My Dollar Densitometer

[dkonigs]

I just saw some new wheels, today I noticed a old equipment utilized TSL2561 or TSL25715 for UV intensity monitoring which has a nominal 100M:1 dynamic range, and checked their updated light sensors. Their new TSL2572 Ambient Light Sensor features two built-in 16-bit ADCs, four programmable gain from 1x to 120x, with integration steps 0~256, results in 800M:1 dynamic range (roughly 29.5-bit), allows dark room to 60K lux sunlight operation——all above in a small DFN chip. With I2C digital interface, it requires almost no external components, yet it’s priced at only $1.91 on DigiKey. I think this would be a good choose for homemade projects.

This thread is mostly a tl;dr for me at this point, but these sensors are absolutely the reason that 90% of the chatter on CDS vs SPD vs ADC choices is a complete waste of energy at this point. Just use one of these sensors, and call it a day. At that point all of the work is entirely around the optical path and any sort of calibration (you often do need to calibrate the gain levels).

For my own devices, the Printalyzer Densitometer uses the TSL2591, the new Printalyzer UV/VIS Densitometer uses the TSL2585, and the various metering instruments in the upcoming Printalyzer Enlarging Timer use the TSL2522.

(The TSL2585 has photopic and UV channels, while the TSL2522 just has photopic channels, but the rest of the sensor is otherwise very similar.)

 

[koraks]

I just saw some new wheels, today I noticed a old equipment utilized TSL2561 or TSL25715 for UV intensity monitoring which has a nominal 100M:1 dynamic range, and checked their updated light sensors. Their new TSL2572 Ambient Light Sensor features two built-in 16-bit ADCs, four programmable gain from 1x to 120x, with integration steps 0~256, results in 800M:1 dynamic range (roughly 29.5-bit), allows dark room to 60K lux sunlight operation——all above in a small DFN chip. With I2C digital interface, it requires almost no external components, yet it’s priced at only $1.91 on DigiKey. I think this would be a good choose for homemade projects.

Yeah, those little TAOS sensors are really neat. Have a look at the AMS product range as well; those are cool if you also want to do spectral analysis. It's become feasible to build a DIY photospectrometer at this point (albeit it with somewhat crude resolution).

At that point all of the work is entirely around the optical path and any sort of calibration (you often do need to calibrate the gain levels).

And esp. the former is actually pretty relevant, as was my contribution in #2 or so.

 

[Yezishu]

This thread is mostly a tl;dr for me at this point, but these sensors are absolutely the reason that 90% of the chatter on CDS vs SPD vs ADC choices is a complete waste of energy at this point. Just use one of these sensors, and call it a day. At that point all of the work is entirely around the optical path and any sort of calibration (you often do need to calibrate the gain levels).

For my own devices, the Printalyzer Densitometer uses the TSL2591, the new Printalyzer UV/VIS Densitometer uses the TSL2585, and the various metering instruments in the upcoming Printalyzer Enlarging Timer use the TSL2522.

(The TSL2585 has photopic and UV channels, while the TSL2522 just has photopic channels, but the rest of the sensor is otherwise very similar.)

I did some searching—great products! Congratulations!

I believe these sensors are part of industrial solutions today. But it’s still fun to discuss SPDs or CdS cells. I know someone who still looking for and disassemble selenium photocells to repair cameras. Their performance wasn't amazing even when they were brand new, and it's even uncertain with time and replacements——they just don’t want to change historical design. The same may also apply to copying a memorable CdS meter.

 

[Alan Townsend]

Thanks for that. Blast from the past. Yes, this was a crude meter. But it worked well enough to be a big help in darkroom printing years ago, and today I use my version to help create enlarged negatives with a certain density range. A super accurate meter wouldn't be much help, because after metering I make a negative and then test print that negative anyway, followed by subjective adjustments to get a desired result. It helps reduce wasted materials and saves time and iterations. I also use a duty cycle type dimmer with mine, which does not change color temperature, to make a crude correction curve when needed. I use my phone lux mater to read that, adjust to half, then red CdS, repeat until minimum reading. Also use the phone meter for adjusting exposure times by enlarger magnification.

 

[Yezishu]

Have a look at the AMS product range as well; those are cool if you also want to do spectral analysis. It's become feasible to build a DIY photospectrometer at this point (albeit it with somewhat crude resolution).

