Silicon Solar Cell response.

[Chan Tran]

I bought a small solar cell and doing a test. I found that the shorted circuit current is proportional to light intensity. But if I want to use it as light meter I will have to convert the linear response to logarithmic response. Can use a log amp but that would make a lot of circuit.

 

[koraks]

There are generally speaking and as far as I can tell 3 options that would work - but only the actual log amp will result in a continuous, analog signal:

1: A log amp. Instead of making one from scratch, I'd opt for one of the integrated types available that use a minimum of external components and offer inherently better temperature compensation than whatever you can cook up at home. Several manufacturers still have a product offering to suit, sometimes even quite recently introduced types.

2: The crude digital approach would be to take an A/D converter of sufficiently high resolution to get the range you need. In practice this means 16 bit+ with overall excellent linearity, low noise etc.

3: A programmable gain linear amplifier will also work, but you need to have some kind of 'auto ranging' logic that determines which gain to set. Then feed the output into an ADC and do the math.

Evidently, 2 & 3 are only useful in systems where you need to have a digital signal. Although arguably, you could do the log & linearization part in the digital domain and then output the linear signal through a DAC (or PWM through an RC filter) to whatever circuitry that wants an analog input.

I'm not aware of any true analog alternative to #1. Log amp IC's are relatively expensive; a quick search suggests that something like TL441 is a reasonable compromise between cost, performance and availability.

Anyway, back to the main question - what kind of output do you need over what kind of range?

 

[bernard_L]

Be aware that there are two (at least) species of log amplifiers. Some are meant for compressing video signals; they accept bipolar input signals (how do they handle log(0)? see below plot for TL441) but have limited dynamic range (decades of input) unless several stages are properly combined. The TL441 belongs to that species.

Another species are DC log amplifiers. They work with a unipolar current source (good match for photocells), and deliver 5 or more decades. An example is the LOG114 from Texas. See below transfer curve.

There are many more aspects to be considered. Is the input zero-bias? Price, availability? Packaging: is it compatible with the soldering options available to you as a DIY-er?

Another issue is the spectral response. A Silicon "solar cell" has a response heavily weighted to red (and near-IR). I looked into this when trying to refurbish a Weston Master and found the most suitable corrective filter was 122 Fern Green from Lee. A color separation green might be better.

Concluding words. Tinkering is fun. But my daily use light meter is a Sekonic 308B that cost me 100€ at a photo fair. Light, compact, accurate, reliable.

 

[Chan Tran]

There are generally speaking and as far as I can tell 3 options that would work - but only the actual log amp will result in a continuous, analog signal:

1: A log amp. Instead of making one from scratch, I'd opt for one of the integrated types available that use a minimum of external components and offer inherently better temperature compensation than whatever you can cook up at home. Several manufacturers still have a product offering to suit, sometimes even quite recently introduced types.

2: The crude digital approach would be to take an A/D converter of sufficiently high resolution to get the range you need. In practice this means 16 bit+ with overall excellent linearity, low noise etc.

3: A programmable gain linear amplifier will also work, but you need to have some kind of 'auto ranging' logic that determines which gain to set. Then feed the output into an ADC and do the math.

Evidently, 2 & 3 are only useful in systems where you need to have a digital signal. Although arguably, you could do the log & linearization part in the digital domain and then output the linear signal through a DAC (or PWM through an RC filter) to whatever circuitry that wants an analog input.

I'm not aware of any true analog alternative to #1. Log amp IC's are relatively expensive; a quick search suggests that something like TL441 is a reasonable compromise between cost, performance and availability.

Anyway, back to the main question - what kind of output do you need over what kind of range?

I think it's a bad idea. My initial thought that I want the output to be from about LV15 down to something relatively dark like LV5 or something. I was thinking of using only a galvanometer with no power but a log amp does need power. Also if I measure voltage instead of current I found that the cell is very sensitive to the load impedance. I can have significant lower reading if the load is 50kohms vs 200kohms.

 

[koraks]

I think you very quickly come to the conclusion that you need a proper transimpedance log amplifier along the lines of what the LOG114 or LOG200 (see @bernard_L 's suggestion above) would make possible. I imagine that defeats the entire purpose of using a cheap & generic little solar cell and no external/additional power, as you might then just as well go through the motions of picking a suitable silicon photodiode etc.

In the end maybe the most sensible question is what the application will be and what the reasonable alternatives are - and whether you're looking for a project to break your teeth on vs. just get the job done. In case of the latter, I'd again side with @bernard_L and just use any old light meter that works and is available...

 

[hospadar]

@Chan Tran are you thinking to build a meter from scratch or are you replacing a selenium cell with silicon? I've modded a couple non-functioning selenium meters with silicon cells with good results so I don't see any reason why cell+galvanometer wouldn't work (that's all a selenium meter is usually, generally with a series resistor for calibration). Probably way easier in general to refurb an old selenium meter anyways since you get the meter movement and housing basically for free.

Not sure exactly how low mine will go but definitely meters down to EV 8ish (well lit interior) with reasonable accuracy. If you really want to get lower with good accuracy you probably need to go powered with a photodiode or CDS.

See over here for a recent job on the built-in meter on a contax: https://www.photrio.com/forum/threads/contax-iiia-meter-repair.213319/

 

[qqphot]

A time-honored way of converting logarithmic signals to linear display is to have the markings on the meter scale be log-spaced.

 

[NedL]

It's been a while since I played with using a solar cell as a detector, but when I did, I found the spectral response was uneven ( the one I had was highly sensitive to UV, which would be mostly blocked by a camera lens. It was also weakly sensitive to IR well below the range that film can cover... ). But the main problem was temperature sensitivity, which is why I gave up on the idea. The output of your cell will be dependent on ambient temperature.... What I was doing involved extremely weak signal detection, so it could be that these effects won't be strong enough to cause you trouble, I'm not sure. Have fun!

 

[Chan Tran]

I get it to work pretty well but then it changed. Somehow it becomes more sensitive than when I first got it. I will have to do test again and see.

 

[bernard_L]

Also if I measure voltage instead of current I found that the cell is very sensitive to the load impedance. I can have significant lower reading if the load is 50kohms vs 200kohms.

With ideal voltage measurement (very high impedance) the voltage goes like log of illumination. Helps achieving dynamic range. Several vintage lightmeters (Selenium) I've seen have a bespoke galvanometer with a variable gap between the pole pieces to achieve [deviation]\propto[\log(lux)]. That is not easy to replicate in the tinkering approach.

 

[reddesert]

I suspect that one could get close enough by approximating the logarithmic curve with some circuit that transitioned between a couple of different slopes at low and high light levels. Similar to how typical log-taper potentiometers aren't really exactly logarithmic. The circuits of older light meters are often very simple with few components, adjusted to get "close enough" over the likely range of application. For example the Yashica-mat circuit shown here: https://www.photrio.com/forum/threads/yashica-mat-124g-meter-ciruit-design-and-adjustment.153186/
I know this uses a CdS photoresistor, not a solar or selenium cell, but I think that's not exactly logarithmic either. It's trimmed using a combination of series/parallel resistance in high and low light so that it approximates the desired behavior.

 

[Chan Tran]

With ideal voltage measurement (very high impedance) the voltage goes like log of illumination. Helps achieving dynamic range. Several vintage lightmeters (Selenium) I've seen have a bespoke galvanometer with a variable gap between the pole pieces to achieve [deviation]\propto[\log(lux)]. That is not easy to replicate in the tinkering approach.

I will try a test with voltage measurement and see.

 

[Chan Tran]

I will try a test with voltage measurement and see.

Did a quick check. The cell is extremely sensitive to loading by the meter. I can see a different voltage reading when I switched between a 10 MegaOhms input impedance vs 4GigaOhms input impedance on the meter.

 

[bernard_L]

Did a quick check. The cell is extremely sensitive to loading by the meter. I can see a different voltage reading when I switched between a 10 MegaOhms input impedance vs 4GigaOhms input impedance on the meter.

I bet, at a quite low illumination level. Was that within your design target?

from about LV15 down to something relatively dark like LV5 or something

 

[Chan Tran]

I think it would work well enough (within 1/3 stop) if I use a digital mulitmeter and not an analog VOM and with a little chart to convert either voltage or current to LV. But that means I need batteries.

 

[Chan Tran]

I tried it with the little Triplett 310 VOM in voltage mode. I made the voltage measurements at different light level and then I did a linear regression with voltage and LV. It came out quite good. I think I can get it to within 1/3 stop.