ISO 100, OMD-EM1II and firmware 3.0?

Started 3 months ago | Questions
katastrofa Senior Member • Posts: 1,010
Re: Basics of digital camera ISO

What are "codes" in a RAW histogram?

 katastrofa's gear list:katastrofa's gear list
PowerShot SX700 Olympus E-M1 II Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko ED 75-300mm 1:4.8-6.7 II Olympus M.Zuiko Digital 25mm F1.8 +6 more
OP Architeuthis Regular Member • Posts: 167
Re: Basics of digital camera ISO

Hi Chris,

I thank you and the others for the valuable information.

I have just one questio at the time:

Chris Noble wrote:

The sensor, analog gain (ISO), A/D, and digital gain (in that signal order) are distinct.

Can one summarize this statement as follows?

#1. Sensor has always same amount of photoelectrons in a given pixel after identical exposure. (Sometimes the readout current can be at different size, when the sensor can be operated at two different supply voltages as e.g. is the case  with the dual gain in the GH5s ???)

#2.: different native ISO settings are produced by different analog amplifications of the signal (Outcome is likely a stable voltage after current to voltage conversion with different gain setting of the I/V converter). Base ISO is when the I/V converter is operated at its lowest gain possible (or reasonable).

#3.: A/D converter has always the same range of input sensitivity (e.g. 0V-1V (according to bobn2 it is more likely around 1V than my intitial 5V guess))

#4.: extended ISO results from postprocessing the digital numbers after A/D conversion. Identical results can be obtained when processinmg the raw file in e.g. LR.

Wolfgang

 Architeuthis's gear list:Architeuthis's gear list
Olympus E-M1 II Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Leica DG Macro-Elmarit 45mm F2.8 ASPH OIS +5 more
Chris Noble
Chris Noble Veteran Member • Posts: 3,172
Re: Basics of digital camera ISO

Architeuthis wrote:

Hi Chris,

I thank you and the others for the valuable information.

I have just one questio at the time:

Chris Noble wrote:

The sensor, analog gain (ISO), A/D, and digital gain (in that signal order) are distinct.

Can one summarize this statement as follows?

#1. Sensor has always same amount of photoelectrons in a given pixel after identical exposure. (Sometimes the readout current can be at different size, when the sensor can be operated at two different supply voltages as e.g. is the case with the dual gain in the GH5s ???)

I don't know about variable sensor bias. Sounds reasonable.

#2.: different native ISO settings are produced by different analog amplifications of the signal (Outcome is likely a stable voltage after current to voltage conversion with different gain setting of the I/V converter). Base ISO is when the I/V converter is operated at its lowest gain possible (or reasonable).

Yes. And you know about I/V! Good. "Lowest reasonable" is when a full pixel well generates a signal that is at the full scale of the A/D input.

#3.: A/D converter has always the same range of input sensitivity (e.g. 0V-1V (according to bobn2 it is more likely around 1V than my intitial 5V guess))

The power rails keep on going down. I think between 1.5V and 2.5V is about right. You want the input noise to be below 1 LSB.

#4.: extended ISO results from postprocessing the digital numbers after A/D conversion. Identical results can be obtained when processinmg the raw file in e.g. LR.

There is often some "light" digital processing (i.e. in the camera) at every ISO step in addition to the Bayer de-convolution. But "extended" or "interpolated" ISO settings refer to digital gain to supplement the analog gain. ("Postprocessing" is on the computer after the Raw file has been recorded, "processing" is in the camera before the Raw file is recorded). By the way "gain" can be unity, >1 or <1.

Wolfgang

 Chris Noble's gear list:Chris Noble's gear list
Panasonic Lumix DMC-GM5 Panasonic Lumix DC-GX9 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 +3 more
MEDISN
MEDISN Contributing Member • Posts: 867
Re: Meaning of base vs. extended ISO

bobn2 wrote:

I'm not sure which DxOmark measurements you're talking about. Whilst they make some errors, overall they are quite good. It's just understanding what the actually signify. Their 'measured ISO' measurement is actually very useful indeed, it's just misnamed. It isn't 'measured ISO'. A better name would be 'saturation exposure', where the measurement of exposure is a non-standard ISO rating.

