4800x2400 dpi. How many ppi is that?

Forums: 
That's the DPI of the Canon 9000 Pro Mk II. It's nice to know how many dots per inch, but I'd really like to know how many pixels that is. How many printer dots per pixel?

Well that may be a variable nuber of DOTS per IMAGE PIXEL sue to the variable dropplet size. Why does that matter at all all long as the printer dithers well enough to be able to print smooth tonalities and dots are not visible at standard viewing distances.

jtoolman wrote: Well that may be a variable nuber of DOTS per IMAGE PIXEL sue to the variable dropplet size. Why does that matter at all all long as the printer dithers well enough to be able to print smooth tonalities and dots are not visible at standard viewing distances.
Yup - any modern printer is very good - if the print looks good I'm happy - I leave the rest to engineers.

probert500 wrote:
jtoolman wrote: Well that may be a variable nuber of DOTS per IMAGE PIXEL sue to the variable dropplet size. Why does that matter at all all long as the printer dithers well enough to be able to print smooth tonalities and dots are not visible at standard viewing distances.
Yup - any modern printer is very good - if the print looks good I'm happy - I leave the rest to engineers.
I AM an engineer.

fft81 wrote:
probert500 wrote:
jtoolman wrote: Well that may be a variable nuber of DOTS per IMAGE PIXEL sue to the variable dropplet size. Why does that matter at all all long as the printer dithers well enough to be able to print smooth tonalities and dots are not visible at standard viewing distances.
Yup - any modern printer is very good - if the print looks good I'm happy - I leave the rest to engineers.
I AM an engineer.
I just play one on TV.

probert500 wrote:
fft81 wrote:
probert500 wrote:
jtoolman wrote: Well that may be a variable nuber of DOTS per IMAGE PIXEL sue to the variable dropplet size. Why does that matter at all all long as the printer dithers well enough to be able to print smooth tonalities and dots are not visible at standard viewing distances.
Yup - any modern printer is very good - if the print looks good I'm happy - I leave the rest to engineers.
I AM an engineer.
I just play one on TV.
In RL 90% of us, the engineers, are not even 10% as geeky as TV personalities are. Also, 90 percent of what we do on regular bases only requires about 10% of what we learn in school.

ThrillaMozilla wrote: That's the DPI of the Canon 9000 Pro Mk II. It's nice to know how many dots per inch, but I'd really like to know how many pixels that is. How many printer dots per pixel?
Don't confuse the two. 4800x2400 is simply the maximum number of ink droplets per inch that the printer will use to reproduce the image on paper, regardless of what the image pixel density is. You could send it a 150 pixel per inch image to print at 8x10 (1200x1500 pixels) and the printer would use 4800x2400 ink drops per inch to produce it. If you send a 300 pixel per inch version of the same image (2400x3000 pixels), the printer will still use 4800x2400 ink drops per inch. The image will be more detailed because the source image has more resolution. Mark

As stated, you are confusing DPI, DOTS/inch, a print resolution density with PPI, Pixel/inch, a camera/monitor resolution measure. To make the point..you could print a SINGLE pixel file, at 1 PPI resolution, a 1 inch square.....at 2880 x 1440 DPI. Figure that one out and you will understand it.

Heres a little more on the same subject. Perhaps the most important reason that the printer has this ridiculously high dpi is that it has so few colors to work with. To simulate MANY MILLIONS of colors and tones on paper, the printer dithers (or mixes) different sized and colored blobs of ink at various spacings(density). And these printers do it in a way to produce very delicate details and textures. This goes off on a slight tangent, but every printer has a native resolution which is vastly different from the maximum addressable ink dot placement of the printer. This native resolution is perhaps as much printer DRIVER specific as anything else. What happens is that an image at whatever PPI is sent to the printer driver and it gets internally resampled up or down to this native resolution. There are those here that might disagree but I believe that it is best to furnish garden variety printers with some evenly devisable fraction of that native resolution. For example my Epson R1900 has a native resolution of 720dpi so I typically feed it 360ppi out of PS. If I’m printing structured graphics that need to be rastered first, it’s typically at 720ppi. It’s not necessary, of course, because the printer driver will upsample from any PPI anyway but in some limited situations it is advisable for us to do so manually. And it also allows us to better optimize sharpness for print. BTW Canons native res is 600ppi. So I’m throwing the OP one more curveball to ponder: 1. PPI at the output image level. 2. PPI as the native printer resolution (this is the new concept) 3. DPI as a printer addressable ink dot placement. Bruce

Folks, I am not confused about the difference between dots and pixels. I really do want to know how many PIXELS per inch this printer will print.

