Is my Bronica SQ-A an 81-MegaPixel camera??

[Daniel Balfour]

I did some basic math. Considering math isn't my strong point, I figured I'd ask the good people of HybridPhoto if I'm doing this correctly.

My brand new Nikon 9000 ED scans at 4000 dpi. If I'm correct, that means 4000 pixels down & 4000 pixels sideways per inch or 16-Million pixels (16-Megapixels). My Bronica SQ-A's 6x6 negative (2.25" square) which comes to 9000 pixels down & 9000 pixels sideways = 81 Mega Pixels??

That right?

 

[lenny]

I'd love to say yes, but unfortunately the answer is no.... That's one of the differences between drum scanners and flatbeds. Flatbeds (almost all) scan a single "raster line" all at once. When they say they are 4000 dpi, that's 4000 for the entire line, going across. You can tell this by going into PhotoShop's size dialog and looking at the count of pixels at the top. Just take one of your scans and report back what pixels you ended up with....

Drum scanners scan a single sample at a time, and when they say they are 4,000 dpi, that is 4,000 for every inch they scan. A 2 1/4 scanned at 4,000 would result in a 9,000 ppi scan. Of course, at EigerStudios, we scan med format film at 8,000 ppi, which results in 18,800 pixels, or a 324 megapixel image. Just over 2 Gigs.

That's why I haven't gone to digital yet.... when they get me a nice, 200 or 300 megapixel camera I can put in the palm of my hand, I'll set aside the view camera...

Lenny
EigerStudios

 

[keithwms]

So, there are megapixels (mp) and then there are meaningful megapixels (mmp).

The difference between the two? Meaningful differences in detail and tone between adjacent pixels. If two adjacent pixels hold mostly the same info then I would not consider them to be meaningfully distinct. So I am imposing a resolution argument, you see.

Based on the highest possible level of scanning currently available to me, drum scanning, I propose the following empirical "meaningful megapixel equivalents":

35mm with top notch prime lens and velvia 100, exposed to the best of my abilities = ~10 mp

medium format 6x6 slide, top-notch mamiya rangefinder glass, velvia, again exposed optimally = ~30mp

5x7 slide, top-notch Schneider glass, good exposure = ~ 100+ mp

And just for the record, the pixels I speak of are ~12 bit per channel pixels, okay?

How do I get these numbers? Well I first came to these conclusions by experimental observation when I first got into MF and LF a few years ago. I played with some flatbeds and a dedicated Minolta scanner. I got things scanned and rescanned and then decided whether there is more detail to be had, and so forth. I had to do a lot of learning. At the end of the day I learned that there is scanning and then there is drum $canning, you see!

But my point is that there is an overriding optical issue here. A top notch MF lens is going to give you ~100 lp/mm across your medium format frame. In my case I use (and adore) the mamiya 6 system, and the 50mm lens for that is quite legendary. That thing really can give ~100 lp/mm right across the frame at an optimal aperture like f/8. So if you do the math.... 100 lp/mm gives you

60mmx60mmx(100 lp/mm)^2= ~36 mp or so.

[yeah there is the issue of line pairs versus pixels, so people can argue over factors of two, but there you have it and it agrees quite well with my simple allegations about optically meaningful megapixels and factors of two are easily attributable to the skill of the photgrapher and his/her ability to execute a good scan]

For 35mm let's see how close I am, the frame is 24x36x(100)^2= 8.6 mp, but lo, some 35mm glass can do better than 100 lp/mm, so you can take that into account if you wish. In any case you see that ~10 mp is in the ballpark. Maybe 16 maybe 8, it all depends. Again I am talking about meaningful megapixels!

What about large format... well, indeed you have a massive slice of film but the lenses are not going to give you 100 lp/mm across it.. more like 80 at best, if you have that kind of lens budget.

Now, obviously there are all kinds of factors that could change my "mmp" estimates by a factor of two. The issue of line pairs versus pixels, for example. The fact that we're usually comparing to interpolated pixels; some don't realize that foveon pixels are different from Bayer pixels, that is another issue. Other major issues: how good the scanner is, not just in terms of resolution but also Dmax/min, and how good the film holder is. And more important than anything else... the competence of the photographer who took the shot and who does the scan.

