Epson printers - 2400 vs. 4800 ??

[Kennedy McEwen]

The dpis aren't there because someone might actually see
the resolution, they are there because the inkjet dot
is not a '24-bit' dot. The printer has to dither - and the
more dpi, the better it can do this.

So you have to divide the stated dpi by the number of inks
and then further by number of 'levels' you want to have
from one ink (depending on whether the inkjet can modulate
the size of the dot or not this really matters or not).
A 5760x1440 dpi printer with 8 inks is in reality 720x1440
for one ink color. Divide the 720 by two and you get 360 lpi -
something that is not far away from what one can see with
bare eye.

Correct in principle, but not in detail. Your estimate of actual
performance is widely inaccurate. What matters is the amount of density
noise that you are prepared to tolerate, which is infinite beyond visual
acuity and steadily reduces to around 40dB at very coarse detail. The
specific dither process used determines the transition between these two
requirements. Even a meagre 1440x720dpi 4 ink printer is capable of
resolving 720ppi detail, albeit at high noise levels. However, 720ppi
is well beyond visual acuity limits so the noise at that extreme is only
relevant when viewing under magnification. At typical visual limits,
the noise can be almost acceptable, even if not quite photo quality,
depending on the dither algorithm used.

 

[Kennedy McEwen]

Kennedy McEwen spake thus:


Like what--stainless steel? Precision-buffed to a high gloss?

I was thinking specifically of chrome on glass at the time I wrote that,
but there are lots of examples. The data on this CD in my PC is more
than 5760dpi and DVDs are even higher. The entire electronic industry
relies on semiconductor media with details on them at higher dpi still.
;-)

It was a slightly tongue in cheek comment in any case because what Mark
appears to have failed to notice is that the whole point of printing at
such resolutions is that the dots are *NOT* resolved on the media - it
is merely a means to achieve multi-tonal results from monotone inks in
resolutions which are visually perceptible.

 

[Nicholas O. Lindan]

I still challenge anyone to produce evidence that ANY media will show the
useful exhibition of 9600dpi from an ink jet type printer.

Don't confuse dpi [random tiny dots] and pixels [resolved information].

300 pix/inch is what most printers actually provide. Try printing a
test target and look with a loupe. For all practical purposes
300 pixels/inch is really all that is needed.

9600 dpi at 300 pix/inch gives 1024 dots/pixel [the number of dots
in a square pixel is the square of the number of dots along each side].
If we print only magenta dots then there can 1024 shades of magenta.
256 shades, the minimum for colors to look smoothly gradated, requires
8 dots/pix-inch resulting in a true 1200 pixels/inch.

With a 1200 dpi printer at 300 pix/inch there are 16 dots/pixel
and only 16 shades of a pure primary color are possible. To overcome
this limitation the printer blends adjacent pixels to make intermediate
shades and the printer is really doing ~75 pix/inch to get saturated
colors. The printers also mix dot colors [pure magenta at 50% will have
yellow and cyan dots in it] to make more shades of a color
but the color is now unsaturated.

As a result, until high dpi printers became available one had a choice
of snappy colors at a low resolution or sharp pictures but blah colors.

This is also why black & white is so hard to do will with an ink-jet
printer with low dpi: 16 shades of grey just doesn't make it. Color
is added to vary the apparent density but then the grays change their
tint when viewed by a different light and the eye is not well fooled
by the color dithering -- yellow + cyan + magenta = dark muddy brown.

As a further complication, an inkjet printer builds linear reflectance:
25% dots => 25% reflectance, 50% dots => 50% reflectance. The eye,
however perceives logarithmically ["God invented logarithms, man
invented the integers." - somebody famous] and the shades are not
equally spaced. If black is 2.0 OD:

printer #black dots/pix OD
9600 dpi/300 pix/inch printer 1024 2.00
1023 1.96
1 0.00042
0 0.00

1200 dpi/300 pix/inch printer 16 2.00
15 1.14
14 0.87
1 0.03
0 0.00

As is easy to see, there is no possibility for shadow detail in a 1200 dpi
printer.

If you use 8 bit software, though, you are often limited to 256 shades.

