Digital BW, The Print

Hi.

This post is to let people know that a new file has been uploaded to 
the Image Processing folder in the Files area of this site. The file 
(testimage_b&w_linearization.jpg) is intended to help users evaluate 
the linearization of monochrome printing setups.

Although I produced the uploaded file (it was a cold, wet Sunday 
afternoon here yesterday) I can't claim to have originated the idea 
behind it : to the best of my knowledge, that honour belongs to Keith 
Cooper of Northlight Images (http://www.northlight-images.co.uk/) who 
included a small version of one of the targets in his own black and 
white test image. Keith has a lot of interesting stuff on his site, 
and I thoroughly recommend a visit to anyone who hasn't been there 
before.

As soon as I printed it, it was immediately obvious how effectively 
Keith's target showed up the problems in my profile, but I found his 
version too small to be really useful and I wanted to be able to 
relate what I could see to actual tonal values. Sad to say, I 
couldn't find anything better to do yesterday than produce a test-
file of my own.

At the risk of stating the obvious, a quick heads-up for anyone who 
decides to make use of the test image . . . .

1) The targets were created by posterising a gradient. Anyone who's 
created their own linear step-wedge by this method will know that the 
posterizations are larger at the extremes than they are in the 
centre. (You can clearly see this effect in the linear 21-step wedge 
included at the bottom of my test file.) In a sense this is helpful 
because it gives additional real-estate to the highlight and shadow 
tones, but it is done at the expense of the midtones. 

Because of the unequal step sizes, and because a circular gradient 
inevitably has a lot of perimeter, the target with white on the 
outside will always (and should) appear lighter than the target with 
black on the outside. In an ideal world they would both appear mid-
grey if viewed at a distance with somebody else's glasses on, but if 
you manage to achieve that effect using this test-image, then you've 
got yourself a seriously warped profile. :)

I don't see this as a serious drawback, so long as users are aware of 
the issue. If in doubt, cross-check with the posterized circular 
targets to see how much space they allocate to 100% black and 0% 
white.

2) Although we're all used to looking at linear gradients, circular 
ones take a bit of mental adjustment. I found that I only really 
began to make sense of what I was seeing when I printed using a 
couple of different profiles and compared the results.

And once I started to understand what my circular gradients were 
telling me, I came to realise that they're not very forgiving. It's 
important to keep in mind that any test image is a tool and not a 
picture : the important thing is that we're happy with our output, 
not that we're producing a perfect linearization.

Finally . . . I genuinely hope that other people find the uploaded 
file useful. And my thanks go to everyone who has made this group 
such a pleasure to belong to.

Happy printing.

Graham Preston

Great job Preston--Thanks!  The circular gradient really shows up the 
problems.  I too find Keith Cooper's image a very useful tool.

I'd also like to reintroduce the group to Tyler Boley's offset grad 
linearization test image, see Files|Image Processing|offsetgrad8.jpg.

Non-linearities show up as faint upper left to lower right diagonal 
bands.  The location(s) can be directly read off the scales.  Any non-
linearities show up best when viewing the print in good light.  Also 
faint bands can be better identified if the print is rotated and/or 
tipped back and forth (enhanced by our primative ability to sense b/w 
motion).

jr

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "gp295" 
<graham@...> wrote:

... a few things that don't make any sense at all. This post is just to 
correct something i said yesterday.

The top left-hand circle looks lighter than the bottom one because of 
the amount of white perimeter, but this has nothing to do with unequal 
steps in the posterization. (The left-hand gradients are not 
posterized.)

What can i say ...? It was late and i was tired.

Apologies for the confusion.

Graham

Hi Graham,

Thanks very much. I did some testing this weekend, naively assuming my 
prints would be near to perfect. I used both your test image and 
Tyler's offset gradient image.

My setup is: Epson 2100 / windows / QTR / resolution: 1440 super / UT3D 
inkset/ Epson Enhanced Matte.

The offset gradient image printed fine. 

But your spherical images, especially the top left (dark center light 
rim), show some slight circular banding. My feeling is that this is not 
a QTR linearization error. The banding is quite regular (several bands, 
almost equidistant). I suspect it is a dithering issue. I played around 
with the QTR dithering settings and (I think) I saw some slight 
changes, but not a real improvement.

Any suggestion how to reduce this? Or just live with it (I never 
noticed it before)?

Another question: for an optimal test I need to convert the image fram 
Gamm 2.2 to Gray Lab. To exclude any artifacts in the transformation 
could you post the original 16-bit Tiff file as well?

Thanks

Joost

Greetings all,

I too did a test using Graham's test file with similar results and would
be interested in any suggestions on which direction to pursue to improve
my profile.

My setup is:  Epson R2400 / Windows / QTR / 1440 resolution UCpk, MIS
LK, LLK, EZW-Yellow ink set,  Hahnemühle Fine Art Pearl.


I printed Graham's test image after my initial linearization of the
profile for this ink/paper combination and was surprised.   The QTR
linearization seemed good and the step tablet printed smoothly as well.
But as Joost noted, I saw some slight banding on the test file spherical
images.