Can we apply this to an example (Em1mkII) Bob?

The "saturation exposure" of DxO is 83

The lowest ISO settings on the camera are now 64, 100, 200, with 64 and 100 labeled "extended" or "low" in camera.

Help me with these scenarios:

  • If I set camera ISO to "64" I am over-saturating what the sensor can likely handle and risk blowing highlights (clipping).
  • If I set camera ISO to "100" I might just avoid blowing most highlights
  • If I set camera ISO to "200" I am providing ample headroom for highlights (wasting exposure latitude, i.e. less signal in the shadows)

If these statements are accurate, it would seem the ideal "Low ISO" setting would be 100...perhaps with even +1/3EV exposure compensation.  I realize all scenes/environments are not equal but IN GENERAL, am I close? 

Thanks in advance.

bobn2
bobn2 Forum Pro • Posts: 62,046
Calling your bluff
4

Chris Noble wrote:

bobn2 wrote:

Chris Noble wrote:

bobn2 wrote:

katastrofa wrote:

My photos look more noisy when I increase the ISO. Why?

Changing the ISO sets the meter so that you choose to use a lower exposure. Lower exposure means less light energy. Less light energy means fewer photons captured in the image. The signal to noise ratio in the image goes as the square root of the number of photons, so if you capture fewer, you get a lower SNR, which you say is 'more noisy'.

You are mixing changes in exposure vs. changes in ISO setting.

With respect, I'm mixing nothing.

Katastrofa's observation is correct even if the exposure does not change (using manual metering for example). The ISO gain introduces additional noise on top of the exposure noise.

That is simply false. It's provably false with suitable examples, which I've posted here several times, but don't have readily available here (if this conversation carries on, I'll post them again, when I can).

Your explanation below that paragraph is full of errors, but rather than explain them to you, I'll take you up on your offer to demonstrate that you are wrong. Post the same exposure, once at ISO 200 and one at ISO 3200. Don't do any noise reduction or sharpening. Adjust the PP brightness to appear the same on the screen. When you look at the shadows, the error of your understanding will be clear.

So, here are two pictures:

So, what's happened here. These are photos from the same camera (a Canon 60D) 100% crop, taken a few seconds apart with the same exposure. One at 100 ISO, one at 1600 ISO. The raw numbers have been read from the raw files using dcraw, then 'normalized' (i.e. stretched to the same range), in Image Magick, using 64-bit floating point arithmetic (52 bits of precision). Thereafter both shots have been subjected to exactly the same processing, all in ImageMagick, all using 64 bit FP. Essentially, demosaicking, white balance, gamma encoding and output to an 8 bit JPEG. You'll notice the 100 ISO one is much noisier, not, as you predict, the other way round. Having been through this discussion with people like you before, I know you'll raise all kinds of spurious objections, one will be that this is a peculiar camera. So here is the same test with a different camera:

Not as extreme, the D810 is close to ISOless, but what we don't see is what you predicted, the 1600 ISO shot being noisier.

Don't like that?

Here's another

(A note here, Panasonic is doing some WB processing on the raws, not so raw). Anyway, still the !600 ISO version is not noisier. Not enough? Here's a lot more:

The photographer here suffered some shake, doesn't change the noise.

If you really want to continue defending the indefensible, I can provide all of the raw files, just in case you felt like accusing me fo fiddling this somehow (I'm speaking from experience here).

Anyhow, the takeaway is this. In none of these cases do actual results support your supposition that the noise increases with ISO. Time to give it up and stop pretending.

-- hide signature --

263, look deader.

bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Basics of digital camera ISO

Architeuthis wrote:

#1. Sensor has always same amount of photoelectrons in a given pixel after identical exposure. (Sometimes the readout current can be at different size, when the sensor can be operated at two different supply voltages as e.g. is the case with the dual gain in the GH5s ???)

That isn't how 'dual gain' works. The 'conversion gain', which is the parameter being changed, is the inverse of the capacitance of the pixel input node (think about it C = q/V, so for a given input charge q, the output voltage V= q/C, i.e., the larger the capacitance, the smaller the output voltage. The dual gain feature builds an additional capacitor in each pixel, and switches it in at low ISOs to lower the pixel output voltage.

#3.: A/D converter has always the same range of input sensitivity (e.g. 0V-1V (according to bobn2 it is more likely around 1V than my intitial 5V guess))

This is an interesting observation. Rather than the provision of variable voltage gain, to produce the effect that Chris says is 'ISO' a much simpler expedient would be to simply reduce the ADC reference voltage (Fr information, almost invariably the 'fall scale' of an ADC is referred to an external or programmable reference voltage. There is a very good reason that cameras are not designed this way, but as far as Chris' story goes, it would be just as good as the VGA, actually better because there would be none of the amplifier noise that he supposes the VGA is adding.

-- hide signature --

263, look deader.

bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Meaning of base vs. extended ISO
3

MEDISN wrote:

bobn2 wrote:

I'm not sure which DxOmark measurements you're talking about. Whilst they make some errors, overall they are quite good. It's just understanding what the actually signify. Their 'measured ISO' measurement is actually very useful indeed, it's just misnamed. It isn't 'measured ISO'. A better name would be 'saturation exposure', where the measurement of exposure is a non-standard ISO rating.

Can we apply this to an example (Em1mkII) Bob?

The "saturation exposure" of DxO is 83

The lowest ISO settings on the camera are now 64, 100, 200, with 64 and 100 labeled "extended" or "low" in camera.

Help me with these scenarios:

  • If I set camera ISO to "64" I am over-saturating what the sensor can likely handle and risk blowing highlights (clipping).
  • If I set camera ISO to "100" I might just avoid blowing most highlights
  • If I set camera ISO to "200" I am providing ample headroom for highlights (wasting exposure latitude, i.e. less signal in the shadows)

That's exactly right. Remember that the ISO standard already builds in headroom above 'white' (i.e. a white reflective object) for light sources and specular reflections. So, even if the 'saturation exposure' was 100, the camera should be able to handle those (as an aside, this means that in the output the white object is not rendered 'white' but light grey. 'white' in the output is reserved for things which are brighter than 'white'). So mostly, you'll be OK. Some of the earlier Panasonic sensors has a very low saturation, so the stock settings actually ate into this headroom, but no-one much noticed (the legacy though is high base ISOs in mFT, it's a cultural rather than technical necessity). Most of the time, is your subject is just normal reflective objects, you'll get away with 64 perfectly OK.

If these statements are accurate, it would seem the ideal "Low ISO" setting would be 100...perhaps with even +1/3EV exposure compensation.

Remember that exposure compensation affects the meter but not the processing, so dial in EC and your JPEGs will be rendered too light. The raws, on the other hand, can be processed however you want.

I realize all scenes/environments are not equal but IN GENERAL, am I close?

Spot on.

-- hide signature --

263, look deader.

katastrofa Senior Member • Posts: 1,010
Re: Calling your bluff

Can you explain why the noise decreased with ISO in the first pair of photos?

 katastrofa's gear list:katastrofa's gear list
PowerShot SX700 Olympus E-M1 II Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko ED 75-300mm 1:4.8-6.7 II Olympus M.Zuiko Digital 25mm F1.8 +6 more
MEDISN
MEDISN Contributing Member • Posts: 867
Re: Meaning of base vs. extended ISO
2

bobn2 wrote:

MEDISN wrote:

bobn2 wrote:

I'm not sure which DxOmark measurements you're talking about. Whilst they make some errors, overall they are quite good. It's just understanding what the actually signify. Their 'measured ISO' measurement is actually very useful indeed, it's just misnamed. It isn't 'measured ISO'. A better name would be 'saturation exposure', where the measurement of exposure is a non-standard ISO rating.

Can we apply this to an example (Em1mkII) Bob?

The "saturation exposure" of DxO is 83

The lowest ISO settings on the camera are now 64, 100, 200, with 64 and 100 labeled "extended" or "low" in camera.

