Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Using Polarisation for Better Profile Creation  (Read 1180 times)
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« on: October 10, 2013, 01:28:36 AM »
ReplyReply

A friend and I have been experimenting extensively over the last week with a Barbieri LFP Spectro with the polarisation option and we are both more than shocked at the results we are seeing. I would really like to hear from others who may have tried this as the results are just simply night and day compared to non polarised profiles with matt paper in the dark tones.

When creating icc profiles with the polariser option on matt papers we are getting incredible shadow reads that are producing profiles significantly better than anything I have ever seen before with matt papers without polarisation. I can clearly demonstrate polarisation is the key to this by re-profiling without the polariser in place. Without the polariser the shadow reads simply revert to what I am used to seeing with other spectro's. Neither worse, not better - just expected.

Using this tonal gradation chart I created in PS I get full tonal gradation down to just 5 with my polarised profile on Moab Somerset Museum Rag with the Canon IPF6350 printer. This is incredible performance from a matt paper profile. With a non polarised profile the best I can get is 20 - under that there is no tonal gradation whatsoever. What this means is that the profile created with polarisation has far great shadow details, requires far less work in preparing the file during soft proofing (less work to restore contrast etc.) and as a result far better prints.

I am curious if others here have tried this system. Polarisation does come with some drawbacks so it isnt all beer and skittles. But from what I can see the benefits far outweigh any increase in noise. I would have perhaps expected to see a drop in the white point as well using polarisation - however I get tonal gradation up to 253 either with or without polarisation.

I should stress that the Barbieri is NOT cheap. It is over $8000USD with the polariser option. So this is not going to be something everyone can try. Additionally, orientation of the polariser is hyper critical. If it is off by even a fraction the results are no where near as good. It takes a LOT of trial and error to get it right.

The other really interesting thing that has come out of our extensive testing is how much better the gamut is of the Epson 7900 compared to the Canon IPF6350. Have a look at this gamut plot showing how much better the 7900 performs. And this was a comparison made after extensive testing to find out what the optimal ink loads where for the Canon. The only place the Canon performs slightly better is the purple spectrum. As the Canon user I felt somewhat miffed at first. Like I was not being allowed to play with the full box of crayons! However, the reality is that in actual prints this makes very little difference to the naked eye. Anyway - the plots say it all. 7900 is the wire frame and the paper was Moab Somerset Museum Rag (which BTW: has the best DMAX of any of the matt papers we have tested - a DMAX of 2.7 on the Epson and 3.0 on the Canon. Again incredible for a matt paper).
Logged

darlingm
Sr. Member
****
Offline Offline

Posts: 291


WWW
« Reply #1 on: October 10, 2013, 04:24:18 AM »
ReplyReply

When creating icc profiles with the polariser option on matt papers we are getting incredible shadow reads that are producing profiles significantly better than anything I have ever seen before with matt papers without polarisation. . .

Curious on how that the profiles are "better".  They certainly have a wider gamut, but "better" to me means accurately characterizing a media.  A smaller gamut profile is better if it's more accurate to the media than an inflated larger gamut profile.  Any reason to say regular photospectrometers and the Barbieri without polarisation are characterizing it wrong?

EDIT: By that, I mean L* x has a defined meaning.  I think of it as that our perception is that a color is x% as bright as the light hitting it would appear through a diffuser.  (Probably not a spot on definition, just how I think of it.)  I'd have to believe that your darkest black using a Barbieri with polarisation profile is visually the same as your darkest black with a no polarisation profile.  In that case, it would just seem to me like all that's happening is that the polarisation is clipping the black.  That is, re-defining what color science would regard perhaps as L* 8 to mean L* 2.
« Last Edit: October 10, 2013, 04:32:11 AM by darlingm » Logged

Mike • Westland Printworks
Fine Art Printing • Amazing Artwork Reproduction • Photography
http://www.westlandprintworks.com • (734) 255-9761
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« Reply #2 on: October 10, 2013, 04:59:38 AM »
ReplyReply

I think you might have misunderstood my post. Firstly the gamut plots I posted are comparing polarised profiles from the two printers. They are not comparing polarised vs non polarised. All they show is the different gamut between the printers.

Secondly I never said the gamut of the polarised profile was greater. I said polarised profiles provide far better tonal gradation deep into the shadows. The evidence is irrefutable. The tonal gradation chart shows it when printed with a polarised profile vs a non polarised profile,

It's not that the polarised Spectro is right and the non polarised wrong. It's just that the polarised  can read much more deeply into the shadows and has the ability to distinguish dark tonalities much better.
Logged

Bullfrog
Full Member
***
Offline Offline

Posts: 168


« Reply #3 on: October 10, 2013, 05:07:04 AM »
ReplyReply


The other really interesting thing that has come out of our extensive testing is how much better the gamut is of the Epson 7900 compared to the Canon IPF6350. Have a look at this gamut plot showing how much better the 7900 performs. And this was a comparison made after extensive testing to find out what the optimal ink loads where for the Canon. The only place the Canon performs slightly better is the purple spectrum.

From the canon wiki:

9/28/10 There is a new page on the Wiki comparing gamuts of Canon 8300 to Epson 9900 on Canson Baryta. The Epson has a slightly larger gamut volume (4.4% larger in Colorthink Pro 3), but a dmax which isn't quite as good (L* of 3.9 for Canon, 4.7 for Epson). The Epson has a larger gamut in the lighter tones, particularly oranges, pinks and reds in the L* range of 60-80, while the Canon wins in the darker tones in blues and purples of L* 10-25. Softproofing of a variety of images with saturated colors showed no difference, except for several images with light, saturated oranges, pinks and reds, where Epson had a modest advantage. However, note that Scott Martin reports seeing better purples and yellows in actual prints, so there are limitations to the gamut comparisons. Click the link to read the details and see an AVI file showing the gamuts at all L* levels, and examples of the softproofing differences noted above.

http://canonipf.wikispaces.com/

Add:  Another quote direct from Wiki which I think explains things really well

Caveat from Scott Martin:

Different applications certainly calculate gamut volumes differently. We've had a bunch of discussion about this on the ColorSync Users list and, basically, ColorThink is considered the gold standard for this. It's really important to only compare profiles made using the same procedure and profiling software though. I did that for the comparison John is illustrating.

I think it's important to look beyond the total gamut volume number. One inkset may have a larger total volume than another but is it larger in ways that are meaningful? The HP Z3100 had a large gamut for it's day but it was all in the highlights - areas that didn't matter to many people. I find the greatest value in having gamut in the darker colors and that's an area where Epson and Canon are focusing on.

Another thing that's super important is to make prints and let them be the final judge! The results of the Perceptual intent aren't shown in these gamut maps - only the Relative Colorimetric intent. Sometimes all the numbers and science of it all doesn't jive with the real world final results with human perception mixed in.

Canon's gamut is really good in the deep purples and yellows. Are those important colors for your work? I've printing some photos of flowers with deep purples and yellows and the 8300 printer were noticeablly better than the 9900. My own photos don't contain those colors so it's a moot point for me. The fact of the matter is that both of these printers rock and it's unlikely that 99% of images have enough saturated color to begin with to stretch the limits of their gamuts. I'm finding the subtle differences in gloss differential and scratch resistance to be actually more important. On a Baryta paper these differences really show up.

When I make reviews I try to be pretty balanced with these observations and not rely to heavily on the tech data alone - as it can actually be misleading and steer people towards the wrong decision at times.

Softproofing with the intent is a pretty good way of doing that. Even so, the perception of pigments and the light scattering qualities with different pigments aren't taken under consideration. Canon's new yellow inks don't contain more yellow pigment than the previous generation, but because the particles are smaller the light scatters differently it has a very real impact on the way we perceive saturation, but not on the way it's documented colorimetrically. The yellows and greens visually appear to have more saturation because of the finer particles, yet that may not be shown in the gamut rendering.

Bottom line - you gotta look at prints. They'll show you all kinds of things that the gamut plots don't - and sometimes the results and conclusions are counter to what you might have determined from the numbers. Sometimes the gamut plots show that one printer has better reds for example but when you make prints with perceptual the other printer might clearly and obviously have much better reds.
« Last Edit: October 10, 2013, 05:13:56 AM by Bullfrog » Logged
aaronchan
Sr. Member
****
Offline Offline

Posts: 279


« Reply #4 on: October 10, 2013, 12:12:03 PM »
ReplyReply

I have used the LFP Spectormeter before and yes, all the data that it has captured was way more accurate than any other devices on the market right now. But I have not used the pola filter. Even that, with more advance lighting system and different size of aperture, it can capture very good data from super heavy textured media such as canvas. I had used it for a computer to build an ICC for Chinese transitional rice paper, it's far more better than using just an i1pro, from data to visual evaluation.

aaron
Logged
darlingm
Sr. Member
****
Offline Offline

Posts: 291


WWW
« Reply #5 on: October 10, 2013, 03:50:36 PM »
ReplyReply

When you say "to just 5", "best I can get is 20", and "tonal gradiation up to 253", what situation are you referring to?  An actual print, I assume?  Printing via relative, perceptual, or absolute?

