papers, it then is as if the image comes to live.
And yes, an image with enough pixels printed at 720 plus fine details looks better at normal viewing distances . Not shockingly different, but visible. This is true for epson and for canon (600 finest).