Adobe Community: Pre-Size Your PPI for Best Print

Noel Carboni

21,326 posts
Dec 23, 2006

Currently Being Moderated

Pre-Size Your PPI for Best Print - Once Good Advice Still Good?

Mar 14, 2012 6:52 AM

In another thread someone mentinoed pre-sizing image data to prepare it for best printing.

For a long time it's been "standard advice" to resize images so that the ppi is an even division of the printer's dpi, because some years ago occasionally one would run across printers that would produce poor results if you didn't - you might see jaggies in straight edges for example.

Thing is, computers have (not so) quietly been getting more powerful over time, and printer makers have been competing with one another to try to make their printers produce better results than the other guys. One way they've done this is by improving the quality of the algorithms in the printer drivers. Use of mega storage and high accuracy math, which was once taxing on older computer systems, is now standard practice.

So it's time to question the old rule of thumb.

Making a few assumptions about the many variables (what printer, what OS, what version of drivers, what application being used to print) , there seem to be several questions here:

1. Can the image resolution be too high, causing the printer driver to make bad decisions about what ink dots to lay down where on the paper?

2. Does it help or matter if the image PPI is an even division of the printer's DPI?

As I have done in the past, I set out to do some actual testing, to see if I can actually SEE anything to help answer these questions.

I created a sharp image to be printed at 3 x 2 inches: http://Noel.ProDigitalSoftware.com/ForumPosts/Ghirardelli.jpg

Then I printed it at 6 different resolutions (1000, 720, 567, 300, 200, and 100 ppi) by resampling the image, labeling it, printing from Photoshop CS5, and feeding the same sheet of HP Premium Plus photo paper through my older HP 932c inkjet printer 6 times. The printer was set to its highest quality settings, including 2400 x 1200 dpi mode. This was the result:

I then looked critically and as objectively as I could at the different images. Here are my observations:

Naked eye:

The four highest resolution images (1000, 720, 567, and 300 ppi) all seemed to have an equivalent high level of crisp detail.
I could not detect the inkjet dots. Smooth objects look smooth.

Jeweler's Loupe:

I could see significant reduction in the finest details in the 300 ppi print vs. the three higher resolution prints, and a slight reduction in the 567 ppi vs. 720.
At no resolution were any jaggies or evidence of aliasing visible.
The inkjet dot pattern was plainly visible, and it does differ between the different prints. But it was not possible to say whether one was "better".
Things seem to have a little more texture in the 1000 ppi print vs. the 720 and 567 ppi prints.

Macro Photo:

Lacking a high resolution scanner, I took photographs of the 6 different prints. Unfortunately, I didn't have the time to set up with my best lighting and lens combination, so I got some reflections off the glossy paper, and and at this resolution I can't really see the inkject dots in the photos. I want to repeat this when I can find more time to do it better. As I did these photos hand-held, I believe the variances between them could be slightly influencing the results. But I'm going to post them anyway, for you to see.

I could see ever so slightly more detail in the 720 ppi print vs. the 1000 ppi print, though from the size of the tiny dust/light reflections I think it may have just been the better focused. Note that this observation is not supported by direct observation through the jeweler's loupe, above.
The 1000 ppi and 567 ppi prints seems to have slightly more noise or texture than the 720 ppi print. Again, this might be issues introduced by the photography process, though I did note a possible increase in texture in the 1000 ppi print with the jeweler's loupe as well.
Beyond just the blurring, I could see some evidence that straight lines are not quite as straight in the lower resolutions (300 ppi and lower). This seemed more apparent than with the jeweler's loupe examination, and I wonder whether the Photoshop downsampling process could have introduced it.

Left to right, top to bottom: 1000, 720, 567, 300, 200, 100:

Conclusions:

Printing to my HP 932c inkject printer on Windows 7 x64

300 ppi is not sufficient to coax the best possible detail out of an inkjet printer. It appears a number in the vicinity of 720 or more is better, and this number could be much higher with modern very high resolution printers (mine's old).
Speed was no different in printing any of these - a modern computer can process a huge amount of data in the blink of an eye.
When a sufficiently high resolution image is printed (in this case 567 ppi or higher) I saw virtually no evidence that a particular ppi value is superior, for example an even division of the printer's dpi, though in hindsight I realize I should have prepared a 600 ppi image (duh). I will add a 600 ppi image before I re-photograph the results.
It's possible ever so slightly more texture becomes visible at 1000 ppi than 720 ppi, but it might be just noise.
Practically speaking, from looking critically at the results I could not see a reason to pre-size the image for a specific ppi value.

I encourage you to experiment and report your results with your particular combination of gear.

Your comments are welcome!

-Noel

1698 Views 32 Replies Latest reply: gator soup, Mar 16, 2012 9:29 AM

Replies

21,326 posts
Dec 23, 2006

Currently Being Moderated

1. Noel Carboni,

Mar 14, 2012 7:00 AM in reply to Noel Carboni

Report

I just printed at 600 ppi and examined the result. I see no evidence whatsoever of improvement of quality over the 567 or 720 ppi prints.

-Noel

|
Mark as:
Currently Being Moderated

2. Jeff_N,

Mar 14, 2012 7:24 AM in reply to Noel Carboni

Report

Noel,

I commend your effort on all of our behalf!!!
thank you...
My comment is not intended to put down your efforts, but to question if a new current model printer would yield different results, therefore alter your conclusions.

You said..
<< and printer makers have been competing with one another to try to make their printers produce better results than the other guys.>>

But you are using, (as you said: and this number could be much higher with modern very high resolution printers (mine's old), an older printer. So the results may not reflect the actual results of what a modern, new printer will show.

Just my thots....

JeffN

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

3. Noel Carboni,

Mar 14, 2012 7:28 AM in reply to Jeff_N

Report

Absolutely true, it could be model-dependent. But it seems unlikely a modern model will work worse than an older one, and to be fair my Windows 7 driver IS pretty modern, even though the printer itself is old.

Hence:
I encourage you to experiment and report your results with your particular combination of gear.

-Noel

|
Mark as:
Currently Being Moderated

4. Jeff_N,

Mar 14, 2012 7:34 AM in reply to Noel Carboni

Report

Actually I was thinking a modern model would work better (not worse) than an older printer.
With new algorithms, advanced dot placement / screening techniques, the results might show that a lower resolution, say 300ppi would look just as good as a higher ppi (600, 720, etc)

JeffN

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

5. Noel Carboni,

Mar 14, 2012 7:59 AM in reply to Jeff_N

Report

You're saying a newer printer driver would be able to make up detail that's not in the image somehow?

-Noel

|
Mark as:
Currently Being Moderated

6. Jeff Schewe,

Mar 14, 2012 9:24 AM in reply to Noel Carboni

Report

Noel Carboni wrote:

I encourage you to experiment and report your results with your particular combination of gear.

Here's my report...The Right Resolution

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

7. Noel Carboni,

Mar 14, 2012 9:34 AM in reply to Jeff Schewe

Report

That's good info, Jeff. I see you reached pretty much the same conclusion I have - that up to 720 ppi is not wasted on a printer.

This thread is more about whether DOWNsampling excess information for printing is needed (perhaps I should have been more clear about that), but yes, providing a true high resolution image to a printer if UPsampling is needed can be important. Did you find that the need for upsampling differs between printer drivers?

-Noel

|
Mark as:
Currently Being Moderated

8. Jeff Schewe,

Mar 14, 2012 9:53 AM in reply to Noel Carboni

Report

Noel Carboni wrote:

clear about that), but yes, providing a true high resolution image to a printer if UPsampling is needed can be important. Did you find that the need for upsampling differs between printer drivers?

Between drivers? No...between printers? Yes. Epson has one style of print head and Canon and HP have another. For Epson the numbers are 360/720 and for Canon/HP it's 300/600. These are the resolutions that the drivers report to the system. If you send data that is different, there is some sort of resampling being done. My rule of thumb is if your image is below 360 (Epson) or 300 (Canon/HP) upsample to 360/300 and then output sharpen. If your resolution is above 360/300 then upsample to 720/600 and then sharpen.

I would never suggest people actually downsample though...why waste the pixels?

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

9. Noel Carboni,

Mar 14, 2012 11:41 AM in reply to Jeff Schewe

Report

Jeff Schewe wrote:

I would never suggest people actually downsample though...why waste the pixels?

Exactly. There was a statement in another recent thread that downsampling to be an even fraction of the print dpi was important to do.

It might have been a misapplied extension of the advice to upsample. It's not been all that long that we've had big enough high resolution data that even makes downsampling a possibility.

-Noel

|
Mark as:
Currently Being Moderated

10. Hudechrome,

Mar 14, 2012 12:43 PM in reply to Noel Carboni

Report

Noel Carboni wrote:

Jeff Schewe wrote:

I would never suggest people actually downsample though...why waste the pixels?

Exactly. There was a statement in another recent thread that downsampling to be an even fraction of the print dpi was important to do.

It might have been a misapplied extension of the advice to upsample. It's not been all that long that we've had big enough high resolution data that even makes downsampling a possibility.

