Is the term ISO “totally fake”?

[dtaylor]

I found in the film days that the native noise in the drum / film / slide scanner has a lot to do with the images final noise. Noise interference patterns occur between the film grain particle size and the CCD pattern noise in the scanner. Iso 200 Fuji Chrome had more noise than 800 iso print scans on my 2700 dpi scanner (approx 11.9 mp equivalent). Iso 50 and 100 looked way better...as did 400 iso. Certain brands of film at certain iso's seems to create mega noise due to the interference patterns. A friend of mine used a slightly higher resolution 4000dpi scanner only to find his beloved Velvia 50 scans were riddled with high iso noise and were unusable. Were as scans from my machine were far smoother at the lower native dpi. So yes...iso noise is a thing.

This used to be referred to as grain aliasing. (http://www.photoscientia.co.uk/Grain.htm) It was less about CCD noise and more about grain size versus scan resolution. But, yeah, certain films on certain scanners would have terrible grain. I'm surprised to hear about Velvia 50 because that always seemed to scan well for me at 2700 dpi or 4000 dpi.

That said, even compared to darkroom prints digital has a huge ISO advantage. ISO 3,200 shots from current APS-C sensors have a noise pattern comparable to tight grained ISO 400 print film. FF is comparable at ISO 12,800. I remember not wanting to shoot film higher than ISO 400 in 35mm format, but I have no reservation about pushing to 12,800 with digital FF.

I used to wonder if anyone would develop film profiles (like the digital modelling in guitar effects). But the reality is that most modern sensors far far exceed the dynamic range, contrast range and noise levels available on slide medium.

Yep. I do miss the color rendering of certain films, and the pre-canned film plugins never seem to get them quite right. But the flexibility I have in ACR/PS/LR to produce the colors and look I want far exceeds anything I ever had with film.

 

[BeenThere]

Yes, and film was also marketed with higher ISO ratings, but was the same exact film but with longer processing times. A similar bunch of comments were made.

Not really. Larger silver Halide grains were used for faster film. This made each grain more sensitive to light and also gave the film a more grainy appearance. The user could boost a low iso film to an apparently higher value by boosting development time, but this would also change the film gamma.

 

[AlanF]

It depends on the camera, 5D Mark IV and 80D are decent, I didn't realize how clean they were. But the 7D Mark II, 6D Mark II, 5DS R are not ISO invariant until at least after 3200. The 6400 number is the one that pops into my head when playing a lot with dual ISO on Magic Lantern, 6400 ISO definitely applies for the 6D and 5D Mark III. This is based on my observations of the noise that appears when bumping the exposures.

I routinely used iso640 on my 7DII and now iso640 on my 5DSR as it is within the isoinvariant region and gives me a fast enough shutter speed for much of my work. I have replotted the data from photonstophotos and you can see that iso640 is in the linear region (from about iso500 for the 5DSR and lower from the 7DII.

 

[LSXPhotog]

Tony Northrup is the YouTube version of Ken Rockwell. He says just enough that is actually correct to make the rest of what he says dangerous.

He's in it for the clicks (money), and he'll say whatever he thinks will get the most clicks. Everyone in any kind of media today (and these days, YouTube is about as financially viable for many folks as what we call the "traditional media") follows the model of Howard Stern: It doesn't matter if they like or hate the way you are constantly stirring the pot, as long as they are talking about you and listening/watching you.

Thank you for speaking my mind. When I was first getting started in the publishing industry I knew a staff photographer just like him. That guy knew -or felt he knew- a lot about photography and would talk your ear off about it. We actually had a handful of good conversations about it over the years. However, he ultimately lost his job to me because he couldn't actually take photos with a camera. He just knew the technical aspects of it but editors would constantly complain about the images he delivered to them. (He had several...MANY reshoots over his tenure there.)

Focusing too much on the technical side of photography can't be fun. I have seen plenty of photos the Northrups share on their channel to know that they are THAT guy I worked with. They know a lot about the technical side of photography - enough to be dangerous - but they may lack the creativity and ability to be successful photographers beyond the stock photography business they had.

 

[magarity]

Just came across this by jrista, who used to make great contributions here (https://photo.stackexchange.com/questions/25917/why-is-sensor-sensitivity-called-iso)
"Just a note. When it comes to digital sensor "sensitivity", the term sensitivity in that context is actually a bit of a misnomer. A digital sensor is a fixed, linear, analog device. It always has the same real sensitivity. When you adjust the ISO setting to a higher level, all that really does is reduce the maximum saturation point. The sensor does not detect more light...it detects the same, so its still just as "sensitive". Its just that instead of pure white occurring at say 40,000 electrons in a pixel (ISO 100), it occurrs at 20,000 electrons (ISO 200), or 10,000 electrons (ISO 400), etc " jrista

A CCD image sensor is an analog device (original 1D, and the old Kodak Canons) and have analog to digital converters. A CMOS image sensor is a digital device (any Canon since the 1D). Amusingly enough CMOS digital goes through a digital to analog conversion and then converted back to digital again later, but they're not analog devices themselves.

 

[Jack Douglas]

Hmm, I thought light was an analog quantity.

Jack

 

[Kit.]

A CCD image sensor is an analog device...
A CMOS image sensor is a digital device...

What do you mean by that?

 

[3kramd5]

What do you mean by that?

I wondered the same.

 

[BeenThere]

Hmm, I thought light was an analog quantity.

Jack

Count the digital photons.

 

[knight427]

Count the digital photons.

Only after the wavelengths have been sorted through the filter array.

What do you mean by that?

I don't get what magarity is saying either. I believe an imagining sensor is simply an array of transducers, transforming light waves into proportional electronic waves (much like a microphone transduces pressure waves into electronic waves). I dunno, maybe the audio analogy doesn’t hold up.

 

[magarity]

What do you mean by that?

Because a CMOS device converts each pixel's output to digital immediately whereas a CCD keeps the pixels' signals analog internally for quite a while (in relative terms). 'Digital' or 'analog' when talking about sensors refer to these internal workings, of which there is a lot in the sensor as a whole in addition to just the light sensitive pixels.

 

[Andy Westwood]

I also watched this on You Tube not long after it was posted, being a studio photographer, I could never work out when I changed camera bodies (All Canon) I would get different exposures body to body using the exact lens and the same settings on my camera and lights.

This explains what could be the only reason for this, I'd never thought of this until I watched Tony's video on the subject, interesting but sad ISO can't be standardised across all cameras.

I like to nail my exposure in camera as much as possible, and it's easy to do that in a controlled environment such as a studio, but it would always take me a week or so using a different camera body, now I know Why!

 

[Quarkcharmed]

Because a CMOS device converts each pixel's output to digital immediately

Nope, it doesn't. More precisely it does in CMOS with so called digital pixels but they're not used in the cameras we compare here.

 

[3kramd5]

Because a CMOS device converts each pixel's output to digital immediately whereas a CCD keeps the pixels' signals analog internally for quite a while (in relative terms). 'Digital' or 'analog' when talking about sensors refer to these internal workings, of which there is a lot in the sensor as a whole in addition to just the light sensitive pixels.

They both generate charge from incident light. They use different means to convert that charge to potential, but both then digitize using ADC.

 

[3kramd5]

Armature disclaimer (you get what you pay for):

1- ISO in-variance still hits the wall of clipping. One you've filled your light bucket, there is no way to recover that information.

Right, but the point those who invented the term “ISO invariant” would make is: the noise is low enough that you can underexpose to prevent clipping highlights, and then selectively brighten only darker portions, recovering a wider range than the same image shot at a higher ISO but equivalent exposure.

Of course:
1) tonality suffers. Yes you can brighten more without noise becoming destructive as soon, but little detail is recorded. It looks flat.

2) returns are diminishing; noise is already fairly low. There is more to be gained by increasing well capacity, whether physically or computationally (e.g., modulo cameras).

3) none of the advantages apply if the scene doesn’t exceed the single exposure limitations.