 

[lenny]

Keith,
Your estimates are very low. It doesn't match up to my experience at all - and I work with this stuff all day long, both scanned files and files from high end digital cameras.. There are plenty of factors. You are using a lightjet, which has less resolution than an inkjet. I don't know which drum scanner you are using, but there are only two that have resolutions in the 7500-8000 range. I looked at your site and some of your images on done with infrared, that's yet another factor...

I think you are missing the point with the large format argument. Even tho' the lenses aren't quite as sharp, you could have an inch of film to describe something that in a smaller format you describe in an 1/8 of an inch. There is a lot more info in there that can be pulled out....

If I thought for a moment I could get the same amount from a 30 meg digital sensor than I can from a 4x5 and scan I would just go out and buy one.

The math isn't really the final arbiter. There are all sorts of numbers floating around out there, some of which are based on false premises. One has to look. I have a number of prints on my wall, from 20 to 40 inches, none of which could have been done with a digital camera. I know because I have tried - and I sold the thing and went back to film. When they get it right, I'll be happy to move over. But they haven't yet - not for the larger print size.

If you have done a drum scan of one of your 8x10's and it isn't better than a digital camera, then something is very wrong.

Lenny
EigerStudios

 

[toyotadesigner]

Scanning a 6x9 @ 4.000 ppi on my Coolscan 9000 @ 16 bit delivers me a 12.900 x 8.600 pixel image. Doing the math the result is 110.94 Megapixel image.

So with a 6x6 (2.5 x 2.5") you'll end up with a bit less than 81 Megapixels, because the image is something around 56mm x 56mm. But your math is correct.

Lenny, Iserious didn't say 'flatbed' but **dedicated** film scanner - the Nikon 9000 is another class than a flatbed.

I've read something about **meaningful** pixels - holy shit, what is this? A new definition to bend the math?

The Nikon 9000 is a fantastic scanner, and the OP didn't ask for a drum scanner resolution - he just asked for the math considering the Nikon 9000 and not for a pixel rat race the the digital guys are always doing. If all of us would have the space to roll in a drum scanner, we would do so, because they are kind of extremely cheap these days - less than a new Imacon.

Iserious, you are correct with your math. Enjoy your Coolscan 9000, and if you need to have an image printed to 6 x 4 meters, give the neg or slide to a drum scanning service. Other than that - use your Coolscan: I'm using our Coolscan 9000 for 6x9 slides to have them printed up to a size of 180 cm x 120 cm or 6 feet x 4 feet. However, most of our images are printed at 90 x 60 cm, that is 3 feet x 2 feet. Pretty hughe for most customers if a frame will be added

 

[keithwms]

Nobody's bending math, my profession relies very strongly on math, thank you Photography is just my hobby, physics is my profession.

Looking at the scans on my site, you will note that I stated that these were moderate res drum scans. Incidentally, the 5x7 scans do probably blow away digital files from any dslr and most pricey digibacks, and they were shot with a 1903 wooden field camera that cost me precisely $250, so don't worry about which camp I'm in I am a certified velvian.

But seriously... just compute what resolution is possible from your lens at a particular aperture across a particular film. It's really that "simple" ...or if you prefer, it is really that complicated.

You can indeed scan a 35mm frame to 100 mp or whatever and it's still got no more detail than is allowed by the MTF of the lens and the film. And the scanner will always give you somewhat less resolution and tone than that, even a drum scanner operated by a genius.

Bottom line... if you really think that they can get a hundred mp out of a medium format slide, and want to store those enormous files, hey go for it, it's your hard drive space not mine Seriously, do what you think is right. I'm always in favour of people proving things to themselves. Don't trust me. Don't trust what anybody says. Think for yourself.

But.... until somebody explains to me why lens and film MTF is not relevant here, I will stick to my story: there are pixels and then there are interpolated pixels... and there are meaningful pixels

 

[lenny]

Lenny, Iserious didn't say 'flatbed' but **dedicated** film scanner - the Nikon 9000 is another class than a flatbed.
D

This is a nomenclature thing... I see flatbed as a CCD based scanner, that scans raster lines. I would definitely rather have the 9000 than all but a few of the best flatbeds - looking flatbed. By this definition, an Imacon is also a flatbed, altho standing up.

