[falconeye]

( I'm assuming ISO100 is Unity Gain and thus ISO80 is a gain reduction. )

Why do you assume so?

The ISO value at no gain is determined by quantum efficieny times fill factor divided by the full well capacity. That shouldn't be a round number.

Gain reduction doesn't exist in current sensors because it wouldn't recover a clipped pixel.

The ISO beyond 1600 thing was discussed September last year. It was revealed as early as first ISO 6400 raw files emerged which was the day before Photokina opened.

---

Last edited by falconeye; 03-06-2011 at 07:07 PM.

[Smeggypants]

Why do you assume so?

The ISO value at no gain is determined by quantum efficieny times fill factor divided by the full well capacity. That shouldn't be a round number.

Gain reduction doesn't exist in current sensors because it wouldn't recover a clipped pixel.

So in the K-5 sensor ISO80 is nearest to Unity gain?

The ISO beyond 1600 thing was discussed September last year. It was revealed as early as first ISO 6400 raw files emerged which was the day before Photokina opened.

A discussion I missed. Sorry about that.

[falconeye]

So in the K-5 sensor ISO80 is nearest to Unity gain?

A discussion I missed. Sorry about that.

Smeggy, I didn't mean it as a critics. Just as a hint that the forum should have more material about the topic some time back.

As for the Unity gain question. I really don't like the idea of "Unity gain". E.g., if the read-out signal is not amplified along its path to the ADCs then the gain is 1. But as long as a full well signal (a voltage) doesn't clip (the ADC input pin) then you are still at the sensor's native ISO. Native ISO is the highest gain where you can still read out all electrons collected into the well (read out a full well). I wouldn't necessarily call that a unity gain.

Therefore, I prefer the term "native ISO" over the term unity gain.

Using an ISO setting other than native has the advantage of lower quantization noise at the ADC and less thermal noise added between the sensel and the ADC. But you can't read-out anymore all electrons thereby reducing the dynamic range. Therefore, not using native ISO is obsolete, assuming the sensor has low quantization noise and thermal transmission noise is smaller than the dark current noise.

The K-5 uses the Exmor with embedded 14 Bit ADC. Probably because it is embedded on chip, there is almost no noise added before the ADC.

The 14 Bits are just shy of covering the full well (requires 15.3 Bit or a 2.4x amplification or ISO 200). But you would see the effect only with sensels with a single electron only and because the K-5 has 3 or 4 electrons read-out noise, it is a non issue here.

If you look at the K-5 DR (DxO screen DR) as a function of ISO, you'll see:

ISO, EV -> DR ratio, well capacity e- => read-out noise e-
70, 13.61 -> 12503, 40000 => 3.2 e-
91, 13.18 -> 9281, 30769 => 3.3 e-
183, 12.30 -> 5043 , 15301 => 3.0 e-
363, 11.53 -> 2957, 7713 => 2.6 e-
717, 10.62 -> 1574, 3905 => 2.5 e-
1417, 9.84 -> 917, 1976 => 2.2 e-


UPDATE:
If I model noise n as n^2 = n0^2 + (n1/IsoAmplification)^2, then a non-linear solver finds n0=2.5e- and n1=2.3e-, where the DxO data at ISO 80 (-0.2e-), 1600 (-0.4e-) and 200 (+0.4e-) slightly deviate from the expectation. This is perfectly within measurements error margins. But maybe, the PGA performs best at min. and max. gain.
END OF UPDATE


So, there are three facts:

1. There is no advantage in increasing ISO from 80 to 200.
2. There is a total of 1 electron less read-out noise when going from native ISO to ISO 1600.
3. There is no advantage in increasing ISO from 1600 to whatever.

So, unamplified signal transmission and quantization errors add a mere 1.0 electrons read-out noise.

This is an almost ridiculously low number to even think about. One electron, or just two photons! Come on, a photographer who isn't physicist must be crazy to care about

OTOH, by DxO's DR definition, the read-out noise can't become smaller than 1 e-. So, a reduction from 3.2 to 2.2 is a 50% reduction of what is feasible. Technically, that's quite significant. Just not from a photographer's perspective


Of course, the firmware exposure controller isn't aware of facts 1-3. So, it wouldn't make good decisions and it would increase ISO rather than underexpose in situations where increasing ISO has no advantage. The K-5 needs a new AUTO-ISO feature: one which only knows 80, 400, 1600 and uses underexposure in all other cases.

[jgmankos]

ISO isn't (and has never been) a fundamental exposure parameter. Shutter and aperture are as they control the light entering the camera. ISO isn't just like WB or contrast isn't. ISO and EV compensation are the same (and actually were mechanically linked in older analog cameras for a reason).

I'll have to search back to the previous discussions you refer to...because it baffles my mind how the sensitivity of the recording medium is not considered to be a fundamental exposure parameter?

There are 3 variables when capturing an image....the amount of light (aperture), the duration of light (shutter speed), and the ability of the recording medium to capture the light (ISO)...this holds true whether you're talking about the light sensitivity of a chemical emulsion (i.e. Film, photoresist, etc), or the gain of digital sensors.

I guess I'm not understanding the point of taking one of them out of the equation...

[falconeye]

the ability of the recording medium to capture the light (ISO)...
[...]
I guess I'm not understanding the point of taking one of them out of the equation...

Fortunately, it is rather easy to digest once one gets the idea

The recording medium (as you call it) is a photon counter. Counters don't have any adjustable parameter. They just count. And count, and count ...

The point is that the ISO setting DOES NOT change the probability of a sensor to enregister an incident photon. This probability (sensitivity) is the product of (CMOS) quantum efficiency, (microlens) fill factor and (Bayer) spectral filter transmission index. And as such it is a constant which can't be adjusted in camera.

Emulsions had differing properties which however is best described as a tradeoff between grain size and grain sensitivity. The same applies to sensors with varying pixels counts. However, a higher pixel count is easily converted to the properties of a lower pixel count after the recording. So even that isn't an exposure parameter anymore.

ISO in a digital camera is nothing but an attempt to work around an imperfect counter implementation. However, counters are simple and the K-5 is pretty close to have an almost ideal counter. Which makes it more obvious that we too easily adopted ISO as a fundamental camera setting (by analog convention) when in fact it was a work around a camera flaw which soon will be a thing of the past.

[Smeggypants]

Thanks for those two posts above Falk



I come from an Audio Engineering background so I tend to think in dBs and also why the unity gain made sense to me.


So looking at ...

So, there are three facts:

1. There is no advantage in increasing ISO from 80 to 200.
2. There is a total of 1 electron less read-out noise when going from native ISO to ISO 1600.
3. There is no advantage in increasing ISO from 1600 to whatever.

In practise I do lot of low light shooting in TAv mode with my K-5. The ISO can auto range between anything from 1600 to 51200.

So, unless i've got you wrong your saying that I could set manual mode and choose shutter and aperture and leave the camera set on ISO1600 ... and then boost the exposure of the picture in post pro once I get home?

This means of course that previewing pics in camera will mean most of them are really dark.

So the question is: is post pro exposure boosting a set of pictures all taken at ISO1600 going to yield better results than using TAv during the shoot and have the exposure roughly correct before you start the Post pro process?

I'm sort of beginning to understand that ISOless is no worse than using ISO, but I'm still failing to see the advantages.

Cheers for your patience.

[dosdan]

I'm sort of beginning to understand that ISOless is no worse than using ISO, but I'm still failing to see the advantages.

Clipping headroom in situations where you would normally use higher than base ISO - solely because you can also fiddle with the TRC when boosting the EV in PP. That's in alignment with the reason for shooting raw: increased flexibility in PP.

Dan.

---

Last edited by dosdan; 03-07-2011 at 01:24 PM.

[Class A]

Which makes it more obvious that we too easily adopted ISO as a fundamental camera setting (by analog convention) when in fact it was a work around a camera flaw which soon will be a thing of the past.

I agree with the notion that variable gain will soon be a thing of the past, but I doubt ISO controls will disappear from cameras because

1. higher ISO settings provide hints to the camera of how much to push exposure for previews (if ISO settings are recorded as metadata (leaving the RAW data alone) to be used by camera previews and later RAW development then one wouldn't lose any clipping headroom)
2. lower ISO settings could either enable some variable neutral density mechanism in front of the sensor, or
3. enable <1 (digital) gain, in order to fit the dynamic range the sensor recorded into a practical image format (even if the full well capacity of sensors will allow ultra-low ISO settings, not everyone will want to move around 20+ bit files. )

[Smeggypants]

Clipping headroom.

Dan.

Ok, but you can have clipping headroom now. As I said you can have an ISOless camera now if you select manual mode at ISO100 ( or whatever base ISO is ), ignore the meter, and simply adjust your shutter and aperture to control your DOF & motion blur.

So where is the benefit in removing the ISO parameter?.

[MrPetkus]

This is a great thread. I've noticed this through my own experimentation exploiting the DR of the K5 and have have shifted my shooting style from exposure right to left. The only problem as mentioned is that preview is difficult. Nice to read the technical explanation.

[Class A]

So where is the benefit in removing the ISO parameter?

I guess a fair statement to make would be "We want an ISOless sensor, but not an ISOless camera". Hope that is self-explanatory.

P.S.: In principle, one could unify the ISO setting with exposure compensation to yield an ISOless camera. The ISO setting is easier to understand for people from the film era, the unified concept might be easier to pick up by younger generations.

[Entropy]

The analog noise is random, while quantisation noise is signal related and thus ot random

Best example of this is "color banding" in low-bit-depth formats/displays - that's an obvious example of signal-dependent quantisation noise.

In terms of actual SNRs, there isn't much difference, however, psychologically, random analog noise is almost always more pleasing than quantization noise. This is the whole concept behind dithering - Noise is intentionally added to the signal (image or audio) with a strength about equivalent to 1 LSB (magnitude of quantization noise) to improve a quantized signal's psychovisual/psychoacoustic properties.

Intelligent dithering can lead to some neat tricks, such as noise-shaping sigma-delta ADCs. (Which have a LOT of quantization noise - but at frequencies you don't care about so it's easy to lowpass filter the shaped quantization noise out.)

So a pure SNR graph does not tell all...

[Smeggypants]

I guess a fair statement to make would be "We want an ISOless sensor, but not an ISOless camera". Hope that is self-explanatory.

P.S.: In principle, one could unify the ISO setting with exposure compensation to yield an ISOless camera. The ISO setting is easier to understand for people from the film era, the unified concept might be easier to pick up by younger generations.

Forgive me here as I'm really trying to get to grips with this ...

.... but surely a sensor is ISOless anyway? After all isn't the ISO setting an applied amplification ( either before or after the AD convertor?

[Smeggypants]

Best example of this is "color banding" in low-bit-depth formats/displays - that's an obvious example of signal-dependent quantisation noise.

In terms of actual SNRs, there isn't much difference, however, psychologically, random analog noise is almost always more pleasing than quantization noise. This is the whole concept behind dithering - Noise is intentionally added to the signal (image or audio) with a strength about equivalent to 1 LSB (magnitude of quantization noise) to improve a quantized signal's psychovisual/psychoacoustic properties.

Intelligent dithering can lead to some neat tricks, such as noise-shaping sigma-delta ADCs. (Which have a LOT of quantization noise - but at frequencies you don't care about so it's easy to lowpass filter the shaped quantization noise out.)

So a pure SNR graph does not tell all...

Yes exactly, which is why I was talking about applying the amplification before the digital conversion rather than afterwards.

And yes dithering. Something I'm au fait with from my audio engineering side. It's interesting that the addition of dither noise can actually reveal detail in the signal that would normally lie below the level of the dynamic range of the converters as these low level signals influence the dither noise.


I'm beginning to learn you can't have a 1:1 analogy between audio recording and light recording though as Falk's explanation of a sensor being a photon counter is a bit different to sampling the level of an audio signal.

And yes a great thread.

[dosdan]

Ok, but you can have clipping headroom now. As I said you can have an ISOless camera now if you select manual mode at ISO100 ( or whatever base ISO is ), ignore the meter, and simply adjust your shutter and aperture to control your DOF & motion blur.

So where is the benefit in removing the ISO parameter?.

Consider a full implementation of this:

1. All shots are taken at base ISO. In "ISOless" AE mode, you adjust aperture & shutter manually.

2. Camera metering decides on a "correct" EV boost parameter from base ISO. It could show this in the viewfinder as a guide or show equiv ISO instead for users who have trouble comprehending the new exposure paradigm.

3. This EV Boost is used to show the LCD preview & embedded JPEG in the raw file at the "correct" EV level.

4. The raw file stores this parameter and the converter is aware of it. Like embedded WB. The raw converter uses it by default.

5. In PP, you can adjust away from the stored EV boost value (no need for EV comp in-camera), so you can compensate in PP for very bright objects in frame. You can also play with the TRC to your hearts content in PP.

6. As long as you are not reaching FS quantization and the signal is sufficiently dithered, you are getting extra clipping headroom that you can work with in PP. This would be a highly effective raw highlight "recovery" - not like the marginal results you get currently.



Dan.