I was not sure how to correct for this problem so I tried changing
densities and limits in different combinations.  Since my QTR
linearization curve was virtually straight I did not make changes there.
Nothing seemed to help.  I did not try changing the dither algorithm.  
Any  ideas?

Thanks in advance,

Kevin



[Non-text portions of this message have been removed]

Kevin Bradford wrote:
> Greetings all,
> 
> I too did a test using Graham's test file with similar results and would
> be interested in any suggestions on which direction to pursue to improve
> my profile.
> 
> My setup is:  Epson R2400 / Windows / QTR / 1440 resolution UCpk, MIS
> LK, LLK, EZW-Yellow ink set,  Hahnem\ufffdhle Fine Art Pearl.
> 
> 
> I printed Graham's test image after my initial linearization of the
> profile for this ink/paper combination and was surprised.   The QTR
> linearization seemed good and the step tablet printed smoothly as well.
> But as Joost noted, I saw some slight banding on the test file spherical
> images.
> 
Kevin

I'd be interested to know if you also see this banding with ABW for 
example or does it only occur with QTR?

Howard

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Joost Horsten" 
<j.h.j.h@...> wrote:

>To exclude any artifacts in the transformation 
> could you post the original 16-bit Tiff file as well?
 
Hi Joost.
Sorry i couldn't reply sooner.

I'll email the above file to you off-list if that's OK (6.6MB), 
rather than add another version to the Files area of DB&WTP at this 
time.

As for your other points . . .
Unlike you, i started out with a very poor linearization and was able 
to significantly improve it using the test image. (In terms of 
technology, i'm still in the dark ages here.) But like you, i could 
still see some faint banding in the left-hand targets. I'd assumed 
this was something specific to my revised linearization (i.e. that it 
wasn't perfect), but i'd improved things so much i was willing to 
live with this.

But you do seem to have highlighted a more generalised problem than 
i'd thought was the case. In an off-list message Jeff Randall reports 
that he can also see some banding, albeit in a larger version of the 
target which i sent him for an opinion. 

Printing this morning directly from the 16bit original (using the 
workspace profile the image was created in) didn't seem to improve 
the situation, so my first thought was that there was something 
fundamentally flawed in the circular targets. But when i looked 
really closely at the linear ramp below the targets, i found that i 
could see some very faint banding there too, and the location of 
these bands did seem to correspond to those that appeared on the 
circular targets. 

Would you mind checking the linear ramp on your own prints carefully? 
If you can also see some slight banding there, then it may be that 
the circular targets are simply doing what they were intended to do 
(i.e. making things more obvious). 

As for what might be causing the banding (linearization, dithering or 
something else entirely), i'm afraid the explanation will need a 
sharper or more experienced mind than mine. And as for what to do 
about it . . . I mentioned in my original post that i found the 
targets quite unforgiving and thought the important thing was to 
produce prints that we're satisfied with. You were clearly happy with 
your profile prior to printing the test-file, so i'd say that's what 
really matters. I'd hate to feel that you were struggling to improve 
a good profile as the result of an image i'd produced.

Regards.
Graham

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw
<glassman@...> wrote:
> I'd be interested to know if you also see this banding with ABW for 
> example or does it only occur with QTR?
> 
> Howard
>

The ABW print did not show any banding, it was very smooth. 

Kevin

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "gp295" 
<graham@...> wrote:
>
> 
> As for what might be causing the banding (linearization, dithering 
or 
> something else entirely), i'm afraid the explanation will need a 
> sharper or more experienced mind than mine. 


Graham:  How did you create the circular gradient (not the circular 
step-wedge)--maybe it's the maths?

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "gp295" 
<graham@...> wrote:

Hi Graham,

> 
> I'll email the above file to you off-list if that's OK (6.6MB), 
> rather than add another version to the Files area of DB&WTP at this 
> time.

I'd appreciate that. I'll mail you another mail adress that is more 
likely to allow such a size. or perhaps you can put it somewhere on a 
website and post the link?

> But you do seem to have highlighted a more generalised problem than 
> i'd thought was the case. In an off-list message Jeff Randall 
reports 
> that he can also see some banding, albeit in a larger version of 
the 
> target which i sent him for an opinion. 
> 
> Printing this morning directly from the 16bit original (using the 
> workspace profile the image was created in) didn't seem to improve 
> the situation, so my first thought was that there was something 
> fundamentally flawed in the circular targets. But when i looked 
> really closely at the linear ramp below the targets, i found that i 
> could see some very faint banding there too, and the location of 
> these bands did seem to correspond to those that appeared on the 
> circular targets. 
> 
> Would you mind checking the linear ramp on your own prints 
carefully? 
> If you can also see some slight banding there, then it may be that 
> the circular targets are simply doing what they were intended to do 
> (i.e. making things more obvious). 

I did and actually I find it hard to tell on the fairly small samples 
I printed (A6) but on very close inspection there seems to be indeed 
some slight banding in the wedges as well/

> 
> As for what might be causing the banding (linearization, dithering 
or 
> something else entirely), i'm afraid the explanation will need a 
> sharper or more experienced mind than mine. 

Actually, when inspecting the test wedges I made with different QTR 
dithering schemes the banding differed as as well. In the coarser 
schemes ("fast" and "very fast") the banding is more obvious than in 
the other cases. This might support the theory that the dithering is 
the issue.


> And as for what to do 
> about it . . . I mentioned in my original post that i found the 
> targets quite unforgiving and thought the important thing was to 
> produce prints that we're satisfied with. You were clearly happy 
with 
> your profile prior to printing the test-file, so i'd say that's 
what 
> really matters. I'd hate to feel that you were struggling to 
improve 
> a good profile as the result of an image i'd produced.

Yeah. I'm not sure myself how much attention to pay to this. We all 
know there is no magic bullet for optimal print quality. It's the 
results of many small steps. This is definitely a small issue, but 
it's difficult to judge the effect in bigger perspective if you can't 
compare "with" and "without". I must say, I have been staring so much 
with a loupe on the test prints these days, that a slight horizontal 
banding (not related to your test image) has started to annoy me 
more :-(

Joost

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Kevin Bradford" 
<bradford@...> wrote:
>
> --- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw
> <glassman@> wrote:
> > I'd be interested to know if you also see this banding with ABW for 
> > example or does it only occur with QTR?
> > 
> > Howard
> >
> 
> The ABW print did not show any banding, it was very smooth. 
> 
> Kevin
>


Howard,

This seems to support your theory we discussed off-line. Perhaps you 
could elaborate here?

Joost

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Jeff Randall" 
<jrandall@...> wrote:
>
> 
> Graham:  How did you create the circular gradient (not the circular 
> step-wedge)--maybe it's the maths?
>
Hi Jeff.
It did occur to me that maybe there was a problem in the initial 
construction of the circular targets, but i don't know enough about 
how Photoshop creates gradients to be able to comment.

In the hope that there's someone out there who does, here's what i 
did . . .

1) Added a few Guides to mark the centre and perimeter of my circle.
2) Used these to make a circular selection of the size i wanted.
3) Selected the Gradient tool with the Radial Gradient option.
4) Dragged from the centre of the circle to the perimeter.

This produced the lighter of the two targets. To produce the darker 
one (black perimeter) i simply used Invert before deselecting the 
circle. Stunningly simple really, but maybe there IS something 
happening in the maths (as you say) that means these gradients are 
less than perfect. They seem to be fine on screen.

To produce the stepped targets, i again used the simplest method i 
could think of - basically all i did was to posterize the gradient 
before deselecting. (Hence the unequal steps, which i mentioned in my 
original post.)

For the sake of completeness, i should maybe add that i used exactly 
the same technique to produce the linear ramp and step-wedges below 
the circular targets. The only difference was (obviously) that i used 
a rectangular selection and the Linear Gradient tool.

So, does anyone out there know why this technique might cause the 
banding people have reported? And (referring back to my post earlier 
today) can anyone see similar (though less distinct) banding in the 
linear ramp on their prints?

Regards.

graham

Joost Horsten wrote:
> --- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Kevin Bradford" 
> <bradford@...> wrote:
>>
>> --- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw
>> <glassman@> wrote:
>> > I'd be interested to know if you also see this banding with ABW for 
>> > example or does it only occur with QTR?
>> > 
>> > Howard
>> >
>> 
>> The ABW print did not show any banding, it was very smooth. 
>> 
>> Kevin
>>
> 
> 
> Howard,
> 
> This seems to support your theory we discussed off-line. Perhaps you 
> could elaborate here?
> 

Joost

I mentioned to you that, at least at 2880dpi, the QTR 100% ink limit 
calibration test produced an anomalous result at the 50% step with the 
density being less than at 45%. It was therefore necessary to 
interpolate this density. As you surmised this appeared to be a specific 
dithering problem that only occurred at this step.

If you then chose an ink limit of 70% say and reprinted the 21-step you 
wouldn't notice it because none of the steps would be at 71.4% (50/70).

The theory was therefore that the series of bands in Graham's test image 
were caused by each ink (for which you had different ink limits) going 
across it's own 50% mark.

If it is caused by this you should be able to work out where each band 
will occur by looking at each ink's limit and calculating where it 
crosses the 50% mark.

There is contrary evidence to this. For example, at 2880dpi my ink 
limits are often beneath 50% yet when I printed the test image I also 
saw banding. Also you would think that the phenomenon would have been 
frequently noticed in images with gentle gradations at which these ink 
limit boundaries are crossed. On the other hand these bands are 
particularly noticeable in Graham's test but not on a linear ramp so 
maybe we wouldn't notice them anyway.

So, inconclusive but I'd be interested to hear if any other QTR users 
have noticed these odd readings during the calibration test.

Howard

I see no banding on my Epson C84.
I am using Paul's recommendations.
Mike J.

