Sounds like either scanner probem – are you SURE those bars aren’t present in the file right after the scan? Look at 100% size.
or
Look at your final ppi in Photoshop before printing. Don’t use a fractional number, like 300.34 or whatever. Convert to whole number 300.0; this can cause abberations in *some* printer drivers/models.
Mac
Thanks for replying.
If the bars were in the original scan, then (a) why are they not visible in Photoshop and (b) why are they not visible when I print at 360 dpi? A bar that is 1/8th of an inch wide would be 45 pixels wide at 360 dpi — very visible. I don’t see how Photoshop could display image info that is absent from the scanned input.
"Look at your final ppi in Photoshop before printing. Don’t use a fractional number, like 300.34 or whatever."
I’ll take another look but I have never tried anything except whole numbers.
Bill
Bill,
The basic file to be printed is about 300 MB if I scan the image at 1200
dpi, and of course, twice that if I scan at 2400 dpi. <<
Just because the printer prints at 2400 dpi doesn’t mean your image file should be at 2400 ppi!!
360 ppi image rez is quite sufficient for printing at any resolution. More than that and you are just creating data that the printer firmware has to throw away because it cannot use it,. The only result is increased printing times.
Let’s look at what that 2400 dpi really means.
Firstly that is the vertical stepping of the print heads and will be twice the ‘real’ printing resolution, so we are really looking at 1200 dpi. Now that 1200 is the SUM of all 4 print heads. Each print head is printing at 1200/4 or 360 dpi. Now even that 360 dpi is only achieved in areas of 100% coverage with that colour. At 50% coverage the printing rez has dropped to 180 dpi.
The bottom line is this. In the vast majority of cases you will see NO improvement in output by printing any image at an image rez of greater than 200 ppi.
I have the Epson 2200 printer and a 3200 scanner. Similar setup. But the hardware isn’t the problem, the brand of paper is. I got those parallel "bars" when using Kodak paper. The Epson ink doesn’t dry fast enough, rubber guides lift the not-yet-dry ink right off the page. You can prove that to yourself by opening the smoked Plexiglas lid and watch the ink applied evenly on the paper, disappear under the platen, then the bars appear as the paper is guided out to the catch tray. Plain paper wicks up the ink, so it doesn’t transfer to the rubber rollers. I use Epson papers almost exclusively (not an endorsement, just too cheap to experiment) and Ilford Gallerie Smooth Pearl Paper.
wrote in message
I am using Photoshop 7.0.1 with an Epson Stylus Photo 2200 printer and an
Epson 4870 scanner. I scanned an 8X10 black-and-white photo; tried to edit it with Photoshop; and tried to print it. I got a strange pattern with vertical bars roughly every inch. these bars are slightly closer together at the left side of the page that at the right side; they are not quite 1/8th of an inch wide; and they appear to be random noise. The bars are visible only in the darker areas of the image. They seem to disappear in the lighter areas, and they disappear entirely in a horizontal band about an inch wide that is toward the bottom of the image. There are no bars at all if I print on plain paper at 360 dpi. The bars appear when I switch to glossy photo paper at 720 dpi, 1440 dpi or 2880 dpi. The basic file to be printed is about 300 MB if I scan the image at 1200 dpi, and of course, twice that if I scan at 2400 dpi. It seems to make no difference whether I make the image black and white or color.
I am speculating that there may be an incompatibility between Photoshop
and the printer software, having to do with the conversion from the dpi in Photoshop and the dpi in the Epson software. However, the bars have looked exactly the same at all dpi setting that I have tried, so it isn’t obvious that there is a computation problem.
Bill Starbuck
why are they not visible when I print at 360 dpi?
Sorry, I missed that…
But looking at Len’s post, I see what else I missed.
I echo his advice….300ppi at actual print size is more than ample for printer output, regardless of the dpi rating of the printer. *Some* Epson folks report very slight quality improvement up to 360ppi, but not beyond. And many say they can see no visible diff with naked eye past 220ppi). Image ppi and printer dpi are different concepts.
You may just be overwhelming the printer driver with too much information once you tell it to print at higher than 360dpi print setting.