ThrillaMozilla wrote: Folks, I am not confused about the difference between dots and pixels. I really do want to know how many PIXELS per inch this printer will print.
None. It prints only dots, not pixels. To underline the helpful "curve ball" post above, the printer goes through two steps. First, it accepts the image file and resizes it to pixel dimensions that equate to the specified print width and height using the printer's native conversion factor, such as 600 ppi for your Canon. Second, the printer performs an elaborate process calculating dots of each color and their placement. Each pixel becomes a perhaps varying number of dots. The printer's dpi specification is an approximate statement of how many dots it lays down along an inch of paper. My own practice is to do step one for the printer, that is, resize it using the native printer conversion ppi. Then I sharpen it as desired (generally, until it verges on oversharpened on the monitor). The justification is that Picture Window Pro, which I use instead of Photoshop, does these better or at least as well as the algorithm in the printer. Step two, of course, is left to the printer.

ThrillaMozilla wrote: Folks, I am not confused about the difference between dots and pixels. I really do want to know how many PIXELS per inch this printer will print.
I'm sorry, but you are confused. Your printer will use 4800 ink droplets per inch in one direction and 2400 ink droplets per inch in the other. It doesn't matter if the image sent to it is one pixel per inch or 300; it will use the same amount of drops per inch to create the image. However a good rule of thumb is to size your source image to a maximum of 300 pixels per inch, so if you're printing an 8x10 the image should be 2400 pixels by 3000 pixels. Mark

ThrillaMozilla wrote: Folks, I am not confused about the difference between dots and pixels. I really do want to know how many PIXELS per inch this printer will print.
The answer to your question is actually fairly simple: either 300 PPI or 600 PPI. If you print photographs, stick with 300 PPI. If you print line art or text, there are advantages to 600 PPI. (With that said, I don't know which Canon printers will actually do 600 PPI and which will not.) If you print photo quality... regardless of the actual pixel dimensions to start with, the driver will in effect resample the image to 300 PPI (it will actually output an entire line of ink dots, not pixels, but it will interpolate at effectively 300 PPI). Note that at 2400 DPI, that means an image will be made using one 8x8 dot matrix for each pixel. The 4800 DPI rate is probably only used for high resolution 600 PPI printing, and still uses an 8x8 matrix except the size will be half for each dot. If you are super critical about your prints... resample the print to 300 PPI and use a 100 percent crop from the image to preview while adjusting sharpening. Send the result to the printer. Anything else means you cannot preview the effects of the sharpening, simply because the pixel dimensions will effectively be resampled by the print driver if it is not at 300 PPI.

apaflo wrote: The answer to your question is actually fairly simple: either 300 PPI or 600 PPI. If you print photographs, stick with 300 PPI. If you print line art or text, there are advantages to 600 PPI. (With that said, I don't know which Canon printers will actually do 600 PPI and which will not.)
I agree. Odds are that this Canon has a native resolution of 600ppi. Lets take a look at what is happening at the paper level. You should view it full size: IMAGE(http://dpzen.com/dpzattaches/dpzattachesm//2014092509075335826.jpg) Here is an old example I did of my now defunct Epson R800 output. This test section was printed at 720ppi (the native resolution of most Epsons) and thus the driver did not up-sample or down-sample the image but massaged it to produce this PRINT that has in-turn been scanned at a very high resolution. You can get a feel for the size of some of the ink dots and how many of them were sprayed on the paper to make this print. That one inch arrow is approximately 2400 pixels across in the file and on your monitor if you look at it full size. You really need higher resolution scans to truly resolve the individual ink dots, especially the smallest ones. There is a HUGE disconnect between the 720ppi sent to the printer and what we actually see here on the paper because the printer driver is doing a lot of Voodoo behind the scenes when it massages that 720ppi image. If I understand your question properly, that native resolution is the number you are looking for. And, again, on Canon it would (probably) be 600ppi Bruce

To put it in simple terms... The printer duplicates color by placing a number of drops of different color ink in close proximity to represent one pixel. A single pixel (ppi) printed could need 6-8 different drops of ink (dpi) to represent it. That is why the printer's dpi cannot be confused with ppi and that is why the printer has such high numbers. General accepted convention is that printing beyond 240 ppi the human eye reaches a point of diminishing returns when it comes to differentiating resolution. Beyond 300 ppi the human eye for the most part can't further differentiate resolution. Printing above 300 ppi technically produces higher resolution prints...you just can't see the difference. Beyond 300 ppi you are just essentially wasting ink. Some printers, due to design and other considerations, get their best performance at 360 ppi (Epson). This is a function of the printer and how it operates. Canon uses 300 ppi and I believe HP does also. You should not exceed the manufacturers recommendation. Large prints do not need 300 ppi. There is a print resolution/viewing distance relationship that needs to be taken into consideration. I.E. - you would view an 8 x 10 at about 10" (300 ppi). You would view a 16 x 20 at a few feet (240 ppi). Both print resolutions would look the same to your eye.

