Adventures in film characteristic analysis

[radiant]

I screwed up the drawing but guess what… you hit the ASA triangle exactly with six minutes. Don’t change the developing time, forget what I said about 30 seconds more!

Some sensitometer designs include feedback for intensity / time with a light sensor / accumulator circuit. I assume you don’t have feedback but have some kind of voltage regulator that assures the light will be the same each time?

You have the ability to tweak the time individually per tube. That’s a unique (patentable!) design feature.

Your scanner seems to be serving very adequately as densitometer!

I calibrate my sensitometer by the ASA triangle and on the assumption that Kodak works very hard to assure that TMY2 is exactly 400. (As Ilford does with HP5+). Using that triangle as a guide, the closer your curve gets to hitting the points of (1,3 run 0,8 rise, from 0,1).

Since you hit the triangle exactly, you can take my paste-up top scale as your calibration (until another test suggests a different calibration, or maybe a discovery that something is “out of control”)

If you can reduce all the exposures by 0.03 log meter candle seconds, then your calibration will coincide with 1 MCS at your “#3 from the top” tube (1,5).

To sum: You hit ASA exactly so now my top scale is your calibration.

This is awesome! From "could this work" to matching TMY2 calibration

I don't have any feedback loop for the light level but I assume LEDs do not fade ever with this kind of use. However I took this into account in exposure to avoid voltage drops on the LEDs.

What comes to scanner I think they key is to scan a Stouffer at the same time. That way one gets pretty close calibration and aren't prone to scanner automatics. I programmed a tool to map the values to densities easily too which is a part of this method. But anyways this proves that you actually can use scanner as densitometer if you know the quirks.

What bugs me a bit that the calculated time for 18% / Zone V is a bit too short as I explained before. But maybe there is some switching delay in the LEDs that actually compensate this I ended up with that time by just eyeballing the previous "failed" strip. Lucky guess.

 

[Bill Burk]

You should be happy, you pretty much nailed it in the first pass.

18% is problematic as Stephen has mentioned.

A better ‘given’ is that a light meter will try to put 10x the exposure of the 0,10 point on film.

On the graph that means it would fall under the ‘40’ instead of the ‘400’ on the top scale. Or -1,7 log MCS should be where a camera exposure of a gray card should land. Is that about 0,72 density? That does sound like 18% transmission… but many people think the averaging meter assumes the scene reflects 18% reflectance but it’s more like 12 something (getting the assumption about the scene reflectance mixed up with the assumption about the film density expected value and thinking it’s 18% all through). We don’t aim for 0.72 even if it kind of works out to be that.

 

[bernard_L]

Bill, could this be a good graph?
The densities are translated from Stouffer (Scanned together with strips and then grayscale mapped to T2115 density).
X-axis is translated by multiplying stops 0.3.
View attachment 297574

And previous graph data zipped.

I did this quick and in Excel. I used a relative log-H and the density is the supplied density minus film base + fog. This is looking to me like a film curve.
View attachment 297590

@Stephen Benskin , @radiant Can you please explain how the improbable HD plot in the first quoted post morphed (using same data??) into the much more plausible plot by Stephen Benskin?

 

[Bill Burk]

@bernard_L They both look good but the vertical scale on radiant’s graph is distorted making the curves appear to indicate higher contrast

 

[Bill Burk]

And it seems I never answered how I can see
1 MCS on the graph.

I work backwards from the curve that meets the “ASA triangle”.

For any speed film you anchor the tip of the triangle where the curve meets 0,10 above base plus fog. Look over 1,3 and up 0,8 (about 0.62 contrast index) and if the curve meets you there (+/- 0,05), you have met the development conditions for speed testing “similar to the way that ISO testing is performed”.

400 speed film should be -2,7 log MCS at that point so I anchor the top scale at that point and call it “calibrated”.

 

[Stephen Benskin]

I've been reading about determining speed. Of course if you know lux-seconds of the sampled data, it is really easy to find out the speed. Just find the speed point and that's it.

Is there some other characteristics rather than how long the film has been exposed to determine the speed?

That's not it. That is where most people get it wrong. If you read up on the ISO standard, you will see that the film has to be processed to a certain degree before a speed rating can be determined. This is because the fixed density point of 0.10 is not the the actual speed point. The idea of speed is to determine a the optimum way to consistently produce the highest quality with the greatest number of materials in most conditions. With black and white film, shadow detail is important to the interpretation of a quality print. This was found not to be a point of Density but where the shadow gradient is at 0.3x the film's overall gradient. This point of minimum useful gradient or limiting gradient is known as the fractional gradient point. It has bee determined to be the most accurate indicator. Once the limiting gradient is determined, any camera exposure that places the shadow at or over that point within a few stop range will tend to yield quality prints. The ASA from the mid-forties to 1960 used the fractional gradient method and once determined added a constant to create an EI that added about two stops of exposure to the Fractional Speed.

There was a problem with the methodology. It was somewhat laborious and prone to experimental error. When an international standard was discussed, many nations didn't want something so complex, yet it yielded superior results. Kodak proposed using the fixed density method popular with DIN but as part of an equation called the Delta-X Criterion. The range of density, ΔD, at 1.30 log-H from the fixed density point of 0.10 over Fb+f can determine how far to the left of the fixed density point (ΔX) the effective fractional gradient point falls. Kodak found that when ΔD equaled 0.80, the fractional gradient point will always fall 0.296 or about a stop to the left of the fixed density point. The Delta-X Criterion enabled keeping the superiority of the fractional gradient method with the simplicity of the fix density method.

But only when ΔD = 0.80. Because the ΔD determines this distance, and the answer is always the same under the stated condition, there is no need to use an equation for the standard, but if ΔD varied from 0.80 the equation was necessary to maintain accurate film speed for film developed to higher and lower gradients. The problem with a fixed density method is it tends to underrate (overexpose) films that are processed to a lower than average contrast and overrate (underexpose) films that are processed to a higher than average contrast. Again, just because you get an answer doesn't mean it's the correct one. What is considered underrating and overrating is based on the judged print speeds from Jones' First Excellent Print tests.

Here is the same ISO diagram with the Delta-X equation included and another showing it applied.

I hope this also emphasizes the importance of producing a film curve that reaches down to at least the film's inertia point with normal processing or one to three stops below the fixed density point of 0.10 over Fb+f.

"What is strange is how you can use CI for determining speed.. That is strange. You must need to know something else of the exposure?"

I hope this answered this question too. Part of the point I've been attempting to make is that drawing a curve is just step one. The important part is interpreting the results.

"It's good that you ask. Yes 0.025lxs for 400. I've measured the lux coming out from the slots and it is pretty accurately 0.2 lux."

I hate to keep harping on this. How? How did you measure the lux? You're not talking about using an exposure meter I hope.

Like I said, interpreting the curve is the important part. I've already shown how most people don't determine film speed correctly. They are not much better with determining development. I think it woujld be helpful for me to address this issue better if you could give me what you consider -2, -1, Normal, +1, and +2 development and why you've determined those numbers.

Stephen

 

[radiant]

I hate to keep harping on this. How? How did you measure the lux? You're not talking about using an exposure meter I hope.

Quick answer to this: I'm using TSL2591 lux sensor for my enlarger analyser so that is very suitable for this. The manufacturer seems to promise it can detect 188 uLux light, so 0.2 lux is peanuts for the sensor.

For the other information I need to chew a bit more ..

 

[Bill Burk]

Your sensor may well be very good, it seems to have gotten you close enough for starters.

An unknown is the Stouffer scale. Since you didn’t spring the extra bucks for T2115c your scanner densitometer could be off a little (and that would affect the ‘calibration’ offered thus far).

If you know anyone with a “real” densitometer then take the Stouffer scale to them and have them read the different patches. If you haven’t got anyone to help let me know. Although I brag about my densitometers, they’re only good enough for “quality control” I do not trust them to be useful for calibration. I do have a fresh T2115c and when I need good readings, I do much the same as you did. I read the nearest Stouffer patch and the sample and interpolate the sample considering how close the densitometer is to the T2115c noted reading.

I love the story of Delta-X every time Stephen tells it.

 

[Stephen Benskin]

Quick answer to this: I'm using TSL2591 lux sensor for my enlarger analyser so that is very suitable for this. The manufacturer seems to promise it can detect 188 uLux light, so 0.2 lux is peanuts for the sensor.

For the other information I need to chew a bit more ..

Great. Thank you, and interesting. How about measuring the exposure time?

 

[Stephen Benskin]

I love the story of Delta-X every time Stephen tells it.

I have to tell it so often because nobody believes me except for you and Michael. I'm the Cassandra of photography.

 

[Nodda Duma]

This thread is like the powerhouse of photrio densitometer wizards. Such awesomeness.

 

[MattKing]

I have to tell it so often because nobody believes me except for you and Michael. I'm the Cassandra of photography.

I too, am a believer.
Although I certainly don't have the same level of understanding as others here.
My math skills are so rfusty!

 

[Stephen Benskin]

I too, am a believer.
Although I certainly don't have the same level of understanding as others here.
My math skills are so rfusty!

Thanks Matt. One of these times, I'm going to tell it where the hero gets the girl and retires rich.

I'm so rusty too but with everything. It's embarrassing. The rust might be the only thing remaining.

 

[albada]

This thread is like the powerhouse of photrio densitometer wizards. Such awesomeness.

Funny you should mention densitometer wizards, as I just posted a new thread about a semi-DIY reflection densitometer here: https://www.photrio.com/forum/threads/accurate-reflection-densitometer-for-120-with-diy-part.189738/ Does that make me a wizard, or just the Royal Fool of Photrio? Actually, I'm still a neophyte when it comes to darkroom work (I've been scanning film for about two decades), but I'm learning...

Mark Overton

 

[Nodda Duma]

Mark, sure! I’ll check it out. I know just enough about sensitometry to characterize the speed of my dry plates.

In fact, I have a spreadsheet that does all the calculations which I put together three years ago and then promptly forgot everything. Not a wizard

 

[albada]

I have to tell it so often because nobody believes me except for you and Michael. I'm the Cassandra of photography.

I've read your description of Delta-X. It's an approximation of the ideal criterion of 0.3 fractional gradient. It seems to me that it would not be difficult to measure the overall gradient by finding the inflection-point of the S-curve. Then finding 0.3 times that slope should be easy after plotting the H-D curve on graph paper with a French curve. What was the difficulty?

Mark Overton

 

[Stephen Benskin]

I've read your description of Delta-X. It's an approximation of the ideal criterion of 0.3 fractional gradient. It seems to me that it would not be difficult to measure the overall gradient by finding the inflection-point of the S-curve. Then finding 0.3 times that slope should be easy after plotting the H-D curve on graph paper with a French curve. What was the difficulty?

Mark Overton

Determining the 0.3X point. Testing involves picking a point, then testing for a very shallow gradient based on an a average gradient of a greater portion of the curve and if it doesn't match the criteria, then checking to the left or right of that spot until it does. Finding the fixed density point of 0.10 over Fb+f requires finding the fixed density point of 0.10 over Fb+f. It's kind of funny how people never seemed to wonder why what they think is the first usable density is found at such a convenient spot. And it's packaged logarithmically too. It was so helpful for the sensitometic gods to have created it in such a way.

In the paper Simple Methods for Approximating the Fractional Gradient Speeds of Photographic Materials, C.N Nelson and J.L. Simonds write, "A frequent criticism of the 0.3 G criterion, however, is that it is more difficult to operate and more subject to random errors than simpler criteria, such as the one based on a density of 0.1 above fog. It is sometimes contended that the increase in ease and precision resulting from the use of the 0.1 density method outweighs the fact that such speeds do not correlate as well with print judgement speeds." The Kodak men of legend would have none of this. They would not succumb to compromise and mediocracy. Thus was born the Delta-X Criterion. Well, also the W Speed Criterion, but that never went anywhere and it gets in the way of a good narrative.

There are a few interesting facts concerning the Delta-X Criterion. It's actually slightly more accurate than the fractional gradient method.

Also, since film speed from the fractional gradient method and the Delta-X Criterion is determined based on gradient and not density, it doesn't change as easily with development as with the fixed density method. This becomes obvious when comparing the two points on films developed to different average gradients.

The second example with a ΔD of 0.80 is the value used in the ISO standard. Remember gradient is the determinant of quality and consequently film speed and not density. The films developed to a higher gradient will produce a negative with overall higher density than at normal development, but the extra density is just printed through. There is no correlation between specific film densities and print densities.

Since the value of ΔX is based on ΔD, a table can be created instead of having to do the calculations with each test.

 

[Bill Burk]

Here’s a Delta-X meter aligned on radiant’s 12 minutes curve. EI that I had noted as 640 based on where 0,1 point falls - by Delta-X indicates closer to 500. For you see the curve meets the right hand side of the scale at 0,19 that instructs you to look at the bottom scale 0,19 which is to the right of the original 0,10 point by about 0,10 (I shifted the scale 0,29 from traditional meter because I want to reveal the film speed lookup instead of the 0,3 gradient) The actual 0,3 gradient falls 0,29 to the left.

You think developing 400 film 12 minutes instead of 6 minutes would give you an effective speed of 640, but in terms of maintaining print quality, you have achieved closer to 500

Not to discount the fact that people will develop HP5+ for 12 minutes and shoot it in their cameras set at 1600 and say it works for them.

But for quality prints, 500 is a better speed to choose with 12 minute development (Rodinal 1+25).

 

[albada]

Finding the fixed density point of 0.10 over Fb+f requires finding the fixed density point of 0.10 over Fb+f. It's kind of funny how people never seemed to wonder why what they think is the first usable density is found at such a convenient spot.

Like most, I had not wondered. I believed what I had been taught. Except I noticed that it's possible to use the curve that's to the left of the ISO speed-point of 0.10 above B+F, especially when scanning.

It appears that Delta-X is based on ISO CI (= deltaD/1.3), and shifts the speed-point left by a quadratic approximation, coming close to the gradient method's point. Or did I misunderstand it? Anyway, Delta-X looks simple. And seeing that it's more accurate, why wasn't it adopted? Committee politics?

BTW, I thought the standard CI was 0.58, but 0.8/1.3 = 0.62. Did I miss something?

Mark Overton

 

[Stephen Benskin]

Here’s a Delta-X meter aligned on radiant’s 12 minutes curve. EI that I had noted as 640 based on where 0,1 point falls - by Delta-X indicates closer to 500. For you see the curve meets the right hand side of the scale at 0,20 that instructs you to look at the bottom scale 0,20 which is to the right of the original 0,10 point by about 0,10

You think developing 400 film 12 minutes instead of 6 minutes would give you an effective speed of 640, but in terms of maintaining print quality, you have achieved closer to 500

Not to discount the fact that people will develop HP5+ for 12 minutes and shoot it in their cameras set at 1600 and say it works for them.

But for quality prints, 500 is a better speed to choose with 12 minute development (Rodinal 1+25).

View attachment 297712

Sorry Bill, I can never make heads or tails out of your graphs, but that speed seems a bit high.

Quick question. Are you including a different constant when calculating the Delta-X speed or just 0.8 / Hm?

 

[Bill Burk]

I am using 0.8 / Hm

It could be the unusual curve shape. Although Delta-X speed doesn’t move as dramatically as chasing 0.10 it still moves somewhat. You gain a real speed increase when you develop longer.

In this case it amounts to a third stop. Not that big a speed gain, but I think it’s real.

 

[Bill Burk]

Here’s how the “real” Delta X meter looks at that 12 minute curve.

In this case the meter reveals the Delta-X speed point, the 0.3x average gradient… right at the edge of my graph paper and corresponding to about -3.1 log mcs… still talking 500 shooting speed vs 400 with 6 minutes development

 

[Stephen Benskin]

Like most, I had not wondered. I believed what I had been taught. Except I noticed that it's possible to use the curve that's to the left of the ISO speed-point of 0.10 above B+F, especially when scanning.

It appears that Delta-X is based on ISO CI (= deltaD/1.3), and shifts the speed-point left by a quadratic approximation, coming close to the gradient method's point. Or did I misunderstand it? Anyway, Delta-X looks simple. And seeing that it's more accurate, why wasn't it adopted? Committee politics?

BTW, I thought the standard CI was 0.58, but 0.8/1.3 = 0.62. Did I miss something?

Mark Overton

It was adopted. It's the ISO standard. Standard's don't have to explain theory. They just need to explain how to do something. In the case of the ISO speed standard explains how to test for black and white negative film speed,. People see the contrast parameter and believe it represents normal development and don't question why it's really there. They also see the fixed density point and believe it's a universal speed point good for all occasions when the standard only says it's good when the contrast parameters are met. I guess it wouldn't have hurt anything if the standard had at least included the Nelson and Simonds paper in their bibliography.

Concerning the contrast parameters of the standard, no you didn't miss anything. Good catch actually. .The answer is, they are not supposed to represent normal development. One of the ideas of speed testing is have method that can be applicable to the range of emulsions on the market. An oranges to oranges comparison. Under the ISO contrast parameters, the approximate fractional gradient point will always fall 0.297 log-H units to the left of the fixed density point. Oranges to oranges. As you can see from the examples in the previous post, the speed is based on gradient and not density. The speed value isn't going to change when the film is developed to the slightly lower CI.

One more thing about the parameters. The results are dependent on the methodology used. Conceptually, it's the same idea that Jones and later Nelson and Simonds faced when determining a speed methodology. Which method works best. With contrast determination, there are basically three different methods generally used. Kodak's CI, Ilford's G-bar, and Agfa's continued use of Gamma. Using each of the different methods on the same curve will most likely yield different results. So the results are based on the measurement.

The ISO contrast parameters according to Nelson and Simonds, "was chosen in order that D2 would be approximately equal to the highlight density in the first excellent negative of an average scene. The difference, D2-D1 is call ΔD." The 1.30 log-H range makes a difference depending on the type of emulsion. While the gradient of 0.80 / 1.30 equals 0.61, it will only match the CI of the film if the film is short toed. Longer toed films need to have a higher CI to match the ISO's ΔD. I've seen it to be around 0.67 or so.

There also isn't a standard CI. There is a CI based on the statistically average Luminance Range, printing on a grade 2 paper on a diffusion enlarger / contact print, and average flare. That would be:

Log Subject Luminance Range: 2.20
Aim LER: 1.05
Flare: 0.40

Flare is interesting. Kodak used to have CI 0.56 as their Normal. Around the time of Dick Dickerson, they started publishing it as 0.58. I believe they changed their value of average flare because of the greater use of small formats which tend to have more lens elements and higher flare. Since his departure, I've noticed that when Kodak retests the films ISO (like when they changed to a new coating alley), the published aim CIs I tended to see for normal processing fell back to 0.56. I haven't really spent much time on it, so maybe it's just a confirmation bias.

I'm glad you touched on this. I'm thinking about writing a post on the subject of developmental models as this thread is partly about curve interpretation. I've found that while the use of the term Normal might be ubiquitous, it is often not well understood or wrongly interpreted using the film curve. Calculating the pluses and minuses can be more involved that first realized too. It's an interesting topic and really what the whole curve plotting is about.

 

[bernard_L]

@bernard_L They both look good but the vertical scale on radiant’s graph is distorted making the curves appear to indicate higher contrast

I don't see how the vertical scale is distorted. Looks linear to me.
On second looks, I understand that Stephen Benskin's curve (post #110) is a re-draw of radian's data (post #106) for 6 minutes devt; with different offsets on both scales; plus, dots alone without connecting lines masks the implausible hump at 4th point from left, which is most prominent on the 12min curve (orange dots) not re-drawn by Stephen Benskin. I'm still definitely not buying the "12min" curve.

 

[bernard_L]

I hate to keep harping on this. How? How did you measure the lux? You're not talking about using an exposure meter I hope.

Quick answer to this: I'm using TSL2591 lux sensor for my enlarger analyser so that is very suitable for this. The manufacturer seems to promise it can detect 188 uLux light, so 0.2 lux is peanuts for the sensor.

I too keep harping. I had the same question in words, but meaning not "how can you go as low as 0.2 lux" but rather: how accurate is your reference illumination of 0.2 lux?

The datasheet for the sensor https://ams.com/documents/20143/36005/TSL2591_DS000338_6-00.pdf mentions the word "lux" only in the introduction:

This digital output can be input to a microprocessor where illuminance (ambient light level) in lux is derived using an empirical formula to approximate the human eye response.

The responsivities (ADC units per µW/cm2) shown in Fig.8 of the datasheet have allowed changes min-max +/-15%. Plus conversion from µW/cm2 to lux is not entirely trivial. Since this thread discusses fine points of sensitometry at the level of 1/3 stop (26%) or less, the measurement should be accurate to 10% or better, total error budget, not just calibration of zero point of illumination. In that respect, I believe that a good exposure meter might actually not be that bad; Gossen used to include with some of its meters a calibration constant in lux.