Adventures in film characteristic analysis

[radiant]

The deviations in light output are hardly measurable, so small that you can use the system as a sensitometer. You can make step wedges for comparisons.

Yes, that is the original idea; to expose film in enough controlled way that I can make comparisons between my own tests.

 

[Bill Burk]

…It's in a way not too different from the first visual densitometers, where one compared the relative brightness of two beams of light, one passing through the sample…

Except for all the electronics and reliance on a calibrated scale that can fade. I have an electronic densitometer that is working but I am never sure when something will go wrong with it. As an alternative, I like ( Marshall Studios ) a visual densitometer takes advantage of the law of inverse squares.

 

[Nikola Dulgiarov]

A purely optical visual densitometer is as accurate as the operator's ability to discriminate fine changes in density. The density tablets are usually silver-gelatin or lampblack-filled gelatin, so it shouldn't fade in normal storage and usage.
The Physikalisch-Technische Bundesanstalt in Germany uses (or used up to the 2000s) a reference densitometer based on the inverse square law. There's a detailed paper on its construction and calibration but it's sadly behind a paywall.

@Steve Goldstein , the temperature-dependent "sensistor" is made up of a +8000 ppm Si temperature sensor in series with metal film resistors (+/- 50ppm), and the weighed average tempco is tuned to 3300. Not ideal but it works well (constant tempco) within normal operating conditions, and was made up from components available locally to me. 0.33% resistors are still available, mostly for audio applications (stability in VCOs, etc).

 

[radiant]

A purely optical visual densitometer is as accurate as the operator's ability to discriminate fine changes in density. The density tablets are usually silver-gelatin or lampblack-filled gelatin, so it shouldn't fade in normal storage and usage.
The Physikalisch-Technische Bundesanstalt in Germany uses (or used up to the 2000s) a reference densitometer based on the inverse square law. There's a detailed paper on its construction and calibration but it's sadly behind a paywall.

Do you mean that they made different exposures by changing the light path distance for each step? And using one light source = same bias for each step?

 

[Nikola Dulgiarov]

@radiant, I believe the PTB densitometer used a very advanced version of the original "grease-spot photometer" - by comparing a source of unknown intensity with a calibrated one and varying the distance until they are equal. If the emitter approximates closely a point light source, the relationship between distance and intensity is ideally an inverse square law.

For a sensitometer, I think it would be impractical to build an instrument on that basis. LEDs are a poor point source until you are quite far away. But the relationship holds in projection systems, so under an enlarger, you can trust that an image at 2x the distance will have 4x the area and respectively 1/4 of the illuminance.

Your time-based approach will work within reciprocity failure limits, as others have pointed out. Controlling the intensity with PWM might cause intermittency effects on the exposure, though IIRC those effects disappear at higher pulse frequencies. Controlling each led with a constant current is fine, but then the current-intensity relationship deviates from linearity and the design is complicated by having to program 3-4 decades of current sources.

 

[bernard_L]

Your time-based approach will work within reciprocity failure limits, as others have pointed out. Controlling the intensity with PWM might cause intermittency effects on the exposure, though IIRC those effects disappear at higher pulse frequencies. Controlling each led with a constant current is fine, but then the current-intensity relationship deviates from linearity and the design is complicated by having to program 3-4 decades of current sources.

Another (luxury) option would be a closed-loop system where the LED current is set by a servo loop to achieve the desired current in a photodiode operating in zero-bias mode. The photodiode becomes the critical component, and, IIRC, it is linear over a range exceeding the 10^3 needed in a sensitometer. No more any worry about intermittency effects in the film under test, transient effects in the LEDs, or thermal effects in the LEDs. Did I write luxury? But the assurance of being rid of spurious effects is priceless.

 

[radiant]

For a sensitometer, I think it would be impractical to build an instrument on that basis. LEDs are a poor point source until you are quite far away.

I get really even illumination from my device, the distance is about 8cm from LED to film surface.

Your time-based approach will work within reciprocity failure limits, as others have pointed out. Controlling the intensity with PWM might cause intermittency effects on the exposure, though IIRC those effects disappear at higher pulse frequencies. Controlling each led with a constant current is fine, but then the current-intensity relationship deviates from linearity and the design is complicated by having to program 3-4 decades of current sources.

This can be overcome with LED based systems with physically dimming down the exposure. Just make sure the shortest times are for example 1/500th second.

I know the PWM factor. I'm running the LEDs at full brightness so PWM shouldn't be in use for preventing this. I've just dimmed the leds physically with thick diffusers. I'm not controlling current either. The exposure is also done one LED/slot at time just to avoid current consumption. I don't want to start measuring what my USB power supply provides on every LED combination.

So many pitfalls are taken into consideration in my design

 

[Europan]

assurance of being rid of spurious effects is priceless.

Fully agree.

 

[radiant]

Ok, adventure continues Making some science that has been done maybe thousand times before!

First "real" test, how does development affect to density. Otherwise identical but varying development time: 6 minutes for box speed and 12 minutes for "pushed" 1600:

 

[radiant]

Lining up densities with 2 stop shift-correction shows that the midtones are pretty close to eachother, so the "overdeveloping" amount is very close to correct:

 

[radiant]

Here is contact print I made from these two strips. Approximately with grade 2-3 :

 

[Bill Burk]

You know you can “drop some strings” in the channel to block some light physically. I don’t know what you call it when you “print” a line in space, but if the 3D material is opaque you can physically block some of the light in the path near the LED

 

[radiant]

You know you can “drop some strings” in the channel to block some light physically. I don’t know what you call it when you “print” a line in space, but if the 3D material is opaque you can physically block some of the light in the path near the LED

Dimming is not a problem and doesn't need 3D printing at all. All FDM materials are more or less opaque. My device has been printed maybe 1.6mm wall thickness of black PLA and the dimmed light is still visible in total darkness - I was suprised about this.