-Noel
I believe I was the one to make that statement, which was based on recommendations by an Epson Print expert at a seminar demoing printers. He showed to prints from the same file, one set at an even multiple of 720 and the other some random number. It was subtle but visible the difference. That was probably 5 years ago.

In the meantime, I have made extensive tests of prints on my Epson 3800 trying many combinations of single pass, hi speed, Super fine print (2880x1440) and down to the basic level.

Everything evened out at 720 dpi. At 360, which is where I output from ACR, I can make an 8x12 print with no resampling whatsoever. Upping that to 720 and pushing the printer hard (2880x1440, single pass on Canson Platine), I see a discernable difference in the smooth tonalities.

As I understood you from past conversations, you employ the maximum output size from ACR which in my case, would double the file size by upsampling, and if necessary, downsample from that. I am not comfortable doing that as a default operation, but perhaps Jeff S might step in here and clarify. After all, ACR does offer that option! But my file size now goes from ~70MP to 143 MP, cutting my storage capability by 1/2. It's not a trivial matter when two of us here can run 600 to 800 images in 1/2 day!

|
Mark as:
Currently Being Moderated

11. Lundberg02,

Mar 14, 2012 12:52 PM in reply to Noel Carboni

Report

On the one hand, recognition occurs at 3 line pairs per object height so that multiples of 360/300 is best, but on the other hand the printer "cells" are based on a 4x3 base ratio with some multiplication so that there might be an improvement in Epson at muliples of 480, and the Canon at 400.

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

12. Noel Carboni,

Mar 14, 2012 12:55 PM in reply to Hudechrome

Report

Yes, I always convert my files to 6144 x 4096 and work with them at that size. I find there are numerous advantages to doing so, with the disadvantage that you pointed out: They require more resources to process and they take more space when you save them.

I guess I have to ask, though... Why do you save 600 to 800 PSD files in half a day? How much work do you put into them that you can't get back just by reconverting from raw, and of what use are they to you in the future?

Speaking for myself, for most prints I just convert the image, sharpen, and print, with little or no touchup. For web publishing I convert the image, dowsample, sharpen, and publish. Into very few general photography images do I actually put any serious editing time, and for those I DO save PSD files (uncompressed no less). I suppose my technique is more like that of a Lightroom user, though I use only Photoshop. I know that I can re-convert a raw image again in the future, then possibly employ a better converter, and apply tools and skills I don't have today...

-Noel

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

13. Noel Carboni,

Mar 14, 2012 1:00 PM in reply to Lundberg02

Report

Lundberg02 wrote:

On the one hand, recognition occurs at 3 line pairs per object height so that multiples of 360/300 is best, but on the other hand the printer "cells" are based on a 4x3 base ratio with some multiplication so that there might be an improvement in Epson at muliples of 480, and the Canon at 400.

Lundberg, have you actually SEEN a difference in downsampling to "multiples of xxx"? I'd sure love to see some evidence that all the theory has an observable reality.

Note that we're talking about DOWNsampling here, from an image with excess detail. Jeff has nicely demonstrated that upsampling has merit.

-Noel

|
Mark as:
Currently Being Moderated

14. Hudechrome,

Mar 14, 2012 1:25 PM in reply to Noel Carboni

Report

Well, they don't all wind up as psd but many do. It still doubles the storage demand.

The volume mentioned is the shoot, not output. Dealing with even 10% as psd or tiff takes far more than 1/2 day unless they can be batch processed. Adding to that is that the images of interest go through DxO as dng outputs so I am rather careful these days!

|
Mark as:
Currently Being Moderated

15. Hudechrome,

Mar 14, 2012 1:28 PM in reply to Noel Carboni

Report

His numbers are not pointing in any direction, simply a math calculation that fits both up and down

|
Mark as:
Currently Being Moderated

16. TLL...,

Mar 14, 2012 1:37 PM in reply to Noel Carboni

Report

Very interesting timing for this test, since I just did something similar to find out the best balance of resolution and print speed for a new Epson 4900.

Per Jeff S "I would never suggest people actually downsample though...why waste the pixels?"

That's what I need to do, make 'check prints' for high res film scans and digital captures. The film scans are 800+mb and digital frames 311mb. I need 'just enough' image data so AT technicians and clients can see what they need to see and no more. For our old, now defunct ML500 that was easy since it printed @300ppi. With the Epson I have all these different quality settings (360, 720,1440 @speed, quality and hi Q). So I ran a test with my film image set up @9x9 x 360, 720 and 1440ppi - sending them to the printer @ speed, quality and HQ settings. The AT guys louped them and determined that the 360ppi @quality setting provided the best image for seeking out image details (like X targets placed along roadways) - at higher resolutions like 720 or 1440 @HQ setting the printer started to introduce 'noise' (?) especially in areas of fine detail or parallel lines (think of a metal roof from 2500 ft up).

