Yes. That is shown on the plot in my post #5 aboveOn a log-log scale, it is nearly linear over some range.
No. See below.The conductance is linear on linear-linear scale.
Thank you; at least I'm not alone having... strong reservations.It's simply not accurate.
This!There is no doubling of anything with twice the light.
First off, what is a linear measurement M of light L? A device that delivers
M = a*L
where a is a constant. In native units of L versus M, the plot is a straight line through the origin.
The plot of conductance --of a CdS photoresistor-- is a straight line in log-log coordinates. This means
log(C) = p*log(L) + q
where p and q are constants; this is equivalent to:
C = L^p * 10^q where "^" has the usual meaning "raised to the power"
Unless p happens to be exactly 1, that relation is not linear. The plot of C versus L (just C and L, not logarithms) is not a straight line. For 0<p<1, it is curving with concavity down, and for p>1, it is curving up. No amount of dividing one value by anther one will change this.
This!
If conductance C (of the CdS photoresistor) were a linear measure of light L, it should double when light is doubled, as in opening the enlarger's diaphragm by one stop. If it does not, your basic assumption "The conductance is linear on linear-linear scale" is falsified. No pesky logarithms involved.
Nope. That already happened long ago.
Clip leads will work OK if both wires are near each other or twisted together. The meter is very high impedence on MOhm scale, so can be noisy. The resistance of a contact connection can be high enough to cause trouble also, so it's best to have soldered connections on both ends of the wires, but especially at the CdS cell end. I used a twin lead cable, and soldered on the CdS end, but have set screw connection on the meter end. This works OK, but sometimes I need to tighten the set screws a bit. Best would be a coaxial cable with soldered ends, but that would be more than a buck. Also, when the meter changes range to MOhm for KOhms, you will need to wait some time before the reading is stable. My meter is not wonderful reading high resistance, so I need to wait about 15 seconds for stable reading in MOhms.I think I'm going to try this tomorrow. I'm pretty sure I bought a couple of photo cells a while ago, but wasn't sure how to use them. I've got a decent multimeter, so it's worth a try. There's certainly a lot of potential there, and it's definitely cheap. Thanks!
We are measuring densities, not light intensities. CdS cells are plus or minus 10% devices at measuring light intensities. Densities are ratios of these intensity measurements, and as long as we make our two readings at almost the same time, these errors cancel out for the most part. Another point is for example, if one resistance is 10% high, the other 10% low when measuring a 3.00 density, the reading will be between 2.97 and 3.03, which is close enough to have value to a darkroom worker. That 0.03 error would be a 1% error in reading a 3.00 density.Yes. That is shown on the plot in my post #5 above
No. See below.
Thank you; at least I'm not alone having... strong reservations.
First off, what is a linear measurement M of light L? A device that delivers
M = a*L
I think you are over complicating things a bit. Sharktooth pointed out I was doing the same. CdS cells are not the most accurate light sensors in the world, but within limits, and being used for film photography, they are accurate enough to give very useful information.where a is a constant. In native units of L versus M, the plot is a straight line through the origin.
The plot of conductance --of a CdS photoresistor-- is a straight line in log-log coordinates. This means
log(C) = p*log(L) + q
where p and q are constants; this is equivalent to:
C = L^p * 10^q where "^" has the usual meaning "raised to the power"
Unless p happens to be exactly 1, that relation is not linear. The plot of C versus L (just C and L, not logarithms) is not a straight line. For 0<p<1, it is curving with concavity down, and for p>1, it is curving up. No amount of dividing one value by anther one will change this.
This is false. Sensors can have a variety of gains, including negative. I just measured my CdS cell, and with a doubling of light the resistance goes down approx. 1/2.1 per fstop over five fstops with no ND filters in place. This about a halving of resistance for a doubling of light, which shows the inverse relationship with resistance, and the almost doubling of conductance therefore. I have been claiming that measuring densities from 0.00 to 3.00 has about 1% error, meaning is "linear" within some range.
This!
If conductance C (of the CdS photoresistor) were a linear measure of light L, it should double when light is doubled, as in opening the enlarger's diaphragm by one stop. If it does not, your basic assumption "The conductance is linear on linear-linear scale" is falsified. No pesky logarithms involved.
As a photographer, I like CdS cells in my camera TTL meter and my hand held meter because they are accurate enough and very robust due to their non-electronic nature. Same in my darkroom Techno-nerds always want the latest technology. CdS cells are very slow in response and mostly green sensitive with very little blue sensitivity but they use extremely low power. Compare my old SRT101 Minolta to my later X700 and two X570's that both failed due to electronics. The early ttl meter cameras with mechanical everything else were extremely reliable. The later electronified models were extremely unreliable due to electronic failures. Without electronics, reliability goes way up. But with it, all kinds of other gadgets and gizmos can be thrown into the mix to confuse everybody about what they really need.If CdS cells were so awesome, why did the industry move to SPC as soon as they became available? Surely, those guys must have been on to something.
??Densities are ratios of these intensity measurements, and as long as we make our two readings at almost the same time, these errors cancel out for the most part.
Here is some reality. Not theory, not argument. Measurements.What the heck, this isn't about reality
The main thing that distorts CdS cell readings is temperature. The current going through the cell will cause it to heat up on higher light levels. That will give an error. The old meters using these cells, including the Science and Mechanics meter I showed, used a mercury cell, which gives 1.3 volts, and gives less heating than the DMM will, which likely uses a 5 volt reference. This will give almost twenty times the heating as 1.3 volts. If I watch my meter when the enlarger is turned on and full aperture, which is where I make all my measurements, the resistance will go up constantly at a slow rate. If I then put a dense negative on the sensor, the reading will go up to the higher readin quickly, but then gradually go down as it cools off with the lower current. This happens more quickly when used under an enlarged negative, since the sensor is open to air. So my normal technique was to turn everything on and read reference value first quickly before heating occurs. This is the sensor looking at the lens with no density. Then I place the dense material on the sensor and the autorange function kicks in and changes the read to the MOhm scale. I then need to wait about 15 seconds for cool down. A better way now, with the help of Sharktooth, is to read the low density first and to not wait very to do that, then remove the dense material and wait only about second before taking the reading. So I can put the resistances in the calculator directly then press LOG. This method gives most accuracy. The way you measured, those conductances relate to much higher current levels than I see, so you would need to quickly make the measure, than turn meter off for cool down, or use a small fan to help cool the sensor.??
Here is some reality. Not theory, not argument. Measurements.
From you charts, your enlarger puts out a lot more light than mine, and the CdS cell is much larger and lower impedance than mine? Also, regression lines are normally drawn to minimize errors, not maximize them. I could easy draw lines through those points that show less error from linearity. Oh, I forgot this is reality and not theory. Yes I would, I have, and will continue to use such a sensor. Also more correct to have increasing values on X-axis, not reducing values.Below: blue line and dots: measurements. Red line: linear extrapolation from the values at low light.
Would you measure your negatives with such a sensor?
Looks to me like someone trying to prove a point using bad science. The second largest source of error for CdS cells is the so called memory effect. Both the very large heating effect and the large memory effect are the same or very similar at a given time, so tend to cancel out in a ratio. It is a bit of a kluge using the DMM more than the CdS cell, but with care it works well enough.Note: I did this for the sake of accuracy of information on the forum. Now anybody can have his/her own opinion. I'm done with this topic.
Plot for the CdS cell from the Philips device.
Plot for the CdS cell from the Ilford device. I forgot to measure the last point.
Thanks for the link. That appears to be a lux meter circuit for a phone. I just got my phone out, and used my lux meter app again to verify my densitometer. I have done this before, and it works to just above a 2.00 density. My enlarger baseboard is only 100 lux, so not enough.Assume the plot on your post #5 is absolutely correct one can devise a formula to calculate the density with the resistance reading but I have played with CdS cell a lot and I found their repeatability is very poor. I had a decent densitometer to compare (X-Rite 810). It's not even good enough for +/-0.20 density tolerance. Something like this would do much better as it already has log output.
Light Sensor 70000 lux - 1143_0 - Phidgets
Measures light intensities of up to 70 kilolux on a logarithmic scale and connects to an Analog Input or VINT Hub port.www.phidgets.com
Thanks for the link. That appears to be a lux meter circuit for a phone. I just got my phone out, and used my lux meter app again to verify my densitometer. I have done this before, and it works to just above a 2.00 density. My enlarger baseboard is only 100 lux, so not enough.
A density I measured with my CdS densitometer was 2.20 using it, and it measures 2.19-2.32 with my phone app. The accuracy is proven again. We need to measure densities, not light intensities to determine accuracy for these. So the smartphone is even less than a buck, since I already had it and the app is free, but it can measure density ranges up to 2.2 with my basement clamp lamp I have. The readings do jump as I indicated, but would be better with a brighter desk lamp.
I just measured again with a much brighter desk lamp and got 2.08 with very stable readings. That's about a 5% difference.
The resistance values changed with differing light levels, so you don't need any special electronic circuit to make this useful.
Density=log(r1/r2) where r1>r2. For phone lux meter Density=log(Lux1/Lux1) where Lux1 is reading the lamp without the density, Lux2 is reading through the density with same lamp.What is your formula to convert your resistance reading into density?
I found a cds cell I bought at an electronics store a year ago for a few bucks. It was in a little plastic baggie, with a sticker that said 5K to 500K, and 100V. I clipped the two wire leads to my multimeter for some seat-o-the-pants testing. Here's what I found.
The resistance values changed with differing light levels, so you don't need any special electronic circuit to make this useful. I next covered the sensor with pieces of film of known density values, and measured the resistance. I took the log of the resistance ratio, and compared that to the density difference of the two pieces of film. The result was nowhere close, which implies that the response is non-linear. That was disappointing, but it just means that it's not going to be as easy as pie.
I could plot out a characteristic curve for the sensor so I could then use that to determine light values from resistance measurements. I'd need an accurate light meter to do that, which I have, but I could just use that to do the measurements and not need the cds cell.
It's still neat that you can get varying resistance values with varying light levels using a cheap sensor and a multimeter. There's lots of potential fun there, but I can't think of a good use for it at the moment.
That's a problem if you want a meaningful density or illumination range.10-bit ADC
That's a pretty low impedance and range CdS cell. The heating by current issue will happen, so more care needed. With meter off, put density on cell, then turn meter on and read withing a second or two. Then remove the density and read again without delay. Turn meter off and leave it off unless making a reading. The meter current will cause the sensor to heat up slightly and increase the resistance.I found a cds cell I bought at an electronics store a year ago for a few bucks. It was in a little plastic baggie, with a sticker that said 5K to 500K, and 100V. I clipped the two wire leads to my multimeter for some seat-o-the-pants testing. Here's what I found.
You should do it the other way to reduce heating effects. All room light must be turned off. Use only the meter internal lamp to illuminate meter scale if it has one. Have the mater turned off for some period of time first so there is no meter current in the cell to allow cool down. Put density on cell, turn meter on and read that immediately, remove densiity and read meter again with no delay.The resistance values changed with differing light levels, so you don't need any special electronic circuit to make this useful. I next covered the sensor with pieces of film of known density values, and measured the resistance.
That's a problem if you want a meaningful density or illumination range.
on paper, but getting 24 usable bits of resolution is another matter. 16 bits is challenging enough already, and that's still a limited range unless you do the log stuff in analog circuitry. But that opens up a whole new can of worms...
I think the idea behind this thread is not so much to mimic a light meter, but a densitometer. The challenge here is the large dynamic range required which typically needs to cover a logD 4.0 range (although arguably logD 3.0 would be fine for all intents and purposes).I think the goal is simply to match the performance of those old analog light meters.
That helps a little, but the gain range is in reality quite limited in the light of what you typically need/want in a densitometer. I think you pretty quickly end up realizing you need to fashion some kind of custom PGA that offers 3 or so major steps; instead of 1-8x as typically offered by the PGA inside a ADC, something like 1x-64x-1024x or so. That's not necessarily very complicated and ultra low noise isn't even required as long as linearity is OK. But it brings us very quickly very far away from "CdS cell plus multimeter makes $1 densitometer" idea.Notable examples of these PGA-integrated ADCs include the MCP3421, and the classic ADS1115.
Yeah, that would be my thought as well.But if goes this far, use an SPD would be better in performance.
I think the idea behind this thread is not so much to mimic a light meter, but a densitometer. The challenge here is the large dynamic range required which typically needs to cover a logD 4.0 range (although arguably logD 3.0 would be fine for all intents and purposes).
That helps a little, but the gain range is in reality quite limited in the light of what you typically need/want in a densitometer. I think you pretty quickly end up realizing you need to fashion some kind of custom PGA that offers 3 or so major steps; instead of 1-8x as typically offered by the PGA inside a ADC, something like 1x-64x-1024x or so. That's not necessarily very complicated and ultra low noise isn't even required as long as linearity is OK. But it brings us very quickly very far away from "CdS cell plus multimeter makes $1 densitometer" idea.
Also, while the PGA-based concept works OK on paper, things tend to get a little tricky if you start implementing it and you want it to be consistent. How to deal with the signal ranges where the gain settings overlap? I.e. how to avoid nasty 'jumps' in response as you switch between gain settings?
Like so often, the concept is easy - the tricky bit is in getting it to work elegantly.
Yeah, that would be my thought as well.
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?