Regarding spectrometers—to expand on that a bit—my personal DIY choice (everyone’s experience varies) would be a line scan sensor(or image sensor of smartphone), a blazed grating(if lower the standard, CD disk in another post), and CNC-machined/3D-printed mechanical positioning, much like the Hamamatsu C12880MA mini-spectrometer. Those single-digit channel sensor works but don't offer enough resolution.

 

[Alan Townsend]

To put my money where my mouth is, I've plotted the Log of the resistance readings from the multimeter vs the density values of pieces of film over the cds sensor. I've done this for two different light levels. Here's the data and the graph.

View attachment 417070

You can see the data is very linear, and well suited to a straight line function of the form Y = mX + b
I'm showing a least squares linear regression fit to both sets of data.

For any set stable light level, you only need two points to define the linear function. With that, you can calculate the density of another piece of film based on the resistance value from the multimeter.

In practical terms, you start with two pieces of film with known density. Place each piece of film over the cds sensor and measure the resistance value of the cds cell for each.
You need to make a calculation to determine the m and b parameters for your straight line function

m is the slope m = (Log(R2) - Log(R1))/(Density2-Density1)

b is the Y intercept b = Log(R1) - m*Density1

Now that you have m and b, you can calculate the density of another piece of film of unknown density. Just measure the resistance of the cds cell with that piece of film over the sensor.

The density of that piece of film will be = (Log(measured resistance) - b)/m

In other words, you need two resistance measurements to calibrate the system to get the m and b parameters. After that, you can find the density of any other unknown piece of film by just measuring the resistance and plugging it in the formula. It's easily set up in a simple Excel file, so you just need to enter the resistance values, and let Excel give you the result.

Thanks for all the explanations. Previously you were basing from linear-linear space as power function. I'm sure you are correct in all you math. I found a manual for my meter so can now give the specs, but it doesn't change anything. My meter and CdS cell, in the simple way I use them is quite accurate measuring densities from 0 to 2.00 because the resistance changes by approx. -1/2 for each factor of 2 increase in light. This is due to the curve of my CdS cell, which is much steeper than yours. The non-linearity of mine at low readings due to temp coefficient and memory just happen to make it work better or more linearly. Mine automatically subtracts a value of about 50KOhms from the reading to correct for contact resistance. Specs claim 0.6% error at 2MOhms, 0.5% on the lower scales.

 

[dkonigs]

Yeah, those little TAOS sensors are really neat. Have a look at the AMS product range as well; those are cool if you also want to do spectral analysis. It's become feasible to build a DIY photospectrometer at this point (albeit it with somewhat crude resolution).

And believe me I've tried. In theory its a drop-in solution to my existing densitometer design to add color and multispectral capabilities. The problem is that I don't get to chose the response curves of the channels, making it almost impossible to actually make an ISO 5-3:2009 compliant Status A/M densitometer out of it. (which is why I've given up on this approach) Technically I could kludge it with some interpolation math, but you'd always find edge cases where that would fail to give desired results.

Regarding spectrometers—to expand on that a bit—my personal DIY choice (everyone’s experience varies) would be a line scan sensor(or image sensor of smartphone), a blazed grating(if lower the standard, CD disk in another post), and CNC-machined/3D-printed mechanical positioning, much like the Hamamatsu C12880MA mini-spectrometer. Those single-digit channel sensor works but don't offer enough resolution.

If I ever take back up the idea of a spectrodensitometer, those Hamamatsu sensors are really the only viable choice I'm aware of. They're kinda expensive, but still cheaper than the alternatives. They're also a lot more complicated to integrate. Might prototype something with one someday, but I currently have enough on my plate that's likely higher priority.

The simplest approach, of course, is the photodiode+filter mechanism X-Rite uses for their older color densitometers. The only problem is that the filters themselves (for the color densitometry standards) are almost impossible to actually find these days (or only possible to find at exorbitant science-lab prices).

 

[Sharktooth]

Thanks for all the explanations. Previously you were basing from linear-linear space as power function. I'm sure you are correct in all you math. I found a manual for my meter so can now give the specs, but it doesn't change anything. My meter and CdS cell, in the simple way I use them is quite accurate measuring densities from 0 to 2.00 because the resistance changes by approx. -1/2 for each factor of 2 increase in light. This is due to the curve of my CdS cell, which is much steeper than yours. The non-linearity of mine at low readings due to temp coefficient and memory just happen to make it work better or more linearly. Mine automatically subtracts a value of about 50KOhms from the reading to correct for contact resistance. Specs claim 0.6% error at 2MOhms, 0.5% on the lower scales.

No, I've never talked about power functions. It's plotting Log R vs Density. It's just like the D vs Log E curves we're all used to using. The X and Y scales are Logarithmic (Density is a log scale), but the plotted relationship is linear.

I can't get anything useful by taking the ratio of the resistance values. Can you provide some actual measurement data, and show how you're doing the calculations. I must be interpreting what you're doing differently, since I can't see how you're able to achieve a result from a simple ratio, without knowing at least one density.

I don't know how linear my cds cell is going to be for densities greater than 2, since I don't have any known film density samples greater than 2.17. The information I could find on the manufacturer's site we're useless in this regard. There's no curves supplied, so the only way to determine linearity is via experiment, which I've done.

 

[Yezishu]

And believe me I've tried. In theory its a drop-in solution to my existing densitometer design to add color and multispectral capabilities. The problem is that I don't get to chose the response curves of the channels, making it almost impossible to actually make an ISO 5-3:2009 compliant Status A/M densitometer out of it. (which is why I've given up on this approach) Technically I could kludge it with some interpolation math, but you'd always find edge cases where that would fail to give desired results.


If I ever take back up the idea of a spectrodensitometer, those Hamamatsu sensors are really the only viable choice I'm aware of. They're kinda expensive, but still cheaper than the alternatives. They're also a lot more complicated to integrate. Might prototype something with one someday, but I currently have enough on my plate that's likely higher priority.

The simplest approach, of course, is the photodiode+filter mechanism X-Rite uses for their older color densitometers. The only problem is that the filters themselves (for the color densitometry standards) are almost impossible to actually find these days (or only possible to find at exorbitant science-lab prices).

Hamamatsu mini spectrometer is, after all, an spectrometer. AMS small color sensor performs well in measuring color temperature, but its limited too much in wavelength measurements.

Ironically, today narrowband filters with 50nm bandwidth at any wavelength are easier to find than color density filters. I noticed that your requirements are different from those XYZ sensor, which is quite troublesome. Schott SFK 100A, SFK 101B, and SFK 102A kits or AS73211 uses CIE 1931/DIN 5033, but.....

 

[Bill Burk]

Do you have manual on how to use this meter?

No, it just had a dial on the back like any other exposure meter, except it really wasn't getting correct readings

 

[Alan Townsend]

No, it just had a dial on the back like any other exposure meter, except it really wasn't getting correct readings

Vezishu just posted this yesterday, so here you are, but this is for the fancier model that I had. Oops, over the 2 mb limit.

 

[Chan Tran]

No, it just had a dial on the back like any other exposure meter, except it really wasn't getting correct readings

I wonder how to use it.

 

[dkonigs]

Hamamatsu mini spectrometer is, after all, an spectrometer. AMS small color sensor performs well in measuring color temperature, but its limited too much in wavelength measurements.

Ironically, today narrowband filters with 50nm bandwidth at any wavelength are easier to find than color density filters. I noticed that your requirements are different from those XYZ sensor, which is quite troublesome. Schott SFK 100A, SFK 101B, and SFK 102A kits or AS73211 uses CIE 1931/DIN 5033, but.....

Well, the AMS sensors I had in mind are the fancier multi-spectral sensors like the AS7343 or the newer TCS3448. These can do a lot more than just color temperature. They just can't match the specific curves for Status A/M in ISO 5-3:2009. (and the difference between Status A and Status M, for each color, is relatively small)

 

[Alan Townsend]

I wonder how to use it.

Here's the link to one of the manuals. Note that the full scale reading is the null with no density at 49 uA and the max density on a single scale is at 1 uA corresponding to 1.69 density which is 49 and the midpoint of the scale is also in the middle with density 0.3 implying that the current scale is linearly interpreted as density over that short range.

http://www.jollinger.com/photo/cam-coll/manuals/meters/SM%20Photo%20Meter%20A-3.pdf

 

[Alan Townsend]

No, I've never talked about power functions. It's plotting Log R vs Density. It's just like the D vs Log E curves we're all used to using. The X and Y scales are Logarithmic (Density is a log scale), but the plotted relationship is linear.

I may have confused sources, but when I looked up converting linear-linear to log-log plots I found this:

Log - log transformations are used to convert power functions to linear plots.

A log-log transformation of y=axb would be log y= log a+b log x

b is the exponent, hard to see in this cut and paste. I thought you were using this?

I can't get anything useful by taking the ratio of the resistance values. Can you provide some actual measurement data, and show how you're doing the calculations.

Very time consuming doing that accurately with my crummy meter. My meter drifts a lot, so I need to make a reading, then shut the meter down for many minutes before making another reading. I also use a dimmer switch to give a controlled range of 1:10 with my enlarger. I use my phone app to set the dimmer to an illumination that gives me 100, 71, 50, 35, 25, 18, 12, and 9 lux. At each of these, I measure the resistance of the CdS cell. The lux app does not give decimal values, so these are closest approximations of 1/2 f-stops. Much easier to plot 100, 50, 25, 10, but last illumination would be plus or minus 10%. My phone app lux meter is more accurate than a stand alone one that I used in the past. It's very repeatable at least. With a much brighter lamp, I have tested it over a broader range, but I measure my densities on my enlarger easel, so I care about those light levels around 1 to 100 lux

I must be interpreting what you're doing differently, since I can't see how you're able to achieve a result from a simple ratio, without knowing at least one density.

I use the density of 0.00 which is lamp with no density. Then we have the unknown density. The ratio of the resistances of these two densities, one known, the other unknown.

I don't know how linear my cds cell is going to be for densities greater than 2, since I don't have any known film density samples greater than 2.17.

You could put those two densities together and read through both. That would give around 3.0 if I remember correctly. For most photography other than alt printing, need for density reading above 2.0 not very useful.

The information I could find on the manufacturer's site we're useless in this regard. There's no curves supplied, so the only way to determine linearity is via experiment, which I've done.

What you've done looks great. Good job. I'm pretty sure you're doings things correctly but I'm not. Why and how my sensor and meter give nearly linear measurements is beyond me based on theory. Happy coincidences rarely occur. My sensor is a very poor one, and I believe my meter is also not good. My readings should be log, but their linear, or very close to linear.

Does your good meter give stable readings on MOhms/KOhms?

 

[Yezishu]

Well, the AMS sensors I had in mind are the fancier multi-spectral sensors like the AS7343 or the newer TCS3448. These can do a lot more than just color temperature. They just can't match the specific curves for Status A/M in ISO 5-3:2009. (and the difference between Status A and Status M, for each color, is relatively small)

I think there's a similar issue here to what discussed before.

If we're aiming for complete match with the ISO 5-3:2009 standard and at least tring to reach the current industry level, as the standard is designed for professional laboratories, a spectrometer is the bare minimum. Without more sampling points, it's impossible to discuss the small difference between the A and M states.

However, if the goal is simply to replicate a color densitometer actually sold decades ago (which, like CdS, not advanced but also isn't impractical), there seem to be many simpler ways. I noticed several methods mentioned in Glafkides' book:

(Behrendt densitometer: red Schott BG12, green Schott VG9, blue Schott RG2)//I think the red and blue colors reversed.

(Evans and Schneider densitometer, later the Kodak Model 17 analyzer, using three colored wedges that just use the dye in the target film. This did not meet international standards, but it was accurate in measuring how much dye was produced on the corresponding Kodak film.)

(Kodak's filter configuration for measuring the M state: red Wratten 92x2 + Corning 9780, green Wratten 52x3 + Corning 9780, blue Wratten 48Ax2 + Wratten 2A + Corning 4303)

Is it possible to achieve them easily (i.e., purchase the corresponding filters, I noticed that many of types still exist today.) or calibrate them using a spectrometer (more precisely, to know how different they are from industry standards today)? E.g, expense of Four TSL2591 sensors and three small pieces of Schott glass (RGBW) would still be less then 20$, within my "DIY" budget range.

 

[Alan Townsend]

No, I've never talked about power functions. It's plotting Log R vs Density. It's just like the D vs Log E curves we're all used to using. The X and Y scales are Logarithmic (Density is a log scale), but the plotted relationship is linear.

I can't get anything useful by taking the ratio of the resistance values. Can you provide some actual measurement data, and show how you're doing the calculations.

Here's some data I just read out from my setup. I used the four apertures my enlarger has that are factors of two. For readings I have the enlarger at f4.5, about half a stop more open than in the data. I take the log of the ratio of two densities, either 0 and an unknown in discussions here, or the two unknowns relative to each other and the zero density from the denser area when enlarging positives. Another point is I have not been subtracting out base plus fog for the most part unless mentioned.

CdS cell data 0203226

Data taken quickly going up and down through fstops


F-stop phone lux CdS cell Ohms

5.6 61 58 56 15.73k 16.6k 16.6k 16.71k
8 30 28 30 30.85k 31.39k 31.04k 32.12k
11 15 14 15 54.38k 56.8k 54.7k 57.4k
16 6 6 6 122.5k 122.0k 124.6k 123.6k

data in each column was taken about five seconds apart. Each column was started where the previous left off and increasing or reducing f stop the opposit direction. Note that the last phone values at f16 are off a fair amount, so adjust the CdS cell data accordingly. Note the highest impedence readings appear most repeatable not sure.

Unable to install open office to put this data in spreadsheet, something I haven't done in many years.

 

[Sharktooth]

Here's some data I just read out from my setup. I used the four apertures my enlarger has that are factors of two. For readings I have the enlarger at f4.5, about half a stop more open than in the data. I take the log of the ratio of two densities, either 0 and an unknown in discussions here, or the two unknowns relative to each other and the zero density from the denser area when enlarging positives. Another point is I have not been subtracting out base plus fog for the most part unless mentioned.

CdS cell data 0203226

Data taken quickly going up and down through fstops


F-stop phone lux CdS cell Ohms

5.6 61 58 56 15.73k 16.6k 16.6k 16.71k
8 30 28 30 30.85k 31.39k 31.04k 32.12k
11 15 14 15 54.38k 56.8k 54.7k 57.4k
16 6 6 6 122.5k 122.0k 124.6k 123.6k

data in each column was taken about five seconds apart. Each column was started where the previous left off and increasing or reducing f stop the opposit direction. Note that the last phone values at f16 are off a fair amount, so adjust the CdS cell data accordingly. Note the highest impedence readings appear most repeatable not sure.

Unable to install open office to put this data in spreadsheet, something I haven't done in many years.

Thanks for the feedback. I think I understand the gist of what you're doing. It seems that you're utilizing the measurements at different f stops to give you a known light differential, which is a good idea. Could you also provide a sample calculation. That would help to clarify what you're doing.

I did some experiment yesterday with my cds cell with an enlarger. I have an old Omega D something with a Chromega colour head. My step wedges are mounted on a card that fits in a 4x5 film holder. Unfortunately, the light levels from the enlarger at the closest setting to the base-board, are too low to get a linear relationship over the full density range (up to 2.17). It's linear up to about 20 KOhms, but the slope increases after that. I can separate it into two distinct linear ranges, and get excellent results, but then I'd need another high density point for calibration.

Intuitively, I think your approach is probably better, but I need to see how the calculation is done, so I can convince myself the math makes sense.

 

[Sharktooth]

O.K., here’s what I think you’re doing. Let me know if this matches.

Let’s say you have an unknown film density, call it Du. The “u” designation is for the unknown point you’re trying to evaluate

This produces a Resistance in the cds cell with a reference light source. Call it Rur

If you reduce the intensity of the light source by 1 stop, that’s equivalent to putting a 0.3 density filter over the light source. If you reduce the intensity of the light source by 2 stops, that’s the equivalent to putting a 0.6 density filter over the light source. …. and so on.

Let’s say we reduced the intensity of the light source by 2 stops, and then measure the Resistance on the cds cell. Call it Rutwo

If the relation between Log(R) and Density is linear, then the slope of the curve (rise over the run) can be calculated. In this case, the “run” is 0.6 density. The “rise” is the difference between the Log(R) values measured with the multimeter.

rise = Log(Rutwo) - Log(Rur)

Slope = rise/run = (Log(Rutwo) -Log(Rur))/0.6

from the Log identities, this can be converted to Slope = (Log(Rutwo/Rur))/0.6

If the curve is linear, then the slope will be the same between any two points on the curve.

We can now solve for the density difference between any two points.

Lets call the lowest density value Da (base + fog). The cds cell resistance with the reference light source will be Rar. The “a” designation indicates that it’s a low density point of the film.

Now consider the slope of the curve between these two points a and u

The “rise” is the difference in the Log(R) values. The rise = Log(Rur) - Log(Rar)

The “run” is the density difference. The run = Du - Da Let’s call this Du-a

The slope will be rise/run = (Log(Rur) - Log(Rar))/Du-a
This can be rearrange by the log identities to be Log(Rur/Rar)/Du-a

Since the slope is always the same, then we can combine the two equations


(Log(Rutwo/Rur))/0.6 = (Log(Rur/Rar))/Du-a


This can be rearranged to Du-a = 0.6 ( Log(Rur/Rar)/Log(Rutwo/Rur))

It’s the ratio of two Log ratios, multiplied by the density equivalent of the difference in stops of the light source.

Rur is the cds cell resistance of the unknown film with the reference light source

Rutwo is the cds cell resistance of the unknown film with the the reference light source reduced by 2 stops

Rar is the cds cell resistance of the lowest density point, with the reference light source.

Du-a is the density difference between the unknown film, and another point. If the other point is the zero density point (no film between the light source and the sensor), then Du-a is actually Du, or the actual density of the unknown piece of film.

Does this match what you’re doing, Alan?

 

[Sharktooth]

Actually, it would make more sense to just measure the resistance with no film over the sensor at two different light levels. Then you can calculate the unknown film density directly. By doing it this way, you only take one restance measurement with the unknown piece of film. Any other unknown piece of film can also be evaluated by just replacing the Ru value in the same formula.

Du = 0.6 (Log(Ru/Rz)/Log(Rztwo/Rz))

Where:
Du = density of the unknown piece of film

Ru = cds cell resistance of the unknown piece of film, under the normal light

Rz = cds cell resistance with no film (zero density), under the normal light

Rztwo = cds cell resistance with no film (zero density), under the normal light reduced by two stops

Note: The 0.6 factor at the start is based on using a two stop light difference for the resistance measurements with no film. If you want to use a three stop light difference, then you'd use a 0.9 factor. A two stop difference should be sufficient if the characteristic curve is linear.

 

[Alan Townsend]

Actually, it would make more sense to just measure the resistance with no film over the sensor at two different light levels. Then you can calculate the unknown film density directly. By doing it this way, you only take one restance measurement with the unknown piece of film. Any other unknown piece of film can also be evaluated by just replacing the Ru value in the same formula.

Du = 0.6 (Log(Ru/Rz)/Log(Rztwo/Rz))

Where:
Du = density of the unknown piece of film

Ru = cds cell resistance of the unknown piece of film, under the normal light

Rz = cds cell resistance with no film (zero density), under the normal light

Rztwo = cds cell resistance with no film (zero density), under the normal light reduced by two stops

Note: The 0.6 factor at the start is based on using a two stop light difference for the resistance measurements with no film. If you want to use a three stop light difference, then you'd use a 0.9 factor. A two stop difference should be sufficient if the characteristic curve is linear.

Thanks for the response. They way I've been measuring density is to measure the resistance through the unknown, and divide that by the resistance of the zero density when we have no density, then take the log of that ratio. When measuring the ratio of two densities, we do not know what either one is, only their ratio. Also, this does not include any density common to both, since this cancels out. I need my high density reading (shadow on positive, highlight on negative) to determine enlarger exposure time, and the ratio of highest to lowest to determine development time to control contrast for ortho litho or Xray film exposure.

This is a link to my photoresistor gm5539 on Amazon, currently $6 for 30:

https://www.amazon.com/dp/B01N7V536K?th=1

I'm attaching the spec sheet for this part. This shows a gamma of 0.8 typical for between 10 and 100 lux which is where I use mine reading low densities typice of my very low contrast positives that I print. Note that the part below the 5539 as a 10 MOhm Vdark and 0.9 gamma. My part has a measured Vdark at 10MOhm, so also likely has that gamma as well. This is very close to 1.

I've read densities as high as 2.92 with my setup, verified using my phone lux app, and within about 3%. This is when measuring Dmax for a given film, which I use as a process control to be sure my developers are working as expected prior to use. I use a high intensity desklamp a few inches above the phone for reading these higher densities.

Cheers

 

[Sharktooth]

O.K., I see exactly what you're doing now. Unfortunately, your ratio only works since the slope of the Log(R) vs D characteristic of the sensor is very close to 1. The slope characteristic of the sensor I'm using is around 0.65, so the simple ratio you're proposing doesn't work at all.

The good news is that the method I proposed will work with any sensor that has a linear relationship for Log(R) vs D. For the method I'm proposing, you only need one more measurement of resistance with a known lower light level, with no film between the light source and the sensor. You also don't need to know anything about the sensor, other than it has a linear relationship for Log(R) vs Log(Lux). Note: Log(Lux) is directly proportional to D, since D = Log(%Transmission).

For any of these methods you'll still need to verify linearity over the range you're trying to measure.