All of this worked out great for me since I'm dealing with very manageable file sizes and print times of 1-2 min per image. I do miss the 10 second print times of the ML but I gotta say this Epson 4900 is a freak'in awesome machine. We'll see how it holds up because flying season is underway and I'll soon be doing 1000's of these darn things

|
Mark as:
Currently Being Moderated

17. Hudechrome,

Mar 14, 2012 2:02 PM in reply to TLL...

Report

The place I see an increase in quality going to 720 is in the smoothness of sky and clouds, (assuming the file is actually smooth!) and I run at 2880 output.

I mean, we are splitting hairs, but clients wanting fine prints are impressed by numbers!

I have a particularly fine shot of wine grapes (Shiraz, IIRC) in fall transition before harvest. The range of color, tonality and surface quality of the fruit was an ideal subject for testing and 720dpi was optimum. And the print size is about 5x7.

If anyone wants to come over, I have the proof! (No, not that kind of proof. The wine hadn't been made yet! )

|
Mark as:
Currently Being Moderated

18. Lundberg02,

Mar 14, 2012 2:15 PM in reply to TLL...

Report

Noel, I NEVER downsample.
TLL, I see you're sort of in the business I used to be in, only I was doing infrared video. At one time , I think I had seen the world in IR more than anyone else. You're evidently sending recon photos to analysts, where I was both shooter and analyst. We had to make significant changes to terminal tracking based on close examination of the last two seconds of video. The IR was not very high res but we needed every pixel. In the central area we used a scanning trick to double the real resolution for correlation tracking.
The Epson likes 360 vertically for sure because that's its vertical base unit.

|
Mark as:
Currently Being Moderated

19. Jeff Schewe,

Mar 14, 2012 2:55 PM in reply to TLL...

Report

TLL... wrote:

So I ran a test with my film image set up @9x9 x 360, 720 and 1440ppi - sending them to the printer @ speed, quality and HQ settings. The AT guys louped them and determined that the 360ppi @quality setting provided the best image for seeking out image details (like X targets placed along roadways) - at higher resolutions like 720 or 1440 @HQ setting the printer started to introduce 'noise' (?) especially in areas of fine detail or parallel lines (think of a metal roof from 2500 ft up).

If you want the finest detail from the 4900, you should downsample to 720PPI and print from the 4900 at 2880 with the Finest Detail button checked. For max quality, also uncheck High Speed. It does double print times because it prints uni-directional, but you'll get the highest detail possible that way. You don't want to use a higher resolution because the driver reports itself as 720 with Finest Detail on so higher resolutions will get downsampled to 720.

|
Mark as:
Currently Being Moderated

20. Jeff Schewe,

Mar 14, 2012 3:09 PM in reply to Hudechrome

Report

Hudechrome wrote:

His numbers are not pointing in any direction, simply a math calculation that fits both up and down

No, if you read the article, you would know that if the originals are under 360, upsample to 360. If over 360 (but under 720) upsample to 720. The only time one would EVER want to downsample is to get to the 720 (or 600 got Canon/HP) because as I said above, the print pipeline resamples the image to the reported resolution of the printer. I've found no real benefit to upsample native rez that starts under 360 to above 720...and I would never downsample a higher native rez down to 360. Not any more.

I do know for a fact that both Epson and Canon/HP printers are known to sometimes produce moiré patterns when the image detail frequency and the dither interfere...and it can happen pretty easily at resolutions other than 360/300 and 720/600. Upsampling eliminates that.

There is actually a sneaky way of printing to the 4900 at a 1440 reported resolution, but it's not really documented. The 49/79/9900 printers were designed for the proofing market and if you have a media settings set to one of the standard Epson proofing papers AND use Finest Detail, then the printer reports as a 1440 device. 1440 is needed to resolve actual halftone dots when making proofs. I've "slightly" tested this–if for no other reason to confirm the 1440 report–but I could see no photographic benefit.

|
Mark as:
Currently Being Moderated

21. Hudechrome,

Mar 14, 2012 6:59 PM in reply to Jeff Schewe

Report

That's reasonable, but the post itself simply mentioned multiples of 360/300. It didn't say anything else, or mention the article (the article is your link?)

I don't like to read more into what a person stated than what's there.

The dithering you mentioned is what happened with the Epson demo.

I have downsampled to 360 (from a multiple shot stitched image and it printed great on the 3800 at 8x10. 720 was better, however, when I went 16x20. IIRC, the resolution was somewhere around 800ppi at 16x20 so I set it to 720.