kenwj wrote: To put it in simple terms... The printer duplicates color by placing a number of drops of different color ink in close proximity to represent one pixel. A single pixel (ppi) printed could need 6-8 different drops of ink (dpi) to represent it. That is why the printer's dpi cannot be confused with ppi and that is why the printer has such high numbers.
For any given configuration there is a fixed matrix of ink dots per image pixel. And regardless of what the original image dimensions were, the image is effectively resampled to the native PPI rate at which the printer is configured. An example would be that an image with pixel dimensions of 1920x2400 if printed with a physical size of 8x10 inches is one you would say has been printed at 240 PPI. But in reality, nothing ran at 240 PPI in the printer! Instead each line of pixels was gathered up as a single data set, and converted and dithered to a line of data for printing at 2400 dots per inch in increments of 300 matrixes per inch. Hence in effect the actual print operation is done at 300 PPI (and instead of 1920x2400 pixels there are 2400x3000 pixels), and each of those pixels is related to an 8x8 matrix of ink dots. Each pixel is made up of 64 dots of ink.
General accepted convention is that printing beyond 240 ppi the human eye reaches a point of diminishing returns when it comes to differentiating resolution. Beyond 300 ppi the human eye for the most part can't further differentiate resolution.
That is roughly true, but it applies to photographs viewed at the "usual" distances. It is also true that text and line art drawings are very visibibly sharper when printed at 600 PPI or even higher. One practical effect of that, noticeable at least with Epson printers, is that Office printers and Home printers, are the ones with extremely high resolution modes (720 PPI and 5760 DPI), while their commercial grade photo printers only print at 360 PPI and use a maximum of 2880 DPI. (And none of the wide carriage units can even feed sheets smaller than 8x10 and even the large sheets have to be hand fed one at a time.)
Printing above 300 ppi technically produces higher resolution prints...you just can't see the difference. Beyond 300 ppi you are just essentially wasting ink.
As noted, true for photographs but not true of other applications.
Some printers, due to design and other considerations, get their best performance at 360 ppi (Epson). This is a function of the printer and how it operates. Canon uses 300 ppi and I believe HP does also. You should not exceed the manufacturers recommendation.
The Canon and HP printers simply print at 300 PPI, and the Epson's print at 360 PPI. Any image that is not at the appropriate dimension for that rate at the requested physical size is effectively resampled, by the print driver, to the appropriate pixel dimensions.
Large prints do not need 300 ppi. There is a print resolution/viewing distance relationship that needs to be taken into consideration. I.E. - you would view an 8 x 10 at about 10" (300 ppi). You would view a 16 x 20 at a few feet (240 ppi). Both print resolutions would look the same to your eye.
That is correct, but it would lead to less confusion if it were stated as not needing 300 original pixels per inch. That is, the number of pixels in the image before the print driver massages it to actually print at the 300 PPI rate used by the printer.

Thank you sir. Good information. Do you have any experience with a Canon Pro 1. I just bought one and haven't opened the box yet. I am not a novice at printing. I am looking for anybody with experience building custom profiles using the 'Matt Photo' media setting with fine art papers to avoid the forced 35mm margins you get with the fine art media setting. I did this with my old Pro 9500 and it worked well. I don't know if I can get away with this on the Pro 1.

kenwj wrote: Thank you sir. Good information. Do you have any experience with a Canon Pro 1. I just bought one and haven't opened the box yet. I am not a novice at printing. I am looking for anybody with experience building custom profiles using the 'Matt Photo' media setting with fine art papers to avoid the forced 35mm margins you get with the fine art media setting. I did this with my old Pro 9500 and it worked well. I don't know if I can get away with this on the Pro 1.
Sorry, I've never used any of the Canon printers, so I have no specific knowledge about any of them.

kenwj wrote: General accepted convention is that printing beyond 240 ppi the human eye reaches a point of diminishing returns when it comes to differentiating resolution. Beyond 300 ppi the human eye for the most part can't further differentiate resolution.
The physiological limit of the eye is expressed here in ppi, and that is what the eye views looking at a monitor. When looking at a print, the eye views dots, not pixels. We do not see a single dot, of course. The point is that the inkjet printer goes through different processes when you send it a 360 ppi file versus a 720 ppi file - and the eye sees different results.

Charles2 wrote: When looking at a print, the eye views dots, not pixels. We do not see a single dot, of course. The point is that the inkjet printer goes through different processes when you send it a 360 ppi file versus a 720 ppi file - and the eye sees different results.
To clarify, the amount of detail is limited by the number of pixels, not the number of dots. If you have an image with fine detail, and you resize it from 4000x6000 pixels to 2000x3000 pixels, you will probably lose detail. You can never recover the lost detail, no matter how many dots the printer puts on the paper (although it doesn't necessarily mean that you will be able to see the difference). You may be able to improve the appearance by smoothing lines and edges, but you cannot truly recover lost detail.

TM, Yes, I think you have got all of that right. Note, your wish to produce the exact BEST print possible is what most of us try to do, depending, most of the time. Your assumtions that up-sampling to 720 ppi or higher are NOT necessarily true even to beyond-human-sight capacity. Set up your on test patches, as I did in my ppi-print resolution thread...and look at them yourself....with a magnifying glass. Also note, you can and SHOULD save your image at the native ppi resolution of the image, which will almost always be SMALLER than the file at PRINT-RESOLUTION to the printer. Then, each time you pull up the image to print...re-size to print resolution, sharpen, and print. If you were to find that your first desired printing works best for an enlarged native ppi of 200.....upsampled to 720....then why save the HUGE 720ppi version of the image. If you later want a smaller image...you then have to further down-sample. Or, if you waned an even larger image...upsample AGAIN. Repeated up and/or down-sampling can only degrade the image. Save the original and you will only, ever, do it ONCE.

kenwj wrote: For any given configuration there is a fixed matrix of ink dots per image pixel. And regardless of what the original image dimensions were, the image is effectively resampled to the native PPI rate at which the printer is configured. Bruce Oudekerk wrote: ...but every printer has a native resolution which is vastly different from the maximum addressable ink dot placement of the printer. ...Canons native res is 600ppi.
Yes, I think that's what it's called. Some might also call it the dot pitch, I believe. If that's the case, then at 4800x2400 dpi, that would be a 4x8 dot matrix per pixel. Of course, the dots might not be laid out in such an easy pattern, and it might not be so easy to tell by looking at a magnified image. I'm going to assume 600 ppi for now, but ultimately I might actually have to test the resolution. In case you're wondering, I have some rather large composite images. I want very high print quality, and I want to get the pixel count right. I want magnificent images if I can get them, and I'm willing to spend pixels to get them. Too good would be better than not good enough, but I don't want to consume a ridiculous amount of storage or memory either. There are many contradictory estimates of visual acuity. A Wikipedia article gives the angular resolution of the human eye as both 1' and 4' within the same article. Assuming 1', that would be 343 line pairs per inch, or 686 pixels per inch at 10 inches. So 600 ppi (or 720 for Epson) might not be excessive. I'm going to guess for now that 1200 is neither available nor necessary. I don't know for sure if there would be a visible difference between 300 and 600, but for now I'm assuming 600. One more thing. Yes, I'm doing this backwards. I don't have the printer set up yet. Because I don't--that's why. Hmm, I wonder if maybe the printer driver might tell me (slapping forehead). Thanks to all who replied.

ThrillaMozilla wrote:
There are many contradictory estimates of visual acuity. A Wikipedia article gives the angular resolution of the human eye as both 1' and 4' within the same article. Assuming 1', that would be 343 line pairs per inch, or 686 pixels per inch at 10 inches. So 600 ppi (or 720 for Epson) might not be excessive. I'm going to guess for now that 1200 is neither available nor necessary. I don't know for sure if there would be a visible difference between 300 and 600, but for now I'm assuming 600.
I'm going to, sort of, side step the issue. While visual acuity certainly comes into play here, there are other factors that are immensely important from a practical standpoint. The most obvious is how sharp the file is in the first place. All pixels are NOT created equal. You can supply a lot of pixels and it might very well not have the same information as an inherently sharper file that has less pixels. Even the term ‘sharp’ is misleading. When we sharpen an image we don’t really make anything sharper. It improves acutance which is generally increasing micro-edge contrast...but it appears sharper. And that appearence is at least as important, within reason, as the base size of the file. As suggested elsewhere, I like to concervatively sharpen a base image, save and resample to final size then sharpen for output. Somewhere in this process another factor comes into play and that is how much 'aliasing' is present in the base file itself. Its been mentioned here a couple of times that we might get away nicely with 360ppi or even 240ppi for a photo but a rastered structured/vector graphic would necessitate 720ppi on an Epson printer. The reason is that the overwhelming majority of cameras have Bayer sensors with an anti-aliasing filter which smoothes out irregularities and ‘jogs’ in curves. Thus, a curve photographed with a good camera setup might look smooth at 300 ppi and WILL look jagged when the same mathematical curve is rastered at 400ppi. I know that is counter intuitive but its all about the aliasing. MF cameras for years have weak or no AA filter and now we even have dSLR that are lacking an AA filter. The best known is probably the Nikon D800e. At any rate I think that these owners will have to approach the print process from a slightly differnt direction
I wonder if maybe the printer driver might tell me (slapping forehead).
Don't feel bad. How to determine the native resolution of a printer or even if there is one was a hotly debated topic on this forum years ago. The problem is that there is little if any documentation regarding this from the actual manufactures of the printers themselves. There were hints in various documentation but if was impossible to determine the actual rationale for those hints. Unless something has drastically changed I doubt you will find this information in the diver or driver docs. It is one step removed from urban myth. Bruce

Bruce Oudekerk wrote: You can supply a lot of pixels and it might very well not have the same information as an inherently sharper file that has less pixels. ...How to determine the native resolution of a printer or even if there is one was a hotly debated topic on this forum years ago. ...Unless something has drastically changed I doubt you will find this information in the diver or driver docs.
These images are composites, with almost unlimited detail and number of pixels. Ultimately I'll probably have to just measure the printer resolution with a resolution chart. My HP driver offers me a setting of 600 dpi, and I figured they were stating dots when they really mean pixels (this happens often). I looked with a magnifier, and sure enough, they really mean dots. Actually, the prints are quite good, which makes me wonder about all this. :\

ThrillaMozilla wrote: These images are composites, with almost unlimited detail and number of pixels. Ultimately I'll probably have to just measure the printer resolution with a resolution chart. My HP driver offers me a setting of 600 dpi, and I figured they were stating dots when they really mean pixels (this happens often). I looked with a magnifier, and sure enough, they really mean dots. Actually, the prints are quite good, which makes me wonder about all this. :\
A lot can be done with little, especially if there are no other choices:) Fortunately you have these higher resolution output levels. To be honest, in real life prints I can't see the difference with my Epson printers between the two highest outputs …if anything sometimes the highest lays down too much ink. But the higher densities (plural) also allow dithering routines unavailable at lower resolutions. These produce subtle tones and colors that are much more elegant than at lesser settings. The cynic in me thinks the only reason for publishing the addressable hardware PPI is for marketing purposes. The dithering algorithms, the ability to accurately vary droplet size, the ink sets and the nature of the LUT all affect the output at least as much. In the immediate future, subtle advancements in printing will not come from higher density resolutions but from increasing the color space, speed, paper handling, elegant switching of black inks (if needed) and the formulations of the inksets. Good luck on your projects. Bruce

And...don't those DPI numbers include all colors? If you are running a 6-ink printer, isn't that really just 600 DPI of each color?

Turbguy1 wrote: And...don't those DPI numbers include all colors? If you are running a 6-ink printer, isn't that really just 600 DPI of each color?
DPI of the printer includes all possible ink colors and for each ink there are multiple droplet size printer cay lay into each "dot". Printers with more inks have wider color gamut, just as printing at more dpi will result in larger color gamut, since printer can mix its ink colors in more ratios to represent 1 pixel. For example if we have 360PPI image printed with multiple settings the printer will lay X dots per pixel: 360 - 1 720 - 4 1440 - 16 2880x1440 -32 2880 - 64 With 4 colors CYMK printer can produce 4^X unique dot patterns, where X is the number i listed above. Eight ink printer, like 3880 can theoretically produce 8^X patterns. I say theoretically, because patterns average out, and so the location of the dots within each pixel is irrelevant, hence the 4x4 block with 1 yellow and 3 red dots will look the same, no matter where the yellow dot is. So for unique colors we have this formula: (N+X-1)!/((N-1)!*X!) where N is the number of inks. So we get this many possible colors per pixel for 4 and 8 ink system: X 4 8 1 4 8 2 3 4 35 330 5.12 8.3 16 969 245157 9.92 17.9 32 6545 15380937 12.67 23.87 64 47905 1329890705 15.54 30.3 Left most column in the number from 1st table i showed with dpi resolution settings. 2nd column is for cymk (4 ink system) and 3rd column is for 8 ink system. Most JPG images are 24-bit so each pixel at most has 2^24 colors (actually a bit less, just like numbers in 2nd table are smaller than 4^X and 8^X). The two columns on the right (4 and 5) show the log2 of the possible colors (2 and 3) based on that rows value of X as well as 4 and 8 ink system. For 8-ink system we get 23.87 bits of information for 32 dots per pixel, which is almost 100% accuracy of 24-bit jpg image color space. The 4-ink system never gets to such large color gamut.

Forgot to add, most printers have many print head nozzles per color depositing different amounts/shades of the ink into a dot. So the number of "inks" should be multiplied by the number of different size nozzles available per color to get the real value of N. In above calculations. I just assumes 1 size nozzles to simplify calculations.

fft81 wrote: Forgot to add, most printers have many print head nozzles per color depositing different amounts/shades of the ink into a dot. So the number of "inks" should be multiplied by the number of different size nozzles available per color to get the real value of N. In above calculations. I just assumes 1 size nozzles to simplify calculations.
The nozzles are all the same size. I assume the different sizes are created by using a different pressure. As an example, an Epson 3880 printer has 8 rows with 180 nozzles per color. When printing with 720 or 360 DPI horizontal resolution it can produce dots in 22.2, 13.8 and 6.6 picoliter sizes. When printing at 1440 DPI is can produce 13.2, 5.9, and 3.5 picoliter sizes. When printing at 2880 DPI is only prints 3.5pl dot sizes. (All of the above is from the Epson service manual.)

fft81 wrote: DPI of the printer includes all possible ink colors and for each ink there are multiple droplet size printer cay lay into each "dot". Printers with more inks have wider color gamut, just as printing at more dpi will result in larger color gamut, since printer can mix its ink colors in more ratios to represent 1 pixel. For example if we have 360PPI image printed with multiple settings the printer will lay X dots per pixel: 360 - 1 720 - 4 1440 - 16 2880x1440 -32 2880 - 64 With 4 colors CYMK printer can produce 4^X unique dot patterns, where X is the number i listed above. Eight ink printer, like 3880 can theoretically produce 8^X patterns. I say theoretically, because patterns average out, and so the location of the dots within each pixel is irrelevant, hence the 4x4 block with 1 yellow and 3 red dots will look the same, no matter where the yellow dot is. So for unique colors we have this formula: (N+X-1)!/((N-1)!*X!) where N is the number of inks. So we get this many possible colors per pixel for 4 and 8 ink system: X 4 8 1 4 8 2 3 4 35 330 5.12 8.3 16 969 245157 9.92 17.9 32 6545 15380937 12.67 23.87 64 47905 1329890705 15.54 30.3 Left most column in the number from 1st table i showed with dpi resolution settings. 2nd column is for cymk (4 ink system) and 3rd column is for 8 ink system. Most JPG images are 24-bit so each pixel at most has 2^24 colors (actually a bit less, just like numbers in 2nd table are smaller than 4^X and 8^X). The two columns on the right (4 and 5) show the log2 of the possible colors (2 and 3) based on that rows value of X as well as 4 and 8 ink system. For 8-ink system we get 23.87 bits of information for 32 dots per pixel, which is almost 100% accuracy of 24-bit jpg image color space. The 4-ink system never gets to such large color gamut.
I’m not so sure it is a simple as this. What you are referring to might be just a first step or ‘they’ might bypass this process altogether. In another portion of this thread I referred to the print rendering as a Voodoo artform and, ‘tongue and cheek’ I believe it to be appropriate. Here is a 3200 DPI scan of a B&W print from my R1900. IMAGE(http://dpzen.com/dpzattaches/dpzattachesm//2014092509075635827.jpg) For good or ill, most inkjets use color for a variety of reasons when printing gray scale. My point here is that when we look at the distribution of color, the color blotches are well outside a fixed matrix that your model infers. My guess is that it has a perceptual as opposed to a strictly mathematical basis and takes a more localized AREA approach. In reality, the printer driver is a black box environment where the various printer manufactures perform their proprietary magic. Bruce

Bruce Oudekerk wrote:
fft81 wrote: DPI of the printer includes all possible ink colors and for each ink there are multiple droplet size printer cay lay into each "dot". Printers with more inks have wider color gamut, just as printing at more dpi will result in larger color gamut, since printer can mix its ink colors in more ratios to represent 1 pixel. For example if we have 360PPI image printed with multiple settings the printer will lay X dots per pixel: 360 - 1 720 - 4 1440 - 16 2880x1440 -32 2880 - 64 With 4 colors CYMK printer can produce 4^X unique dot patterns, where X is the number i listed above. Eight ink printer, like 3880 can theoretically produce 8^X patterns. I say theoretically, because patterns average out, and so the location of the dots within each pixel is irrelevant, hence the 4x4 block with 1 yellow and 3 red dots will look the same, no matter where the yellow dot is. So for unique colors we have this formula: (N+X-1)!/((N-1)!*X!) where N is the number of inks. So we get this many possible colors per pixel for 4 and 8 ink system: X 4 8 1 4 8 2 3 4 35 330 5.12 8.3 16 969 245157 9.92 17.9 32 6545 15380937 12.67 23.87 64 47905 1329890705 15.54 30.3 Left most column in the number from 1st table i showed with dpi resolution settings. 2nd column is for cymk (4 ink system) and 3rd column is for 8 ink system. Most JPG images are 24-bit so each pixel at most has 2^24 colors (actually a bit less, just like numbers in 2nd table are smaller than 4^X and 8^X). The two columns on the right (4 and 5) show the log2 of the possible colors (2 and 3) based on that rows value of X as well as 4 and 8 ink system. For 8-ink system we get 23.87 bits of information for 32 dots per pixel, which is almost 100% accuracy of 24-bit jpg image color space. The 4-ink system never gets to such large color gamut.
I’m not so sure it is a simple as this. What you are referring to might be just a first step or ‘they’ might bypass this process altogether. In another portion of this thread I referred to the print rendering as a Voodoo artform and, ‘tongue and cheek’ I believe it to be appropriate. Here is a 3200 DPI scan of a B&W print from my R1900. IMAGE(http://dpzen.com/dpzattaches/dpzattachesm//2014092509075635827.jpg) For good or ill, most inkjets use color for a variety of reasons when printing gray scale. My point here is that when we look at the distribution of color, the color blotches are well outside a fixed matrix that your model infers. My guess is that it has a perceptual as opposed to a strictly mathematical basis and takes a more localized AREA approach. In reality, the printer driver is a black box environment where the various printer manufactures perform their proprietary magic. Bruce
What you are seeing in the scan is how the printer actually reproduces RGB colors using CYMK color space. Your findings do not actually contradict my calculations, but rather my calculations are the "asymptotic" behavior of what actually happens, while your microscope scan is the "actual" ink deposition. In real world the ink dots may overlap a bit, they may "miss" the intended location, they may smudge or vary in the amount of ink absorbed by the paper, even on the same sheet of paper. As the result of all that, the print driver does add other colors to "average out" to a grey on B&W print. For example, to get a 50% grey one might think that printer will use black on every other dot and leave other dots white. In reality the "black ink may be not black enough and the pure white paper may be not glossy enough, so printer might add some yellow to the mix.

I BET you are so CONFUSED now, that you be afraid to say so..I know what I'm doing when it comes to printing but this lot on here has Tottaly confused me..LOL If you use photoshop or elements then you are nearly there. All you need to do First, is Size your image to whatever size you want to print it. i.e--8x10--12x8---10x12---11x14, Hope you understand that so far as that bits easy..Choose 300ppi (NOT dpi as most tell you). Choose Print and follow the print box. Plenty of tutorials on Google...that will NOT confuse you.. HAPPY NEW YEAR.

MrScary wrote: I BET you are so CONFUSED now, that you be afraid to say so..I know what I'm doing when it comes to printing but this lot on here has Tottaly confused me..LOL If you use photoshop or elements then you are nearly there. All you need to do First, is Size your image to whatever size you want to print it. i.e--8x10--12x8---10x12---11x14, Hope you understand that so far as that bits easy..Choose 300ppi (NOT dpi as most tell you). Choose Print and follow the print box. Plenty of tutorials on Google...that will NOT confuse you.. HAPPY NEW YEAR.
LOL +1

fft81 wrote: What you are seeing in the scan is how the printer actually reproduces RGB colors using CYMK color space. Your findings do not actually contradict my calculations, but rather my calculations are the "asymptotic" behavior of what actually happens, while your microscope scan is the "actual" ink deposition. In real world the ink dots may overlap a bit, they may "miss" the intended location, they may smudge or vary in the amount of ink absorbed by the paper, even on the same sheet of paper. As the result of all that, the print driver does add other colors to "average out" to a grey on B&W print. For example, to get a 50% grey one might think that printer will use black on every other dot and leave other dots white. In reality the "black ink may be not black enough and the pure white paper may be not glossy enough, so printer might add some yellow to the mix.
I understand this. Its the SCALE of the dot pattern that, in my mind doesn't jive with the type of simple matrix you suggest. Bruce

Here is how a skin tone looks like printed on 3880 and viewed under the microscope. The width of the picture is 1.5mm http://magicmrv.com/Pictures/MISC/Img00003.jpg Photo printed on 3880 using 1440x2880 setting. Skintone, Width =1.5mm

fft81 wrote: Here is how a skin tone looks like printed on 3880 and viewed under the microscope. The width of the picture is 1.5mm http://magicmrv.com/Pictures/MISC/Img00003.jpg Photo printed on 3880 using 1440x2880 setting. Skintone, Width =1.5mm
Nice example...love the detail. I’m in a hurry so maybe I really botched this but I took your image and drew a white grid that represents roughly a 1/720 inch grid. Thus a block of 4 squares represents 1/360 inch per side. This is a random section of your image and the grid is randomly placed also. To discuss the implications of what we are talking I thought we need some common frame of reference.. IMAGE(http://dpzen.com/dpzattaches/dpzattachesm//2014092509075935829.jpg) It still seems to me that there is a disconnect between what you proposed and what I see here. Bruce

Bruce Oudekerk wrote:
fft81 wrote: Here is how a skin tone looks like printed on 3880 and viewed under the microscope. The width of the picture is 1.5mm http://magicmrv.com/Pictures/MISC/Img00003.jpg Photo printed on 3880 using 1440x2880 setting. Skintone, Width =1.5mm
Nice example...love the detail. I’m in a hurry so maybe I really botched this but I took your image and drew a white grid that represents roughly a 1/720 inch grid. Thus a block of 4 squares represents 1/360 inch per side. This is a random section of your image and the grid is randomly placed also. To discuss the implications of what we are talking I thought we need some common frame of reference.. IMAGE(http://dpzen.com/dpzattaches/dpzattachesm//2014092509075935829.jpg) It still seems to me that there is a disconnect between what you proposed and what I see here. Bruce
This is a section of a 36.3MP photo printed on a 5x7 with printer driver doing the required resampling from 980 PPI i am feeding it. Would the driver down-sample to 360 PPI and print at 8x4 Dots per Pixel? Or down sample to 720 ppi and print at 4x2 dots per pixel? What i see from this real world photo is: 1. printer overlaps the dots to produce an average color. 2. Printer varies size of the dots. This suggests a difference from my initial "dot matrix per pixel" model, however the color alphabet calculations should be still reasonably "close" since it still boils down to averaging of discrete colors to produce another color. Due to overlap of ink dots i suspect printer deposits the 8 dots into the same "pixel cell" but the location of these dots is pseudo random within that cell, in some cases even overlapping with next cell.This results in smother color transition from one pixel to the next.

fft81 wrote:
Bruce Oudekerk wrote: IMAGE(http://dpzen.com/dpzattaches/dpzattachesm//2014092509075935829.jpg)
This is a section of a 36.3MP photo printed on a 5x7 with printer driver doing the required resampling from 980 PPI i am feeding it. Would the driver down-sample to 360 PPI and print at 8x4 Dots per Pixel? Or down sample to 720 ppi and print at 4x2 dots per pixel? What i see from this real world photo is: 1. printer overlaps the dots to produce an average color. 2. Printer varies size of the dots. This suggests a difference from my initial "dot matrix per pixel" model, however the color alphabet calculations should be still reasonably "close" since it still boils down to averaging of discrete colors to produce another color. Due to overlap of ink dots i suspect printer deposits the 8 dots into the same "pixel cell" but the location of these dots is pseudo random within that cell, in some cases even overlapping with next cell.This results in smother color transition from one pixel to the next.
Your guess is as good as mine as to the native res of the 3880. With Epson printers there is a transition from prosumer to professional printers whereby the native res switches form 720ppi to 360. When the 3800 came out InkJetArt posted a blog stating it was 360ppi but in the intervening time i have read (can't remember where) that many of the newer big printers have a 720 mode. This really doesn't surprise me as printing non-aliased vector graphics at a resolution of 360ppi is crude at best when viewed remotely close by. Stair-stepping can even be discerned (just barely) at 600ppi or 720ppi in this specific case For anyone that has one of these bigger printers (or any printer)a test is easy. Create a few diagonal lines and some circular objects in Corel Draw or Adobe Illustrator and raster it out with no aliasing to two separate files, one at 360ppi and the other at 720ppi Print the two separately at highest printer resolution. If there is a significant difference between the two you know the printer is resolving more than 360ppi. Canon and HP should use 300ppi and 600ppi. In recent years, Epson has been touting a new LUT (look up table) that purportedly offers more elegant rendering of colors. What that tells me is that what you were doing is valid to a point but that they have taken it just that much further in terms of using these combinations and permutations of color, dot size and placement to fine tune the color rendering as well as using very sophisticated and proprietary algorithms to look at neighboring areas to determine how the ink is going to be laid down. Thus I would think that your ‘pseudo random” is really less random and more a carefully crafted result calculated by the immediate color environment surrounding your ‘pixel cell’. To us it looks completely random. To me this is very interesting to see the actual output and how it really manifests itself on the paper. Your illustration if far superior to my crude scan. Thank you. Bruce

Bruce Oudekerk wrote: For anyone that has one of these bigger printers (or any printer)a test is easy. Create a few diagonal lines and some circular objects in Corel Draw or Adobe Illustrator and raster it out with no aliasing to two separate files, one at 360ppi and the other at 720ppi Print the two separately at highest printer resolution. If there is a significant difference between the two you know the printer is resolving more than 360ppi. Canon and HP should use 300ppi and 600ppi.
I'm thinking of doing something like that, maybe with a resolution chart. One thing I noticed is that the dots are much larger than one would expect from the dpi. Thanks for taking an interest in this, guys. I'm embarrassed I haven't gotten to it yet myself.

ThrillaMozilla wrote: One thing I noticed is that the dots are much larger than one would expect from the dpi.
It is my understanding this section of the print is a photo-micrograph at the equiv of 35,000 DPI scan. ...AND this is a fairly large format printer. I suspect my R1900 has the capabiltiy of even smaller dot structure. Bruce

Add new comment

Image
More information
  • Files must be less than 2 MB.
  • Allowed file types: png gif jpg jpeg.
Attachment
More information
  • Files must be less than 2 MB.
  • Allowed file types: zip rar.