The Maths of a Fine B&W Print

[Roger Hicks]

Ah; psychophysics -- the art of the opinion poll. The first photographic application of psychophysics of which I am aware was the original First Excellent Print research at Kodak in about 1940.

Basically, it consists of asking 'What looks best?' and trying to relate this to measurable criteria: in the Kodak case quoted, of course, exposure.

PE has already described a number of the psychophysical tests done at Kodak, freely admitting that MTF was involved in some, and pointing out that there are LOTS more criteria.

Focusing solely on MTF is about has helpful as focusing solely on grain -- 'fine grain is better' -- or of course that fine old criterion of the amateur 'tester', the resolution in lp/mm -- of which MTF is a jumped-up version.

Cheers,

R.

 

[Jed Freudenthal]

Focusing solely on MTF is about has helpful as focusing solely on grain -- 'fine grain is better' -- or of course that fine old criterion of the amateur 'tester', the resolution in lp/mm -- of which MTF is a jumped-up version.

Cheers,

R.

Roger, I wrote before that MTF is a tool to solve certain things, but it will not solve everything. MTF can be used to select the proper lens and diafragm setting e.g. That is why the lens manufacturers give that kind of information.
But the photographer is the one setting the contrast and the tone in the print. The parameters contrast and tone can never be quantified, and nobody will try to quantify that. That is per definition impossible. The photographer determines what grain size is acceptable and not..
The photographer has certain tools to control certain parameters in the photographic process. MTF is one of them. Other photographers use a densitometer, and think it is a useful tool.
The reason, I introduced the MTF in this thread, is because somebody asked if there is a mathematical way to describe the difference in a analog and digital print.

Jed

 

[Les McLean]

Can we get back to talking about Images?

 

[Bruce Osgood]

Can we get back to talking about Images?

LOL, I knew you couldn't keep still on this.

 

[gainer]

I don't recall seeing in this thread that anyone had stated the following facts.

The MTF is not a conformal (I think that is the proper term) mapping. That is, you cannot recover from the MTF the data that went into it. The MTF, like the more familiar frequency and phase transfer functions in the audio world, is taken from the statistics of the photo when you want to know the spacial frequency content. The MTF is not necessarily uniquely associated with only one photograph. You can no more tell visually which photograph belongs to which MTF than you can tell whether an audio frequency content graph belongs to a Beethoven symphony or to filtered white noise.
The process of determining an MTF is usually done by sampling the object periodically. Aliasing can occur if the sampling rate is not at least double the highest frequency of the data. The same precaution must be taken with spacial frequencies as with temporal frequencies. Now you may see the conundrum involved in using MTF to compare analog and digital photographs. If you scan the analog photo at high enough rate to resolve all its spacial frequencies, then you can build a digital camera with the same spacial resolution. Your raw data from the digital photo of a scene may in fact show a better MTF than the one obtained from the scan of the analog photo simply because it does not need to got through the extra printing step, providing you can establish what "better" means.

 

[Alan Johnson]

It appears the digital camera cannot be built yet.
For a 50%MTF the EOS 5D has 40lpm,the D60 39 lpm and 100 TMX from which to make the print has 130lpm,from the sources I mentioned previously.
Of course ,there is a lot of small print with these comparisons.

 

[Ed Sukach]

.... you cannot recover from the MTF the data that went into it. The MTF, like the more familiar frequency and phase transfer functions in the audio world, is taken from the statistics of the photo when you want to know the spacial frequency content. The MTF is not necessarily uniquely associated with only one photograph. You can no more tell visually which photograph belongs to which MTF than you can tell whether an audio frequency content graph belongs to a Beethoven symphony or to filtered white noise.

Ah, success! That is exactly my point. The "MTF" of a lens IS of INTEREST, but only in a wildly technical situation will it really be anything like of critical importance.

All three of the MTFs I posted were of Zeiss lenses for the Hasseblad: In order, the first is of the Distagon CF f/3.5 60mm; the second, of the Sonnar CF f/4 150mm; and the third, the Planar CF f/2.8 80mm - the lens used. All of these lenses are, IMHO, Superb. The MTF is far down the list of attributes I consider in making the choice of which to use in real time.

 

[Photo Engineer]

Well, I'll add to this that that is what is important here Ed and I'm glad you and Patrick made that point. MTF is of interest, but only in a technical situation, not in normal artistic applications.

All of our tests at Kodak were done 2 ways. One way was 'artificial' in the sense that all the pictures were taken with the same lens, but then the prints themselves were 'doctored' to give different MTF values, contrast, toe, shoulder, grain, etc.... The other was done by taking photos using the same subject, but different lenses.

