Inkjet Printers: Really 8-bits? Long Question

[Alan Smithee]

OK this is a dumb question, or maybe not. Why do people say printers are
"8-bits"? It seems that a printer with 3 or 4 inks probably can't produce
the same range of tones that a printer with 6 inks or even some printers
these days that sport 12 inks (e.g. that new Canon). New printers also can
makes variable drop sizes too. Doesn't increased gamut mean increased tones?
Is 8-bits just a limit of the OS or Photoshop? I understand that printers
are classified as RGB devices or 8-bits per each channel. Doesn't this mean
that when printing in colour that the printer is actually capable of 256
tones for each colour channel? How about that fact that printers are using
CMYK inks in most cases. To make RGB Red you need two dots to overlap
(Yellow and Magenta) to make Green you need Cyan and Yellow, to make Blue
you need Cyan and Magenta. Can, or do, inkjet printers "dither" tone
information between these tones? Is 8-bits an accurate term? Thx.

 

[zakezuke]

OK this is a dumb question, or maybe not. Why do people say printers are
"8-bits"? It seems that a printer with 3 or 4 inks probably can't produce
the same range of tones that a printer with 6 inks or even some printers
these days that sport 12 inks (e.g. that new Canon). New printers also can
makes variable drop sizes too. Doesn't increased gamut mean increased tones?
Is 8-bits just a limit of the OS or Photoshop? I understand that printers
are classified as RGB devices or 8-bits per each channel. Doesn't this mean
that when printing in colour that the printer is actually capable of 256
tones for each colour channel? How about that fact that printers are using
CMYK inks in most cases. To make RGB Red you need two dots to overlap
(Yellow and Magenta) to make Green you need Cyan and Yellow, to make Blue
you need Cyan and Magenta. Can, or do, inkjet printers "dither" tone
information between these tones? Is 8-bits an accurate term? Thx.

From my understanding, an inkjet produces only one "tone" per color

channel. There is no means of making a given tone lighter or darker.
You can vairy the size, or have different sizes, but ultimatly it's
just a stark dot. So one could percieve a given image as being 4 bit,
1 bit per color channel, or in the case of that new Canon 12bits. But
we are talking about bits in the neighborhood of 4800x1200 to
9600x2400. DPI is somewhat meaningless because these dots of dithered
in such a way you tend to not notice giving the illusion of a
continuious tone... but up close they are very much like a Monet
painting. The effective dpi is much lower than the advertised DPI...
given the choice between a 9600x2400 dpi inkjet or a 600dpi dye sub
offering continious tone... odds are the continious tone will better,
though we are reaching the point were one would need a loop to really
tell the difference. Or given the choice between a 4800x1200 8 channel
printer and 9600x4800 4 channel one, odds are the 8 channel will win,
though odds are you'd need a loop to really tell the difference.

But whether bits is an accurate term is hard for me to say... as you
are definatly feeding information to the printer in terms of bits per
color, and it's translating it into dithered paterns. But you could, I
imagine, percieve a 9600x4800 printer as being 1200x600 if you look at
it in terms of 8 dot by 8 dot segments. You effectivly have the
possiblity of 256 paterns by the number of channels. I honestly don't
know exactly how the printers work at this low level, this was just a
means of illistrating a point.

 

[Arthur Entlich]

It is a good question Alan.

The answer is longer than the question, however ;-) (Isn't it usually?)

As you surmised, the actual colors in terms of one dot, that the printer
can produce is limited to the ink colors plus any direct overlapped or
overprinted color.

So, a simple CMYK printer can basically product, in terms of pure color,
assuming transparent inks, only the following colors:

Cyan, Magenta, yellow, green, red, blue and black.

The light colorant load inks allow for more variations of shades, and
yes, some printers also include other ink colors such as blue, red,
green or orange to improve gamut, but the process is still the same.

All inkjet printers reply on some form of dithering to create the colors
that we perceive. They also reply upon the use of white paper space or
gaps to produce shades. All inkjet printers use some type of matrix of
dots to make up a "cell". In pure language this isn't quite correct
because the "cell is continually being redefined as the next line of ink
is laid down.

The use of dot size is still another way of controlling the perceived
color density of a dot. With a high gamut ink, the ability to produce a
very small dot (2 or even 1 picolitre) makes for a nearly impossible to
see "grain" pattern. In my opinion, a well designed driver should allow
for nearly perfect gradients with just a CMYK printer as long as the
dots can be made small enough, vary in size and be positioned well
relative to the other dots. Probably the main weakness in these CMYK
printers is the K, which is so dark that integrating it into the color
areas often creates graininess, which is why some companies have begun
offering printers with two or three gray shade inks, to lessen that
problem. If the black dot could be made .5 picolitre it may become no
problem, but it is often hard to fractionate black that small without
clogging.

As to bit depth of the color, inkjet printers probably aren't really
even 8 bit printers. One would be hard pressed to find 256 levels of
distinguishable color values in one tone. However, the drivers are
usually designed around creating data values of 8 bit, and then the
output printer driver does its best to make that many distinctions,
albeit very likely unsuccessfully.

Then again, most continuous tone image processes (even photographic
images made via silver halide) don't have much if anything above 8 bit
accuracy. The printer driver's ability to accurately move images back
and forth between the analog and digital creation processes become of
more interest when color matching is more critical. It is in this area
that some manufacturers or third party products can make a difference.

Art

 

[Alan Smithee]

Thanks Art that's good info to know. What about that new Canon Printer, the
iPF5000, which claims to be a 16-bit printer (by-passes the OS driver). So,
16 bit would be 65536 per Red, Green and Blue channel? I've also read that
the GIMP photo editor for Linux has been able to send 16-bit images to
Gutenprint for a little while already. Seems like aficionados would be
flocking to a system that could do this (the Linux learning curve probably
slows that down). Also, what's the point of the manufacturers adding the new
inks like Red and Green if those colours can be produced by the CMY inks? Do
the new Red and Greens just print when the colour called for is close enough
to the colour of those ink? I don't quite understand how this would increase
the gamut. Also, re bit depth I read that if an image was printed at 1440
resolution at 360 DPI that would be 4 dots per pixel, the writer equated
this to 3 or 4 bits, this must have been an old printer.
Thx.

 

[Arthur Entlich]

The process is not quite that directly equitable. Canon, or GIMP can
speak about 16 bit printing design, but can the printer/ink/paper
mixture actually define that many levels of color? In general, no. The
addition of several extra ink densities and colors plus variable dot
improves the apparent bit depth of the final image, but I really doubt
they can squeak even 12 bit, let alone 16 bit. That would imply, as you
state, that about 65K different discernible levels of each primary would
be reproducible. Sorry, I just don't buy it even if the driver could
work with the data. And if the output can't reproduce it, it just means
a heck of a lot of data data to crunch through, probably vastly slowing
down the spooling process.

Maybe they will eventually develop a method to get extra bit depth from
the printer, but right now, I think it a tough act to bring forward.

The only printing methods that *might* be able to take advantage of a 16
color bit depth might be a LED digital light system that uses real
photographic paper. Although I somehow doubt they could get 16 bit
there either, but maybe 10 bit (1024 levels) or twelve bit (4096 levels).

Personally, I think this is some more specsmenship than anything showing
up on the paper.

Art

 

[SteveB]

HP claim 1.2 million colours without halftoning using PhotoREt IV
technology.

http://h10025.www1.hp.com/ewfrf/wc/...&docname=bpu02182&lc=en&jumpid=reg_R1002_USEN