Help me with these scenarios:

  • If I set camera ISO to "64" I am over-saturating what the sensor can likely handle and risk blowing highlights (clipping).
  • If I set camera ISO to "100" I might just avoid blowing most highlights
  • If I set camera ISO to "200" I am providing ample headroom for highlights (wasting exposure latitude, i.e. less signal in the shadows)

That's exactly right. Remember that the ISO standard already builds in headroom above 'white' (i.e. a white reflective object) for light sources and specular reflections. So, even if the 'saturation exposure' was 100, the camera should be able to handle those (as an aside, this means that in the output the white object is not rendered 'white' but light grey. 'white' in the output is reserved for things which are brighter than 'white'). So mostly, you'll be OK. Some of the earlier Panasonic sensors has a very low saturation, so the stock settings actually ate into this headroom, but no-one much noticed (the legacy though is high base ISOs in mFT, it's a cultural rather than technical necessity). Most of the time, is your subject is just normal reflective objects, you'll get away with 64 perfectly OK.

If these statements are accurate, it would seem the ideal "Low ISO" setting would be 100...perhaps with even +1/3EV exposure compensation.

Remember that exposure compensation affects the meter but not the processing, so dial in EC and your JPEGs will be rendered too light. The raws, on the other hand, can be processed however you want.

I realize all scenes/environments are not equal but IN GENERAL, am I close?

Spot on.

Thank you for taking time to respond and explain Bob.  These are the sort of valuable exchanges I visit these forums for (as I tip-toe around the usual landmines).  Appreciate all your efforts over the years to distill the complex and technical into something even I can understand and put into practice.  

Chris Noble
Chris Noble Veteran Member • Posts: 3,172
Re: Basics of digital camera ISO

bobn2 wrote:

Rather than the provision of variable voltage gain, to produce the effect that Chris says is 'ISO' a much simpler expedient would be to simply reduce the ADC reference voltage (Fr information, almost invariably the 'fall scale' of an ADC is referred to an external or programmable reference voltage. There is a very good reason that cameras are not designed this way,

Actually, there are not one but several good reasons, for cameras and other mixed-signal circuits with multiple analog ranges. That is why your "much simpler expedient" approach is not feasible.

but as far as Chris' story goes, it would be just as good as the VGA, actually better because there would be none of the amplifier noise that he supposes the VGA is adding.

The VGA does not add appreciable noise. But (a) it amplifies the sensor noise and (b) if the sensor output range is not scaled to the A/D input range, A/D quantization noise rises, eventually above the input noise floor. Those are the reasons for the variable (actually fixed-step) VGA.

 Chris Noble's gear list:Chris Noble's gear list
Panasonic Lumix DMC-GM5 Panasonic Lumix DC-GX9 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 +3 more
Chris Noble
Chris Noble Veteran Member • Posts: 3,172
Not a bluff but yes
2

bobn2 wrote:

Anyhow, the takeaway is this. In none of these cases do actual results support your supposition that the noise increases with ISO.

Your analysis is good! Thanks for taking the time to take and process these photos.

 Chris Noble's gear list:Chris Noble's gear list
Panasonic Lumix DMC-GM5 Panasonic Lumix DC-GX9 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 +3 more
bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Not a bluff but yes

Chris Noble wrote:

bobn2 wrote:

Anyhow, the takeaway is this. In none of these cases do actual results support your supposition that the noise increases with ISO.

Your analysis is good! Thanks for taking the time to take and process these photos.

You're welcome. However, I can't take credit for taking the time. I didn't take the photos. They were taken by participants at workshops I've run on the topic, and the processing is automated via a script, so I can do it in real time. It always creates a stir, they take the photos with their own cameras, and a few seconds later, there are the results.

And sorry about the comments about what you 'would' do, you turned out to have a whole load more about you than most of the people I have these discussions with.

-- hide signature --

263, look deader.

bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Basics of digital camera ISO

Chris Noble wrote:

bobn2 wrote:

Rather than the provision of variable voltage gain, to produce the effect that Chris says is 'ISO' a much simpler expedient would be to simply reduce the ADC reference voltage (Fr information, almost invariably the 'fall scale' of an ADC is referred to an external or programmable reference voltage. There is a very good reason that cameras are not designed this way,

Actually, there are not one but several good reasons, for cameras and other mixed-signal circuits with multiple analog ranges. That is why your "much simpler expedient" approach is not feasible.

It's feasible, just not very sensible, because lowering the reference voltage whilst it will =multiply the digital numbers produced by the same input voltage, won;'t to anything with respect to what the voltage gain is actually for, which is boosting teh sensor signal up above the ADC noise.

but as far as Chris' story goes, it would be just as good as the VGA, actually better because there would be none of the amplifier noise that he supposes the VGA is adding.

The VGA does not add appreciable noise. But (a) it amplifies the sensor noise and (b) if the sensor output range is not scaled to the A/D input range, A/D quantization noise rises, eventually above the input noise floor. Those are the reasons for the variable (actually fixed-step) VGA.

Sorry, that still isn't right. The VGA amplifies the sensor signal and noise equally, since the  noise is not something separate, it is just a variation on the signal. So the VGA makes no difference to the sensor SNR. What it does is boost the signal level up above the (fixed) ADC noise floor. Sensors with column ADCs tend to be much more ISOless, because the thousands of ADCs can operate much slower, so can be much quieter. Thus there is less need for variable gain. A/D quantisation noise is not the culprit, because the ADC has a big enough enob (there's a phrase to conjure with) to completely encode the DR given by the sensor at those exposure levels.

-- hide signature --

263, look deader.

bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Calling your bluff
1

katastrofa wrote:

Can you explain why the noise decreased with ISO in the first pair of photos?

ADCs are noisy. The actual number of available bits is given by a parameter called 'ENOB' (effective number of bits) and is rarely the same as the full bit width of the ADC. For a fast 14 bit ADC it might be 11-12 bits. The bits you don't get are essentially full of noise. The role of the variable voltage gain is to boost the signal above this noise. So, lets compare the two images, both have the same exposure, but the 100 ISO one is using just the lower bits of the ADC, which are noisy. In the second, the sensor signal has had gain applied, which means it is using the upper bits of the ADC, which aren't noisy.

The D700 was the first camera to appear with Sony's column parallel 14-bit ADC (or actually the first one to make use of it, the Sonys at the time only use 12 bit conversion). The ADC has an ENOB of a bit over 13 bits (due to there being thousand operating in parallel, allowing them to work slowly - noise and speed are inextricably intertwined), it shows very little benefit from the gain, since the ADCs aren't noisy in the first place.

With respect to mFT, one of the advantages of its small pixels is that 12 bits is enough to completely (or nearly, with some modern sensors) encode the DR of the pixel output, so it actually has no need for 14 bit ADCs. The modern column parallel ADCs on both Sony and TPSCo FT sensors are something around 11 ENOB, so again the voltage gain doesn't do a lot.

-- hide signature --

263, look deader.

VideoPic
VideoPic Senior Member • Posts: 1,363
Interesting replies.....great post, thanks.... :-)
-- hide signature --

See my Blog for short articles on the Olympus and Panasonic cameras.
https://myolympusomd.blogspot.com/
Olympus has a history of innovation, they setting the pace on technology and camera functionality. Those in the know realize this trend will continue. Olympus is in the game for optical advances and a passion for the best possible equipment to record the image. Study the Olympus corporate portfolio and you will be amazed by this diverse and interesting company....

 VideoPic's gear list:VideoPic's gear list
Olympus PEN-F Olympus E-M1 II Olympus M.Zuiko Digital ED 12-50mm 1:3.5-6.3 EZ Panasonic 12-35mm F2.8 Panasonic Lumix G X Vario 35-100mm F2.8 OIS +4 more
katastrofa Senior Member • Posts: 1,010
Re: Calling your bluff

Got it. What if we applied the variable gain AFTER the ADC?

 katastrofa's gear list:katastrofa's gear list
PowerShot SX700 Olympus E-M1 II Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko ED 75-300mm 1:4.8-6.7 II Olympus M.Zuiko Digital 25mm F1.8 +6 more
spike29 Senior Member • Posts: 1,790
Re: Basics of digital camera ISO

bobn2 wrote:

spike29 wrote:

bobn2 wrote:

.....

That's a common problem in this discussion, although also an irrelevance. When you realise what the photographic process is, and what ISO actually does, you realise that whilst where the sensor starts and stops is of academic interest, it's irrelevant to the discussion. ISO says what lightness should result from a given exposure.

This is one of the clearest , most understandable short definitions i read about this subject .🙂

Simplified but accurate enough.

Thanks for the feedback. I'll remember it and use that text again

I will explain why i find this to the point, most of the time ISO value is connected to "gain" of voltage or current or digital gain in threads and webpages which brings your mind to a "amplifier". Which is a trap of misconceptions and assumptions. (I did) most internet searches don't help to untangle the confusion either, on the contrary it tighten the knots even more.

That's why this simple text don't distract you and don't sent you in a maze of half truth's. and keep you on track for the next information.

So as far is i remember all the basics:

While ISO isn't a Gain/Amplifier but more a (wanted) result of the several gains (to get the signal in range of correct processingvalues for readout and conversions) along the road of capturing photons towards a image represented in a rawfile. This in order to get the image correct(lightnes) represented on screen or paper. By getting a underexposed sensor/pixel , low amount of photons on a given shuttertime, and by setting a ISO value you define the lightnes you want to correct that (under)exposure in the camera to get a correct lightness in your jpeg/tiff. (edit: Correct?)

new question:

What if you alter a rawfile of a 1600iso image and change the stored value of 1600iso in the exif in lets say 200iso? (lets say this is "base iso") drops the lightnes of the image in a rawdeveloperapplication 3stops down? so it looks like a -3EV Exposurecorrection?

(highlights stayed clipped so those are gone and black is black so you scrunch the DR of the tonecurve. )

and this altered exif shows the "real image" what's captured by the sensor showing the real lightnes it would show on base ISO? (except the clipped data)

Or are some "gains" non reversible? (i think simplified that all gain before ADC is irreversible, after the ADC it's just "software calculation" and fully reversible. )

To determine the lightness in the output file, all the camera needs to be able to do is measure the exposure in each pixel. 'Gain' doesn't change the exposure in the pixel, but it can change how accurately it is measured.

Is this the reason that dynamic range changes when ISO raises? less accurately means less steps between black and white (full exposure)

So far as DR is concerned, the problem is that if you apply voltage gain before the ADC you boost the higher signals up above the ADC maximum, the larger exposures can't be digitised. Thus each stop of gain means a loss of a stop at the high end. Once that loss stops lowering the noise floor, there is no point applying any more voltage gain. There's another question of just why manufacturers continue unnecessarily clipping the raw file, which we can have if you want.

People have difficulty with this concept because they tend to call both the input and output 'exposure', in which case they think of 'gain' as giving a bigger 'exposure' from a smaller one, but that is not what is happening. Exposure and lightness are different things, which is why two words are needed.

Extended ISO at the high end i can think of: Just add measurement's at the lowest part of the sensor output voltages until noise overgrows mesurable signal and higher iso values are redicules because you can't separate random noise from the actual latent image capturing enough to get a reasonable clear image at the end anymore.

The real problem is that the noise is in the signal, which is why photography is somewhat different from common electronic applications. The SNR goes as the square root of the number of photons captured. The problem with very small exposures is just not many photons.

But at the low end below base iso. Why don't use that as base iso?

See upthread. The thing,about th Lo settings is you can meter as normal, get higher exposures, thus more photons, thus less noise.

Got it. metering is for the determination of the shuttertime to get a proper exposure and shuttertime increases when lowering iso so more photons are able to reach the sensor/pixel with in the shuttertime.

More then enough light to "fill" pixels at base iso so when extend to a lower iso, you get some longer ss and or smaller diafragma at the cost of what?

Whether pixels are 'full' or not isn't important. It's the absolute number of photons that matters.

(Sorry to bump in but this is interesting, most things i read is that extended iso at the low end is extended dynamic range , so less overexposed pixels.)

Bumping in is what a public forum is for

-- hide signature --

knowledge is addictive, every time i get some i want more.....
-----------------------------------------------------------------------------------------------------
(If i can remember 1/1000 of everything i learned/read in the past i will be happy as a monky with........)

 spike29's gear list:spike29's gear list
Fujifilm FinePix F70EXR Panasonic Lumix DMC-FZ200 Panasonic G85 Panasonic Lumix G Vario 14-140mm F3.5-5.6 O.I.S Panasonic Leica 12-60mm F2.8-4.0 ASPH +1 more
bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Calling your bluff

katastrofa wrote:

Got it. What if we applied the variable gain AFTER the ADC?

You can't, then you're in the digital domain, dealing with numbers. What you apply, should you want to, is a mathematical operation, multiplication, not 'gain'. The raw numbers coming out from the ADC represent a count of photons in the pixel, which in turn is the exposure (given that exposure is light energy per unit area, photon number tells you the energy and you know the area of the pixel). It's in a non-standard, device and setting dependent unit, rather than the SI unit, the lux second, but that doesn't change what it is or what it means. Multiplying it doesn't change the exposure it represents, it just changes the unit it's expressed in. Counting your savings in cents rather than dollars makes the numbers 100 times bigger, but doesn't change the cash you have. The job of translating to the lightness values dictated by the ISO could be done by a mathematical function, which might involve multiplication, division, addition and subtraction, but is more commonly done with a 'look-up table', which involves none of those operations.

There an interesting reversal of terminology that has gone one with the whole fiction that ISO is 'gain'. Then, 'multiplication' becomes 'digital gain, which actually is a reversal of the original meanings from the old days of analog computation, in which analog multipliers were variable gain amplifiers. It's much easier to get the ideas straight considering mathematical functions and 'multiplication' rather than spurious electronic ideas such as 'gain'. In terms of the imaging process, there is nothing being 'gained'.

-- hide signature --

263, look deader.

bobn2
bobn2 Forum Pro • Posts: 62,046
Re: Basics of digital camera ISO
1

spike29 wrote:

bobn2 wrote:

spike29 wrote:

bobn2 wrote:

.....

That's a common problem in this discussion, although also an irrelevance. When you realise what the photographic process is, and what ISO actually does, you realise that whilst where the sensor starts and stops is of academic interest, it's irrelevant to the discussion. ISO says what lightness should result from a given exposure.

This is one of the clearest , most understandable short definitions i read about this subject .🙂

Simplified but accurate enough.

Thanks for the feedback. I'll remember it and use that text again

I will explain why i find this to the point, most of the time ISO value is connected to "gain" of voltage or current or digital gain in threads and webpages which brings your mind to a "amplifier". Which is a trap of misconceptions and assumptions. (I did) most internet searches don't help to untangle the confusion either, on the contrary it tighten the knots even more.

That's why this simple text don't distract you and don't sent you in a maze of half truth's. and keep you on track for the next information.

So as far is i remember all the basics:

While ISO isn't a Gain/Amplifier but more a (wanted) result of the several gains (to get the signal in range of correct processingvalues for readout and conversions) along the road of capturing photons towards a image represented in a rawfile. This in order to get the image correct(lightnes) represented on screen or paper. By getting a underexposed sensor/pixel , low amount of photons on a given shuttertime, and by setting a ISO value you define the lightnes you want to correct that (under)exposure in the camera to get a correct lightness in your jpeg/tiff.

A word of caution there, the term 'underexposed' is something of a trap, too. The pixels's job is to measure exposure, whatever it is. 'Underexposed' is a relative term, relative to the exposure you want, and what exposure you want depends on what you're trying to do and towards what end you're trying to manage exposure. It doesn't relate to any requirement in the pixel or any need for some absolute level of exposure. A low exposure doesn't 'need' boosting, it just needs the desired lightness value assigned to it.

new question:

What if you alter a rawfile of a 1600iso image and change the stored value of 1600iso in the exif in lets say 200iso? (lets say this is "base iso") drops the lightnes of the image in a rawdeveloperapplication 3stops down? so it looks like a -3EV Exposurecorrection?

Think of the ISO development process as a table of lightness values against exposure values. (What a processor is doing is like 'painting by numbers'). By changing the ISO you shift all the lightness values up or down the table. If you set the ISO three stops lower. then all the lighness values end up against three stops larger exposures, so if you dindn't actually change the exposures, the image will be rendered three stops darker.

(highlights stayed clipped so those are gone and black is black so you scrunch the DR of the tonecurve. )

The sensor tone curve is pretty much a straight line, unlike film. It doesn't get 'scrunched'

and this altered exif shows the "real image" what's captured by the sensor showing the real lightnes it would show on base ISO? (except the clipped data)

There isn't a 'real image' in the sense you mean. What the sensor captures is what in film days would be called a 'latent image'. The image you see is a perceptual illusion, tailored to human vision, not a direct funelling through of the light from the original scene. It doesn't exist until you've processed the latent image.

Or are some "gains" non reversible? (i think simplified that all gain before ADC is irreversible, after the ADC it's just "software calculation" and fully reversible. )

I think it's best to forget about 'gain'. It has to do with optimising the camera electronics, and nothing to do with photography. If you don't optimise the capture electronics for low exposures (which is what happens when you raise the ISO) you won't recover the shadow information lost in the resultant noise. If you optimise for low exposure, then use a high one, you won't recover the clipped highlights.

To determine the lightness in the output file, all the camera needs to be able to do is measure the exposure in each pixel. 'Gain' doesn't change the exposure in the pixel, but it can change how accurately it is measured.

Is this the reason that dynamic range changes when ISO raises? less accurately means less steps between black and white (full exposure)

So far as DR is concerned, the problem is that if you apply voltage gain before the ADC you boost the higher signals up above the ADC maximum, the larger exposures can't be digitised. Thus each stop of gain means a loss of a stop at the high end. Once that loss stops lowering the noise floor, there is no point applying any more voltage gain. There's another question of just why manufacturers continue unnecessarily clipping the raw file, which we can have if you want.

People have difficulty with this concept because they tend to call both the input and output 'exposure', in which case they think of 'gain' as giving a bigger 'exposure' from a smaller one, but that is not what is happening. Exposure and lightness are different things, which is why two words are needed.

Extended ISO at the high end i can think of: Just add measurement's at the lowest part of the sensor output voltages until noise overgrows mesurable signal and higher iso values are redicules because you can't separate random noise from the actual latent image capturing enough to get a reasonable clear image at the end anymore.

The real problem is that the noise is in the signal, which is why photography is somewhat different from common electronic applications. The SNR goes as the square root of the number of photons captured. The problem with very small exposures is just not many photons.

But at the low end below base iso. Why don't use that as base iso?

See upthread. The thing,about th Lo settings is you can meter as normal, get higher exposures, thus more photons, thus less noise.

Got it. metering is for the determination of the shuttertime to get a proper exposure and shuttertime increases when lowering iso so more photons are able to reach the sensor/pixel with in the shuttertime.

Not shutter time, exposure. Remember, exposure is controlled by shutter, aperture and the light coming from the scene. The ISOI that you set provides your meter with a target exposure, that's the exposure that the meter measures when it's centred. You can get that target exposure by changing any of the shutter speed, aperture or light coming from the scene (if you have control over that). This is hard to conceptualise if you're thinking of 'exposure' as the output of the process, how light or dark the final image is. Many people do this, including supposed experts that write web sites (for instance, I've caught the Northrups doing this, left a comment, not responded to). Exposure is what you control at the input. Lightness is what you get at the output.

-- hide signature --

263, look deader.

katastrofa Senior Member • Posts: 1,010
Re: Calling your bluff

Ok, so what if we applied multiplication to the digital signal instead of gain to the analog one? That wouldn't add any noise.

 katastrofa's gear list:katastrofa's gear list
PowerShot SX700 Olympus E-M1 II Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko ED 75-300mm 1:4.8-6.7 II Olympus M.Zuiko Digital 25mm F1.8 +6 more
Keyboard shortcuts:
FForum MMy threads