Try downrezzing you image to 360, 300, 220 ppi renditions from whatever you have now, and test print. See if:
1. the banding disappears
2. at what ppi can you no longer see a quality improvement.
M
Len,
These weary old arguments might spoil your day, so I’ll begin with non-arguable points:
"…1200/4 dpi or 360 dpi…" Ummmhhh, 1200/4 = 300.
"…At 50% coverage [360dpi] would drop to 180 dpi…" At 50% coverage 360 dpi would drop to 360 x sqrt2 = 254.
Now for the contentious stuff:
"…1200 dpi…is the sum of all 4 printheads…" I still think it’s per CMYK channel.
"…360 dpi is only achieved in areas of 100% coverage…" The space reserved for a printer dot counts as a printer dot whether it’s filled or not. That’s how you get tones.
Respectfully, George
George,
OK,so my brain was working on the more usual 1440/2880 <g> I give you that one.
I still think it’s per CMYK channel.<<
I think your wrong.
The space reserved for a printer dot counts as a printer dot whether it’s
filled or not.<<
Only true for amplitude modulation, George, not for F.M.. Take a 50% flat pure colour (let’s say yellow since that avoids 4/6 ink questions. The printhead will only spray ink over 50% (approx) of that area. Count the dots. That’s the dpi the eye sees.
Hi Len,
I AM (no pun) referring to FM. Ink jets are primarily FM devices. Yet most ink jets (Epsons, for sure) are hybrids in that they have several discrete dot sizes to choose from, providing some AM capability to ease the resolution/tone-depth trade-off.
Despite being basically FM, ink jets still use an array of dots to form a pixel. Although the array is filled randomly, there is nothing random about the percentage of a pixel’s area that gets filled. One array serves all channels because the printer’s C,M,and Y dots do not need to be in registration. It is irrelevant whether or not the different colored dots overlap each other. The array "real estate" required is consequently more reasonable.
George
Thanks for trying to explain things, but you seem to be speaking a jargon that I do not understand.
"Just because the printer prints at 2400 dpi doesn’t mean your image file should be at 2400 ppi!!"
Why not? Your exclamation points imply that this should be obvious or that everyone but me knows this. But I certainly do not know it and I do not see a reason for it. In particular, it would seem to me that I should get the most information into the data file if I scan the original at the highest resolution that the scanner can produce. And whatever the limitations of the printer, I should get the best image by using the highest resolution that the printer can produce. In this case, that would have been a scan at 2400 and printing at 2880.
"360 ppi image rez is quite sufficient for printing at any resolution. More than that and you are just creating data that the printer firmware has to throw away because it cannot use it,. The only result is increased printing times."
Again, this may be obvious to you but it is not obvious to me. I tried to use a resolution higher than 360 because I thought a higher resolution would be better. And in fact, when I compare an image printed at 360 with one printed at 720, the one printed at 720 looks better. So why should I not prefer the image that looks better?
Moreove, I simply do not care how long it takes the printer to produce this image. I can start the printing in the morning and come back to it in the evening if need be.
"Let’s look at what that 2400 dpi really means."
"Firstly that is the vertical stepping of the print heads and will be twice the ‘real’ printing resolution, so we are really looking at 1200 dpi."
What do you mean by "the ‘real’ printing resolution"?
"Now that 1200 is the SUM of all 4 print heads. Each print head is printing at 1200/4 or 360 dpi."
To me, 1200/4 is 300, not 360. I obviously do not understand this arithmetic.
"Now even that 360 dpi is only achieved in areas of 100% coverage with that colour. At 50% coverage the printing rez has dropped to 180 dpi."
This certainly does not make sense to me, although I can see that it would work that way if the ink flowed evenly all over the paper. If the dots of ink are very small, 50% coverage would produce alternating colored spaces and blank spaces, which is very different from colored spaces that are twice as large and overlap each other.
"The bottom line is this. In the vast majority of cases you will see NO improvement in output by printing any image at an image rez of greater than 200 ppi."
I don’t really care about "the vast majority of cases". I was trying to copy a rare photograph that cost $52,000 and so I wanted the very highest quality print I could possibly make.
Bill
Yesterday, I ran some experiments, and I think I have discovered that one central issue has to do with file size. A second issue may be the arithmetic ability of Epson’s printer driver, which may have trouble with ratios that are not very simple, such as 2 to 1.
Photoshop could print an 8X10 image at 720 dpi with the printer set at 720 dpi.
Photoshop could not print an 8X10 image at 1200 dpi with the printer set at 720 dpi. (This was asking the printer to produce 1 dot 1.666667 in the Photoshop image.)
Photoshop could not print an 8X10 image at 1440 dpi with the printer set at 1440 dpi (471 MB file). But when I cropped the image to approximately one fourth (125 MB file), printing was OK. Likewise, I could print this cropped image at 1440 dpi with the printer set at 2880 dpi.
I also tried the Windows Picture and Fax Viewer. The image was 1200 dpi and I told the Viewer to print at 360 dpi (a ratio of 1 to 3.33333). Either Viewer or the Epson driver was unable to display a preview, which I interpret as a sign of arithmetic problems.
Bill
wrote:
Yesterday, I ran some experiments, and I think I have discovered that one central issue has to do with file size. A second issue may be the arithmetic ability of Epson’s printer driver, which may have trouble with ratios that are not very simple, such as 2 to 1.
Photoshop could print an 8X10 image at 720 dpi with the printer set at 720 dpi.
Photoshop could not print an 8X10 image at 1200 dpi with the printer set at 720 dpi. (This was asking the printer to produce 1 dot 1.666667 in the Photoshop image.)
Photoshop could not print an 8X10 image at 1440 dpi with the printer set at 1440 dpi (471 MB file). But when I cropped the image to approximately one fourth (125 MB file), printing was OK. Likewise, I could print this cropped image at 1440 dpi with the printer set at 2880 dpi.
I also tried the Windows Picture and Fax Viewer. The image was 1200 dpi and I told the Viewer to print at 360 dpi (a ratio of 1 to 3.33333). Either Viewer or the Epson driver was unable to display a preview, which I interpret as a sign of arithmetic problems.
Bill
bill,
the way i see it, just find the quality that you are looking for that your equipment will produce and go with it. in the long run, all the math and other cowpatties don’t matter as long as your end result satisfies you.
however, if you want to out source printing to a commercial printer…now that’s another subject in itself.
best,
dwain
Bill,
Before agonizing further on the big picture, I think you need to ponder more deeply on the more fundamental difference between PPI and DPI.
A printer dot is not a pixel. A pixel is made up from many printer dots. The number of pixels will always be smaller than the number of printer dots.
It is not inconsistent to state you don’t need more than, say, 360 PPI (that first letter is a P) while stating you should print at, say, 2400 DPI (first letter is a D).
Go to Google and search for "A Few Scanning Tips" by Wayne Fulton for an excellent primer on resolution.
George
when I compare an image printed at 360 with one printed at 720, the one printed at 720 looks better.
Yes, the higher DPI setting of the PRINTER will look better. But you don’t need to have image PPI match it.
See George’s follow, and refer back to my post also.
I encourage you to peruse scantips.com also.
Equally relevant for all types of digital images re
ppi/dpi/print size/interpolation/etc.
M
Just because the printer prints at 2400 dpi doesn’t mean your image file should be at 2400 ppi!!
that’s referring to the combined print head resolution. all printer manufacturers seem to do it. it’s kinda misleading in the way that monitor size was misleading a few years back until the law was changed. they used to say 20" monitor and people thought that’s what they were getting. turns out the 20" monitor only had 17" of viewing area. now monitor manufacturers are required to list that little fact.
so with printers, if you printer gets "2400 dpi" and you have 4 print heads, you’re getting a max of 600 dpi out of each print head for a combined 2400 dpi. if you have 6 print heads, the max to get to 2400 is 400 dpi per head.
at least that’s the way i understand it. does that help a little?
dave
Dave,
if you have 6 print heads, the max to get to 2400 is 400 dpi per head.<<
Not that bit – as you can’t print C and c or M and m at the same point (the two are mutually exclusive) it’s still a divide by 4
As Spock would say "Fascinating!!!"…
Another problem with 2200’s is that they only seem to work well with Epson and some of the higher end art paper (Crane etc.). They do not play nice with HP and Kodak lower end paper. They don’t play nice with Epson’s lower end papers (I only use the Professional Media in the 2200. (I have seen the same banding problems discribed by the original poster when trying to print with All HP, or low end Kodak or Epson papers.) So it may not be the printer but rather the substrate that you are printing on.