Then, each attribute was assigned a JND (Just Noticable Difference) as judged by the people who viewed each set of prints.

So, from my practical experiencs over the years, for our aerial photos, the MTF of the film and lens were important when taking shots from up to 100,000 ft altitude, and they were also important for what we termed "Metric Photography" at Cape Canaveral, but for ordinary professional use it was less critical and for amateur use, contrast and saturation (in color) were more important. MTF is just a small tip of the iceberg when considering the whole subject of photographic system design.

PE

 

[gainer]

It appears the digital camera cannot be built yet.
For a 50%MTF the EOS 5D has 40lpm,the D60 39 lpm and 100 TMX from which to make the print has 130lpm,from the sources I mentioned previously.
Of course ,there is a lot of small print with these comparisons.

50% MTF means little. The MTF is a distributed function and should be carried at least over the complete range of human visual acuity. An analogy in sound reproduction would be an MTF of 50% at any one frequency tells nothing about the response at any other frequency. It could be the response of an amplifier sharply tuned to that frequency.

If we cannot design a scanner with the necessary resolution to obtain the MTF curve of an analog photograph, then we cannot obtain that MTF. If we can devise such a scanner, we can make a digital camera of it that will have an MTF at least as good. All I can say is that using a digital camera (in effect) to prove the superiority of an analog camera by photographing an analog photo with the digital camera is not the logical way to go about it. That is exactly what you would be doing unless you know of some way other than periodic sampling to get the data you need for calculating the MTF.

Sooner or later there will be a digital camera with the spatial resolution to excel the EOS 5 D. We will still say its prints are not as satisfying to the viewer. As it stands, by the time you take the 50 lpm of the analog lens, the 140 lpm of fine grain film, the 50 lpm or so of the enlarging lens, and the resolving power of the printing paper into account, you will be below the 40 lpm of the digitizer + lens. In fact, by the usual rule of reciprocal of sum of reciprocals, you are already down to about 36 lpm with the analog at the negative stage. I say let us look for a more compelling reason for wanting analog over digital for our photography.

 

[gainer]

Sorry. It just occurred to me that no one has mentioned bokeh.

 

[Sparky]

Hi all,

I am giving a talk on B&W prints to people who are digital photographers who know nothing about traditional B&W process. So I need to know some of the maths involved to enable them to understand how a B&W print can have so much detail in the shadow and highlight areas. If anyone knows the maths involved here of the difference between the DMAX of B&W compared to digital I would appreciate it greatly. Many Thanks Paul

Why don't you just draw up a simile between that and audio, for example - the way the CD was, at one time, considered the 'ne plus ultra' of audio - and now it has finally been commonly accepted that quantization, even at seemingly high frequencies results in a situation whereby CDs create a lot of 'listener fatigue' - and that vinyl is also able to carry FAR more information than could ever be digitized... at least conveniently, if at all. That most of the 'spatial' information and subtlety present with vinyl - is simply missing with CD. In short, one could refer to this as the 'soul'...

But I think the whole reason digital photo and audio got hyped in the first place was that engineers looked at overly simplistic models of what the problems were - and paid attention only to parameters that would fit easily on a spec sheet.

Another analogue for this problem is the problem of genetic engineering and cloning. Despite having gotten the sequence roughly correct for Dolly the Sheep (remember THAT one) - it was assumed that all you had to do was get a nucleotide sequence correct and then you could produce a sheep. Well, Dolly died of massive cancers shortly after birth, cancers which were supposedly brewing since 'conception' if we can call it that.

The reason? The model was overly simplistic - 99% of the most important factors were missing (i.e. amino acid environmental factors, hormonal influence, blah blah blah...).

Well - I suppose that'd be MY take on it - were it MY talk.

 

[kevs]

To Paulcop if you're still reading,

As others far more knowledgeable than myself have commented, it could be a mistake to start throwing numbers around unless your audience will expect a mathematical thesis. Instead, you can consider the unique qualities of the silver-gelatin monochrome print.

Different papers produce different results. Toning silver-gelatins in certain ways results in unique colour and density effects. There are the archival qualities - frame an inkjet print and a silver-gelatin together in a sunny window and see which lasts longest! There's the hand-crafted aspect too - the printmaker must know how to handle and engage with his/her materials to produce fine results. And don't forget to mention the satisfaction of emerging from the darkroom with a wet masterpiece. You can't get that from an ink-squirting machine!

Oh, and welcome to APUG btw.

 

[Michel Hardy-Vallée]

You know sometimes I have a nagging feeling if some people are just drive-by-shooting:

1) go to APUG and ask which of condenser or cold light is better
2) watch the discussion get heated
3) ???
4) Profit!

Let's see if the original poster is interested in the responses he's getting from here and Pnet.

 

[gainer]

To Paulcop if you're still reading,

As others far more knowledgeable than myself have commented, it could be a mistake to start throwing numbers around unless your audience will expect a mathematical thesis. Instead, you can consider the unique qualities of the silver-gelatin monochrome print.

Worse yet, you could start throwing numbers around and find that the audience has some who are more knowledgable about information theory and its math than you are and will take great delight in pointing out any fallacies in your argument. I have seen it happen at technical symposia, fortunately not to me. If you become interested, a good book on the general subject is:
Information Transmission, Modulation, and Noise by Mischa Schwartz.

 

[reub2000]

You know sometimes I have a nagging feeling if some people are just drive-by-shooting:

1) go to APUG and ask which of condenser or cold light is better
2) watch the discussion get heated
3) ???
4) Profit!

Let's see if the original poster is interested in the responses he's getting from here and Pnet.

Good point. Now that you mention it, I wouldn't be surprised if the OP was a troll.

 

[Jed Freudenthal]

Ah, success! That is exactly my point. The "MTF" of a lens IS of INTEREST, but only in a wildly technical situation will it really be anything like of critical importance.

All three of the MTFs I posted were of Zeiss lenses for the Hasseblad: In order, the first is of the Distagon CF f/3.5 60mm; the second, of the Sonnar CF f/4 150mm; and the third, the Planar CF f/2.8 80mm - the lens used. All of these lenses are, IMHO, Superb. The MTF is far down the list of attributes I consider in making the choice of which to use in real time.

It is a good example; I am using the wide angle Distagon 40/40, used on Hasselblad and Rollei. And I have a Schneider Super Angulon 90/6.8. Both are very good lenses; however from the MTF data I can conclude: the Super Angulon is to be preferred for landscape (fine details and good off-axis behaviour) above the Distagon. I tested it on a wide landscape with a lot of detail ( Dead horse point in the South West) and found that the Super Angulon is far superior to the Distagon. However, the Distagon is the choice in case of near-by (fashion) photography. Therefore, for the purpose of selecting the most appropriate lens, MTF is a nice tool.
In the same way, a lightmeter is a nice tool. However you have to know where the limits are, and how the readings are to be read.
The value of a tool depends on the qualifications of the person using it.
Jed

 

[Cor]

..Off Topic, neverteheless relevant I guess..

W
Another analogue for this problem is the problem of genetic engineering and cloning. Despite having gotten the sequence roughly correct for Dolly the Sheep (remember THAT one) - it was assumed that all you had to do was get a nucleotide sequence correct and then you could produce a sheep. Well, Dolly died of massive cancers shortly after birth, cancers which were supposedly brewing since 'conception' if we can call it that.

The reason? The model was overly simplistic - 99% of the most important factors were missing (i.e. amino acid environmental factors, hormonal influence, blah blah blah...).

Well - I suppose that'd be MY take on it - were it MY talk.

Discussions about cloning are ofcourse legitmate and important, and a lot can be said about Dolly. But I think you should get the facts right to make a good example:

Dolly died when she was 6 years old of lung cancer, not un-common for sheep. Whether or not it her dead had to do with her "birth" is hard to say with only 1 example..and a far OT discussion here anyway..

anyway see oa: http://en.wikipedia.org/wiki/Sheep_dolly#_note-0

Best,

Cor

 

[Ed Sukach]

It is a good example; I am using the wide angle Distagon 40/40, used on Hasselblad and Rollei. And I have a Schneider Super Angulon 90/6.8. Both are very good lenses; however from the MTF data I can conclude: the Super Angulon is to be preferred for landscape (fine details and good off-axis behaviour) above the Distagon.

I am VERY interested!!

Would it be possible to scan and post both MTF charts, so that I - we - might compare them?

BTW - The "DISTAGON 40/40"? - are you referring to the Distagon CF f/4 FLE or the older lens without the two "floating" front elements... and are you comparing a 40mm lens to a 90mm?

 

[Photo Engineer]

Discussions about cloning are ofcourse legitmate and important, and a lot can be said about Dolly. But I think you should get the facts right to make a good example:

Dolly died when she was 6 years old of lung cancer, not un-common for sheep. Whether or not it her dead had to do with her "birth" is hard to say with only 1 example..and a far OT discussion here anyway..

anyway see oa: http://en.wikipedia.org/wiki/Sheep_dolly#_note-0

Best,

Cor

There is a timing sequence at the end of each DNA strand that determines age. When Dolly was cloned, she ended up with the timing sequence of a 'parent' that was already old and mature and therefore was 'born old'.

