Upsize image or reduce dpi/ppi?

[Vulcan]

I have an Epson 3880 and I understand that the native resolution of the printer is 360 ppi. If I want to make a 17" print I have the option to either resize the image (using NX2, CS5 or a plug-in like Perfect Resize 7 (I'm on a Mac so I can't use QPrint)) or to reduce ppi. Assuming that one would look at the print from a "normal" distance for that size of a print, what would be the best option:

1. resize the image based on 360 ppi
2. resize based on 300 ppi, or
3. reduce ppi to, e.g. 200 ppi and avoid resizing?

Is there a point (size/viewing distance) where either option takes advantage over the other or are they as good/bad?

Regards

Hans

 

[Growltiger]

I would resize the image so you have enough pixels for the native resolution of the printer. You can do this better in Photoshop in advance than the print driver can manage.

 

[TonyW]

You can resize the image to get the maximum ppi although for this size print you should not really need to - unless of course you are going to have pixel peepers that want to view your image from a few inches away :smile:.

As you asked about a size/viewing distance :smile:.....

Viewing distance will/can dictate your choice of printer resolution. Two things need to be established. First is the diagonal of your image to ascertain the normal viewing distance. Then the number of pixels per inch at that distance to give a sharp image with smooth gradation.

Be aware that any formula/calculations are only intended as 'rule of thumb' and as such a certain amount of interpretation of 'the rules' is certainly acceptable. These figures based on the assumption of viewer having 20/20 vision.

Viewing Distance
To find the viewing distance first find the diagonal dimension of the print and multiply by 1.5, for example a 20"x16" print:

20" squared = 400" + 16" squared = 256" Total = 656"
Square root of 656" = 25.60" (rounded) diagonal of print
x 1.5 = 38.4"

Therefore Viewing distance = 38"

Minimum ppi
Now to find the minimum ppi there needs to be enough pixels per inch into fooling the eye into seeing a smooth image at the viewing distance. This can be calculated by dividing a value of 3438 by the viewing distance anything above this should look acceptable at the calculated viewing distance

So for a viewing distance of 38.4" minimum ppi = 3438/38.4 = 89.53ppi

Maximum ppi
These figures from the Digital Workflow group gives and indication of maximum resolution at various distances. It is important to realise these figures are based on optimal lighting at exhibitions etc and not for a print hanging on the wall at home in average lighting conditions either daylight or artificial. The figures could be halved for these average situations

Viewing Distance (").... Resolution ppi
6 .................................. 1145
10................................... 687
24................................... 286
36................................... 191
60................................... 115

 

[gchappel]

I would never downsize. If you are above 180ppi you could upsample to 360 and, at least in my experience with the 3880 you MIGHT see a difference if you look for it. You might have to get out a loupe. The difference is not enough that if you look at one print today and another tomorrow you would remember the difference. Now if you are below 180ppi the wheels start to come off. I have used perfect resize, and find that works better than photoshop using bicubic smoother. With bicubic smoother I find it no better to upsample 10%at a time, like they used to talk about- just do it in one step. To be honest though, you cannot recreate what is not there and unless I am pushed for size I no longer print with a native resolution below 180.
That is my experience, hope it helps. Play around some yourself, your mileage may vary. Let us know what you find out.
Gary

 

[Vulcan]

Thank you all! I really like this forum, people are very quick in responding, knowledgable and nice (you don't see that very often on the internet). Just one additional question to Tony, where does the figure "3438" come from (i undertstand the reference to the 20/20 vision).

//Hans

 

[TonyW]

...Tony, where does the figure "3438" come from (i undertstand the reference to the 20/20 vision)...

Not my work (the math usually leaves me cold :smile Extract from Full article
[FONT=&quot]Where did "3438" come from? Given the visual acuity angle for ordinary viewing, the equation can be written as:[/FONT]
[FONT=&quot]1 ÷ PPI = 2 × V × tan( 0.000290888 ÷ 2)
1 ÷ PPI = V × 0.000290888
PPI = 3437.746747 ÷ V [/FONT]
[FONT=&quot]To show how this works in practice, consider a standard 6 by 4 inch print. It depends on the person, but these might typically be viewed at a distance of 15 inches.[/FONT]
[FONT=&quot]Basing the calculation on ordinary (typical) viewing conditions where the visual acuity angle is 0.000290888 radians (θ), the distance (D) between two details such that they are just distinguishable when viewed from a distance of 15 inches (V) is:[/FONT]
[FONT=&quot]D = 2 × V × tan( θ ÷ 2 )
D = 2 × 15 × tan( 0.000290888 ÷ 2)
D = 0.004363323 [/FONT]
[FONT=&quot]The reciprocal of that is the "details per inch" = 229. This would be the minimum "pixels per inch" required for printed details to be at the limit of visual resolution under ordinary conditions.[/FONT]
[FONT=&quot]A larger print, intended to be viewed at a distance of, say 30 inches, would require a minimum PPI of 115.[/FONT]
[FONT=&quot]A quick way to determine the minimum PPI resolution to print your photos is to divide 3438 by the intended viewing distance in inches. This is for ordinary conditions; double the number for ideal conditions.

[/FONT]Worth testing yourself to see if it suits your needs. Perhaps consider printing a section of your large print at different ppi and view all at the calculated viewing distance.

My experience with Perfect Resize different to Gary's. In versions prior to CS5 I used to use Genuine Fractals which certainly seemed to give improvement over PS. However with the introduction of CS5 I tried Perfect Resize and came to the conclusion that there were no benefits to be had. My testing:
Started with a RAW image (approx 4000x3000pix) I duplicated the image and cropped both to 800x600. Then applied PR to one set for 9000x12000pixels at 300 ppi which yielded a print size of 30"x40". Then applied image this took a long time to complete (cup of tea and biscuit!)

CS5 applied the same size crop, pixel count and ppi to end up with the same print size of 30"x40" (Bicubic smoother). The processing time for this so much faster than PR!! Both images ending up around 618Mb.

Examined side by side on screen both images looked very similar. [FONT=&quot]PS[/FONT] version just a little softer which was brought back with a modest application of levels layer. My feeling was/is that there was very little difference. When printed could not tell images apart. Incremental increases of 10% I could not see any real benefits although it may depend on image content.


All IMHO of course yours and others mileage may vary :smile:[FONT=&quot]
[/FONT]

 

[Vulcan]

All of a sudden it became very clear :smile:

 

[gchappel]

I am still on cs4 and find perfect resize a little better than bicubic alone. I do not find that in lightroom 3, so I guess adobe improved their algorithms. Looking forward to CS6. Printing through lightroom, I do not print native resolution below 180. At 180 or above I do upsample to 360, and if native resolution is above 300 I upsample to 720. Printing on epson 3880 I can tell the difference- usually need a loupe- noone else can tell so maybe it is not real. Only takes a second, so I still do the above.
Gary

 

[Retief]

Wow, Tony, thank you so much for the details. I do have a question, though.

In your example below you use a 20x16 print which yields 89.53 minimum PPI for a viewing distance of 38". Yet in your chart, it shows 191 for 36".

Oh, oh, I just answered my own question, I think. That is 191 in OPTIMUM lighting, so for us normal "living room lighting" folks, divide that by 2 and we get:
191/2=90.5, pretty darned close to 89.53.

Am I understanding this correctly?

This, then, is why so many online print services are happy with anything over 200PPI or so?

An education, I love it!

ps. My 3880 is scheduled to arrive next Monday.

As you asked about a size/viewing distance :smile:.....

Viewing distance will/can dictate your choice of printer resolution. ...
Viewing Distance
To find the viewing distance first find the diagonal dimension of the print and multiply by 1.5, for example a 20"x16" print:

20" squared = 400" + 16" squared = 256" Total = 656"
Square root of 656" = 25.60" (rounded) diagonal of print
x 1.5 = 38.4"

Therefore Viewing distance = 38"

Minimum ppi
Now to find the minimum ppi there needs to be enough pixels per inch into fooling the eye into seeing a smooth image at the viewing distance. This can be calculated by dividing a value of 3438 by the viewing distance anything above this should look acceptable at the calculated viewing distance

So for a viewing distance of 38.4" minimum ppi = 3438/38.4 = 89.53ppi

Maximum ppi
These figures from the Digital Workflow group gives and indication of maximum resolution at various distances. It is important to realise these figures are based on optimal lighting at exhibitions etc and not for a print hanging on the wall at home in average lighting conditions either daylight or artificial. The figures could be halved for these average situations

Viewing Distance (").... Resolution ppi
6 .................................. 1145
10................................... 687
24................................... 286
36................................... 191
60................................... 115

 

[TonyW]

...In your example below you use a 20x16 print which yields 89.53 minimum PPI for a viewing distance of 38". Yet in your chart, it shows 191 for 36".

Oh, oh, I just answered my own question, I think. That is 191 in OPTIMUM lighting, so for us normal "living room lighting" folks, divide that by 2 and we get:
191/2=90.5, pretty darned close to 89.53.

Am I understanding this correctly?

Bill, yes your understanding is correct, or perhaps I should qualify that by saying that our understanding is the same :wink:

As with anything of this nature a formula can only calculate a theoretical limit when certain parameters are applied. In this case the parameters rely heavily on the 'correct' viewing distance and 'normal or known' lighting conditions present. Formulas are unable to make aesthetic decisions and also fail to take into account what happens when the formula conditions are not adhered to.

For instance in my case I do confess to being a pixel peeper so when I see a displayed print I cannot help myself and want to get a close up view. So unless you can prevent a viewer (like me :smile from getting too close to a print in your home environment you may want to consider getting close to the higher ppi for this eventuality. Alternatively a small sign in you image that when seen up close reveals the words 'you are too bloody close! :biggrin:

I think Gary hit the nail on the head "Printing on epson 3880 I can tell the difference- usually need a loupe- noone else can tell so maybe it is not real. Only takes a second, so I still do the above".
The difference observed up close or with the aid of loupe is satisfying perhaps to us but may not be important when the print is being viewed under more 'normal' conditions.

Good luck with the new printer and have fun :smile:

 

[Retief]

Bill, yes your understanding is correct, or perhaps I should qualify that by saying that our understanding is the same :wink:

Good luck with the new printer and have fun :smile:

Thanks, Tony, much appreciated. While I am not a pixel-peeper, I'm way too lazy :wink:, I do believe that the higher PPI will obviously yield a smooter image, so whenever possible I agree that is the way to go. What your post really did for me was to quantify the "why" of the whole thing, in a way that I could understand.

There is so much "urban legend" about this stuff that having some hard resource to point to really helps.

The other side of this is that having more PPI than the printer will natively handle is obviously not going to do you any good either, correct? Just like color space and gamut. So hooray for soft-proofing!

 

[TonyW]

...The other side of this is that having more PPI than the printer will natively handle is obviously not going to do you any good either, correct?..

Yes, that is I believe the case but needs to be qualified as PPI is not really the correct term for a printer resolution. Many of these terms get mixed up and misused even by manufacturers who should know better! I am as guilty as the next person and frequently seem to change without good reason ppi for dpi or vice versa :redface::smile:.

So in an effort to clarify based on my understanding:

Pedant Mode [On]
DPI is most commonly used referring to printer resolution. As applied to the printer one dot or droplet means the smallest possible ink droplet the printer can lay down on paper.

A droplet in printer terms is not actually a pixel or even a halftone dot size but just a blob of ink. It can take several ink droplets of each colour to make up a full colour tone pixel. That is the reason you need a printer with a much higher dpi than your image files measurement in ppi to get a good continuous tone.

Technically dpi is not really the correct terminology either for an inkjet printer. Ink is laid down in volume usually measured and quoted as picolitrers (a picoliter is a trillionth of a liter, which can be represented numerically as 0.000000000001/liter. - Had to look that one up! :smile
Pedant Mode [Off]

In the case of your Epson 3880 the smallest droplet size is 3.5 picoliters and it can produce variable droplet size of up to 3 different sizes per print line. The quoted max resolution being 2880x1440 dpi.

In contrast to this your old HP B9180 best mode 4800 dpi x 1200 dpi droplet size unknown! So what have you done swapping for an inferior resolution printer :tongue::smile:

Only joking but this is a bit like the megapixel myths that float around e.g. more is better!
From what I understand you cannot really measure IQ based on just the dpi count and I did see a comparison somwhere stating to get similar IQ to Epson you needed to double the dpi count when printing via the HPB9180 - wish I knew where I saw that.

You may find in fact that there is no point in printing past 1440 dpi as the advantage in going to 2880 dpi is not worth bothering about but it will keep Epson happy with ink sales :smile:. I do not know this as a fact but I am sure that your printing experiments will highlight what is needed

 

[Free Range Otter]

Sorry to bring up a zombie thread, but I just found Tony's post above and it is very informative. I have one question, though - does the minimum PPI needed depend on the media? In other words, how would the minimum PPI change if I were ordering a canvas print from WHCC versus a paper print? I'm assuming canvas requires fewer PPI but would that be a correct assumption?

 

[7006]

re inkjet dots.
I seem to remember that there are different technologies at play too.
HP's multilayers the the colours on one dot and claims to have the ability to generate wider colour gamut per dot.
Others (but I do not remember if this applies to Epson) fool the eye by placing different drops of different primary colours next to each other, as opposed to HP's on top of one another.

 

[7006]

Here is the HP description
http://h10025.www1.hp.com/ewfrf/wc/document?lc=en&dlc=en&cc=us&docname=bpu02182

 

[TonyW]

..I have one question, though - does the minimum PPI needed depend on the media? In other words, how would the minimum PPI change if I were ordering a canvas print from WHCC versus a paper print? I'm assuming canvas requires fewer PPI but would that be a correct assumption?

I do not print on canvas but as canvas is usually far from a smooth media then any lack of detail by printing at a lower ppi would be lost in the surface texture - to a degree.

So I think there is likely to be a point where at the 'correct' viewing distance the difference between a 300 ppi print on smooth or glossy media would seem to match a lower ppi canvas print. I do not know the point at which this will happen but it maybe that 300ppi on smooth fine art paper equates to around 180-240 ppi on canvas. The texture of the canvas will also have a bearing some seem much heavier than others

But whatever the size of print consideration must be given to the common distance that this print is likely to be viewed at (ignoring the pixel peepers like me :wink and also the lighting conditions.

As it looks like you would order canvas prints from an outside source it is probably preferable to seek their advice first