Drum scanners use PMT... which is a different technology.

Lenny

 

[lenny]

Nobody's bending math, my profession relies very strongly on math, thank you Photography is just my hobby, physics is my profession.

Looking at the scans on my site, you will note that I stated that these were moderate res drum scans. Incidentally, the 5x7 scans do probably blow away digital files from any dslr and most pricey digibacks, and they were shot with a 1903 wooden field camera that cost me precisely $250, so don't worry about which camp I'm in I am a certified velvian.

But seriously... just compute what resolution is possible from your lens at a particular aperture across a particular film. It's really that "simple" ...or if you prefer, it is really that complicated.

You can indeed scan a 35mm frame to 100 mp or whatever and it's still got no more detail than is allowed by the MTF of the lens and the film. And the scanner will always give you somewhat less resolution and tone than that, even a drum scanner operated by a genius.
But.... until somebody explains to me why lens and film MTF is not relevant here, I will stick to my story: there are pixels and then there are interpolated pixels... and there are meaningful pixels

People can make lousy - or moderate - drum scans easily. You can not use these to compare real results. I have a drum scan made on a Tango from an 8x10 that makes it look like it was done on a consumer level flatbed. Operator was obviously off that day..

The number that represents lens resolution and aperture alone DO NOT tell the story. There is way more detail in there than there should be. That's why operators actually make a difference. Further is a Goerz Dagor a better or worse lens than a Rodenstock lens at the same size. You can't answer that, and no one can. One is sharper, the other has more "character". It's not about resolution, or even color accuracy. How about the ability to render tonality? Atmosphere? What is the effect of a coating that is on there by oxidation? Of course, it's impossible to know.

By rights, I should be able to scan a 4x5 at 4000 dpi and scan an 8x10 at 2000 dpi and the file should be identical. Same size, same number of pixels, the scanner can do either resolution easily. Problem is, it doesn't work. The 8x10 is still better. I wish it weren't so - the 4x5 is a lot light to carry around.

Your contention that there are "meaningful pixels" at this low level simply doesn't hold. I know it would make sense to a physicist, but it doesn't make sense in real world application. I don't think the folks on the "hybrid" list here are just a bunch of fools. They actually know something. There is a reason to do things with a scanner, and museum quality print after museum quality print bears this out.

Lenny
EigerStudios

 

[keithwms]

C'mon Lenny the original post basically asked if a Bronica SQ-A is an 81 mp camera, and I simply made the point that there are pixels and then there are pixels. You can take any image and interpolate it to any size you want and that does absolutely nothing to the actual information contained in the starting image. So I am drawing a distinction between interpolated pixels and what I am calling meaningful pixels. Call them noninterpolated pixels if you prefer.

Also I suspect that you glossed over most of what I wrote; I explained that there are various ways to get factor of two differences in my simplistic estimates. I also said that there are 35mm lenses that support resolution beyond 100 lp/mm. But I correctly asserted that 100 is pretty damn good for MF and that 80 is pretty damn good for LF, across the frame. I also said 5x7 film gives 100 mp+.

Now, your assertion that there is more information there needs to be clarified. You surely do not mean more spatial information than can be suported by the MTF of the lens and the film. What you mean is more tonal information, and on that I concur. There are 10 mp 24 bit files, and then there are 10 mp 42 bit files and so forth. There is a difference. We all know that. That is obvious.

Seriously, I think you had an allergic reaction to what I wrote. After considering it more carefully, you will see that I offered a very simple and sound argument for ballpark estimates of what realistic mp numbers are.

As for the character of a lens, I won't touch that with mathematics. Of course we like the characters of different kinds of lenses and there are wonderful, mathematically indefinable reasons why some lenses appeal more for certain applications. I am fully aware of the unending threads attempting to define bokeh. But that is also subterfuge, it has nothing to do with the orginal post here, which (as I understand it) was just trying to get to grips with what we really mean by pixel count in the context of scanned film vs. digital.

And by the way I do not think that I insulted anyone's intelligence at any stage.... if I did, then sorry. But I assert that my thinking is laid out right there in the open for anyone to consider / refute etc. There is no fancy math, no hidden agenda, just simple order-of-magnitude logic.