Using the 5/20/253" values based on what was in the file (hence, how the square is labeled on the print), or doing another reading with the Barbieri with polarisation and coming up with RGB 5/5/5 20/20/20 and 253/253/253 under a certain lighting condition and converting to RGB using which working space?

Printing 8 or 16 bit?  (Not how many bits your test file is, but how many bits is the printer pipeline?)  As you probably already know, the Windows printer system limits to 8 bits, unless using a RIP that bypasses the stock driver and printing system and communicates with the printer in 16 bit.  (Thank you Microsoft.)  Mac is certainly capable of 16 bit printer pipeline printing, not sure if it has to be enabled or if it does so by default.
« Last Edit: October 10, 2013, 03:54:07 PM by darlingm » Logged

Mike • Westland Printworks
Fine Art Printing • Amazing Artwork Reproduction • Photography
http://www.westlandprintworks.com • (734) 255-9761
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« Reply #6 on: October 10, 2013, 04:51:20 PM »
ReplyReply

When you say "to just 5", "best I can get is 20", and "tonal gradiation up to 253", what situation are you referring to?  An actual print, I assume?  Printing via relative, perceptual, or absolute?

Using the 5/20/253" values based on what was in the file (hence, how the square is labeled on the print), or doing another reading with the Barbieri with polarisation and coming up with RGB 5/5/5 20/20/20 and 253/253/253 under a certain lighting condition and converting to RGB using which working space?

Printing 8 or 16 bit?  (Not how many bits your test file is, but how many bits is the printer pipeline?)  As you probably already know, the Windows printer system limits to 8 bits, unless using a RIP that bypasses the stock driver and printing system and communicates with the printer in 16 bit.  (Thank you Microsoft.)  Mac is certainly capable of 16 bit printer pipeline printing, not sure if it has to be enabled or if it does so by default.

Yes, sorry for the confusion. When I refer to 5 and 253 I am referring to the printed patches on the target I created and attached to my first post. This is using the polarised profile with relative colorimetric with the Canon 16 bit plug in. So I am bypassing Apple's horrible Colorsync and have a true 16 bit pipe. The squares labelled on the target are the relative densities. So it could be said that with the polarised profile printing with Rel Col from the Canon 16 bit plug in in PS that the relative density I am getting on Somerset Museum Rag is '5' and '253' Below 5 there is no tonal gradation. Above 253 there is no gradation. The non polarised profile only shows graduation down to patch 20 (which is the best I can achieve with either the Barbieri or the i1IO without polarisation).
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 8030



WWW
« Reply #7 on: October 10, 2013, 06:02:56 PM »
ReplyReply

When creating icc profiles with the polariser option on matt papers we are getting incredible shadow reads that are producing profiles significantly better than anything I have ever seen before with matt papers without polarisation.
To be clear, you're saying better with polarization on the same measuring device?
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« Reply #8 on: October 11, 2013, 01:32:07 AM »
ReplyReply

To be clear, you're saying better with polarization on the same measuring device?

YES. Significantly Better - In fact its game changing. I have spent most of the last two days re-profiling my favourite matt papers as a result. I will caveat this comment with the statement I made above however. Orientation of the polariser is hyper critical to the end result. If its out by even a smidge (a technical term Grin) the results are nowhere near as good in the shadow reads.

Attached screen shot showing the non polarised profile on the left and the polarised on the right applied as soft proof in CS6 with Rel Col intent with simulate paper and ink. The results speak for themselves and are even more obvious in print. In print there is no tonal gradation whatsoever under 20 in the non polarised profile. On the polarised I have printed tonal gradation down to just 5.

I have not used Black Point Compensation (and never do). The reason is, I have no idea what Adobe is doing under hood with BPC in terms of how much 'lift' it provides. I am sure it is documented somewhere but I cant be bothered looking for it and I prefer to adjust my Black point manually once I know the exact point at which tonal gradation begins. In the case of the new polarised profile I can apply a curve to the image with a black point lift of '5' (since anything under that is not going to have any visible tonal variation anyway). If I want a solid black in the print I can set it to less than 5.

As an aside Black Point compensation is not an option in the Canon 16 Bit Plug in anyway - so any adjustment for Black point needs to be done manually if one is printing with this plug in.
« Last Edit: October 11, 2013, 01:42:44 AM by Josh-H » Logged

Rhossydd
Sr. Member
****
Offline Offline

Posts: 1862


WWW
« Reply #9 on: October 11, 2013, 02:50:24 AM »
ReplyReply

Orientation of the polariser is hyper critical to the end result.
How are you choosing to set this ? trial and error ?


An interesting thread. Especially as my old Specrascan T has a pl filter I've never bothered to try.
Logged
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« Reply #10 on: October 11, 2013, 02:52:39 AM »
ReplyReply

How are you choosing to set this ? trial and error ?


An interesting thread. Especially as my old Specrascan T has a pl filter I've never bothered to try.

Yes, unfortunately trial and error has been the only way to really get an ideal orientation. A process which is extremely time consuming.
Logged

Rhossydd
Sr. Member
****
Offline Offline

Posts: 1862


WWW
« Reply #11 on: October 11, 2013, 03:10:25 AM »
ReplyReply

Have you tried using a normal polarizing filter first to try to get a rough idea of the orientation or is it impossible to see this sort of subtly with the naked eye ?

Any idea what sort of tolerance we're discussing here in angle ? +/- 1°, 5°, 10° ?
Logged
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« Reply #12 on: October 11, 2013, 03:34:30 AM »
ReplyReply

Have you tried using a normal polarizing filter first to try to get a rough idea of the orientation or is it impossible to see this sort of subtly with the naked eye ?

Any idea what sort of tolerance we're discussing here in angle ? +/- 1°, 5°, 10° ?

Hard to quantify the tolerance as there are no measurement markings to put a number against. However, I would guess its in the order of +/- 1º. Since very subtle adjustments quickly yield results either side of the lowest shadow reads. The Polariser itself is very small and a ridiculously expensive option at over $700 for its size.

Have not tried using a normal polarising filter although I guess you could use this to get a sort of starting ballpark. Tough to eyeball it though.

I have been trying to find something online that jells with my own results with the polariser and found this from an Australian company using the same product.

We also use a polarising filter to subtract flare, providing much more open shadow tones. SOURCE I have no idea about this company - never heard of them before, just stumbled on them whilst googling for others who are using this instrument with the polariser. Their offices are closed now as its 7:33pm here in Melbourne at the moment, but I am going to give them a call tomorrow and see if I can find someone there knowledgeable about this and to find out in more detail about their results.

Logged

darlingm
Sr. Member
****
Offline Offline

Posts: 291


WWW
« Reply #13 on: October 11, 2013, 12:25:45 PM »
ReplyReply

In the attached TestChart.jpg, the RGB values of each patch don't match up with the labeled number.  Patches 0 & 1 are both 0/0/0, 2 & 3 are both 1/1/1, 4 is 2/2/2, 5 is 3/3/3, etc.  252-248 are all one high, ranging 253-249, etc.

The attached JPEG is in sRGB.

I see in your recent screenshot there's a TestChart.tif.  Perhaps that one's correct, and the conversion to JPEG using compression may have thrown things off.

In any result, what working space is the test chart that you're printing?  sRGB/Adobe/ProPhoto?
Logged

Mike • Westland Printworks
Fine Art Printing • Amazing Artwork Reproduction • Photography
http://www.westlandprintworks.com • (734) 255-9761
Josh-H
Sr. Member
****
Offline Offline

Posts: 1905



WWW
« Reply #14 on: October 11, 2013, 04:02:24 PM »
ReplyReply

In the attached TestChart.jpg, the RGB values of each patch don't match up with the labeled number.  Patches 0 & 1 are both 0/0/0, 2 & 3 are both 1/1/1, 4 is 2/2/2, 5 is 3/3/3, etc.  252-248 are all one high, ranging 253-249, etc.

The attached JPEG is in sRGB.

I see in your recent screenshot there's a TestChart.tif.  Perhaps that one's correct, and the conversion to JPEG using compression may have thrown things off.

In any result, what working space is the test chart that you're printing?  sRGB/Adobe/ProPhoto?

Yes, there is errors in the attached jpeg test chart as a result of color space conversion so you wont get an accurate read from it. I just posted it so that you could see the type of chart I am using to test for gradation. The test chart Tiff is in Pro Photo. I can put it in a drop box if you want to have a look at it (It will just have to wait until I get back from New Zealand next week).
« Last Edit: October 11, 2013, 08:45:44 PM by Josh-H » Logged

Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad