Paper reciprocity-failure after only 30 seconds

[albada]

Ralph Lambrecht was right. Starting on page 336 of his excellent book, Way Beyond Monochrome 2nd ed, he states that papers exhibit significant reciprocity-failure even after a short time (a few seconds). I was testing my new LED controller/timer, and one test was a strip of exposures of 1, 2, 4, 8, 16, and 32 seconds, while dropping LED PWM-levels to keep actual exposure the same. The rightmost tile was a repeat of the 1-second exposure. The first strip below is the result using Ilford's latest RC paper:

Density steadily decreases from left to right, that is, from 1 to 32 seconds. And as the rightmost (1-second) tile shows, density has dropped significantly.

To rule out a problem in my PWM-calibration, I created a second strip of the same test, but changing light-level using the aperture instead of PWM-levels. In the second strip above, you see I got the same result, but a little uneven due to mechanical error of the aperture.

Practical effect: If you make 5x7 work prints, and then make a 16x20 enlargement using 9x the exposure time, it will be a little lighter than you expect, possibly ruining the print. Lambrecht states that his testing revealed that reciprocity-loss ranges from 1/12 to 1/24 stop-loss per doubling of time. My 9x example is about 3 time-doublings, and 3/12 is a 1/4-stop loss. That will hurt, especially if your contrast (grade) is high. My strips above are at grade 4, and the tiles encompass 5 time-doublings.
Mark Overton

 

[bernard_L]

Thank you for the heads-up!
So, need to use diaphragm to stay close to constant time?

 

[Nicholas Lindan]

That's strange - my tests resulted in no perceptible reciprocity effects over an 8-stop range. They can be found here:
http://www.darkroomautomation.com/support/appnotereciprocityandintermittency.pdf

The two tests used different equipment and probably different methods and that may explain the difference in results.

 

[ic-racer]

You calibration is off for light intensity. How are you measuring it?

 

[radiant]

I think this could be measured with step wedge? Change exposure time (and aperture) in stops. Exaine if the step wedge has "reprocity" changes. If you change time and aperture, each test strip should be identical, I think.

 

[Nicholas Lindan]

So, need to use diaphragm to stay close to constant time?

That was the case with the old Cibachrome/Ilfochrome color process for making prints from slides. Changing the time produced color shifts. The Ilford EM-10 enlarging comparator was designed as an aid in setting the lens aperture so the light falling on the easel was the same for each print - and thus assuring that the time was the same for each print. [Aside: Cibachrome prints were stunning - I went to an exhibition in Palo Alto some years back - 4x5 prints (4x5 in feet) - of redwoods and rhododendron and such; funny how size has a quality all its own, I felt in among redwoods.]

 

[albada]

You calibration is off for light intensity. How are you measuring it?

The first strip above relied on the light-levels from PWM being accurate. I calibrated the PWM-table using the enlarger meter sold by DarkroomAutomation.com, which is our own @Nicholas Lindan. The meter has a resolution of 0.01 stops, and perhaps Nicholas will tell us how he calibrates that meter. I double-checked intensity with a Luna Pro, which agreed with the DA meter within the limit of its accuracy.

My second strip only used one PWM-level, so its calibration was irrelevant, but accuracy was limited by the click-stops on my EL-Nikkor 80mm.

Anyway, why do you believe my calibration is wrong?

 

[Vaughn]

Changing time with RA4 also could change color. Very noticable if one tries to burn in the sky. I was taught to do color 'test strips' using aperture instead of time.

Never worried about it printing on silver gelatin, but I kept my base exposures around 20 seconds. So are we talking a different kind of RF than with film? With film, RF does not significantly affect the mid-tones to highlights -- RF causes the shadows to lag behind...hence one gives more exposure, and less development (because the highlights were acting normally and you gave more exposure in correcting for RF.) It seems like we are looking at a changing paper speed relative to exposure time, since we are dealing with approx. middle grays.

Your tests do not show any relationship between high and low values, and I am not sure we can infer from them what the differences will be in the print. One would need to have a negative that gives you a Zone VII and a Zone II (or thereabouts), and expose several prints to make the Zone VlI at Zone VII using different apertures and times. Then check where the Zone II actually falls at each time.

If paper acts like film with RF, we should be seeing some contrast changes.

 

[Andrew O'Neill]

Why would reciprocity matter in the darkroom, when we make test strips for the exposure? I can see it mattering though if we want to make paper negatives incamera.

 

[Vaughn]

I use to use graded papers -- I suppose if I needed just a touch more contrast to the Grade 3, I might opt for a longer exposure?

How about burning -- is this why it seems to take forever to burn in that dang light area? (with the blacks seeming to take off faster than they should).

But we print to what we want to see, not to what the paper/filters/lamps/chemicals are 'suppose' to do.

 

[koraks]

To rule out a problem in my PWM-calibration

Have you verified the pulse width at low duty cycle and the actual light output of your leds? I say this because nonlinearities occur at very low duty cycle, depending on component choice, circuit design and pwm frequency.

 

[albada]

Have you verified the pulse width at low duty cycle and the actual light output of your leds? I say this because nonlinearities occur at very low duty cycle, depending on component choice, circuit design and pwm frequency.

It's a coincidence that you asked now. A few minutes ago, I made a test strip that uses the same 4-second exposure for all tiles, and for each successive tile, the aperture is opened one stop and the LED power-level is reduced one stop (using PWM). Here is the result:

The rightmost tile is the same as the leftmost for ease of comparison. The PWM duty-cycles ranged from 100% on the left down to 3.125% in the 2nd-to-rightmost tile (=1/32, a 5-stop reduction). The density is almost the same in all tiles, with the 2nd-to-rightmost being slightly lighter. This strip tells me two things:

1. My PWM calibration is accurate enough.
2. Ilford MGRC V paper does not exhibit reciprocity-failure due to PWM. I hope this is true of all papers.

The second observation (paper not affected by PWM) is important. Anyone building a LED-head using PWM will be interested to know that paper responds to pulsed light the same as continuous light. FYI, my PWM is 10 bits with a frequency of 122 Hz (ATmega328p in an Arduino Uno).

@koraks: Can you tell us the specific problems that occur at low duty-cycles? My calibration procedure was simply to find the PWM that gave intensity closest to every tenth of a stop, ranging from full power down to 1/32 power. I was hoping this calibration procedure would compensate for rise and fall times at low duty-cycles. And reducing those rise/fall effects was why I kept the frequency fairly low at 122 Hz. I'm thinking of dropping it to 61 Hz, but I'm afraid that it might produce discreet lines instead of smooth blur if the user moves a dodge/burn tool quickly.
Mark Overton

 

[Nicholas Lindan]

Darkroom Automation's meter calibration, even if it is wildly off, should make no difference in this test as all that is looked for is the ratio of light intensity between the patches. FWIW, the Darkroom Automation meters are calibrated to an enlarger whose light output is set with a pair of reference meters. The reference meters are calibrated to and periodically checked against a closed-loop Ulbricht sphere light source.

I have no explanation for the difference in the two outcomes. One possibility is a difference in reciprocity between MGIV and MGV papers - a long shot but interesting if this is the case.

 

[Nicholas Lindan]

Why would reciprocity matter in the darkroom...?

Reciprocity failure in the range of normal printing light intensities would be significant. Reciprocity failure isn't a function of time but of light intensity falling on the media. But in photography the light intensity falling on a given media is proportional to the shutter speed/exposure time: we want the integrated exposure to be constant, so as the light intensity on the media falls the exposure time increases. And so time becomes a convenient stand in for the intensity of the light falling on the media (there should be a GDA for that - ILM?).

Reciprocity failure when taking pictures results in loss of shadow detail/increase in shadow contrast while the highlights may not be affected at all. Reciprocity failure when making prints results in a loss of highlight detail/increase in highlight contrast.

Assuming the paper does exhibit a 1/12 stop of RF over a 32:1 (5 stop) intensity range as Mark Overton's experiments indicate, I'm not sure this would cause much of a problem. I print 2-stops down from wide open and make all exposure changes by changing the exposure time. Going from a 5x7 proof print to a 16x20 final print is a 9:1 (3 stop) reduction in printing light intensity and shouldn't cause any significant change in the highlights. Well, not entirely true, what may be a small sparkly highlight in a 5x7 may turn into a large disturbing blank area on a 16x20 print without any help from reciprocity failure; there are, as we all know, many factors at play.

 

[koraks]

Can you tell us the specific problems that occur at low duty-cycles?

Rise-time issues, mostly. But at your pwm frequency and the kind of duty cycles you use, they're unlikely to be an issue, unless there's a lot of (stray) capacitance in the circuit.
Another potential issue is short exposure reciprocity failure. At your low duty cycle, your pretty close to where short exposure reciprocity issues are likely to manifest in b&w paper.

 

[Craig75]

why do the left most tile at 1 second and the right most tile at 1 second look different?

 

[albada]

why do the left most tile at 1 second and the right most tile at 1 second look different?

I think it's due to mechanical hysteresis. When moving an aperture from closed to open, the click-stops will yield slightly smaller apertures than when moving from open to closed, due to space between parts. I measured my lens a few minutes ago with my DarkroomAutomation enlarger-meter, and the difference was 1/20th of a stop. That small difference is magnified in the print because the high contrast of grade 4 increases small differences in exposure. Nonetheless, I want to reprint these strips, but using a Stouffer wedge to see how shadows, midtones, and highlights are changing as time and PWM are changed. @Nicholas Lindan mentioned that RF causes highlights to be lost first, so I want to explore that.

A couple of posters mentioned changing exposure using aperture. I avoid that because my Micro-Sight grain-focuser shows that sharpness is best between f/8 and f/11, and quickly dropping above f/11 due to diffraction. So I think f/11 is the best compromise between sharpness and adequate depth of field to cover misalignment of the enlarger. That's part of the reason I like LEDs so much: I can stay at f/11 and change the brightness of the LEDs instead.

@koraks: I'm using the BuckBlock A009 as my LED-drivers, and its maker recommends a maximum PWM frequency of 200 Hz. But the MeanWell LDD-700L has a 1000 Hz maximum, implying a much faster rise-time, so I'd like to experiment with it. I'm almost certain that at low duty-cycles, the BuckBlock outputs a triangular curve that does not reach the target voltage or current, thus dimming the LEDs, which could worsen RF based on Nicholas Lindan's statement that "Reciprocity failure isn't a function of time but of light intensity falling on the media".
Mark Overton

 

[koraks]

I'm using the BuckBlock A009 as my LED-drivers, and its maker recommends a maximum PWM frequency of 200 Hz. But the MeanWell LDD-700L has a 1000 Hz maximum, implying a much faster rise-time, so I'd like to experiment with it.

Unfortunately, for neither of these drivers I can find a datasheet that includes useful dynamic operation characteristics. So things like rise-time etc, we'll have to take a wild guess at. Maybe they both work fine at low duty cycles. Maybe only at lower frequencies. Maybe there are inherent non-linearities within the range you're likely to use them. There are only two ways to find out:
* Locate a better datasheet that gives more detailed information, preferably some scope plots for PWM scenarios. To get a feeling for what a datasheet could/should look like in this respect, see e.g. that of the MP24894: https://www.monolithicpower.com/en/.../document_type/Datasheet/lang/en/sku/MP24894/ refer to page 5 and then the first two plots on the bottom row.
* Test it. And not only verify if they appear to work by printing on paper, but actually doing measurements. It's not horribly difficult to set up a test rig to check how the current through the leds is modulated using a pwm signal. A fairly simple scope and a series pass resistor go a long way.

I'm almost certain that at low duty-cycles, the BuckBlock outputs a triangular curve that does not reach the target voltage or current, thus dimming the LEDs, which could worsen RF based on Nicholas Lindan's statement that "Reciprocity failure isn't a function of time but of light intensity falling on the media".

Yes, well, obviously a triangular current curve would be pretty bad at the kind of pwm frequencies and duty cycles we're looking at here. Naturally rise times aren't instantaneous and there's always a bit of smoothing on the leading edge - how much depends on circuit, components etc. This in itself has little to do with reciprocity failure, but the two effects can of course occur together.

 

[138S]

Ralph Lambrecht was right. Starting on page 336 of his excellent book, Way Beyond Monochrome 2nd ed, he states that papers exhibit significant reciprocity-failure even after a short time (a few seconds). I was testing my new LED controller/timer, and one test was a strip of exposures of 1, 2, 4, 8, 16, and 32 seconds, while dropping LED PWM-levels to keep actual exposure the same. The rightmost tile was a repeat of the 1-second exposure. The first strip below is the result using Ilford's latest RC paper:

This effect is seen mainly in the scene highlights, highlights (high density in the negative, low intensity in the paper) have way lower light intensity in the scene highlights, (1/100 for negative density 2.0D, compared to Zone 0). If you repeat your test with base exposure (say) 1/30 simulating the density of some highlights then you will see a way more intense reciprocity failure, LIRF (Low Intensity Reciprocity Failure)

In fact the failure does not come directly from the long exposure, but from the Low Intensity, still low intensity is a usual situation in long exposures, but effect is from the low intensity. In a long exposure in the negative the highlights have less Failure than the shadows. In the paper printing it's the counters the highlights have way more reciprocity failure, because dense areas provide less intensity.

 

[albada]

@koraks: At 1/32 power with 10-bit PWM, and no rise/fall time, the PWM setting should be 1023/32 = 32. My calibration-tables contain values of 82 to 103 for the three colors. The height of the rectangle with the same integrated area but a width of 100 would be .32 of the perfect-PWM height. Assuming zero fall-time, we can make that a right triangle whose height is .64 of perfect height. That tells me the BuckBlock's current curve is a triangle that reaches about 2/3rds of the desired current. Not impressive.
The MP24894 you mentioned above is tempting! At my frequency, it will have an infinitesimal rise/fall time. Have you used it in a LED controller?

 

[koraks]

The MP24894 you mentioned above is tempting! At my frequency, it will have an infinitesimal rise/fall time. Have you used it in a LED controller?

Currently only in a prototype that hasn't exposed any paper yet. In my existing enlarger I just use LM350T's, but that is a bit of a hacksaw solution, even though it does work fine. The mp24894's will be far more efficient.

 

[albada]

Currently only in a prototype that hasn't exposed any paper yet. In my existing enlarger I just use LM350T's, but that is a bit of a hacksaw solution, even though it does work fine. The mp24894's will be far more efficient.

This is a constant-voltage design which works fine for you, so you have no problems with thermal drift? I assumed an open-loop design would suffer from thermal drift. If not, what do you think of the following design?:

This is an open-loop design that relies on having good regulation in the supply and no significant thermal-drift in any component. I envision the PWM frequency being around 50-100 KHz, and the inductor L large enough so that LED-current will have only a small ripple.
I see two advantages of this design:
1. The LEDs will probably last longer because they are generally not driven at full power (unlike pulsing). This is a minor advantage.
2. The LEDs will provide nearly constant light, thus simulating the behavior of a tungsten bulb, and thus this light-head should cause papers to behave the way they were designed and tested, with no surprises due to pulses.

And I see some weaknesses in this design:
1. Susceptibility to thermal-drift and age-drift because it's open-loop.
2. If the diode D fails, the flyback voltage from L is likely to destroy the MOSFET switch Q.
3. Calibration must be done carefully to avoid a PWM that would overdrive the LEDs. I would temporarily insert a sense resistor to measure current.
4. A bug in software or any mistake that leaves the switch Q closed will crowbar the supply (if we're lucky), or burn out the LEDs.

Is this design feasible? If so, can you think of additional advantages or disadvantages of it?

 

[koraks]

Is this design feasible?

Yes, it's essentially what all the buck drivers like the mp24894 do. The main drawback of your approach is that you're solving the engineering challenges that are already tackled in those buck ic's.

 

[albada]

The main drawback of your approach is that you're solving the engineering challenges that are already tackled in those buck ic's.

With one difference: In my design, the high frequency PWM is used to control brightness. Commercial drivers use high frequency PWM at a fixed 50%, with a much lower frequency PWM superimposed from an external source (MCU or PWM chip), which thus PWMs the LEDs. My approach yields a constant current through the LEDs that is not pulsed, so the paper will see constant light (like tungsten) instead of pulsed light. I hope this will avoid reciprocity artifacts from pulsing.

Many commercial ICs support analog dimming, which I suspect is done as above, but the dynamic ranges I've seen are 2-3 stops, which is insufficient.
Do you know of any LED controllers that can produce near-constant (non-pulsed) current with a dynamic range of 4+ stops?

 

[koraks]

Do you know of any LED controllers that can produce near-constant (non-pulsed) current with a dynamic range of 4+ stops?

I havent looked specifically for this property, but all devices I surveyed didn't provide linear dimming at very low levels. I'd probably go for a opamp driven mosfet or bjt with a current sense resistor for this.