# black dots/pix OD Zone Shades/Zone change
255 2.0 0
254 1.86 I 1
251 1.59 II 3
245 1.31 III 6 - detailed black
232 1.00 IV 13 - dark grey
1 0.0017
0 0.00

Easy to see why so many digital photos have no shadow detail.

 

[measekite]

Mark² wrote:

I still challenge anyone to produce evidence that ANY media will show the
useful exhibition of 9600dpi from an ink jet type printer.

Don't confuse dpi [random tiny dots] and pixels [resolved information].

300 pix/inch is what most printers actually provide. Try printing a
test target and look with a loupe. For all practical purposes
300 pixels/inch is really all that is needed.

9600 dpi at 300 pix/inch gives 1024 dots/pixel [the number of dots
in a square pixel is the square of the number of dots along each side].
If we print only magenta dots then there can 1024 shades of magenta.
256 shades, the minimum for colors to look smoothly gradated, requires
8 dots/pix-inch resulting in a true 1200 pixels/inch.

With a 1200 dpi printer at 300 pix/inch there are 16 dots/pixel
and only 16 shades of a pure primary color are possible. To overcome
this limitation the printer blends adjacent pixels to make intermediate
shades and the printer is really doing ~75 pix/inch to get saturated
colors. The printers also mix dot colors [pure magenta at 50% will have
yellow and cyan dots in it] to make more shades of a color
but the color is now unsaturated.

As a result, until high dpi printers became available one had a choice
of snappy colors at a low resolution or sharp pictures but blah colors.

This is also why black & white is so hard to do will with an ink-jet
printer with low dpi: 16 shades of grey just doesn't make it. Color
is added to vary the apparent density but then the grays change their
tint when viewed by a different light and the eye is not well fooled
by the color dithering -- yellow + cyan + magenta = dark muddy brown.

As a further complication, an inkjet printer builds linear reflectance:
25% dots => 25% reflectance, 50% dots => 50% reflectance. The eye,
however perceives logarithmically ["God invented logarithms, man
invented the integers." - somebody famous] and the shades are not
equally spaced. If black is 2.0 OD:

printer #black dots/pix OD
9600 dpi/300 pix/inch printer 1024 2.00
1023 1.96
1 0.00042
0 0.00

1200 dpi/300 pix/inch printer 16 2.00
15 1.14
14 0.87
1 0.03

You are one one thousandth off

 

[rafe b]

Easy to see why so many digital photos have no shadow detail.

As dot-placement resolutions go up, I would think
the "effective contone bit-depth per unit area" would
follow, in proportion to dots/unit area.

Hey, I like that term... "effective contone bit-depth
per unit area." Has a ring to it.

Halftoning and error diffusion are not trivial topics.
Lots of PhDs working on that stuff nowadays. There
are a couple of firmware engineers at my workplace
dedicated to just that.

Plus there are some nice contone digital imaging
systems available to all, eg. LED or laser imaging
on photo paper and thermal (dye-sub.) The latter
is already used in household printers, albeit usually
for very small prints.


rafe b
www.terrapinphoto.com

 

[Mark²]

You didn't say anything about injet printers either! ;-)

What is the Canon in question if not an ink-jet type printer (a poorly
selected generic term for ink-based printers, since "ink-jet" tends to refer
only to HP printers...).

 

[Mark²]

I was thinking specifically of chrome on glass at the time I wrote
that, but there are lots of examples. The data on this CD in my PC
is more than 5760dpi and DVDs are even higher. The entire electronic
industry relies on semiconductor media with details on them at higher
dpi still. ;-)

It was a slightly tongue in cheek comment in any case because what
Mark appears to have failed to notice is that the whole point of
printing at such resolutions is that the dots are *NOT* resolved on
the media - it is merely a means to achieve multi-tonal results from
monotone inks in resolutions which are visually perceptible.

No, I quite understand that each dot is not intended as an individually
perceived unit, but rather a basis through which ink is combined to form
perceived color. Still...can anyone point to the realization of a
visually-perceived benefit of 9600dpi over, say, 4800?

 

[David Nebenzahl]

Mark² spake thus:

What is the Canon in question if not an ink-jet type printer (a
poorly selected generic term for ink-based printers, since "ink-jet"
tends to refer only to HP printers...).

