Native ISO vs Boost ISO. What kind of amplification corresponds to each of them?

Question

As I understood, there's a native range for ISO in DSLR cameras and a boost mode which is usually a bigger range and contains the native ISO. I'm lost a bit here. Which of the following processes is correct?

1- After photons hit the sensor, sensor produced a number of photoelectrons which are collected by a circuit and sent to ADC (no amplification). This is for Native ISO which is intrinsic for each sensor and depends on the material. Now if we want to change the ISO, we simply use a gain circuit which changes the input as: (Input*gain=output).

2- We have two amplifications: One before ADC for all ISO settings and one after ADC unit which is used in boost mode.

Answer 1

You're getting a little confused.

Un-boosted ISO

Or what you're calling "native range." There's a native ISO setting, which is what the sensor uses without any amplification, and you can consider this the "base" ISO setting of the camera. All of the unboosted other ISO settings are achieved by amplifying the gain on this native signal.

Boosted ISO

Boosted ISO settings, otoh, aren't done with signal amplification. They are achieved by in-camera digital processing of the data after the image is taken. If the boost is in the higher-ISO direction, the unboosted setting used is the highest one available (e.g., if you have 'native range' from 200-6400, and 12800 and 25600 are your boosted levels, then you're really using 6400 amplification on the sensor), and the image is underexposed by the additional stop(s) for the boosted level. Then the image's exposure is adjusted 1 or 2 stops brighter in the in-camera processing. But the cost is that you will have increased noise--more so than with only amplification.

If the boost is at the lower levels (e.g., iso 100 or iso 50), then the signal used is the native/base ISO, the image is overexposed by the stop(s) needed, and the exposure is adjusted darker in post. But the cost here is decreased dynamic range.

This distinction is why they are careful to label the digitally boosted settings as such--unlike the amplified signals, you have additional tradeoffs to consider before using them (like digital zoom).

Canon vs. Nikon

The other thing to keep in mind is that Nikon uses amplification for the entire unboosted range, but Canon doesn't. Canon also uses the digital push/pull processing for their 1/3-stop settings--that means you're losing dynamic range on the -1/3EV settings, and gaining noise on the +1/3EV settings. This is why you'll sometimes run across the advice for Canon shooters to keep your iso on the full stop settings or use the -1/3EV settings.

Answer 2

The answer is 2, hardware amplification only for native ISO, hardware with additional software amplifications/reduction for extended ISO.

Extended ISO is like taking a RAW image and then changing ISO in Lightroom, only it's done is camera. Extended ISO also only works in jpeg, not raw for obvious reasons.

Meaning physical exposure is done in native ISO and then additionally cut or amplified in camera by software. For instance 50 ISO is recorded at base ISO and then simply cut short to mimic ISO 50. Expanded ISO of let's say ISO 12800 may be done at ISO 6400 as an underexposed shoot and then boosted in camera software to make is equal of ISO 12800. Different camera makers have different approaches, but they are all have this is common, extended ISO is done by manipulation and it's differently than native ISO.

Answer 3

Strictly speaking, "Boost ISO" means nearly as much as "HD" written on the objective. It is a buzzword without formal definition and may be misused as such. Consequently, it is not possible to give precise answer about it's usage.

Neither 2 nor 1 is completely correct.

• the pixel does not send the number of photons (it is read as a voltage, not an integer number)
• no camera uses two ADCs for each pixel (to my knowledge)
• there cannot be any amplifier after ADC. After quantisation occurs, it is digital manipulation, not amplification.

  ---

This could be the only answer to the question but I will try to dive deeper.

ISO number for sensitivity is well defined, read more about it there.

Sensor read circuit consists from:

• pixel (capacitor)
• voltage amplifier (either per-pixel (Active Pixel) or single/sequential (Passive Pixel))
• Analog To Digital Converter (which accepts voltage lying inside it's dynamic range)

Having two amplifiers means returning bottleneck to Active Pixel Sensor (and also probably affects the shadow noise but I am not competent enough to claim it). Also, having two amplifiers instead of one affects noise too.

Native ISO of a sensor corresponds to the lowest amplification level at which ADC uses all available levels for output. Native ISO range most oftenly corresponds to the range of possible amplifications (I do not hear that oftenly).

Note: ADC outputting the rough number of photons is more of a coincidence than a rule and there may be no ISO setting to obtain that.

Camera OEMs may use the word "boost" for whatever they want to. However, the somewhat smart (if something about using generic word in complicated sphere ("boost") may be smart at all) usage of this word means:

• in case of recording JPEG/TIFF files in automatic exposure mode: it means that camera will use more exposure (in case of "extending" ISO range downwards) or less exposure (in case of "extending" ISO range upwards) than it does for "unboosted" modes with same analog amplification what will result in either undesirably looking highlights (for "extending" down) or somewhat (mostly unconsiderably) more noise than an image could probably have with bigger analog amplification (when extending upwards). So, "boosted" in this case could mean "camera spoils my highlights"
• in case of recording RAW: actually, there are both cameras which may produce identical RAW files at "extended" ISO and at "normal" ISO setting and those that may produce different RAW output even if the only exposure parameter which you changed was ISO changed from "normal" to adjacent "boost" value. So, saying "boosted ISO" in this case could mean "camera throws data out of my precious photos for some elusive profit".

That is not whole truth. Both Canon and Nikon are modifying RAW data (effectively removing some information which was present initially)

• to hide the drop of sensor sensitivity for fast objectives caused by bigger angle of light incidence (yes, it is real. Unfortunately, article does not say whether that amplification occurs in digital or analog domain)
• Nikon (including all those expensive models like D3*) unconditionally scales individual channel data upwards (dunno the purpose really, it is a headache in specific cases) and also excludes very fine shadow noise by manipulating black point (what is important for astrographers)
• some Fujifilm and Pentax cameras are applying noise reduction either at all ISO settings or at high ISO settings only

It gets even more complicated if one tries to achieve same result with different cameras. Different cameras may produce different numerical values even at same exposure settings because OEMs do not choose same exposure for strictly measured ISO sensitivity value. So, one will end with two photos, one of which needs to be manipulated digitally ("boosted") to be the same as other. If one knows the bias, they can correct it the exposure in the field.

Even more complex: 2 channels of output of most Bayer cameras should be scaled up to match the most sensitive channel to avoid highlight reconstruction - effectively requiring "boost". There will unavoidably be colour artifacts in highlights even with this scaling applied - because if only one channel is saturated it efectively prevents accurate colour conversion (which is required for every camera to produce natural image).

To put a final nail: big portion of recent cameras does not benefit from bigger aplification past certain point. There is neither any detectable noise benefit nor any tonal depth increase, and the effect might actually be negative (one won't see any noise improvements when only ISO sensitivity is changed. In other words: one stop increase of ISO sensitivity will yield more than one stop of dynamic range loss.). See the graph for Dynamic Range of Canon 6D for an example, pay attention to difference between nominal ISO 12800 and ISO 25600.

To sum it up: even if you obtain an image which was not digitally manipulated you will need to do "boost" yourself to get the final image.

---

I will quote answers and say what is wrong.

All of the unboosted other ISO settings are achieved by amplifying the gain on this native signal.

Canon vs. Nikon

This camera uses only one gain for all ISO settings. I have seen more but I cannot recall them, just believe me that there are some more.

Additionally, many cameras report higher ISO setting than one being used, and Canon cameras do that too.

Boosted ISO settings, otoh, aren't done with signal amplification.

There is nothing in the Nikon D300 saying "ISO6400", one may only set "HI1" (it is natural to suggest that it should be called "boosted" or "extended") to get equivalent exposure behaviour, yet Nikon D300 does additional amplification at this setting compared to nominal ISO3200.

There are more examples of this.

---

As you may see, having boosted ISO or not is not what defines the quality of output.

There are two points which I want to express. 1) I cannot remember a camera which produces output bad enough to distinguishably affect the quality of output at any ISO setting. 2) If you want to do serious science or astrography, use neither Nikon nor Bayer sensor for that.

I hope that I answered all your questions.

Answer 4

Native ISO is the camera ISO sensitivity that requires the least amplification to be read by the A/D.

It's a precise attribute of the design of the camera, does not have to match the standard ISO values (and often it doesn't) and doesn't imply the absence of any amplification -the A/D converter needs a readable signal, after all. The camera knows internally which is the native ISO and amplifies the signal in both directions according to the ISO selected for the shot; this means that, if a camera has a native ISO of 200, when shooting at ISO 100 the signal is actually reduced, not amplified.