-----Original Message-----
From: DigitalBlackandWhiteThePrint@yahoogroups.com
[mailto:DigitalBlackandWhiteThePrint@yahoogroups.com]On Behalf Of gp295
Sent: Tuesday, December 05, 2006 3:22 AM
To: DigitalBlackandWhiteThePrint@yahoogroups.com
Subject: [Digital BW] Re: New linearization test image uploaded

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "gp295"
<graham@...> wrote:

... a few things that don't make any sense at all. This post is just to
correct something i said yesterday.

The top left-hand circle looks lighter than the bottom one because of
the amount of white perimeter, but this has nothing to do with unequal
steps in the posterization. (The left-hand gradients are not
posterized.)

What can i say ...? It was late and i was tired.

Apologies for the confusion.

Graham

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw 
<glassman@...> wrote:
>
> 
Re : Message # 82082

Howard.

I don't know if this will help, but I've just asked the list-owner to 
upload two more circular gradients to the Files > Image Processing 
folder on this site : (…)_precise_light and (…)_precise_ dark. All 
being well, they should appear there within the next couple of days.

I produced these images a few days ago after an off-list discussion 
with Jeff Randall in which it became clear to me that my original 
test file wouldn't be accurate enough for people who are able to work 
more precisely than I am (using QTR and a 51 gray patches, for 
example). The idea was that these bigger images would make it 
possible for people playing hunt-the-reversal (or similar DB&W 
pastimes) to accurately locate where a problem occurs, and enable 
them to tweak the exact value that needs adjustment. The images 
themselves include an explanation of how they operate.

Jeff was kind enough to offer to try out the images for me, but 
because of some printing problems he's having, won't be able to do so 
for a while. I'm slightly uncomfortable about uploading the images 
without them being tested on the kind of system they were intended 
for because I can't be certain that they will actually do the 
business. But it seemed to me that you might be struggling to do what 
you want with the less-accurate (original) test file, so in the end I 
decided I might as well upload them. I'll ask the list-owner to 
remove them later if they prove not to be helpful. 

Best.

graham

gp295 wrote:
> --- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw 
> <glassman@...> wrote:
>>
>> 
> Re : Message # 82082
> 
> Howard.
> 
> I don't know if this will help, but I've just asked the list-owner to 
> upload two more circular gradients to the Files > Image Processing 
> folder on this site : (\ufffd)_precise_light and (\ufffd)_precise_ dark. All 
> being well, they should appear there within the next couple of days.
> 
> I produced these images a few days ago after an off-list discussion 
> with Jeff Randall in which it became clear to me that my original 
> test file wouldn't be accurate enough for people who are able to work 
> more precisely than I am (using QTR and a 51 gray patches, for 
> example). The idea was that these bigger images would make it 
> possible for people playing hunt-the-reversal (or similar DB&W 
> pastimes) to accurately locate where a problem occurs, and enable 
> them to tweak the exact value that needs adjustment. The images 
> themselves include an explanation of how they operate.
> 
> Jeff was kind enough to offer to try out the images for me, but 
> because of some printing problems he's having, won't be able to do so 
> for a while. I'm slightly uncomfortable about uploading the images 
> without them being tested on the kind of system they were intended 
> for because I can't be certain that they will actually do the 
> business. But it seemed to me that you might be struggling to do what 
> you want with the less-accurate (original) test file, so in the end I 
> decided I might as well upload them. I'll ask the list-owner to 
> remove them later if they prove not to be helpful. 
> 
> Best.
> 
> graham

Okay, thanks Graham.

Howard

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "gp295" 
<graham@...> wrote:
> It did occur to me that maybe there was a problem in the initial 
> construction of the circular targets, but i don't know enough about 
> how Photoshop creates gradients to be able to comment.

Graham, I'd also be interested in hearing from others whether or not 
the banding might just be inherent in the gradient structure of PS.  
Your original post happened at a time I was building a profile with 
PFP 2.0 and UT-3D and the idea looked good as a test.  I'd built a 
larger (8x8) version to print thinking I'd be able to get areas wide 
enough to use a Colorvision Spectro on and I'd kill two birds: a 
visual confirmation of the ramp and a printout I could check for the 
L*a*b axes.  What I got at the 8x8 size before posterization showed 
very visible banding on my monitors and I gave up on printing at that 
point.  I use a ViewSonic 19" VP930b and a Mac laptop lcd, both 
calibrated with a Colorvision Spyder.  The Viewsonic shows in the 
neighborhood of 15 distinct bands and the laptop shows about 8 or 
so.  Any linear ramp I use doesn't show any banding, it's just in the 
circulars

Really don't know if it's a function of the gradient or my monitors.  
I be interested in hearing if anyone gets the same banding in a large 
size on a better LCD or on CRT monitors.  Or another method of 
building the circular gradient?  Robert.

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw 
<glassman@...> wrote:
>
> Joost Horsten wrote:
> > --- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Kevin Bradford" 
> > <bradford@> wrote:
> >>
> >> --- In DigitalBlackandWhiteThePrint@yahoogroups.com, Howard Shaw
> >> <glassman@> wrote:
> >> > I'd be interested to know if you also see this banding with ABW for 
> >> > example or does it only occur with QTR?
> >> > 
> >> > Howard
> >> >
> >> 
> >> The ABW print did not show any banding, it was very smooth. 
> >> 
> >> Kevin
> >>
> > 
> > 
> > Howard,
> > 
> > This seems to support your theory we discussed off-line. Perhaps you 
> > could elaborate here?
> > 
> 
> Joost
> 
> I mentioned to you that, at least at 2880dpi, the QTR 100% ink limit 
> calibration test produced an anomalous result at the 50% step with the 
> density being less than at 45%. It was therefore necessary to 
> interpolate this density. As you surmised this appeared to be a specific 
> dithering problem that only occurred at this step.
> 
> If you then chose an ink limit of 70% say and reprinted the 21-step you 
> wouldn't notice it because none of the steps would be at 71.4% (50/70).
> 
> The theory was therefore that the series of bands in Graham's test image 
> were caused by each ink (for which you had different ink limits) going 
> across it's own 50% mark.
> 
> If it is caused by this you should be able to work out where each band 
> will occur by looking at each ink's limit and calculating where it 
> crosses the 50% mark.
> 
> There is contrary evidence to this. For example, at 2880dpi my ink 
> limits are often beneath 50% yet when I printed the test image I also 
> saw banding. Also you would think that the phenomenon would have been 
> frequently noticed in images with gentle gradations at which these ink 
> limit boundaries are crossed. On the other hand these bands are 
> particularly noticeable in Graham's test but not on a linear ramp so 
> maybe we wouldn't notice them anyway.
> 
> So, inconclusive but I'd be interested to hear if any other QTR users 
> have noticed these odd readings during the calibration test.
> 
> Howard
>

Howard,

I'm not sure what printer you are using.  The one thing that gives anomalies
is too much ink. I can't imagine much else especially at 2880dpi.  But if you 
have some other than that I'd like to see what it is.

Roy

Roy Harrington wrote:
...
>> I mentioned to you that, at least at 2880dpi, the QTR 100% ink limit 
>> calibration test produced an anomalous result at the 50% step with the 
>> density being less than at 45%. It was therefore necessary to 
>> interpolate this density. As you surmised this appeared to be a specific 
>> dithering problem that only occurred at this step.
>> 
>> If you then chose an ink limit of 70% say and reprinted the 21-step you 
>> wouldn't notice it because none of the steps would be at 71.4% (50/70).
>> 
>> The theory was therefore that the series of bands in Graham's test image 
>> were caused by each ink (for which you had different ink limits) going 
>> across it's own 50% mark.
>> 
>> If it is caused by this you should be able to work out where each band 
>> will occur by looking at each ink's limit and calculating where it 
>> crosses the 50% mark.
>> 
>> There is contrary evidence to this. For example, at 2880dpi my ink 
>> limits are often beneath 50% yet when I printed the test image I also 
>> saw banding. Also you would think that the phenomenon would have been 
>> frequently noticed in images with gentle gradations at which these ink 
>> limit boundaries are crossed. On the other hand these bands are 
>> particularly noticeable in Graham's test but not on a linear ramp so 
>> maybe we wouldn't notice them anyway.
>> 
>> So, inconclusive but I'd be interested to hear if any other QTR users 
>> have noticed these odd readings during the calibration test.
>> 
>> Howard
>>
> 
> Howard,
> 
> I'm not sure what printer you are using.  The one thing that gives anomalies
> is too much ink. I can't imagine much else especially at 2880dpi.  But if you 
> have some other than that I'd like to see what it is.
> 
> Roy

Roy

I'm using a 2100. This happens with the 100% ink calibration at the 50%
step for all ink positions. For example these are the density readings
for the latest ink/paper combination I've been making curves for:

	K	C	M	Y	LC	LM	LK
0	0.04	0.04	0.04	0.04	0.04	0.04	0.04
5	0.42	0.32	0.27	0.08	0.17	0.15	0.08
10	0.72	0.53	0.44	0.12	0.27	0.24	0.11
15	1.01	0.74	0.61	0.16	0.38	0.33	0.15
20	1.23	0.91	0.75	0.2	0.47	0.42	0.19
25	1.4	1.03	0.86	0.23	0.55	0.49	0.22
30	1.51	1.13	0.96	0.27	0.63	0.56	0.26
35	1.58	1.2	1.02	0.3	0.69	0.61	0.29
40	1.63	1.25	1.08	0.33	0.74	0.66	0.31
45	1.68	1.3	1.14	0.36	0.8	0.72	0.34
50	1.68	1.31	1.13	0.36	0.8	0.71	0.33
55	1.72	1.35	1.23	0.41	0.87	0.8	0.39
60	1.73	1.37	1.26	0.43	0.9	0.82	0.4
65	1.74	1.38	1.29	0.45	0.93	0.85	0.42
70	1.75	1.39	1.3	0.47	0.95	0.87	0.43
75	1.75	1.39	1.31	0.48	0.98	0.89	0.45
80	1.75	1.4	1.33	0.5	1	0.92	0.46
85	1.76	1.4	1.33	0.51	1.02	0.94	0.48
90	1.75	1.4	1.34	0.52	1.05	0.95	0.49
95	1.75	1.4	1.34	0.54	1.07	0.97	0.5
100	1.75	1.4	1.34	0.56	1.08	0.99	0.52

I'm sure I've read other posts about this in the past but for the moment 
can only find message 69526 which relates to a R200 and 1440dpi and may 
or may not be related:

http://tech.groups.yahoo.com/group/DigitalBlackandWhiteThePrint/message/69526

The exact settings I use are 2880dpi, Better, Adaptive Hybrid.

It's not a big deal - as I said, I've never noticed the effect in an 
actual image and its easy enough to interpolate the 50% step at the 
calibration stage. I just thought this might account for the bands that 
are apparent in the circular test image.

Howard

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "robert49brake" 
<robert49brake@...> wrote:
>
> 
> I'd also be interested in hearing from others whether or not 
> the banding might just be inherent in the gradient structure of PS.

Robert.

I think you may have nailed down the problem in your post. I went 
back and looked at the targets more critically on-screen, and when i 
zoomed to Actual Pixels i could see the banding quite clearly. 
(Pretty much as clearly as i could see it in my prints.) 

Although it was less obvious, unlike you i thought i could see some 
banding in the linear ramp when it was also viewed at 100%. Thinking 
it might be less obvious simply because my linear ramp was narrower, 
i quickly made a full-page linear ramp, and on this image the banding 
was as clear as it was on the circular ones.

So it does seem as if the banding problem is something inherent in 
the way PS creates gradients. And in my view it occurs in both Radial 
and Linear Gradients. I'm surprised i never noticed this before, but 
i guess i just never inspected PS gradients on-screen at that 
magnification until now.

In fact, when i zoomed to 500%, i realised that the banding also 
occurs at a smaller scale as well, though this is not really obvious 
to the naked eye because the steps are only about half a millimeter 
wide.

Like you, i'd be interested to know if other people confirm what we 
think. I'd also be interested to know what people feel about the 
usefulness (or otherwise) of my test images, given this (fairly 
significant) drawback. If the test-files encourage people to see 
problems where there aren't any, it would probably be best to remove 
them from the site. (Who's going to remember this explanation in five 
year's time?)

Thanks for passing on the explanation, Robert. For the record, i'm 
using a calibrated LaCie Electron Blue IV CRT monitor.

Best.

graham

The PS gradient does not band.  I believe you are looking at a less than
optimal monitor and/or calibration.  Your crt should not be that bad; you
are setting the white point of the monitor with it’s controls, and
calibrating to native white point, yes?

I assume you are working with the 16-bit generated gradient, not the jpeg.
View at 800% and select a small area of the gradient where you see banding
(gradient reversals), and do an “auto levels” on the selection.  You should
see that the gradient steps are not banded.  You can also set the info
palette to show 16-bit values, and “go-by-the-numbers” to see that there are
no reversals.

I would convert to 8-bit before printing with QTR.

Best regards,
John Moody

-----Original Message-----
From: DigitalBlackandWhiteThePrint@yahoogroups.com
[mailto:DigitalBlackandWhiteThePrint@yahoogroups.com]On Behalf Of gp295
Sent: Wednesday, December 13, 2006 6:36 PM
To: DigitalBlackandWhiteThePrint@yahoogroups.com
Subject: [Digital BW] Re: New linearization test image uploaded

--- In DigitalBlackandWhiteThePrint@yahoogroups.com
<mailto:DigitalBlackandWhiteThePrint%40yahoogroups.com> , "robert49brake"
<robert49brake@...> wrote:
>
>
> I'd also be interested in hearing from others whether or not
> the banding might just be inherent in the gradient structure of PS.

Robert.

I think you may have nailed down the problem in your post. I went
back and looked at the targets more critically on-screen, and when i
zoomed to Actual Pixels i could see the banding quite clearly.
(Pretty much as clearly as i could see it in my prints.)

Although it was less obvious, unlike you i thought i could see some
banding in the linear ramp when it was also viewed at 100%. Thinking
it might be less obvious simply because my linear ramp was narrower,
i quickly made a full-page linear ramp, and on this image the banding
was as clear as it was on the circular ones.

So it does seem as if the banding problem is something inherent in
the way PS creates gradients. And in my view it occurs in both Radial
and Linear Gradients. I'm surprised i never noticed this before, but
i guess i just never inspected PS gradients on-screen at that
magnification until now.

In fact, when i zoomed to 500%, i realised that the banding also
occurs at a smaller scale as well, though this is not really obvious
to the naked eye because the steps are only about half a millimeter
wide.

Like you, i'd be interested to know if other people confirm what we
think. I'd also be interested to know what people feel about the
usefulness (or otherwise) of my test images, given this (fairly
significant) drawback. If the test-files encourage people to see
problems where there aren't any, it would probably be best to remove
them from the site. (Who's going to remember this explanation in five
year's time?)

Thanks for passing on the explanation, Robert. For the record, i'm
using a calibrated LaCie Electron Blue IV CRT monitor.