They're all inkjets, even those (like Canon) called "bubble jets" or
some such other. Inkjet is a generic term for printers what squirt ink
onto paper (or other substrate), regardless of what the marketroid types
say. (Including "giclée", the ultimate $2 snob-appeal term.) Even
includes some non-consumer types that use solvent-based (as opposed to
water-based) inks.

Not poorly-selected at all; describes how they work admirably.

 

[Mark²]

Mark² spake thus:


They're all inkjets, even those (like Canon) called "bubble jets" or
some such other. Inkjet is a generic term for printers what squirt ink
onto paper (or other substrate), regardless of what the marketroid
types say. (Including "giclée", the ultimate $2 snob-appeal term.)
Even includes some non-consumer types that use solvent-based (as
opposed to water-based) inks.

Not poorly-selected at all; describes how they work admirably.

Yes.
And oops... -My tired little brain was thinking of "desk-jet" -which has
been nabbed by HP...not ink-jet.

 

[Arthur Entlich]

The real question about companies that claim 9600 dpi inkjet resolution
or what have you is can this be accomplished with the minimum dot size
they offer? In other words, using inkjet media, even with the best
available in terms of minimum dot gain or bleed, could that many dots
occupy the space provided (1/9600th of an inch).

In general, this spec is not stating that this resolution can be created
in a real world situation (such as a photographic image) but that the
printer could "place" any one dot with that accuracy.

Art

 

[Kennedy McEwen]

What is the Canon in question

An Epson! (read the subject) ;-)

 

[Kennedy McEwen]

No, I quite understand that each dot is not intended as an individually
perceived unit, but rather a basis through which ink is combined to form
perceived color. Still...can anyone point to the realization of a
visually-perceived benefit of 9600dpi over, say, 4800?

Reduces tonal noise at the visual acuity limit, which is all that a
finer dither matrix is intended to achieve.

 

[Mark²]

An Epson! (read the subject) ;-)

Yes... But the 9600dpi comment came from a Canon printer comment.
Threads change.
Here's the quote for ya:

 

[Mark²]

Reduces tonal noise at the visual acuity limit, which is all that a
finer dither matrix is intended to achieve.

Yes.
We know that.
But is the difference between 4800 (for example) and 9600 able to be
visually perceived without a loupe?

 

[rafe b]

Yes.
We know that.
But is the difference between 4800 (for example) and 9600 able to be
visually perceived without a loupe?

That would depend on a lot of factors, such as
the relationship of the final ink-dot size (after
diffusion into the paper surface) and the dot pitch
and the number of colors used.

I think the main benefit at that point would be
improved tonality (bit-depth, in effect) rather
than detail.


rafe b
www.terrapinphoto.com

 

[Kennedy McEwen]

Yes.
We know that.
But is the difference between 4800 (for example) and 9600 able to be
visually perceived without a loupe?

Well since 4800 still results in visible noise below the eye resolution
limit, albeit made more obvious with a loupe, that is clearly due to
dither then even without seeing the results the answer must be a clear
"yes".

 

[Nicholas O. Lindan]

The real question about companies that claim 9600 dpi inkjet resolution
or what have you is can this be accomplished with the minimum dot size
they offer?

Well, 9600 dots/inch can be done with 1 inch dots [splats], allowing for
some overlap. Sounds dumb but it would be legitimate if the printer
was adding density with each splat.

Dot size / dot density / pixel density are all independent; excepting
the pathological case of dot size >> pixel size.

The relationship between dot density and pixel density sets the number
of shades that can be produced by the printer. And this has _nothing_
to do with gamma correction.

 

[Mark²]

Well since 4800 still results in visible noise below the eye
resolution limit, albeit made more obvious with a loupe, that is
clearly due to dither then even without seeing the results the answer
must be a clear "yes".

I'd love to see a real world comparison...

 

[Nicholas O. Lindan]

I'd love to see a real world comparison...

Go to the Epson dealer and see for one's self? Everybody
will have a different opinion: "I can't see it.", "Plain
as day.", "Costs too much", "A bargain." ....

 

[Mark²]

Go to the Epson dealer and see for one's self? Everybody
will have a different opinion: "I can't see it.", "Plain
as day.", "Costs too much", "A bargain." ....

Good idea...except Epson doesn't do 9600dpi.