 

[epsiloneri]

I agree that the non-standard ISO numbers used by camera manufacturers are bad. Yes, "gain" would have been an appropriate label, but I can see the use of a standardised ISO - non-standardised ISO just gets confusing.

About the other point, that the ISO setting is just digital gain, I highly doubt that as it would indeed be useless and only result in a loss of dynamic range. I've read about the tricks made by manufacturers to fake sub-stop ISO increases by adjusting the digital gain, but data (as shown by e.g. AlanF above) contradicts the ISO setting in general being a purely digital gain setting. I agree that the gain setting is of minor importance relative to other exposure decisions (focal length, aperture, and time) but to say that it is meaningless is just wrong.

Increasing the gain has two positive effects: improving the digital resolution of the signal, and reducing the relative importance of the (downstream) read-out noise. The downside is that the dynamic range is reduced, but as long as the detector is not saturated, nothing is lost by increasing the gain. That's why it makes sense to use the highest gain that doesn't saturate your exposure. You may not win much by increasing the gain to the saturation limit, but you don't loose anything either. Unless you intentionally want to overexpose or have good control over the lights, "auto-iso" therefore makes good sense.

If you're into extreme technical photography like astrophotography, there are other considerations that may restrict you (as characterising the detector becomes more important) but then you hopefully know what you're doing.

 

[epsiloneri]

Because a CMOS device converts each pixel's output to digital immediately whereas a CCD keeps the pixels' signals analog internally for quite a while (in relative terms). 'Digital' or 'analog' when talking about sensors refer to these internal workings, of which there is a lot in the sensor as a whole in addition to just the light sensitive pixels.

Perhaps you are referring to photon counting cameras, that can resolve single photon events? Neither CMOS nor CCDs are in general photon counters. CMOS and CCDs both record the photo-electric charge, but are read out differently. CCDs are read out by moving the charges row- or column-wise to the detector edge where they are A/D converted, while CMOS are A/D converted on site. This is the main reason why CMOS are so much faster at reading out.

 

[hollybush]

I wish camera manufacturers would call sensitivity and gain by their names, and use proper units.

Since the sensitivity is constant for a particular camera, it should live in the instruction manual. I'm happy with the combination (currently denoted "ISO") being the setting on the camera, because I want to use it with my flash meter or the Sunny 16 rule.

And use a logarithmic scale instead of ridiculous large numbers.

It already exists, in the form of "DIN film speed", but has fallen out of use since the 1970s. DIN 21 = ISO 100, DIN 24 = ISO 200 etc. (it works like the dB scale).

Originally there was ASA (the American standard, Kodak) and DIN (the German standard, Agfa). When the international (ISO) standard was agreed, speeds were originally supposed to be written like this: "ISO 100/21°" and I remember film boxes came like that. Later on, only Agfa persisted with the dual numbers. I forget what Ilford (British) used.

Arguably a scale starting at 1 and proceeding in exact integers for stops might be more understandable, since no-one uses tables of logarithms base 10 any more, but the DIN numbers (if the ISO standard still permits them) would be a lot friendlier in the high ranges than the usual ISO scale.

 

[Kit.]

Perhaps you are referring to photon counting cameras, that can resolve single photon events? Neither CMOS nor CCDs are in general photon counters. CMOS and CCDs both record the photo-electric charge, but are read out differently. CCDs are read out by moving the charges row- or column-wise to the detector edge where they are A/D converted, while CMOS are A/D converted on site. This is the main reason why CMOS are so much faster at reading out.

Not "A/D converted", but "converted to voltage range". A/D converters (which then convert a voltage value from this range to a binary code) are usually still one per row (or column), one per chip or even off-chip.

 

[justaCanonuser]

Armature disclaimer (you get what you pay for):

1- ISO in-variance still hits the wall of clipping. One you've filled your light bucket, there is no way to recover that information.

Digital & film shooter disclaimer: that's still a nice thing about negative film, it is much more forgiving in the highlights, there is no harsh clipping.