Can't believe someone would say this with a straight face

Eric Stevens wrote:
On Tue, 4 Sep 2012 15:10:18 +0200, Wolfgang Weisselberg
Eric Stevens wrote:
On Thu, 30 Aug 2012 23:32:18 +0200, Wolfgang Weisselberg
Alan Browne wrote:
On 2012.08.26 16:19 , R. Mark Clayton wrote:

The dot pitch of my printer is 1440 and the "standard" resolution is 360
dpi. At any normal viewing distance this is imperceptible.

Not when it dithers over blocks of 64 or more.

What do you mean? If I print at 360 on the Epson 3800 (head is 1440)
the dithering for a print dot should be no more than 4x4 (16 dithered
dots). [not counting dot 'edge' to dot 'edge' dithering]. Why would it
dither to 64 blocks?

So how many colour steps and gray steps do you get at 16 dithered
dots?

I understand where you are coming from but the evidence is that the
Epson 3800 printer (at least) is not confined to the regular
deposition of drops in simple rectangular cells. See
http://gerryeskinstudio.com/ABW_sept.../image002-.jpg
Not only is it hard (?) to see any visible evidence a rectangular cell
containing dot patterns, but it is evident that there need be no
simple pattern of dot depositions.

All I see is that the head passes over the paper several times[1],
with different start timings (which explains the horizontal
overlaps) and I guess the head is slightly tilted (so e.g the
magenta rows are slightly shifted compared to the other rows).

The diagnostic test procedure doesn't suggest the head is tilted. It's
more likely to be a subtle difference in timing of the droplet spray.

Horizontal, yes, vertical, no. That's either tilt or not perfect
alignment of the nozzles. However, the mismatch may be tiny
(less than 1/2 diameter of an ink drop), and I guess you'll not
see that in the diagnostic test procedure.


with the
result that overprinting causes colour-mixing and the formation of
colours other than those of the raw ink. And then there is the ability
of the printer to deposit drops of different sizes, although there is
little evidence of this in the particular example.

I don't know what the dithering algorithm is that is used by Epson but
I suggest it renders moot any analysis based on simple dot patterns
deposited in rectangular blocks or cells.

Still, printer manufacturers give out dpi ...

Buyers expect them, just as they expect megapixels in cameras.

So you invent some number?


'dpi' do make some sense. The 3800 has a basic pitch of 360/dpi

360 divided by dots per inch? What's that?

which
can also manifest itself as 720/dpi. They also claim 1440/dpi and
2880/dpi. Some of the more recent Epson printers built around the same
mechanical print engine even claim 5760/dpi.

The print engine can move the paper in small increments so that
you get 5760 horizontal dot 'lines' per inch (with the
specific printer head).
The print head can't print as dense, so the horizontal dpi
value is smaller, 1440 dpi (or 1440 vertical dot lines).

Presumably these
subdivisions of the basic pitch are managed by software and timing in
some way.

Everything's managed by software these days.

Once software comes into it there is no reason why dots
should be deposited in positions which are simple integer fractions of
the basic pitch.

The limits of hardware also come into it.


Let's make a simple model:
Any spot on the paper can take only a given amount of ink
before it starts smearing and running. Any dot from the
printer has a minimum size and minimum amount of ink.

Thus: there's only so many dots that can be printed in any
given area on paper before there's too much ink. That gives
us a good idea of dpi.

There is nothing in the example I cited above which suggests that the
ink/paper is anywhere near saturation. The print-pattern is
predominantly white space.

Irrelevant. It's a 5% gray swatch, so of course you don't
get saturation YET. But what you *could* get as saturation
and what you could get as an MTF depends on the dpi.

[1] The ink nozzle firing repeat rate is too low to place dot
on dot in a single pass.

Agreed, but we are talking about the print pattern, and the number of
colours it can generate. I don't think we are really much concerned
about how the print pattern is generated.

To understand the print pattern, we need to understand how it's
generated, otherwise we'll get a "analog film has infinite
resolution as it doesn't have a rectangular pixel pattern"
argument.

-Wolfgang