Jeff on my pot 10 I mentioned upsampling from ACR, which I believe Noel does but I generally don't. If one anticipates the higher resolution to be needed, is it better to do it then?

Another question which is a bit OT: when an image opens in PS from ACR, the extension still says nef or dng, but I assume that it is no longer RAW . It may be moot since I'll save it as a psd or tiff anyway.

Thanks.

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

22. Noel Carboni,

Mar 14, 2012 7:06 PM in reply to Hudechrome

Report

Hudechrome wrote:

Another question which is a bit OT: when an image opens in PS from ACR, the extension still says nef or dng, but I assume that it is no longer RAW . It may be moot since I'll save it as a psd or tiff anyway.

"It is no longer raw" is not really meaningful - Photoshop only edits documents. It may be that a document has come from some source, such as a file, and may be saved to one or more work products - files, but while you're editing it it's a document.

You're seeing the raw file name as the document name, which is entirely arbitrary. It just helps you know where the document came from. You'll note that if you hit File-Save it prompts you for a filename.

-Noel

|
Mark as:
Currently Being Moderated

23. Hudechrome,

Mar 14, 2012 7:14 PM in reply to Noel Carboni

Report

You san also not save it in which case the RAW remains . I assumed it is as you say but it's the ol' ***-U-ME!

|
Mark as:
21,326 posts
Dec 23, 2006

Currently Being Moderated

24. Noel Carboni,

Mar 14, 2012 7:20 PM in reply to Hudechrome

Report

Of course you can save it, but you have to save it AS something that Photoshop knows how to write.

-Noel

|
Mark as:
Currently Being Moderated

25. Jeff Schewe,

Mar 14, 2012 7:56 PM in reply to Hudechrome

Report

Hudechrome wrote:

It didn't say anything else, or mention the article (the article is your link?)

I responded to Noal with my comments reffering to an article I did for Digital Photo Pro magazine...and yes, that title was a link to the article.

|
Mark as:
Currently Being Moderated

26. Jeff Schewe,

Mar 14, 2012 7:58 PM in reply to Hudechrome

Report

Hudechrome wrote:

Another question which is a bit OT: when an image opens in PS from ACR, the extension still says nef or dng, but I assume that it is no longer RAW .

Of course not...it's been demosiaced and processed into Photoshop. The fact that is still has a .nef or .dng is to inform you where the file came from. When you save, it'll save as a PSD or TIFF depending of you choice...

|
Mark as:
Currently Being Moderated

27. Hudechrome,

Mar 14, 2012 9:04 PM in reply to Jeff Schewe

Report

I remember it from some time back. It's a good article.

|
Mark as:
Currently Being Moderated

28. Hudechrome,

Mar 14, 2012 9:06 PM in reply to Jeff Schewe

Report

As I said to Noel, that's what I expected to hear, but clarity has it's virtues.

Anyway, about the upsampling from RAW......

|
Mark as:
Currently Being Moderated

29. Lundberg02,

Mar 14, 2012 9:46 PM in reply to Hudechrome

Report

From what Jeff says, Epson has increased its baseline resolution to 720 vertical from the 360 they used in their first generation high dpi printers. That's good to know. They apparently use a 6x4 cell of dots to form each resolution element at the highest setting. Their five dot sizes and the number of dots in a cell give them excellent tonal and color control using the seven inks.

|
Mark as:
Currently Being Moderated

30. Hudechrome,

Mar 14, 2012 9:53 PM in reply to Lundberg02

Report

As far as I know, this applies to the Epson 3800 as well.

I'll have to check.

|
Mark as:
Currently Being Moderated

31. TLL...,

Mar 15, 2012 7:54 AM in reply to Jeff Schewe

Report

Thank you very much for the tips Jeff, I saw you @PSW 2011 in Vegas and you do know your prints. For day-to-day production of this particular product print times must be as short as possible, hence the downsampling and lower print Q settings. As long as AT and clients are happy and I don't fall asleep waiting for prints its all good. I'm looking for opportunities to let the Epson shine in the future but for now its my 'proofer'.

Lundberg02 - IR imagery editing and manipulation continues to be a bigger part of what I do, mainly in conjunction with traditional RGB as multi-channel imagery. (LiDAR is a big deal too). This presents its own set of challenges, hopefully Adobe will recognize the growth potential in helping the geospatial community with their products.

|
Mark as:
Currently Being Moderated

32. gator soup,

Mar 16, 2012 9:29 AM in reply to Jeff Schewe

Report

brilliant article, Jeff, it proves there's no substitute for actually knowing what you are talking about... (a jab at my own pitiful rants, no one elses)

|
Mark as: