Photosites have capacitors...OK

[Abitconfused]

So I have read that photosites have capacitors. If so, they must be the tiniest capacitors imaginable. These capacitors discharge their data on "demand" but the timing must be controlled by the ISO setting to some extent...right?

[george013]

So I have read that photosites have capacitors. If so, they must be the tiniest capacitors imaginable. These capacitors discharge their data on "demand" but the timing must be controlled by the ISO setting to some extent...right?

No.
The readout of the sensor is analogue. It's measured in Volts or Ampere. That value is digitized. And here is the magic. If you want to digitize a value you must specify 2 values: the analogue range and the amount of steps. In base iso the analogue range is the acceptable maximum analogue value of the sensel. That value is divided in 2^12 or 2^14 parts, depending on camera and/or setting. When shooting jpg this will be transformed to 2^8.
When you change your iso, the analogue range is changed. If you half that range it will look as the sensor is more sensitive. It isn't. When dividing the input range by x but not the output range, your monitor or whatever, you get a result as if the sensor is more sensitive with a factor x.

That's how I understood.


George

[dje]

So I have read that photosites have capacitors. If so, they must be the tiniest capacitors imaginable. These capacitors discharge their data on "demand" but the timing must be controlled by the ISO setting to some extent...right?

Yes the capacitance of a photosite plays an integral part in the operation of a CMOS sensor pixel. This capacitance is very small, of the order of FemtoFarads, a fF being 1F x 10^(-15) or 1 pF x 10^(-3). The basic operation involves applying a reset voltage to the capacitor and then as the pixel is exposed to light and charge is developed, the voltage across the capacitor reduces towards a minimum voltage which corresponds to the full well point. As George says, this is all analogue at this point.

The ISO sensitivity or base ISO of the sensor is basically related to how much sensor exposure is required to reach full well capacity. For ISO settings above base ISO, the exposure of the sensor is reduced and the sensor doesn't reach full well capacity.

Dave

[Manfred M]

The photosites themselves consist of both a photodiode to capture the light and a very basic amplifier (which will include all of the standard circuit elements of transistors, capacitors and resistors) required to provide the gain for the additionaly circuitry on the sensor chip to manage the data. This is all analogue data.

The sensor has a single "ISO" sensitivity. Anything that we view as ISO is amplification (gain) of the data captured by sensor.

[xpatUSA]

So I have read that photosites have capacitors.

Not quite, Ed. A photo-diode has the property of capacitance, not "a capacitor". The amount of capacitance varies with the fixed 'bias' or 'reset' voltage. The only photo-sensor that I know of with actual additional capacitors is by Aptina in pursuit of more DR, IIRC.

http://www.photonstophotos.net/Aptin...WhitePaper.pdf

If so, they must be the tiniest capacitors imaginable. These capacitors discharge their data on "demand" but the timing must be controlled by the ISO setting to some extent...right?

Again, they are NOT "capacitors" per se, they are part of the diode, known as the 'depletion region'. Each side of that region is analogous but not equal to a capacitor plate.

Page one here says it all:

http://www.osioptoelectronics.com/application-notes/an-photodiode-parameters-characteristics.pdf

As to the effect of the ISO setting, I'll leave that to others.

George and Dave both answered you very well, I reckon.

[xpatUSA]

. . . The basic operation involves applying a reset voltage to the capacitor and then as the pixel is exposed to light and charge is developed, the voltage across the capacitor reduces towards a minimum voltage which corresponds to the full well point.

Well said, Dave, not many people know that. Only found it out myself a few months ago. Was quite shocked, as matter of fact.

One could almost say "as charge is dissipated" and "well empty", ho hum.

It's a pity that photography continues with the bucket "filling up with ping-pong balls" analogy which effectively prevents anyone new from understanding how it really works!

[Manfred M]

It's a pity that photography continues with the bucket "filling up with ping-pong balls" analogy which effectively prevents anyone new from understanding how it really works!

Yes, capturing and holding a photon in a container is implausible, so the analogy is not a particularly good one. However the term "well capacity" would tend to lead the casual observer to suspect that this is what is happening.

I had not realized that the capacitor discharge is how this is measured, but that approach certainly makes sense. I had suspected it was the other way around, i.e. a charge was built up

[dje]

Well said, Dave, not many people know that. Only found it out myself a few months ago. Was quite shocked, as matter of fact.

One could almost say "as charge is dissipated" and "well empty", ho hum.

It's a pity that photography continues with the bucket "filling up with ping-pong balls" analogy which effectively prevents anyone new from understanding how it really works!

and

I had not realized that the capacitor discharge is how this is measured, but that approach certainly makes sense. I had suspected it was the other way around, i.e. a charge was built up

Yes it took me a fair bit of digging to wake up to this. I think this "capacitance discharge" process came in with 3T and 4T APS (Active Pixel Sensor) technology. Before that with single gate transistor CMOS (Passive Pixel Sensor) and also probably CCD, I think it was a charge accumulation process but I haven't delved into those processes much. With APS technology, the "Full Well Point" is determined by the range over which the charge - voltage transfer curve is fairly linear.

Dave

[Abitconfused]

So setting the ISO is like the teacher who graded on a curve. The highest test score was 50% (like an ISO of 600 I imagine) so that became the new “A”.

[Manfred M]

So setting the ISO is like the teacher who graded on a curve. The highest test score was 50% (like an ISO of 600 I imagine) so that became the new “A”.

I don't think the analogy works.

Think of ISO more like the volume control of your television, radio, phone, etc. with the volume turned down. You might just barely hear it in a very quiet room. The moment you step out of that extremely quiet room, you end up having to turn up the volume, dependent on how noisy things are; the noisier your environment, the higher you have to set the volume in order to hear it reasonably well. Exactly the same principle applies with ISO; you are just "turning up the volume". Turn the volume up really loud, and you will start noticing distortion and that is somewhat consistent with digitial noise getting to be too high to be useful.

[george013]

So setting the ISO is like the teacher who graded on a curve. The highest test score was 50% (like an ISO of 600 I imagine) so that became the new “A”.

Maybe this will help.

Light is hitting the sensor. The amount of light is catched as an analogue signal. That signal is digitized in the A/D converter. Calculations in the pc are done on that digital values. When send to the monitor that signal has to be converted to an analogue signal. Where this exactly happens is for now unimportant.

As mentioned before a lot of magic is in the A/D converter. In the drawings below the sizes of the sensor and the monitor are representing the dynamic range. In base iso the lightest part of the sensor corresponds with the lightest part of the monitor. The same for the darkest parts. The A/D converter assigns the value 255 to the brightest part of the sensor and the D/A converter to the brightest part of the monitor/


When it's dark in drawing 2, the maximum capacity of the sensor is not used, let's say only half. That part of the sensor will produce a digital value of let's say 128. And so it will be sent to the monitor by the D/A converter: halfway.

But in the third drawing we used a higher iso. As mentioned before, in a A/D conversion we need to declare a range and an amount of parts in which we divide it. Both in drawing 1 and 2 that range is the maximum sensor range and the amount is 255. But in drawing 3 we changed that range, it's only half of the original. But we keep that division in 255 steps. Now we see that that part half way the sensor corresponds with the lightest part of the monitor. This is the situation of double the base iso.

That's how I understood it.

George

[Manfred M]

George:

In your first diagram, you list the computer monitor as an analogue device. Every one of the "modern" flat screen style I have owned was described by the manufacturer as either an 8-bit or 10-bit display in the spec sheet. There is nothing analogue about that. The old CRT (cathode ray tube / picture tube) computer screens were analogue.

In the second diagram, #3, you show the output from sensor to screen as a sloping line. In the digital world, that would be a step-function, not a constant slope output.

[xpatUSA]

We could also regard the sensor as digital. When a photon is captured, the photo-diode output changes by a fixed amount per capture (electron), e.g. 9uV. That amount can not be more and it can not be less: therefore the sensor output can not be analog because it changes by steps.