Even though she appeared to be aging normally, she was running out of 'end cap' which is kind of like losing the little plastic ends on shoelaces.

When they ran out, here DNA effectively unravelled and she died of the diseases of old age which in this particular case was as you describe. The cancer was hastened by this unravelling effect which led to a major breakthrough in the causes of aging.

People with long "end caps" tend to live longer than those with short "end caps". And don't ask... I'm having a senior moment and cannot remember the correct name for these biological time bombs.

PE

 

[Cor]

There is a timing sequence at the end of each DNA strand that determines age. When Dolly was cloned, she ended up with the timing sequence of a 'parent' that was already old and mature and therefore was 'born old'.

Even though she appeared to be aging normally, she was running out of 'end cap' which is kind of like losing the little plastic ends on shoelaces.

When they ran out, here DNA effectively unravelled and she died of the diseases of old age which in this particular case was as you describe. The cancer was hastened by this unravelling effect which led to a major breakthrough in the causes of aging.

People with long "end caps" tend to live longer than those with short "end caps". And don't ask... I'm having a senior moment and cannot remember the correct name for these biological time bombs.

PE

..we're getting farther and farther on Off-Topic road.. First hats off to you PE, you truly have a vast knowledge...! Although this is more "my field" I do not dare to make firm statements before looking up the literature.

The phenomena you are referring to is the ends of the chromosomes, called telomeres. Indeed after evey cell division we loose a bit of these ends, and eventually we/ sheep/etc. run in to problems. The mechanisms are (obviously) quite complicated.

Anyway probably she did not do too bad; she became 6, average age of sheep seems to be around 11. (that would make her 17 (well actually the age of the egg donor plus 6)..;-)..)

Best,

Cor

 

[Photo Engineer]

Yes, thanks Cor. I remembered the word telomeres about a half hour after he post. Big deal that was then. Oh well, old age isn't all it is cracked up to be.

Thanks for the comment. My minor in graduate school was Biochemistry and I have kept up an active interest in it ever since.

PE

 

[gainer]

Not to kick a dead horse, but another thing no one has mentioned is the characteristic curve comparison of analog to digital. We can pick a film for its response curve as well as its speed and granularity. These qualities enter into the individuality of photographs of the same subject by different photographers on different days and with different cameras and so on. Perhaps the magic of PhotoShop can achieve these effects, but it would only be for the sake of simulating the analog photo.

 

[Photo Engineer]

Patrick, you are right. But when using a digital camera, the characteristic curve (called V Log E or Voltage developed in the sensor as a function of Log Exposure) is quite a bit poorer than the analog curve. This is due to the sensor, aliasing and noise among others. It only looks good due to software enhancement.

I have run enlargment and manipulation tests on scans of analog images and on direct digital images and can see both the enhancement in sharpness and the loss of detail caused by digital sensors in cameras.

If the sensor was about 4x5, and as dense as present sensors, then a digital image might just approach a 35mm negative in quality, as the artifacts would be rather supressed by the size of the image. And, I guess a sensor would have to be about 11x15 to come anywhere near an MF negative or a 4x5. IDK, but the differences on magnification and manipulation are quite astounding.

PE

 

[DrPablo]

If the sensor was about 4x5, and as dense as present sensors, then a digital image might just approach a 35mm negative in quality

But what kind of output would you need to see this difference in practice? I mean there are a LOT of former medium format users who swear by their high end DSLRs these days, and it seems their output is fine for their applications incl large prints.

In other words, is there some other equalizing or mitigating factor between the two different media?

 

[vet173]

Ah!! An expert on "Modulation Transfer Function"!!!



Here are three .jpgs of MTFs, and an image taken with a lens represented by one of them. Can you tell me which MTF applies to the image, and which lens I SHOULD have used?

?

Looking at this from a different standpoint, the attendee. Unless you can have graphs and prints that you can point to this spot on the graph and this spot on the print VS this spot the graph and this spot on the hand made, distinctly unique every one, toned fiber base print. How can I ever be able to make an unbiased assessment. Be prepared to give a class on how to read the curves. For all I know it could just as well be a profit and loss graph. I did understand the picture though. I just can't see the correlation. Is my view inside or outside the curve of the audience? "Academics" or "do you want fries with that". " When he saw the judge with the seeing eye dog, he knew he wasn't gonna look at the 8x10 glossies of Alices' restaurant."