Also Epson has put out a new set of ICC profiles for each paper with profiles for each ink level 720/1440/2880 (hold on to the old icc profiles that shipped with the printer for your 720 profiles). These profiles are significantly improved over the original icc profiles that shipped with the 2200. Try downloading those, Convert your file to the paper profile and use these profiles as the print source when you print your document. You may notice significant improvement in the higher 1440/2880 ranges. (Color is much improved) Disregard any information given to you by the Epson print preview–It’s not color managed.
By the way do you have Epson’s Color Management turned off?
it’s still a divide by 4
thanks for clarifying len. not really my area of expertise, i was just parroting what i’ve learned from you! <SQUAWCK!> 🙂
Christine —
"I have seen the same banding problems discribed by the original poster when trying to print with All HP, or low end Kodak or Epson papers. So it may not be the printer but rather the substrate that you are printing on."
I was in fact trying to us HP paper. It never occurred to me that the paper could couse such an effect.
"By the way do you have Epson’s Color Management turned off?"
I haven’t gotten that far. I saw the FAQ about this but ran into the banding and that took my attention. The actual image I was trying to copy was a black and white photo that had been gold-toned. The colors that printed were MUCH too blue, so I am going to have to establish the coloring manually.
Bill
Go get yourself some Epson Professional paper (the only place I can get it in Denver is at CompUSA. Generally I order it off the internet because buy.com has it cheaper.) You will see the banding disappear!
HP papers are not formatted to work with the Epson UltraChrome Inks and this should resolve your banding problem. (What I think is happening is that the ink is not setting on HP and low end papers or is not being distributed correctly on the paper.)We also use the Epson 2200 in our college lab and once a semester we get a couple of students who don’t take our recommendations on paper and cry when their projects come out horrible!
I have had the 2200 for almost a year and for the first 2 months it gave me fits until I turned off the Epson Color Management and let Adobe drive! Download the new profiles because they make a huge difference in color fidelity! Remember to profile for input and then output and set your source in the print with preview to the paper and ink you are using in the Epson!
Let us know what happens!
And keep up the dpi/ppi discussion…I always learn something when these get started!!
And keep up the dpi/ppi discussion…
there’s no such thing as ppi/dpi on the web… 🙂
wrote:
Thanks for trying to explain things, but you seem to be speaking a jargon that I do not understand.
"Just because the printer prints at 2400 dpi doesn’t mean your image file should be at 2400 ppi!!"
Why not? Your exclamation points imply that this should be obvious or that everyone but me knows this. But I certainly do not know it and I do not see a reason for it. In particular, it would seem to me that I should get the most information into the data file if I scan the original at the highest resolution that the scanner can produce. And whatever the limitations of the printer, I should get the best image by using the highest resolution that the printer can produce. In this case, that would have been a scan at 2400 and printing at 2880.
Printer dots are not equivalent to scanner samples.
The printer needs more than one of its ink dots to create the impression of the color of one pixel.
In the usual PC environment you do not have a lot of control over your printers resolution: You tell the printer driver what kind of paper you use and the printer driver optimizes the amount of ink it can place on this kind of paper. Sometimes you have the option to ask for fast or draft quality – but that only makes sense for standard paper.
"360 ppi image rez is quite sufficient for printing at any resolution. More than that and you are just creating data that the printer firmware has to throw away because it cannot use it,. The only result is increased printing times."
Again, this may be obvious to you but it is not obvious to me. I tried to use a resolution higher than 360 because I thought a higher resolution would be better. And in fact, when I compare an image printed at 360 with one printed at 720, the one printed at 720 looks better. So why should I not prefer the image that looks better?