Best.

graham



[Non-text portions of this message have been removed]

On banding:

I think that you should first make the target larger (use half 
the circle) and use 360 PPI or 720 PPI in the image at 1:1 
scale on Epson systems. The last to reduce any banding as a 
result of resampling in driver or application you print from. 
The targets are now at 300 PPI, good for Canons and HPs. Set 
no print sharpening in for example Qimage. There may still be 
enough other factors than density shifts that influence the 
result.

After that you could try to add a minimum of noise to the 
image in order to reduce banding while keeping the tone range 
representation. Had to do that too often on gradations I had 
to print for customers.  Depending on the quality of the total 
printing system you have to scale (but keep 360 PPI) and/or 
add noise to get rid of banding. The last may make the test 
less discriminating. There will be some relation in printer 
quality between printing 256 grey shades perfectly separated 
and good detail representation, both on any time of the day. 
If not this test would be over the top anyway.  Wonder how 
many printers actually can pass your test right now without 
banding + having the 256 steps perfectly separated.

Print like you would with the normal images later on, so using 
the perceptually correct printer profile.



Met vriendelijke groeten,  Ernst


|  Dinkla Grafische Techniek  |
|     www.pigment-print.com    |
|             ( unvollendet )            |

I just made two gradients ( a linear and a radial) with same length,
specs, etc. I did them at 360ppi and if you zoom all the way in at
1600% the area around 47-48 L is quited interesting. This is where I
saw a noticeable band when I printed the radial out via QTR, K7 inks,
on my 2200.
If you look at the two gradients at that magnification you can see a
bigger shift than other spots. However, the radial seems to be more
prominent. This is most likely because of how square pixels have to
achieve an 'arc'. Those pointed edges may be visually more
distinguishable than a flat edge.  But does it matter in a real image?
I'd be interested in what others think. 

Scott

> 
> I think you may have nailed down the problem in your post. I went 
> back and looked at the targets more critically on-screen, and when i 
> zoomed to Actual Pixels i could see the banding quite clearly. 
> (Pretty much as clearly as i could see it in my prints.) 
> 
> Although it was less obvious, unlike you i thought i could see some 
> banding in the linear ramp when it was also viewed at 100%. Thinking 
> it might be less obvious simply because my linear ramp was narrower, 
> i quickly made a full-page linear ramp, and on this image the banding 
> was as clear as it was on the circular ones.
> 
> So it does seem as if the banding problem is something inherent in 
> the way PS creates gradients. And in my view it occurs in both Radial 
> and Linear Gradients. I'm surprised i never noticed this before, but 
> i guess i just never inspected PS gradients on-screen at that 
> magnification until now.
> 
> In fact, when i zoomed to 500%, i realised that the banding also 
> occurs at a smaller scale as well, though this is not really obvious 
> to the naked eye because the steps are only about half a millimeter 
> wide.
> 
> Like you, i'd be interested to know if other people confirm what we 
> think. I'd also be interested to know what people feel about the 
> usefulness (or otherwise) of my test images, given this (fairly 
> significant) drawback. If the test-files encourage people to see 
> problems where there aren't any, it would probably be best to remove 
> them from the site. (Who's going to remember this explanation in five 
> year's time?)
> 
> Thanks for passing on the explanation, Robert. For the record, i'm 
> using a calibrated LaCie Electron Blue IV CRT monitor.
> 
> Best.
> 
> graham
>

For what it's worth--

I had downloaded the JPEG image only and on my calibrated LaCie 21" 
the banding was quite visible in the upper left target at 100%.  So I 
printed the target through QTR with three profiles.  The first was 
Cone's linearization (for the 4800/NK7) with no other profile; the 
second with the Cone profile and my own 21x4 random QTR ICC; and the 
third with the Cone profile and my own 51 random QTR ICC.  With the 
first two the printed target looked quite smooth.  With the third, 
the banding seen on the screen was quite visible in the print.

So I'm not quite sure what this speaks to.

