ChartThrob V1.01

[Digidurst]

Very happy with both, as it turns out!

If you want to see a Ziatype, look here: (warning - contains nudity) http://www.eljay.org/ebay/ziatype5.jpg

Wow! The print is gorgeous on screen - bet it looks 10 times better in person. Thanks for sharing the fruits of your labors

 

[bjorke]

Very happy with both, as it turns out!

If you want to see a Ziatype, look here: (warning - contains nudity) http://www.eljay.org/ebay/ziatype5.jpg

That's fantastic! Thanks for sharing your results. Very lovely tones in that image, I'm glad they printed well.

I can only guess that your scanner is pinning on the high values? Never thought of that issue, hmmm, will have to add to the FAQ...

 

[sanfrancisconudes]

I can only guess that your scanner is pinning on the high values? Never thought of that issue, hmmm, will have to add to the FAQ...

Something along those lines seems likely. Now that I have an approach that works, if I get a chance I'll try scanning again and seeing if I can spot what's different.

 

[Katharine Thayer]

Setting (or not) black and white points?

I'm a gum printer; I know a lot about printing gum but not very much about this curve stuff, nor have I been very interested in it up til now. But the mention of this ChartThrob got me curious and I've tried it out a bit.

It seems like it works best for me not to set the black and white points, but rather to leave them where they are, at the ends of the scale, outside (in some cases, such as a very light pigment like yellow, WAY outside) the printed values. I gather from Kevin's comments that the main thing is to be sure that the black and white points aren't cutting off any of the printed values. When I set the black and white points manually, as described on the FAQ, I get a curve that I don't like as well. I'd be interested in comments or suggestions. I couldn't find guidelines for including images in posts here, so just put them on my website:

http://www.pacifier.com/~kthayer/html/Curve.html

Katharine Thayer

 

[Katharine Thayer]

Nonlinearity

P.S. I'm also curious about the nonlinearities in the verification charts. I would assume that the nonlinearity in the original calibration chart could be due to something with the printer, but I would also assume that such nonlinearity would be corrected by the curve. So I'm curious; does the nonlinearity in the verification chart simply mean, as someone said, that the curves as generated sometimes need further tweaking, or could the nonlinearity be due to some other cause?
Katharine Thayer

 

[mkochsch]

Coloured Negatives

Katharine,
Are you using coloured negatives or Black and White (CMYK all colours)? Ideally you should end up with a curve that starts in or close to the lower left corner and finishes in the upper right corner. Clay's example curve is pretty close to what I consider a "normal" looking curve.
View attachment 28
In general, if you set the starting colour (the colour that becomes your whitest of highlights) correctly you avoid having to "curve out" the excess ink/density at curve building time.
Most negatives I've used for iron process usually have to make up density in the mid-tones and shadows. Hence you see a bulge from the median in the centre and upper right areas of the curve. I've heard of the weird gum curves, however I don't have enough experience to say what a "typical" curve would look like for gum. I suspect most are close to the example but some are very unique (the yellows).
As for non-linearity, it could be a number of things. Inconsistent coating, lighting source etc. The ChartThrob script is just going to send back whatever gets put into it.

 

[sanfrancisconudes]

P.S. I'm also curious about the nonlinearities in the verification charts. I would assume that the nonlinearity in the original calibration chart could be due to something with the printer, but I would also assume that such nonlinearity would be corrected by the curve. So I'm curious; does the nonlinearity in the verification chart simply mean, as someone said, that the curves as generated sometimes need further tweaking, or could the nonlinearity be due to some other cause?
Katharine Thayer

If your printer is doing something weird, it should do the same weird thing to both the test chart and the negative and therefore should be compensated for.

I played with my scanner some more with a variety of settings and as far as I can tell it's just incapable of capturing the highlight end of the curve accurately. Very disturbing, frankly. It's a relatively inexpensive Canon, for the record - I mostly use it to send faxes.

The fix for me was to simply photograph the curve with a digital camera in RAW mode, and then extract that out with a linear gamma and making sure there was no clipping on either end. The results were very nice - I've now made 11 ziatypes with different images and they all look like I would expect.

I should also add that I used to write software that would test scanners on the factory floor to see if they met specifications and there was an amazing unit to unit variation. The software I wrote could spot all kinds of interesting problems - banding, other noise - and all adjustable. But instead of having a predetermined quality level and failing whatever percentage of scanners failed, they had a predetermined failure level and adjusted the parameters to fail that percentage of scanners.

I also had access to the driver code and there was quite a lot of hard coded logic to adjust things to "look good", not be accurate, partially because HP was cleaning their clocks because people preferred their warm, oversaturated images. Marketing made everybody throw out their test charts and stop trying to do things accurately and instead tweak the output according to focus group studies.

Anyway, if you can borrow a scanner from a friend, I would try scanning your chart with it, generate a curve, and see if they're pretty close. If not the scanning would seem rather suspect.

 

[mkochsch]

I played with my scanner some more with a variety of settings and as far as I can tell it's just incapable of capturing the highlight end of the curve accurately. Very disturbing, frankly. It's a relatively inexpensive Canon, for the record - I mostly use it to send faxes.

I'm not sure I understand why you're having problems. So you're saying your scanner is incapable of properly capturing/representing a reflective target which contains a black square. Are you scanning the transparent media (the ChartThrob target) or the final product (the print)? Could you give a step by step of your calibration workflow for critique?

 

[sanfrancisconudes]

I'm not sure I understand why you're having problems. So you're saying your scanner is incapable of properly capturing/representing a reflective target which contains a black square. Are you scanning the transparent media (the ChartThrob target) or the final product (the print)? Could you give a step by step of your calibration workflow for critique?

I think you can assume I'm not doing something stupid given that a) I used to do this stuff for a living and b) I got it to work by photographing it.

If you look at the histogram produced while scanning, it looks like the one from the photgraph except the tail of the highlight end is chopped off. If you adjust contrast or whatever to make sure it's well within the range you're supposed to be scanning it just leaves a gap - there's nothing recorded past a certain point.

I know our scanners had some code designed for the black text on white paper scenario - they would basically purposfully chop off some data on the white end so that you wouldn't get any speckles in the white area. Now, in our case it only ran in "text mode" not "photo mode" but who knows what some other driver's doing?

Actually that's a thought - if some driver has a "photo mode" versus a "photo and text mode" that could be one difference between them.

 

[mkochsch]

I know our scanners had some code designed for the black text on white paper scenario - they would basically purposfully chop off some data on the white end so that you wouldn't get any speckles in the white area. Now, in our case it only ran in "text mode" not "photo mode" but who knows what some other driver's doing?

If your scanner driver/software does not support turning off "auto" exposure maybe you have to look at other options. ie. Is your scanner supported by VueScan? Problem solved if it is. You can "lock" the exposure at nominal 1.0 value and turn off all histogram and white balance settings. Good on you for figuring out how to do it with a camera though. Are you metering through an external light meter or just using the camera's meter to figure out exposure?

 

[sanfrancisconudes]

If your scanner driver/software does not support turning off "auto" exposure maybe you have to look at other options. ie. Is your scanner supported by VueScan? Problem solved if it is. You can "lock" the exposure at nominal 1.0 value and turn off all histogram and white balance settings. Good on you for figuring out how to do it with a camera though. Are you metering through an external light meter or just using the camera's meter to figure out exposure?

I'm just eyeballing the histogram. I guess one thing I should point out is that when I sell prints they're too big to fit on my scanner so I photograph them anyway. I put a diffuse light source up at the ceiling pointing down, put the print on the floor, get up on a chair with a 100mm lens and lean over the print. I find doing it that way prevents any kind of geometry issues (if you use a wide angle it's hard to get everything square). So it's really no big deal to ditch the scanner and just use my regular workflow. Chances are I have a print I'm supposed to be photographing anyway and I can do them at the same time.

It does look like VueScan supports my scanner, and if I get the urge to play around some more I'll try that - thanks for the suggestion.

 

[Katharine Thayer]

Thanks...

Thanks for responses to my questions, none of which answered the questions I thought I was asking, as is often the case. So I deduce from the answers that in general you think it's not terribly important how I set or don't set the black and white points, because a bigger problem is that my curves are the wrong shape either way, which could be (a) because I'm printing the negative using the wrong set of color inks or (b) because my scanner isn't accurately capturing the highlight end of the scale.

I'll go back and re-read Clay's page on curves, and think about the color issue.

Re nonlinearity, I was referring to unequal intervals between tones in the calibration chart, jumps in the tonal scale. I wasn't referring to nonlinearities that could be attributed to unevenness of coating or light, which would be very unlikely to correspond exactly to the size and shape and location of individual squares in the calibration chart, as do the nonlinearities I'm talking about.

Thanks, everyone, for input.
Katharine Thayer

 

[sanfrancisconudes]

(b) because my scanner isn't accurately capturing the highlight end of the scale.

Well, I think my driver software is doing some strange things to the high end of the scale, and given that and my history working on scanner software and knowing how dodgy it is, I think anybody who is getting odd results ought to at least consider that they might want to try another scanner - preferably of another brand so you get different software entirely.

I think assuming some arbitrary user is able to have a reproducable result in terms of their printing technique - if they print a negative from the same file twice and coat and so forth and in the end they get basically the same picture - then it seems to me that the scanning process has both the most potential for user error, hardware errors, software not doing what you think it is...