Oh and thanks for commenting on my work, that was nice of you.

 

[keithwms]

I don't know which drum scanner you are using, but there are only two that have resolutions in the 7500-8000 range.

The drum scanner the lab uses is an Aztek. People tell me it's pretty good.

I looked at your site and some of your images on done with infrared, that's yet another factor...

The infrared crops on my site were done with a flatbed, which is plainly stated. It's also irrelevant because if you look a few posts above, you will see that I am referring to velvia 100.

 

[toyotadesigner]

Keith, I'm well aware of different techniques to scan images. I didn't mean interpolation - I'd never use it. For us the Coolscan 9000 is a perfect machine when considering the return on investment. Sure enough I would like to scan all the slides with a drum scanner, but it makes much more sense to send only the most important slides for hughe prints to a service and have them scanned by an excellent operator.

MTF and all the nine yards: I know what it means but I never really cared about it because I just use the best lenses I can get (Zeiss, Fuji, Nikon, Schneider) to avoid any hassle. All of them have at least a very good MTF, and I think it won't make such a big difference in the result if I end up with 5 line pairs less than another lens would offer. In addition I know the limits of 35mm and MF, but our concept doesn't allow for larger slides or negs - the costs would be prohibitive, and there won't be a lab in our area to process these formats.

Fact is: A Bronica is a lot better in resolution and tonal range than any digital camera currently on the market, so to answer the question of the original poster I'd say: yes, it is more or less a 81 MP camera. Just use it on a tripod with a good metering and a Velvia 50 for landscapes or Provia 100F for portraits and you'll blow away any kinky Canon D1-dust-in-the-eyes image.

 

[sanking]

But seriously... just compute what resolution is possible from your lens at a particular aperture across a particular film. It's really that "simple" ...or if you prefer, it is really that complicated.

I agree with you on this point. If you want to determine the best resolution for scanning, assuming scanning for archival purposes (which means you want the best possible scan), just calculate the best possible resolution possible from your camera/lens system at the aperture in question. If that resolution is 100 lppm you really need to scan with an *effective* resolution of slightly more than 5000 ppi. Mind you , this is not the stated optical resolution of the scanner but the real, *effective* resolution. With a drum scanner *effective* and optical resolution are virtually identical. With consumer flatbeds the true *efffective* resolution will be less than 50% of the stated optical. The LS-9000, a dedicated flatbed, has true resolution of 65-70% of stated optical. Very high end flatbeds have true resolution of from 75-95% of stated optical.

Sandy King

 

[lenny]

Also I suspect that you glossed over most of what I wrote;
Seriously, I think you had an allergic reaction to what I wrote. After considering it more carefully, you will see that I offered a very simple and sound argument for ballpark estimates of what realistic mp numbers are.

Keith,

I did have an allergic reaction to your post, altho' I did read it carefully and I understand every bit of it.

I get sick and tired of this general argument. While some of what you say is applicable, some of it I find is not. There are so many folks who just say - hey, digital cameras are just as good as anything you can do with a scanner. There is only so much resolution, etc. This is actually true for a certain level of photographer, or even a certain size of print. (I'm not trying to put words in your mouth regarding these comments that I hear from others - all too often. I will respond generally.)

It is clearly not true for people who are at the next level of photography than that. If you are in Seattle, then zip on over to Tyler's and take a look what's possible... with a good piece of film. Or come on over when you are down in the SF area and I'll show you a few things you can't do with a less expensive scanner - or any digital camera there is, currently.

I happen to be an experienced printer. Someone who is steeped in the different styles of printing over the History of Photography and who did what with what tools. I have studied this stuff my whole life, I know the formulas most of the great photographers used to develop their film. We are in a different time - now everyone who has a camera wants to call themselves a photographer, and wants to imagine they print well. The overall level of photography, when compared with Walker Evans, or any of the greats, is atrocious. People are after "the shot" which is a ridiculous concept and everything ends up centered in the frame. In addition, what people consider great printing is getting watered down to the expedient, Walmart level photo. Most of the photographers I talk to don't know any other photographer other than Ansel Adams, who printed in one of many different styles. As much as I respect his style, and all he did for all of us, I don't like that particular style. I much prefer a softer style... Of course that's all personal taste.