As said above: Your printer needs several dots of ink and lots of plain white paper between them to create a full color pixel. If your printer has 2880dpi this specifies how near two adjacent dots can be placed. BTW: The ink drop is a lot bigger than 1/2880"
So if the printer uses a 8×8 pattern with the restriction that not more than two of three neighbouring dots may be filled to avoid soaking of the paper, the printer can create about 170 shades for each ink. The given restriction is also valid for use of different inks in the same pixel, thus effectively reducing the number of possible colors to something in the same range as the 24bit screen colors. But still most of the possible colors are in the dark brown to black range. But the 2880 dpi being used for an 8×8 pattern have already been reduced to 2880/8 PIXELS per inch or only 360ppi.
This certainly does not make sense to me, although I can see that it would work that way if the ink flowed evenly all over the paper. If the dots of ink are very small, 50% coverage would produce alternating colored spaces and blank spaces, which is very different from colored spaces that are twice as large and overlap each other.
You still assume that an ink drop only covers a 1/2880" square. Printer manufacturers claim ink drops of 2-3picolitres for their newest model printers.
To work with picolitres we need to convert the 2880dpi to dots per millimetre: 2880/25.4=113
Now one liter is a 10cm cube. A millilitre is a 1 cm cube. A microlitre is a 1mm cube, 0.1 mm cubes are nanolitres and 0.01mm cubes are picolitres.
Here we see, that the 2-3 picolitre ink drop does not fit flat in the 1/113mm square. If we transfer it into sizes we can handle, take an
8.8cm square and pour 2-3 litres of water on it. You will see that even
on a 10cm futon (paper has to grow too) you will soak a much larger area.
50% coverage is just not possible.
I don’t really care about "the vast majority of cases". I was trying to copy a rare photograph that cost $52,000 and so I wanted the very highest quality print I could possibly make.
The maximum resolution the best inkjet printers can achieve on the best paper available is somewhere in.the 360ppi range. Your 2880dpi printer is by far not in that class of printers. If you actually create a print that has 180ppi call yourself happy.
Michael
wrote:
Yesterday, I ran some experiments, and I think I have discovered that one central issue has to do with file size. A second issue may be the arithmetic ability of Epson’s printer driver, which may have trouble with ratios that are not very simple, such as 2 to 1.
Ignore the filesize.
Create a test pattern of black and white lines, one pixel wide each. 200×200 pixel for the whole thing.
Now print it at different resolutions – you need photo paper to be able to see anything, but you can use 4×6 pieces more than once as the prints will be very small.
The resolution that clearly shows the white lines is the maximum your printer can achieve in black and white. Color is worse.
Michael
Et tu, Dave?? You are going along with Len and accepting that 2400 dpi is the sum of dots from four print heads? Balderdash!! I don’t know what Himalayan guru we need to get an audience with, but you identify him and I’ll make the trek to get his blessing and win a round of drinks at a pub of Len’s choosing.
George
You are going along with Len and accepting that 2400 dpi is the sum of dots from four print heads?
makes sense to me… why, is that wrong? i said that printing wasn’t my strong suit. if you know different, like, fill us in daddyo!
<snapping fingers> 🙂
Dave,
"…why, is that wrong?…"
I thought you’d never ask ;-)!!
To begin with, the printer DPI must be sufficient to give the desired PPI AND—repeat AND—a reasonable number of tones. If you will be satisfied with just 64 tones, divide the printer DPI by 6 (2 to the 6th power = 64) to get the resulting resolution. 2400/6=400. If I must then divide that number by 4 for the 4 printheads, I’m left with a mere 100 PPI. And that’s for 2400 printer DPI. Were I to use 1200 printer DPI, I’d be looking at a res of 50 PPI and, God forbid, 25 PPI at a printer res of 600DPI. If I want 256 tones per channel instead of 64, divide all those resolutions by another 4.
A real printer is not quite that bad, but you should get the general idea. In a real printer, part of the tonal variety load is eased by adding the AM capability I cited but, at best, this recovers a factor of 4. That leaves you with a final res in PPI WAY below a decent level. No, Dave, you don’t assess an additional resolution penalty for printheads. A printer "dot" is a composite of 3 colors and you don’t dissect it further—even to make your printer performance look good!
George
Whoa Timothy,
Messing around with inks other than CMY changes the whole ballgame. It doesn’t matter (well, almost but that would be a digression clouding the point being made)if cyan falls on top of magenta or yellow or these three colors pile up on each other in any other sequence. That’s because each of them take out one and only one of the RGB primaries. Using an ink that filters out more than one primary makes a huge difference and you cannot allow such inks to lie on top of each other—a severe design constraint requiring that the dots be in registration. Heck, that’s the very reason red, green and blue inks are NOT used—they all filter out TWO primaries. That is, a red ink takes out blue and green, a blue ink takes out red and green, and a green ink takes out red and blue
George
maybe i’m not understanding but
2400/6=400.
isn’t that what i (from len) said?
Dave,
Dividing by 6 is just the FIRST degradation of resolution invoked to give 64 tones (arbitrary number of tones taken for illustration only). IF you must ALSO divide by another factor of 4, as Len says you must, that’s an overall division of 24 and now the 2400 printer dpi yields only 100PPI. My argument is that a decent PPI does not calculate out with the additional degradation Len imposes, especially if you start with the more customary printer resolutions of 1200 or even 600 DPI. Of course, you can always scarifice tone depth. Going with only 32 tones doubles the resolution achievable with 64 tones.
George
George,
You are attempting to use AM theory in an FM situation.
I’m not getting it. What additional levels of degredations is len proposing. He says (and it sounds probable) that each head is printing 600dpi for a total of 2400. what’s wrong with that statement? it sounds plausable to me.
please make any replies in the QUEENS ENGLISH ONLY! 🙂
Dave
PPI = DPI/sqrt(n) where n= no. of tones
If n= 36, sqrt(n) = 6
PPI = 600/6 = 100
A PPI of 100 is not very good. But it gets worse. The 2400 DPI number is inflated (Len ‘s reply #3 re vertical stepping). Use 1200 DPI instead of 2400. In fact, use the 1440 DPI which he really had in mind. Then each head would be printing at 360 DPI. Now for 36 tones:
PPI = 360/6 = 60
Are you happy with 60 PPI and only 36 tonal variations?
George
Len,
"…You are attempting to use AM theory in an FM situation…"
Exactly. Current ink-jet printers are hybrids.
George
George,
PPI = DPI/sqrt(n) where n= no. of tones<<
NO. DPI = LPI x sqrt(n) where n= no. of tones and that is specific to a conventional linescreen ie A.M.
However, even if we were to accept that formula, 256 levels per channel is the accepted figure for ‘half-tone’ printing to be seen as continuous tone (and is therefore the PostScript limit).
So taking your formula:
ppi = 360/16 – [sqrt(256)=16]
or 22.5. even IF you were to take each print-head as being capable of 1440 dpi it would STILL only work out at 90 ppi!
I’m sure you will agree that both figures are blatantly absurd.
In fact the required ppi of an image when printing using F.M. is a function of the dot SIZE, but since manufacturers only quote figures in picolitres, and the dot size suffers a lot of dot-gain (even on glossy papers), it would not be a particularly helpful figure anyway.
Len,
The printer doesn’t give you 256 tones per channel. It doesn’t need to. You cannot differentiate anywhere near that number. Maybe 25 if you’re under age 40. The printer compromises. 36 tones are more like it. You may even get fewer.
Yes, of course, dot size is related to required resolution. Regarding it’s cloaked appearance as picoliters in printer specs, the dot size is inherent in the DPI spec. And dot gain—a function of the ink and paper—is compensated for in the printer/paper/ink profile as well as specifically in the color management settings.
George
Not quoting but paraphrasing my personal notes from a May 03 conversation with an Epson consultant:
Epson printers use "multi-level half-toning". At each addressable location, multiple drop sizes are available. Depending on the printer, these may be 0,4,8, and 13 picoliters. That provides 2 bits (4 levels) of tonal resolution per ink. Only 6 bits (64 levels) of spatial resolution are then needed to get 256 tonal levels. You can think of them as an 8×8 half-tone cell, although that’s not typically how they are generated. Epson printers do modulate drop volume, effectively modulating drop size, not continuosly but in these discrete quanta.