 

[radiant]

I try to compare my results to other graphs but there is some bump in low tones. Could only Rodinal as developer explain this? I'm comparing to Adrian Bacons graph: https://i0.wp.com/adrianbacon.com/w...nished-XTOL-24C-7_30-JOBO-Agitation.png?ssl=1

I converted stops to lux seconds (0.3 lux seconds = 1 stop in exposure) and tried to plot same range as Adrian has plotted. I don't know if there is some kind of standard what is the zero of X-axis? Probably explained in the book next to me

Any thoughts?

 

[Bill Burk]

Dimming is not a problem and doesn't need 3D printing at all. All FDM materials are more or less opaque. My device has been printed maybe 1.6mm wall thickness of black PLA and the dimmed light is still visible in total darkness - I was suprised about this.

At the lowest exposures dimming with a physical “screen” may give you better consistency. Especially when you make an exposure test for 1000 speed film. You can run the bulb at a slightly longer time (e.g., 1/1,000 second instead of 1/10,000) if you block some of the light going through the tube.

 

[radiant]

For me the most interesting test has been overdeveloping + its effect. I've always wanted a controlled way to see what actually happens to contrast if one overdevelops.

I'm analysing the two strips (box speed) and then +2 stops overdevelopment (2x more time).

I chose +3 .. -2 stops range from 18% for my analysis area. A bit low in contrast for "normal" scene but lets start easily.

f-stop difference:
box = 2,84 / overdeveloped: 2,84

ISO-R:
box = 48 / overdeveloped: 54

This would require grade 4-5 printing for both. No dramatic change in grade here.

Ok, lets pick more dynamic area (8 stops) (+4 / -4 from 18%):

f-stop difference:
box = 6,13 / overdeveloped: 5,25

ISO-R:
box = 87 / overdeveloped: 105

"Box" would need grade 2-3 filter and overexposed grade 1-2 filter.

My conclusion: by overdeveloping HP5 by double time you need one grade lower filter. If this is true, then we can stop all discussion on these 10-20% increased development.

I must have made a mistake here ..

 

[Bill Burk]

I must have made a mistake here ..

Haaa. Thinking any test will settle any question is the mistake. But don’t worry you’ll get used to it

 

[Bill Burk]

@radiant, I believe the PTB densitometer used a very advanced version of the original "grease-spot photometer" - by comparing a source of unknown intensity with a calibrated one and varying the distance until they are equal. If the emitter approximates closely a point light source, the relationship between distance and intensity is ideally an inverse square law.

The principle of a visual densitometer is simple.

You balance the light from two sides with no sample. That’s 0.00. Then put a sample in place and move the light closer to it until the spots match again. The distance you moved the light directly relates to density based on the law of inverse squares. The densitometer can’t go out of calibration.

In this design there is one bulb. Light from the back of bulb hits two mirrors and then comes through a tube to the front on top. Sample goes on the little light table. The bulb racks back to front over about one foot distance. Effectively light travels about a foot from each side for zero. But near density of 3.00 you roll the light bulb all the way to the front where it might be 2 1/2 inches from the sample, while the light from the back now has to travel about two feet.

 

[radiant]

"Box" would need grade 2-3 filter and overexposed grade 1-2 filter.

My conclusion: by overdeveloping HP5 by double time you need one grade lower filter. If this is true, then we can stop all discussion on these 10-20% increased development.

Anyone? This bugs me, it cant be this way ..

 

[Bill Burk]

Anyone? This bugs me, it cant be this way ..

Your 6 minute development time is near 0.5 contrast and your 12 minute development time is near 0.66 contrast.

With 6 minutes development the negative is so flat that 8 1/3 stops of subject luminance range fits Grade 2 paper. And with 12 minutes development you have a more contrasty negative that accommodates only 6 1/3 stops subject luminance range.

It’s like you found N-1 and N+1

http://beefalobill.com/imgs/Practical Flare Model b.jpg

 

[radiant]

Your 6 minute development time is near 0.5 contrast and your 12 minute development time is near 0.66 contrast.

Looking at the table you provide this would mean for 6 minutes the grade would be 2 and for 12 minutes grade would be 0.5? That is 1.5 grade change. Did you calculate the contrast from my data with 8 1/3 SBR?

What is wrong with my development if 6 minutes and 12 minutes only resolve to N-1 and N+1 (assume that is something like 10-15% under/over)?

 

[radiant]

I measured the densities with scanner (two strips + Stouffer wedge on single scan) and mapped the measured densities to Stouffer numbers. The overdeveloped one saturates at top becauase of scanner limits probably (too dense).

 

[radiant]

Looking at previous image I wonder does HP5 really react to over-development that way after toe? I mean toe isn't that much denser than "box" speed developed one, but then it jumps up (to position 11).

I've read somewhere else that over-development doesn't really affect to toe and doesn't increase contrast in midtones. That is visible from the previous graph.

 

[Bill Burk]

I don’t think there’s anything wrong with your results. I could wish for more densitometer readings under 0.10

That’s why I suggested a barrier screen (or more thick diffusers) on the low end.

 

[radiant]

I don’t think there’s anything wrong with your results. I could wish for more densitometer readings under 0.10
That’s why I suggested a barrier screen (or more thick diffusers) on the low end.

Ah, yes now I understand. I could go in half stops down from 18% of course. That way maybe there would be more resolution and maybe few under 0.1.

That is a great addon, I will change the software in future that way. That bump in low tones feels a bit strange and I was suspecting my device or analysis methods.


Also nice to know that my graphs are probably accurate.