Walt


--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "John Moody" 
<moodymz3@...> wrote:
>
> The PS gradient does not band.  I believe you are looking at a less 
than
> optimal monitor and/or calibration.  Your crt should not be that 
bad; you
> are setting the white point of the monitor with it's controls, and
> calibrating to native white point, yes?
> 
> I assume you are working with the 16-bit generated gradient, not 
the jpeg.
> View at 800% and select a small area of the gradient where you see 
banding
> (gradient reversals), and do an "auto levels" on the selection.  
You should
> see that the gradient steps are not banded.  You can also set the 
info
> palette to show 16-bit values, and "go-by-the-numbers" to see that 
there are
> no reversals.
> 
> I would convert to 8-bit before printing with QTR.
> 
> Best regards,
> John Moody
> 
> -----Original Message-----
> From: DigitalBlackandWhiteThePrint@yahoogroups.com
> [mailto:DigitalBlackandWhiteThePrint@yahoogroups.com]On Behalf Of 
gp295
> Sent: Wednesday, December 13, 2006 6:36 PM
> To: DigitalBlackandWhiteThePrint@yahoogroups.com
> Subject: [Digital BW] Re: New linearization test image uploaded
> 
> --- In DigitalBlackandWhiteThePrint@yahoogroups.com
> <mailto:DigitalBlackandWhiteThePrint%
40yahoogroups.com> , "robert49brake"
> <robert49brake@> wrote:
> >
> >
> > I'd also be interested in hearing from others whether or not
> > the banding might just be inherent in the gradient structure of 
PS.
> 
> Robert.
> 
> I think you may have nailed down the problem in your post. I went
> back and looked at the targets more critically on-screen, and when i
> zoomed to Actual Pixels i could see the banding quite clearly.
> (Pretty much as clearly as i could see it in my prints.)
> 
> Although it was less obvious, unlike you i thought i could see some
> banding in the linear ramp when it was also viewed at 100%. Thinking
> it might be less obvious simply because my linear ramp was narrower,
> i quickly made a full-page linear ramp, and on this image the 
banding
> was as clear as it was on the circular ones.
> 
> So it does seem as if the banding problem is something inherent in
> the way PS creates gradients. And in my view it occurs in both 
Radial
> and Linear Gradients. I'm surprised i never noticed this before, but
> i guess i just never inspected PS gradients on-screen at that
> magnification until now.
> 
> In fact, when i zoomed to 500%, i realised that the banding also
> occurs at a smaller scale as well, though this is not really obvious
> to the naked eye because the steps are only about half a millimeter
> wide.
> 
> Like you, i'd be interested to know if other people confirm what we
> think. I'd also be interested to know what people feel about the
> usefulness (or otherwise) of my test images, given this (fairly
> significant) drawback. If the test-files encourage people to see
> problems where there aren't any, it would probably be best to remove
> them from the site. (Who's going to remember this explanation in 
five

> year's time?)
> 
> Thanks for passing on the explanation, Robert. For the record, i'm
> using a calibrated LaCie Electron Blue IV CRT monitor.
> 
> Best.
> 
> graham
> 
> 
> 
> [Non-text portions of this message have been removed]
>

In the photoshop gradient editor, there is a setting for “smoothness”.  Set
that to 0%, and the gradients are truly linear, it makes equal size steps
when you posterize.

Best regards,
John Moody

-----Original Message-----
From: DigitalBlackandWhiteThePrint@yahoogroups.com
[mailto:DigitalBlackandWhiteThePrint@yahoogroups.com]On Behalf Of gp295
Sent: Monday, December 04, 2006 7:26 PM
To: DigitalBlackandWhiteThePrint@yahoogroups.com
Subject: [Digital BW] New linearization test image uploaded

Hi.

This post is to let people know that a new file has been uploaded to
the Image Processing folder in the Files area of this site. The file
(testimage_b&w_linearization.jpg) is intended to help users evaluate
the linearization of monochrome printing setups.

Although I produced the uploaded file (it was a cold, wet Sunday
afternoon here yesterday) I can't claim to have originated the idea
behind it : to the best of my knowledge, that honour belongs to Keith
Cooper of Northlight Images ( http://www.northlight-images.co.uk/
<http://www.northlight-images.co.uk/> ) who
included a small version of one of the targets in his own black and
white test image. Keith has a lot of interesting stuff on his site,
and I thoroughly recommend a visit to anyone who hasn't been there
before.

As soon as I printed it, it was immediately obvious how effectively
Keith's target showed up the problems in my profile, but I found his
version too small to be really useful and I wanted to be able to
relate what I could see to actual tonal values. Sad to say, I
couldn't find anything better to do yesterday than produce a test-
file of my own.

At the risk of stating the obvious, a quick heads-up for anyone who
decides to make use of the test image . . . .

1) The targets were created by posterising a gradient. Anyone who's
created their own linear step-wedge by this method will know that the
posterizations are larger at the extremes than they are in the
centre. (You can clearly see this effect in the linear 21-step wedge
included at the bottom of my test file.) In a sense this is helpful
because it gives additional real-estate to the highlight and shadow
tones, but it is done at the expense of the midtones.

Because of the unequal step sizes, and because a circular gradient
inevitably has a lot of perimeter, the target with white on the
outside will always (and should) appear lighter than the target with
black on the outside. In an ideal world they would both appear mid-
grey if viewed at a distance with somebody else's glasses on, but if
you manage to achieve that effect using this test-image, then you've
got yourself a seriously warped profile. :)

I don't see this as a serious drawback, so long as users are aware of
the issue. If in doubt, cross-check with the posterized circular
targets to see how much space they allocate to 100% black and 0%
white.

2) Although we're all used to looking at linear gradients, circular
ones take a bit of mental adjustment. I found that I only really
began to make sense of what I was seeing when I printed using a
couple of different profiles and compared the results.

And once I started to understand what my circular gradients were
telling me, I came to realise that they're not very forgiving. It's
important to keep in mind that any test image is a tool and not a
picture : the important thing is that we're happy with our output,
not that we're producing a perfect linearization.

Finally . . . I genuinely hope that other people find the uploaded
file useful. And my thanks go to everyone who has made this group
such a pleasure to belong to.

Happy printing.

Graham Preston



[Non-text portions of this message have been removed]