George
wrote:
I’m not getting it. What additional levels of degredations is len proposing. He says (and it sounds probable) that each head is printing 600dpi for a total of 2400. what’s wrong with that statement? it sounds plausable to me.
please make any replies in the QUEENS ENGLISH ONLY! 🙂
plausable is Queens English?
But back to printing: The printer can position its ink drops for each color at a minimum distance of 1/2880th inch. This is not what causes the difference between the printers dot distance (dpi) and the number of full color pixels it can print per inch (ppi).
The printer needs to create shades between a combination of several full color inks and the white from the underlying paper – this is done by printing not each position.
If you have square that can have an ink drop or not you only have two colors. If you have a four square pattern, you can fill from none to four of the squares, resulting in five shades of the color – but you can print your five color pixel at only half the original dpi. Actual patterns are larger than 2×2, usually somewhere between 6×6 and 10×10, reducing ppi to a sixth or tenth of the dpi.
Michael
wrote:
Dave
PPI = DPI/sqrt(n) where n= no. of tones
If n= 36, sqrt(n) = 6
PPI = 600/6 = 100
A PPI of 100 is not very good. But it gets worse. The 2400 DPI number is inflated (Len ‘s reply #3 re vertical stepping). Use 1200 DPI instead of 2400. In fact, use the 1440 DPI which he really had in mind. Then each head would be printing at 360 DPI. Now for 36 tones:
The situation is not that bad with modern printer drivers. Colors are not only mixed by printing to adjacent dot positions but also by placing two differently colored ink drops on the same position. It is also possible to use the spaces left by one color pattern to place ink dots of other colors. All this reduces the space needed for mixing colors but tends to create darker tones.
The main reason for lower resolution (especially for non photo papers) is the bleeding of the dots in combination with the need for free, white space between the ink drops. Bleeding is reduced on color papers, thus the ink dots stay smaller, more ink can be applied before a generalized state of muddy brown is reached – or more white is visible between the dots.
Actual resolutions on copying paper may be as low as 60ppi for a printer that can print 300ppi on good photo paper – the dpi for that printer may go up to 4800 to specify the precision of printhead and positioning mechanics.
Michael
George,
The printer doesn’t give you 256 tones per channel. It doesn’t need to. << Only 6 bits (64 levels) of spatial resolution are then needed to get 256
tonal levels.<<
Please, at least keep your arguments consistent.
So back to your argument of PPI = DPI/sqrt(n) where n= no. of tones:
You are now saying that required ppi =360/8 [sqrt(64)] i.e 45 ppi Even were we to take YOUR per-head figure of 1440 we’re still only at 180 ppi!
I’m sorry but all this is just ‘junk science’ which will succeed only in the total confusion of those who are interested in what image resolution they require when printing to their desktop inkjet. Consequently I will be ending my contributions to this topic with this post, as I have no wish to encourage you further in spreading confusion.
Len,
Sorry for the confusion, but that’s part of the learning process for me and I regret having apparently exasperated you. For others whom I may also have perturbed, I offer this explanation:
The printer must provide spatial AND tonal resolution. It trades one against the other. It normally can’t provide the ultimate in both. The most innoccuous compromise is tonal. The printer will cut back on the 256 tones as necessary to provide at least a minimally acceptable spatial resolution.
The fundamental reason for the conflict is that the printer achieves tonal variety by using more or less white space mixed in and around the ink dot. It allocates a percentage of a pixel’s area to the ink and leaves the remainder blank. Limited by the finite size of the dot it produces, it can provide multiple tones only by increasing the area allocated for one pixel.
Crudely, the number of tones the printer can provide is given by the ratio of the dot’s area to the pixel area. The larger that ratio, the more tones. So, large pixels give more tones. But enlarging pixels reduces spatial resolution. When the image is defined by larger and larger pixels, it loses definition. Hence, the trade-off. The printer makes the decisions for you.
The "softest" parameter is the number of tones. 256 is great, but yielding to a lower number is least likely to be noticed because the eye has its limits. Spatial resolution is not unassailable either. More isn’t necessarily better. Here I am on shaky ground, but I have seen it said that 360 PIXELS per inch is as much as your eye needs and as far as Epson printers go. Any printer dot resolution beyond that is used to extend tonal range toward 256 tones.
Epson and likely other printers as well give the tonal/spatial trade-off some breathing room by providing multiple dot sizes to moderate ink area coverage, and this takes some of the burden off the pixel matrix—it doesn’t need as many cells.
In all of this, the concept of a round printer dot in a square matrix hole is not necessarily literal, but it serves as a model for understanding the action. When you come down to it, the printer is handling data to guide the deployment of ink mixed with white space in proportion to the tone desired.
Suppose your printer can lay down 2880 DOTS per inch. If you can’t use more than 360 PIXELS per inch, then one pixel can extend over 8 dots horizontally and vertically (2880/360 = 8 ). and still provide the maximum usable spatial resolution of 360 PIXELS per inch. The 8×8-pixel array gives you 64 tones. You only need another factor of 4 to extend that to 256 tones. You pick up the factor of 4 via the dot size control (Remember, current Epson printers allow 0,4,8,and 13 picoliter dot volumes.
There is how you get MAXIMUM tonal AND spatial resolution. You use a printer with enough capability and with settings properly selected. With printers of lesser capability or set to lower dpi, you accept lower spatial and/or tonal resolution.
Time and again, users tell you they can’t see ANY difference between images printed at 2880, 1440 or 720 dpi. It undoubtedly is because the image doesn’t require the tone depth provided at the higher printer resolutions. Try an image more challenging than the Sunday comic strip.
Hey, I’m trying my best to sort this stuff out. I may well be off base. Critiques are welcome. As I said at the outset, you learn by sticking your neck out. In some parts of the world that gets your head chopped off, but certainly not in this forum. Len, are you there?
George
OOPS…Obvious error in my last post:
"…the number of tones the printer can provide is given by the ratio of the dot’s area to the pixel area. The larger that ratio, the more tones…"
Invert the ratio to read "…the ratio of the pixel’s area to the dot area…"
wrote:
Suppose your printer can lay down 2880 DOTS per inch. If you can’t use more than 360 PIXELS per inch, then one pixel can extend over 8 dots horizontally and vertically (2880/360 = 8 ). and still provide the maximum usable spatial resolution of 360 PIXELS per inch. The 8×8-pixel array gives you 64 tones. You only need another factor of 4 to extend that to 256 tones. You pick up the factor of 4 via the dot size control (Remember, current Epson printers allow 0,4,8,and 13 picoliter dot volumes.
A four picoliter dot volume for a 1/2880th inch square of paper is still like pouring a gallon of ink on a 4×4 inch square of a dry 4" thick futon.
My impression is that they use the larger dot volumes to cover multiple dot positions with one shot for reasons of speed, not of precision.
There is how you get MAXIMUM tonal AND spatial resolution. You use a printer with enough capability and with settings properly selected. With printers of lesser capability or set to lower dpi, you accept lower spatial and/or tonal resolution.
Time and again, users tell you they can’t see ANY difference between images printed at 2880, 1440 or 720 dpi. It undoubtedly is because the image doesn’t require the tone depth provided at the higher printer resolutions. Try an image more challenging than the Sunday comic strip.
You can only see the difference if you da all the prints on glossy photo paper – normal copier paper will not show any difference.
Michael
I hope this conversation will not be lost to posterity.
I have zoomed through it, but would like to spend a bit of time trying to follow the arguments which have been tossed hither and thither like a shuttlecock in a game of badminton.
For me, if you have the right scanner/monitor/printer profiles loaded, everything works perfectly.
I didn’t even have to intervene myself: each device that I use loaded its profile and the workflow just works.
Am I lucky, or is it a fact that modern peripherals know how to get on with other without human intervention?
For the record: Epson 4870 scanner, Sony lcd monitor, Epson Stylus Photo 890 printer.
I follow the Len Law: 150 dpi is good enough for nearly everything on an inkjet printer, maybe 200, the proof of the pudding is in the eating (as we Brits are prone to say).
Anyway my customers like what they get. And don’t say we Brits are easily pleased!
