Digital BW, The Print

Steve Kale wrote:

>
>  
>
>>From: Ernst Dinkla <E.Dinkla@...>
>>    
>>
>
>  
>
>>I know that the profiles are generic and that the underlying (custom)
>>linearisation per printer/inkset should take care of a basic layer to
>>use the generic profiles on top off. That it is still a compromise is
>>also correct and custom profiling on top of a custom
>>ink-limited/linearised-ink/paper/resolution set printer should be
>>better, especially as the Dmax + the white point per printer/paper could
>>vary and the generic profile will not adapt to that. On the other hand
>>we shouldn't forget that many do not have the extra hardware to get the
>>best custom calibrated and profiled output. So a bit more choice in
>>generic profiles may be good for them while QTR should develop to even
>>better control where possible.
>>    
>>
>
>
>I guess people could share profiles for a particular printer/paper/ink
>setup.  I guess I would think it easier in the long run if we didn't have to
>kep coming back to source if, say, MIS released a new and improved Eboni ink
>or a new printer was released.  Roy does enough as it is....
>
>  
>
With custom linearised printers (and I think that that can be achieved 
with cheaper hardware) we probably get much closer to the desired 
quality if there were just 3 to 6 choices of generic perceptual profiles 
or a very simple profile creator that started from a linearised print of 
a short, dedicated target with greys/points fitting the perceptual curve 
creation better.  21 selected greys is probably more than enough for the 
last. But in many cases the generic profiles will be good as well. 
Custom linearising by the user taking away Roy's work .......

>  
>
>>Another thing that could be added to the B&W profiling software is the
>>possibility to use data from 3 readings, either from 3 targets printed
>>or from 3 readings from the same target. Increasing the number of
>>greyscale steps in the target may give less gain in precision than
>>multiple readings from more targets. Depends on the accuracy of the
>>instrument, the handling and the software interpretation. To be honest
>>producing a 256 step target goes beyond the capacity of many printers I
>>think and what has to be done afterwards in getting a profile from so
>>many shades may be worse as a result.
>>    
>>
>
>It is actually very easy to read even 256 patches with Quickread.  It takes
>very little marginal time.  If the patches are shuffled (as my test target
>is) an Eye One can get good readings but I agree there is a potential issue
>there.  I expect measurement devices to continue to improve.  As to the
>printer capability (ie separation), perhaps that is actual.  I guess the
>more observations going into the kTRC tag the more likely you are to iron
>out issues in the original linearization (eg black only is woefully
>non-linear).  As for multiple readings or multiple prints and measurement of
>a test chart, this is likely best left up to the user - the data can be
>easily averaged prior to being read into the ICC profile generator.
>  
>

There are easy ways for multiple readings, multiple sampling in the 
scanner, merging in PS etc.  Averaging the readings in a repeatable way 
with error control is important though and it could well be a source of 
endless threads where it went wrong if that isn't addressed. I'm looking 
for the cheap linearisation and good results, that has to be done as 
foolproof as possible. 3 scans of 35 grey steps targets scanned at 
different angles in a scanner, scanner, averaged (merged 33/33/34) with 
a PS action will probably work though. It is possible to have a better 
scanner greyscale profiling in Vuescan based on an IT8 target profiling 
first or something B&W dedicated (tables or greyscale calibration of the 
scannner) should be developed. If that part of the linearisation is done 
then an extra interation step of 128 grey steps linearising could be 
added if it brings more.

>
>  
>
>>What would be interesting is an
>>iteration feature, where a new target printed with the custom
>>linearisation + custom profile delivers data that can be added to the
>>profile for fine tuning. Then the printer setup has to have more
>>consistency than the iteration feature delivers. I wonder whether that
>>is possible with the desktop printers. Increasing the target steps in
>>both linearising and profiling to 70 steps at most + an iteration of a
>>200-256 step tablet to one of them may be a solution. For the densito-
>>spectrometer and the software it should be better/easier if that was
>>done in the linearisation. The profile could then probably be based on
>>far less than 70 readings.
>>    
>>
>
>Unfortunately this assumes that people are able to linearize.  This is not
>possible if one wanted to use the Epson driver eg Adv B&W or black only.  My
>suggestion would be to leave QTR as is and build the next stage - a program
>to profile its results (or the results of any other workflow).
>  
>
I already thought your aim is drifting away a bit of just QTR 
workflows.  I understand that but wonder why Jon Cone went for QTR after 
his ICC profile method. There must be an advantage in QTR and I think it 
is the linearisation and the flexibility of the inksets/printers it can 
drive. My preference is to bring linearisation at a better and 
affordable level in the first place. Building on that makes it easier 
for everyone.


Ernst