Datacolor User to User Support Group.

Forgive me if this topic has been discussed; couldn't find it in the 
last 6 months.  The issue has come up on the Canon IPF5000 Wiki re: 
Printfix and 8 bit vs. 16 bit profiles.  Thought we could get some 
clearer answers here :-)

Questions:

1) I understand that the 8 bit vs. 16 bit refers to the bit depth of 
the LUT, but don't understand what the practical significance of this 
is in terms of the profile precision.  Why would one choose 8 bit 
only in some circumstances?  What advantage to 16 bit?  Does this 
have anything to do with the bit depth of the target printed to make 
the profile?

2) When looking at profiles in Colorthink, those made with GMB 
Profilemaker (with Atkinson 1728 patch target) show:

16 bit table, 33 grid points
16 bit table, 33 grid points

while those produced by Canon for the IPF5000 show:

8 bit table, 9 grid points
8 bit table, 33 grid points

What does the number of grid points mean, and how does it relate to 
bit depth (if at all) and number of patches in the target used to 
make the profile (if at all)?  I haven't seen any of the printfix 
profiles, but wonder how they compare here.

Any help appreciated on this confusing topic.

Thanks.

--John

>>1) I understand that the 8 bit vs. 16 bit refers to the bit depth of
the LUT, but don't understand what the practical significance of this
is in terms of the profile precision.  Why would one choose 8 bit
only in some circumstances?

Saving space or time would be the only reasons, and with fast computers 
and large drives, its really not an issue. If you were to send profiles 
via low speed connections to other locations, that might be a reason... 
but printer profiles are not tagged to images for distribution, as they 
are printer specific. So there isn't a lot of reason to build low bit 
profiles.

>>What advantage to 16 bit?  Does this
have anything to do with the bit depth of the target printed to make
the profile?

None at all, it just specifies the number of digits each value is 
calculated to.

>>2) When looking at profiles in Colorthink, those made with GMB
Profilemaker (with Atkinson 1728 patch target) show:

16 bit table, 33 grid points
16 bit table, 33 grid points

while those produced by Canon for the IPF5000 show:

8 bit table, 9 grid points
8 bit table, 33 grid points

>>What does the number of grid points mean, and how does it relate to
bit depth (if at all) and number of patches in the target used to
make the profile (if at all)?

No direct relation to bit depth, a coarser grid on the return side of 
the profile (used only for previewing, not printing in most 
circumstances) is just a space saver... the grid is how many 
measurements there are, the bit depth is how many digits in each 
measurement, so you would ideally want more in both... within reason.

>> I haven't seen any of the printfix
profiles, but wonder how they compare here.

Our low bit profiles also contain a coarser grid return table, on the 
theory that those wishing to save space will want to save it there as 
well, like the Canon profiles.  Our high bit profiles are fine grid in 
both directions, like the Atkinson profiles.

>>Any help appreciated on this confusing topic.

I think that covers it; you would have to look very closely to see any 
difference, but we added the highbit/fine return grid option to PFP for 
those who would prefer the larger, and at least theoretically more 
accurate, option.

C. David Tobie
Product Technology Manager
ColorVision, Inc.
CDTobie@...
www.colorvision.com

________________________________________________________________________
Check out the new AOL.  Most comprehensive set of free safety and 
security tools, free access to millions of high-quality videos from 
across the web, free AOL Mail and more.

--- In colorvision_group@yahoogroups.com, CDTobie@... wrote:

> >>2) When looking at profiles in Colorthink, those made with GMB
> Profilemaker (with Atkinson 1728 patch target) show:
> 
> 16 bit table, 33 grid points
> 16 bit table, 33 grid points
> 
> while those produced by Canon for the IPF5000 show:
> 
> 8 bit table, 9 grid points
> 8 bit table, 33 grid points
> 
> No direct relation to bit depth, a coarser grid on the return side
> of the profile (used only for previewing, not printing in most 
> circumstances) is just a space saver... the grid is how many 
> measurements there are, the bit depth is how many digits in each 
> measurement, so you would ideally want more in both... within 
> reason.

Thanks, that answers most of my questions.  Only have two more:

1) I take it that the "coarser grid on the return side" would be used 
for softproofing?  Or does it have some other use?  If so, seems that 
the softproof would be less accurate with the coarser grid.  

2) Would the "coarser grid" cause the profile gamut to display 
incorrectly in Colorthink, or is that unrelated?

3) How do you get 33 grid points out of 1728 patches?  Or are most of 
the grid points interpolated from the actual measured patches on the 
target?

--John

>>Thanks, that answers most of my questions.  Only have two more:

>>1) I take it that the "coarser grid on the return side" would be used
for softproofing?

Yes...

>>Or does it have some other use?

Well, if you hard convert to the printer profile, and then need to 
reconvert to some other profile, there is no other way to do that than 
to convert through that return side... its not particularly ideal to 
convert to one output profile then another; unless you are proofing one 
device to another. So the biggest reason for the high rez grid in the 
return side would be device proofing, but its useful for softproofing 
as well.

  If so, seems that
the softproof would be less accurate with the coarser grid.

In theory... but going to a screen its not as much of an issue, doesn't 
hurt to have the high rez grid in there though, so we chose to do that 
on our high bit profiles.

>>2) Would the "coarser grid" cause the profile gamut to display
incorrectly in Colorthink, or is that unrelated?

ColorThink must use that return side for its graphing, so yes, it would 
effect ColorThink's (or any other such tool's) gamut graphing. I 
wouldn't expect the difference to be large, however, and any such 
representation must be taken as an approximation.

>>3) How do you get 33 grid points out of 1728 patches?

You don't...

>>Or are most of
the grid points interpolated from the actual measured patches on the
target?

Correct...

C. David Tobie
Product Technology Manager
ColorVision, Inc.
CDTobie@...
www.colorvision.com
________________________________________________________________________
Check out the new AOL.  Most comprehensive set of free safety and 
security tools, free access to millions of high-quality videos from 
across the web, free AOL Mail and more.

Thanks for adding to the knowledge base for the Wiki.  Will be adding 
an FAQ, and I know we have several users of Printfix there who will be 
happy to have the additional information.

--John