The real thing I am after is professional quality results, or more specifically, museum quality results. I work very hard at this, test all the time, mix my own b&w inks to my own targets, etc. To have someone come along and be dismissive, suggesting that some cheap scanner, or some digital camera would do just as well - well it is insulting. Truth, is they are right for a certain level. Just not this one.

I used to print in platinum, which I still love and appreciate. One day a client asked me how long it took me to make one of my prints. He was thinking about the process of platinum printing and expecting an answer like an hour, etc. It triggered a different answer in my brain and I responded "three years". I had printed it a hundred times over a three year period before I really understood the image. There's a different level here, that's not about resolution, lines per mm, or any of that. There is more in play.

Lenny
EigerStudios

 

[keithwms]

Thanks for the clarification, Sandy!

I get sick and tired of this general argument.

Yup, me too!

Let me just point out my response in the other post on a similar theme. Like I said there, "I don't think that you really want to think about this in terms of pixels." If nothing else, this meandering thread shows that while pixels are the convenient language of dSLR capture, that language is quite inconvenient for analogue photography.... even when a piece of film is scanned.

Next time I am in lovely Seatte or San Fran, I will indeed stop by. Likewise when you are in DC or northern VA...

 

[Daniel Balfour]

Just take one of your scans and report back what pixels you ended up with....

In the Nikon scanning utility, the scan's pixel length/height is roughly what I guessed it at, about 9000 pixels each way (8874x8922).

Drum scanners scan a single sample at a time, and when they say they are 4,000 dpi, that is 4,000 for every inch they scan. A 2 1/4 scanned at 4,000 would result in a 9,000 ppi scan.

So far, I don't see any (size) difference. I should point out, I'm hardly an expert operator, doing things mostly by intuition. I'm simply stating my observations.

 

[sanking]

In the Nikon scanning utility, the scan's pixel length/height is roughly what I guessed it at, about 9000 pixels each way (8874x8922).



So far, I don't see any (size) difference. I should point out, I'm hardly an expert operator, doing things mostly by intuition. I'm simply stating my observations.

I think the total pixel count is not the relevant issue with the LS-9000. The more important consideration is that you will only get about 65-70% of the stated optical resolution as true and effective resolution. In that sense, yes, there are megapixels and there are other megapixels. The fact that you can record 9000 pixels does not mean that you have 9000 pixels of resolution.

I am not kicking the LS-9000. It is a very good scanner, much better than anything in its price range new. But you will only get about 3000 pppi of real effective resolution with it, not 4000 ppi. Lenny is definitely correct in that a drum scanner at 6000 ppi would give much better results. About 50% more real resolution in fact.

A high end flatbed like the Screen Cezanne, EverSmart Supreme, or IQSmart3 would also give much better results than the LS-9000, or any Imacon for that matter. But the cost difference is in the range of $2K for the LS-9000 to $20-25K for a scanner than is clearly superior.

We are of course talking new prices. For many reasons there are many used drum and high end flatbeds on the used marked today, at a fraction of their original selling price.

Sandy King

 

[Daniel Balfour]

With much respect to those posting to this thread who clearly dwarf my cursory grasp of the factors, if I can even call it that; I think we've gotten sidetracked. The purpose of my post was to compare the *useful* resolution ("useful" in function of "output") of a capture made using a professional D-SLR (Canon 5D & 85mm/1.8) to one made on a Bronica SQ-A using any one of their PS-series (pro) lenses. This, in order to gain perspective on the *practical and applied* elements in play as well as much-needed practical direction.

To simplify things (keeping with "apples to apples") lets consider the following parameters:

I am an *amateur* photographer with *average* experience and technical skill, trying to decide on an image capture method. I shoot exclusively fine-art portraiture in black & white for traditional output to (digital negatives) silver-gelatin, alt-processes (ambrotypes/tintypes & liquid light on canvas & wall murals) and some, albeit very minor, inkjet prints. Output size ranges from 8x10 to 16x20 and my printer of choice (or availability) is the Epson 3800. I can scan my negs on the Nikon 9000 ED (some of you may have read my post bragging about the deal I got on this scanner last week!).

I have no natural talent for photography with at best, average aptitude. My pursuit of a hybrid workflow stems from a love of images while recognizing my own inability to achieve the desired results using an all-traditional route. I've shot hundreds of rolls over the last decade with little to show for it (at least by my standards - I don't believe I'm being modest), having had a heck of a time testing film-to-paper, understanding scene contrast, effective printing (or lack thereof) and the like. I'm just not a quantitative thinker by nature, or at the very least not in the photo-technical sense, but am I learning.

In approaching the question of which medium to use, I'm concerned with image sharpness, tonal range, ease of use and of course, cost. Cost being a one-time bullet, isn't really the determining factor (I am of course referring to the acquisition of a DSLR and NOT a drum scanner!).

Next, ease of use. A few years back I toyed with a Nikon D1x. I took my best images ever with that camera, don't ask me why. The ability to see the image as it was captured coupled with elimination of film development seemed to reduce some of the variables. Add an Olympus P400 dye-sub printer to the equation and you get vibrant colors (it sucked for b/w prints) smooth, flowing tones and a sharpness not to be had from traditional enlargement (in my experience). A digital camera is certainly easier to use then a traditional film camera in this regard. However, in film's defense, I will say that not seeing you image and having to develop film aren't really obstacles. After all, one can hone capturing skills and film development is just a matter of testing (for the desired output medium). Still, having to scan your films or print contact sheets is a lousy substitute for Aperture 1.5 or Adobe Lightroom. I'll cut it short here before I go on yet another ramble!

Finally, sharpness & tonal range (in function of output medium & size) and on that note I'll return the discussion to the floor, having refocused the thread and seeing as I have little to put forth. The parameters have been set forth, objectivity should be assured. One final request - please try to keep information pertinent to the process & equipment outlined in this post. On the subject of "useful resolution", statements made about film should be presented in comparison to digital captures possible on the 5D. As always, in practical function of intended output as I have outlined in this post.

One final note - as descriptions only go so far, I've posted 2 reference prints. The first is an straight scan made with the LS9000, the second is an edited version printed as a positive on OHP white film (very high tonal fidelity). At my current level of proficiency, the edited version is something I would consider an "acceptable" print.

Here's the EDITED IMAGE

Here's the (low-res)ORIGINAL SCAN

This is one thread I'll be keeping an eye on.

 

[Daniel Balfour]

I am not kicking the LS-9000. It is a very good scanner, much better than anything in its price range new. But you will only get about 3000 pppi of real effective resolution with it, not 4000 ppi.

Sandy, what settings would you recommend I use with the LS 9000 when scanning b/w negatives so as to maximize EFFECTIVE resolution without useless file bulk?

I don't know what bit-depth means. So far I've been using it at 8x-superfine, 8-bits, 4000dpi. Is this correct? What recommendations would you suggest I follow?

 

[sanking]

Sandy, what settings would you recommend I use with the LS 9000 when scanning b/w negatives so as to maximize EFFECTIVE resolution without useless file bulk?

I don't know what bit-depth means. So far I've been using it at 8x-superfine, 8-bits, 4000dpi. Is this correct? What recommendations would you suggest I follow?

I would recommend that you do all of your scans in high bit. That would be 16-bit for grayscale, or 48-bit for RGB. You can set the resolution in ppi for the size print you want to make, although I think it makes a lot of sense to scan to archive, to use Lenny's term. In other words, just make the best scan you can do, and save it, so you never have to do it again.

This is what i do even with 5X7 negatives. I scan at the highest possible resolution in ppi, and save the archive on a DVD.

Sandy King

 

[keithwms]

Bit depth is directly related to tonality.

To compute the number of tones you raise 2 to the power of the bit number.

Example: imagine a simple pixel with only two possible tones: white or black. The bit depth is 1. That's because 2^1=2.

Another example: 8 bit. Then you have

2^8=2x2x2x2x2x2x2x2=256 possible tones. Better!

The higher the bit depth, the more continuous will be a tone gradient. Imagine you have a gradient going from white on one side to black on the other. If you rendered that gradient with a low bit depth then there would be few intermediate shades and the rendering would look posterized. If you render it with high bit depth then it looks smooth and continuous...and natural.