[Manfred M]

We could also regard the sensor as digital. When a photon is captured, the photo-diode output changes by a fixed amount per capture (electron), e.g. 9uV. That amount can not be more and it can not be less: therefore the sensor output can not be analog because it changes by steps.

In that case we could also look at film as "digital". When a sufficient number of photons hit a silver salt crystal, it would would undergo a photochemical change; i.e. the "latent image". Areas that did not, there would be no change in the silver halide structure. A case where the development process would turn these crystals into silver metal.

So the process is "digital" too. Metallic silver areas and non-metallic silver areas that were removed in the fixing process.

[george013]

George:

In your first diagram, you list the computer monitor as an analogue device. Every one of the "modern" flat screen style I have owned was described by the manufacturer as either an 8-bit or 10-bit display in the spec sheet. There is nothing analogue about that. The old CRT (cathode ray tube / picture tube) computer screens were analogue..

The monitor emits light and that's analogue, just as the sun or your lamps in the kitchen. What you're referring to is the communication to the pc. That's why I wrote

When send to the monitor that signal has to be converted to an analogue signal. Where this exactly happens is for now unimportant.

In the second diagram, #3, you show the output from sensor to screen as a sloping line. In the digital world, that would be a step-function, not a constant slope output.

To digitize a signal you need a range and an amount of steps or bitdepth. In figure 3 the range is halved but not the steps. If you send that digital signal to the monitor it acts as if it's multiplied.

George

[xpatUSA]

The monitor emits light and that's analogue, just as the sun or your lamps in the kitchen. What Manfred is referring to is the communication to the pc.

I think not. My "modern flat screen style" monitor is 8-bit inside but my PC connection is 0-5V which is analog in the sense of this thread. So, Manfred could not have been talking about the PC connection.

[dje]

So setting the ISO is like the teacher who graded on a curve. The highest test score was 50% (like an ISO of 600 I imagine) so that became the new “A”.

Ed I'm not sure what you mean by that but increasing the ISO setting puts extra amplification on the signal coming out of the sensor. This means that you need less exposure of the sensor to get the same digital output. This is useful to get a faster shutter speed or smaller aperture but it has a downside - degraded noise performance, specifically lower dynamic range and lower S/N ratio. The trick is to know what your camera is capable of ie what is the highest ISO setting that gives manageable noise performance.

Dave

[xpatUSA]

Originally Posted by Abitconfused
So setting the ISO is like the teacher who graded on a curve. The highest test score was 50% (like an ISO of 600 I imagine) so that became the new “A”.

Ed, I'm not sure what you mean by that
Dave

Me neither, Dave.

but increasing the ISO setting puts extra amplification on the signal coming out of the sensor.

By coincidence, this past week I bought a Sigma SD15 DSLR, my first that has an AFE which works like the above. I'll compensate for that by always using 100 ISO.

All my others to date have been truly ISO-less, doing the amplification digitally during conversion to RGB.

[dje]

By coincidence, this past week I bought a Sigma SD15 DSLR, my first that has an AFE which works like the above. I'll compensate for that by always using 100 ISO.

All my others to date have been truly ISO-less, doing the amplification digitally during conversion to RGB.

Yes Ted I think there are quite a few cameras that use “digital amplication“ these days, at least above a certain ISO setting eg ISO 1600

Dave

[xpatUSA]

Yes Ted, I think there are quite a few cameras that use “digital amplication“ these days, at least above a certain ISO setting eg ISO 1600.

Dave

Indeed there are, Dave.

However, from the beginning, c. 2002, most Sigmas are a bit different in that regard - their whole range of ISO settings, not just above 1600, are done digitally in the converter. I find it quite refreshing to look at a high-ISO raw histogram and observe how badly under-exposed the sensor was.

OTOH, all their Quattro models, plus the DP2x and SD15, have AFE's like "normal" cameras.

Sigma :- "Consistency 'R' Us"