digital camera ISO setting

same as the film. it changes the sensitivity of the capture medium. on digital as the iso goes up up so does the picture noise (looks much like the same effect as film but film is caused by grain size)

Changing the ISO also changes the dynamics of the focus (lens aperture/speed) which changes focus depth. A higher ISO allows smaller aperture settings for a given speed, and a deeper focus depth of field.

--
I'm never going to grow up.

e

With all due respect...This is electronics discussion group.... I am not asking about photography issues (aperture/speed/focus depth, etc). I have an idea how CCD works (interfaced Sony ilx553b) - do not know how to change CCD sensor "ISO" by factor of, say 64 (from 100 to

6400). Never dealt with CMOS sensors... Does anybody know what happens inside of digital camera body when ISO setting is changed?

Changing ISO is changing the gain of the amplifier,

The higher ISO puts the amplifier in higher gain, hence you see more noise

CCDs and CMOS imagers have a certain noise floor, usually a few electrons RMS per pixel. (The last one I looked at was around 5.) They also have a maximum number of electrons that each pixel can hold, called the _full_well_charge_, which goes roughly as the pixel area, and is near 50k electrons/pixel for camcorder type sensors. (Antiblooming, also called inverted mode operation, gets rid of a lot of nasty artifacts from overexposed areas at the cost of about half the full well capacity.)

Since photographic film isn't nearly as linear as a silicon image sensor, the sensitivity comparison is a bit fraught, but the usually quoted number is that a frame-transfer CCD is roughly equivalent to film at ASA 400. (Since nobody cares about film nowadays, the ISO police will probably let me slide on that one.)

Since the SNR is equal to the RMS number of photoelectrons per pixel(*), if you reduce the exposure, you get a lower SNR, as David said.

Cheers

Phil Hobbs

(*) Since the noise photoelectrons are stochastic as well, it should really be N**2/(N+5), but the two are very close in the brighter areas.

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

e

hey

snip

really only 5e- rms?

I've used some CCD's for xrf and to get those down to that level, you'd have run them slow and cold

-Lasse

Must be speed of scan and integration times.

Greg

e

Wrong. Changing ISO does NOT affect the focus or the focal length of the lens. Therefore, ISO does not change depth of field.

That said, adjusting ISO does alter the latitude one has to obtain a given exposure - and THAT is what affects the depth of field. I think that's what you meant to say, but somehow ended up saying something else. (?)

For example, bumping the ISO up one or two stops would allow a corresponding reduction in aperture for a given exposure and shutter speed. Or a reduction in shutter speed for a given aperture, etc... (You can play with these three variables all day. Four if you want to include flash.) But ISO by itself is only going to affect the gain of the sensor (not enough gain =3D black, too much =3D white image). But whatever that image happens to be (black, white or preferably in- between) for the ISO you've set, it has an unchanging depth of field.

Linear gain.

The analog amplifier gain between the sensor and the ADC.

Increasing the gain and the Poisson noise will become more noticeable.

Essentially the gain of the readout amplifier and/or the scaling of the raw CCD data during post processing.

Regards, Martin Brown

Nope. Changing the aperture changes the depth of field period. You can compensate for lower iso with longer exposure - they are separate effects. The photographer makes the decision as to what combination is "best" but the actual affects of changing depth of field and iso are independent

--
We have failed to address the fundamental truth that infinite growth is 
impossible in a finite world.

Go away, then... You started a photography thread then complain about the replies...

--
I'm never going to grow up.

Speed of scan? Unlikely. Number electrons in the well (pixel) does not depend how fast it's drained. Integration times? Unlikely. The pixel is exposed to the light for given amount of time only once. ... Phil Hobbs mentioned pixel well size of ~50k pixels and noise floor of about 5 pixels. This is the kind of info I was looking for, thank you Phil. It means that by cranking ISO from 100 to 6400 one changes SNR from ~10k to ~150, doesn't it? Doesn't one need cool the sensor to get that low noise floor? Nikon says nothing about waiting for sensor to cool down .... I highly doubt they cool camera sensors (at least the ones that costs less than 5 digits).

s/pixels/electrons/ !

The noise has typically two or three components - unavoidable readout noise, stray IR photons from the readout electronics, and thermal noise.

The first is ever present and so you choose your exposure to maximise available signal to noise without overflowing wells. The second could be annoying in very early CCDs with long exposures not so much now. The last becomes a real nuisance on very long time exposures which is when you start to get into active cooling. Active Peltier cooled CCD sensors are common now in amateur astronomy and all professional astronomy gear.

For short exposures that do not use the full dynamic range of the sensor the extra gain makes the noise floor more obvious and so some cameras will bin 4x4 or apply noise reduction filtering to the output.

Four digits will get you a cooled CCD low light camera eg.

A few years back some of the Sony chips were exceptionally quiet even at room temperature. I don't know if that is still the case.

Most practitioners have given up on film ISO ratings for CCDs in favour of metrics more suited to the medium like QE vs wavelength response.

Regards, Martin Brown

It depends on the well size. Yours probably had 10x the full well capacity of a commodity CCD, which means roughly 1/10 the gate voltage per electron. (Disclaimer: I've never actually built a CCD camera, I'm just going from Janesick et al.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

Both the readout noise and the photoelectrons are stochastic, so in terms of electrical power, if a given pixel has a mean of N photoelectrons and M noise electrons per frame, the SNR is

SNR = M**2/(M+N).

It gets better in the brighter areas and worse in the dimmer ones, so unless the scene has very low contrast, you get a somewhat more complicated situation than this. Photo quality is very often limited by detail in the shadow regions, where the readout noise is proportionally more important.

You can get cooled CCD cameras for much less than that--see the Astrovid StellaCam III for instance. (

) Of course that one's monochrome, because if you care about sensitivity, b&w is the way to go.

Cooling is not at all easy in a consumer device. It eats power, requires a fan or water cooling to get rid of the waste heat, takes at least tens of seconds to equilibrate, and worst of all, is likely to lead to water condensing inside the camera.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

n

ike

=A0They

ed

ll

it was something like 500ke-/pixel and 1.5uV/e-

I have, kinda, it didn't really take pictures it just captured frames and then made a histogram of single pixel values

-Lasse

u y 00) =A0Of course that one's

I am surprised that

is so inexpensive... How many do they have to sell to be profitable? It's a rhetorical question ;o)