An important thing to realize is that the pixel count and bit depth are orthogonal terms. It's possible to have a high number of megapixels but poor tonality (i.e. poor tonal range). But obviously you want optimal resolution and optimal tonality. Your film has a continuous tone scale so you want to scan to a high bit depth to try to pull out as much as you can. The downside: you have to store much bigger files!

 

[Daniel Balfour]

Bit depth is directly related to tonality.

To compute the number of tones you raise 2 to the power of the bit number.

Example: imagine a simple pixel with only two possible tones: white or black. The bit depth is 1. That's because 2^1=2.

Another example: 8 bit. Then you have

2^8=2x2x2x2x2x2x2x2=256 possible tones. Better!

Wow.. I think I almost understood that! But if a given pixel is "light-grey" then why do "possible" tones make a difference? After all, doesn't that pixel have just 1 tone to it?

- As an aside, I made a gradient in photoshop in a 16-bit document, then reduced it to 8-bit and observed no change. Word on this?

 

[Daniel Balfour]

On a separate note, I took to heart Sandy's earlier statements about the "effective resolution" of a Nikon LS 9000 being roughly 3000 ppi. That having been said, the "effective resolution" of the scan would then seemingly be around 9mp (3000 x 3000) per inch.

Given that most of my images only use about 1/2 the frame for the actual subject matter, that would mean I'm using roughly 1.25 x 1.25" of my total frame-space. At 3000 dpi (height-wise & width-wise) that translates to 1.25 x 3000 = 3750 x 3750 = ~ 14mp of useable resolution.

If so, then "apples to apples", a Canon 5D with an "Effective Resolution" of 12.7 million pixels (Canon's specs - actual "recorded resolution" of almost 13.8 mp) would seem to have comparable resolving power. Considering the PITA of film/processing/scanning and factoring the generational gap, the choice is getting more and more clear.

How's this looking?

 

[keithwms]

On screen you probably won't be able to tell an 8 bit gradient from 16. The reason why it still makes sense to work on the file (do curves etc.) in 16 bit is that that you'll introduce fewer artifacts into the tone scale.

Try this experiment. Open a image and work on it in 8 bit mode. First open up the histogram. Then do something to the image involving the tonality. For example, use curves and do some transformation, or do a levels adjustment. Then look at the histo again, you will probably see artifacts that look like discontinuities in the histogram.

If you do the same edit in 16-bit mode you will not see as many of those artifacts in the histogram.

So why does it matter what the histogram looks like if you can't really see much difference between an 8 and a 16 bit image on your screen or in print? Well, each time you do some tonal manipulation you can have artifacts e.g. banding and posterization. So the idea is to try to carry as much information through the editing process as possible. The final output will truncate the tonality so you want to carry as much info as you can through to the point of print to avoid artifacts.

 

[keithwms]

But if a given pixel is "light-grey" then why do "possible" tones make a difference? After all, doesn't that pixel have just 1 tone to it?

But there is grey and then there is grey! With my 1 bit example, there are zero greys. There is only white or black, so no grey. But the 8 bit example gives you a true white, a true black, and then a lot of greys in between. In other words possible tones = different tone values that the pixel can take on.

Let me suggest googling posterization, that will show you visually what the consequences of low bit depth are.

The hangup that a lot of people have with this is thinking, wait, a photosensor is just a light detector and that detector reports how much light arrives and that's it, so the grey you record is the grey that was there in the scene or on the film you're scanning. Nope, not quite! What happens is that the light hits the detector, that signal is amplified, and that signal is digitized. There is an analogue to digital conversion (ADC) step. It's a necessary evil of digital amplifier electronics Anyway, at the point of detection, the tone scale is continous, but after the ADC, you have a truncated tone scale.... As a consequence, the continuous tone scale is made discontinuous.

Bottom line: the higher the bit dpeth, the more continous the tone scale.

I am simplifying things a hair, but that's about it.

 

[Daniel Balfour]

Bottom line: the higher the bit dpeth, the more continous the tone scale.

I am simplifying things a hair, but that's about it.

Keith,

Googled "posterization" (banding) and I think I understand the concept. I can see how this would apply to color images, but do you still feel that 16-bit depth is necessary for continuous tones using b/w film?

Thanks for all the input here btw...

- Daniel