Piezography: Talk me into/out of it

[Michael R]

I've been pouring through the Piezography site at the same time as reading through some basic introductory stuff - probably not a great idea. Anyhow, although one must keep in mind the Piezography material is trying to sell you something, I am intrigued by a few things. It is also alluring specifically because Cone seems to personally be very into detailed low values. This is extremely important to me. But I'm slightly off topic here.

What I really wanted to do is run my very preliminary understanding of the overall issues of linearity and calibration by everyone to see if I'm even roughly on the right track - keeping in mind I don't yet know much about the various steps in a digital workflow and am struggling with the computer lingo. I'll do this by way of analogy with "analog"/darkroom tone reproduction.

In analog, one way of studying tone reproduction is the well-known 4-quadrant diagram. Essentially what we are doing is tracking the original subject luminance values as they transition from step to step in the workflow, and ultimately map to print values. For a given scene and chosen set of materials, the transitions in the 4-quadrant diagram represent what is imposed upon us, and each is a non-linear function:

1. Flare curve
2. Film curve
3. Paper curve

In the context of a hybrid workflow where we are scanning B&W negatives, items 1 and 2 above still apply, but item 3 is replaced with two new, potentially non linear (or worse, non-linear and choppy) transitions – scanner/scanning and printer/printing.
In the analog workflow we use printing controls to produce an expressive print with the tonal relationships we envision.
In the digital/hybrid realm, “editing” in Photoshop or similar – I think – is analogous to printing controls in the analog/darkroom realm, and that we ideally want those controlled edits to be the only transition between the negative and the print. For that to be the case, we want the scanning and printing transitions to be linear. In other words, the scanning and printing (printer) transitions should not apply tonal relationship changes of their own. Since scanning comes before editing, we can live with some distortion and correct it implicitly in editing. However we are still left with the printer.

It seems to me what we are basically talking about in Piezography (besides grey scale hues, selective toning, d-max, archival concerns etc.) is modifying the printer/driver so that this final transition is linear – ie it does not introduce any curves of its own to distort the tonal relationships we finalized in editing. My understanding is that this fix is essentially what a “profile” is.
Further to this, a calibrated display is desirable so that, assuming we achieve linearity at the printing stage, what we saw in the editing software is as close as possible to the printed output.

I suppose my first question (knowing little about computers/software) is – why, after all this time, do we need to jump through hoops to get linearity from the printer? I understand the B&W linearity problem arises at least in part from printers/drivers being optimized for colour printing, but it just naively seems to me this should have been solved by now. Related to this, I’m having trouble understanding the nature of banding, and how Piezography profiles get rid of that. To be continued.

Anyhow, just some thoughts.

[Ken Lee]

I suppose my first question (knowing little about computers/software) is – why, after all this time, do we need to jump through hoops to get linearity from the printer? I understand the B&W linearity problem arises at least in part from printers/drivers being optimized for colour printing, but it just naively seems to me this should have been solved by now.

When most people see an apple or an orange in a publication, they don't notice or care what Pantone color it is. But printers and art directors do. Imagine how important precise color rendition becomes when we're printing an advertisement for women's makeup or house paint in a fancy magazine. Big money is at stake.

So linearity is an issue in color too. In the world of commercial printing, there's all kinds of instrumentation at work, at high-end prices. Just as a guitar needs constant tuning, so papers, pigments, printers are all subject to variation. Even monitor calibration software gives an option for re-running every X days, since electronics are subject to drift and pros want freshly calibrated equipment, just like concert pianists. Nothing is linear, but we strive towards it.

If we work in monochrome and tonality is important to us, we are more likely to notice linearity, since our palette is so restricted. Someone like Jon Cone, as a photographer and highly experienced printing professional, is routinely able to associate tones with numbers. We do that with the Zone System, but at his level, he deals with 256 tones. We don't have to work that way, but he suggests this approach to help us develop an awareness of our craft at the level which modern imaging tools already support.

Methods like Piezography represent an effort to bring the same degree of control as we might find in high-end printing processes, to B&W photographers. If there's a learning curve in going from 1950 to 2020, we shouldn't be too surprised

Related to this, I’m having trouble understanding the nature of banding, and how Piezography profiles get rid of that.

Profiles don't prevent banding. Working in an adequate bit-depth does that.

A good profile ensures that our printed values match the values we specify in our image file. We want our apples to look like apples, and our oranges to look like oranges.

[h2oman]

For that to be the case, we want the scanning and printing transitions to be linear. In other words, the scanning and printing (printer) transitions should not apply tonal relationship changes of their own. Since scanning comes before editing, we can live with some distortion and correct it implicitly in editing.

Here is an experiential, totally non-scientific observation of mine - to put it in context, I like fairly high-key images like this:



I've been using an Imacon Flextight scanner, and it gives me a choice of adjusting what is called something like "Shadow Depth Recovery" when I scan. (I don't remember the exact language, and it would take me 5 minutes of computer firing up to find out.) I have found that I get results that are more pleasing to me by boosting that to various extents, depending on the negative, and taking those tones back down in editing, rather than doing a straight scan and then trying to bring those tones up. I don't know how linear a straight scan is, but if it is, I am destroying that linearity!

[Ken Lee]

Here is an experiential, totally non-scientific observation of mine - to put it in context, I like fairly high-key images like this:



I've been using an Imacon Flextight scanner, and it gives me a choice of adjusting what is called something like "Shadow Depth Recovery" when I scan. (I don't remember the exact language, and it would take me 5 minutes of computer firing up to find out.) I have found that I get results that are more pleasing to me by boosting that to various extents, depending on the negative, and taking those tones back down in editing, rather than doing a straight scan and then trying to bring those tones up. I don't know how linear a straight scan is, but if it is, I am destroying that linearity!

People don't care much how we got our image from the subject to the film to the darkroom to the scanner to the editor and beyond. If we're lucky, they care about the final image.

We care of course. We'd like to do a better job and have better mastery of our materials.

Linearity is not some ultimate goal of its own: it just comes in handy when we want to print something and have what we see on the monitor, appear on the print.

Otherwise, it's like the old exercise in translation, where after a few iterations, "The spirit is willing but the flesh is weak" becomes "The ghost agrees but the meat is rotten".

[Michael R]

Thanks for the additional feedback, guys. Digital is brand new to me so this whole thing is going to be quite interesting - I hope.

In reading about the Piezography digital negative product/solution I became overly excited imagining I could create masks perhaps good enough to sandwich with negatives in the enlarger (analogous to silver masking but potentially much more powerful). Then I came back down to earth.

Lovely image, h2oman!

[Ken Lee]

Speaking of linearity, I dug through some of my old test prints and found this rendition of a 50-step bulls-eye. On my monitor it looked fairly smooth, but when I printed it on my Epson 2400 using Epson paper and the corresponding Epson profile, the results were far from... linear.

Click on this smaller image to view a larger version. This photo of the print was made with my iPhone but even so, you can see that the "ramp" is far from continuous. At one point in the lower values, there's a gray strip where the values suddenly get lighter rather than darker. After that the low values drop off precipitously and are clipped to pure black. Oops.

... and then there's the issue of the non-uniform magenta color cast, even though I was printing a B&W image.

I have used the same Epson profile to make color prints of this image and they're fine for snapshots, but what works for "millions of colors" isn't always good enough for monochrome-only.

[Michael R]

Very interesting example. That band of greys going the wrong way really jumps out.

Even on the bulls-eye test pattern, it occurs to me aside from relatively glaring deviations from linearity, I might have trouble seeing less obvious deviations without a linear reference version to compare with. For example, even if all 50 steps were discernible and progressive, it still might not be linear. A curve can still look smooth. Or maybe I’m overstating the case. Perhaps I could use a densitometer to measure the 50 reflection densities, but even that process has its issues.

[Ken Lee]

Very interesting example. That band of greys going the wrong way really jumps out.

Even on the bulls-eye test pattern, it occurs to me aside from relatively glaring deviations from linearity, I might have trouble seeing less obvious deviations without a linear reference version to compare with. For example, even if all 50 steps were discernible and progressive, it still might not be linear. A curve can still look smooth. Or maybe I’m overstating the case. Perhaps I could use a densitometer to measure the 50 reflection densities, but even that process has its issues.

You're right. A reference is needed. They measure these things with Spectrophotometers, as Sandy mentioned. But I wonder if a spectrophotometer is merely a portable reflection densitometer.

And you're not overstating the case. As you suggest, the Epson profile I used was good enough to make a 10-step wedge that looked progressive, but the result was nowhere near linear. That became evident as we included more steps, and it was evident to me as I tried in vain to make decent prints of images with gradients, like blurry flowers, clouds and skin tones.

It’s quite ironic that with digital methods we struggle to make things appear analog.

[Tin Can]

Eyes vary, some see way more

1% of women may see a lot more

All the Colors We Cannot See

[Michael R]

You're right. A reference is needed. They measure these things with Spectrophotometers, as Sandy mentioned. But I wonder if a spectrophotometer is merely a portable reflection densitometer.

And you're not overstating the case. As you suggest, the Epson profile I used was good enough to make a 10-step wedge that looked progressive, but the result was nowhere near linear. That became evident as we included more steps, and it was evident to me as I tried in vain to make decent prints of images with gradients, like blurry flowers, clouds and skin tones.

It’s quite ironic that with digital methods we struggle to make things appear analog.

In this context a spectrophotometer is essentially a colour reflection densitometer.