 

[bjorke]

I gather from Kevin's comments that the main thing is to be sure that the black and white points aren't cutting off any of the printed values. When I set the black and white points manually, as described on the FAQ, I get a curve that I don't like as well.

Iy your defaults still include the entire range of the printed and scanned tones, it's all good. EpsonScan seems to deliberately clip. Using Vuscan with zero whitepoint/blackpoint values should not clip (if I understand VueScan correctly).

I couldn't find guidelines for including images in posts here, so just put them on my website:

http://www.pacifier.com/~kthayer/html/Curve.html

This URL 404's for me.

In your second post you ask about "nonlinearities" -- the ChartThrob base chart is exactly linear and should cover all tones from 0 to 100 K. What is it you are seeing?

 

[bjorke]

...it seems to me that the scanning process has both the most potential for user error, hardware errors, software not doing what you think it is...

Could be! Using a good quality camera as an alternative isn't bad (with the usual caveats about even lighting, glare, and so forth). There's also the danger that the printing, being hand-work, is not super consistent either -- but in general that's the DESIRED effect

 

[rippo]

hey bjorke, i just wanted to say thanks for charthrob! i've used it in my cyanotype and kallitype making, and just tonight made my first silver gelatin contact print, with CT in the mix. works very nicely!

one thing i noticed, with both cyano and kalli, is that no matter what exposure i give the chart, i get either an under or overexposure notification from CT. the processes can be pretty contrasty, so i guess that's the problem. mildly disturbing, but it does generate a working scan. meanwhile on the silver gelatin it didn't give me an error...i guess that's the magic of 20th century printing processes.

very useful, and you can't beat the price! thanks again.

-matt

 

[bjorke]

i get either an under or overexposure notification from CT....

Those are only suggestions -- what it does is really a bit simpleminded, it counts up the number of dark cells that are completely black and the number of light ones that are completely white, and if the counts vary by much it makes that suggestion, based on the (possibly erroneous?) notion that the best straight-line portion of any medium's characteristic curve will be near the middle -- but it's only a suggestion, and if your prints are getting a full range, then great!

 

[rippo]

ok then i won't worry about it. the curve seemed to be valid anyway, i was just wondering if there was something i was doing slightly wrong. but the results are great, so i can't complain.

 

[andy8x10]

For some strange reason I cannot get chartthrob to analyze my positive chart. When I click "analyze", an error pops up...I believe I have done everything correctly. scanned in positive, cropped it. activated chartthrob (file-scripts)...can't figure out the problem...

 

[sanfrancisconudes]

For some strange reason I cannot get chartthrob to analyze my positive chart. When I click "analyze", an error pops up...I believe I have done everything correctly. scanned in positive, cropped it. activated chartthrob (file-scripts)...can't figure out the problem...

And what does the error say?

 

[andy8x10]

it works now...no idea what I did right this time. I think I'll go back in my darkroom...this digital stuff is making my blood boil.

 

[rippo]

and are you using CS2? won't work in the older CS.

 

[bjorke]

Andy, if this error occurs again please post details here, and better yet send them to me!

Maybe you weren't really seeing an error, just a diagnostic mesage?

 

[dwross2]

it works now...no idea what I did right this time. I think I'll go back in my darkroom...this digital stuff is making my blood boil.

 

[jag2x]

ChartThrob and Color?

Hi guys,
Thanks to the author of ChartThrob, I havent as yet used it but will do so in the near future. I'm mainly going to be printing via an imagesetter.

Though I understand how ChartThrob use for B&W alternative printing can work; because the chart itself is composed of K and 0/255 Photoshop Values of B&W. Then use the negative of the chart and contact print say via all B&W Silver, Van Dyke, Platinum type process etc. Scan that in and get ChartThrob do the work for you to come up with a curve. Then go through the motions of applying the curve, reprinting and getting the end result by contact printing. Your done.
From my understanding the SOURCE Chart which is B&W will eventually match a B&W TARGET print Chart because all the methods of silver, vandyke and platinum(more or less black) gives me a B&W result.

Now I'm unsure about this and I have a question in regards with colour alternative printing. Now with a colour process such as cyanotype or gum printing (say you use a watercolour of green), how can you use the same B&W chartThrob chart? Doesnt the chart have to equate to the same colour values as your print? Such as cyan in cyanotype or the green in the gum print? If you use the B&W chart and contact print via cyanotype and scan the result in, then chartThrob produces a curve. Now my thinking is that the curve that results would be wrong? Mainly for the fact that the SOURCE chart which is B&W is not in any way related to the COLOURED cyan result, as the K and 0/255 photoshop values would be way off when it has to do its calculations to make a CURVE? Same logic goes for my green gum print.

Is my point valid or have I got myself confused here, any help on steering me on the right track would be great!
Thanks
Jag2x