Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
02-14-2008, 05:21 PM   #1
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
At same aperture/shutter speed, does ISO have impact on IQ?

This question has been raised in another thread https://www.pentaxforums.com/forums/177023-post72.html and I would be interested in the reponse:

At same aperture/shutter speed, does ISO have impact on IQ?

THEORY #1:
The exact same amount of light is captured, the image imformation only is multiplied by a constant factor which can be undone in photoshop. So, same quality.

THEORY #2:
ISO somehow primes the sensor cells (i.e., makes them more or less sensitive to build up charge with light, maybe by changing a resistor). So, different quality.

Somebody out there knows what ISO is at the electronical level?

02-14-2008, 05:33 PM   #2
Veteran Member




Join Date: Jun 2007
Location: California
Posts: 426
To obtain higher sensitivities from the sensors, more power is sent to it (amplifying the signal). When you amplify a signal, interference is introduced and the A/D converters interprets this as slightly different colors. These random color inaccuracies are what we call noise

https://www.pentaxforums.com/forums/pentax-dslr-discussion/19317-native-iso-1...en-mood-2.html

Easier for me to copy & paste than to type it out again. So it's your theory #2. The sensor is actually more sensitive to light.

The stronger the amplification, the more noise. So I would say yes to IQ as colors are now less accurate concerning details. Different ISOs also have different dynamic ranges. It won't affect things like sharpness or blur.

Last edited by AVANT; 02-14-2008 at 05:50 PM. Reason: Added more
02-14-2008, 05:40 PM   #3
Pentaxian
Lowell Goudge's Avatar

Join Date: Jan 2007
Location: Toronto
Photos: Gallery | Albums
Posts: 15,312
QuoteOriginally posted by falconeye Quote
This question has been raised in another thread https://www.pentaxforums.com/forums/177023-post72.html and I would be interested in the reponse:

At same aperture/shutter speed, does ISO have impact on IQ?

THEORY #1:
The exact same amount of light is captured, the image imformation only is multiplied by a constant factor which can be undone in photoshop. So, same quality.

THEORY #2:
ISO somehow primes the sensor cells (i.e., makes them more or less sensitive to build up charge with light, maybe by changing a resistor). So, different quality.

Somebody out there knows what ISO is at the electronical level?
You can't change ISO and not change either apature or shutter speed, because you need to modify the amount of light when you change ISO.

I did a simple test when I first got my *istD, and played with shutter and ISO a little. Same shot, same lighting, went 200, 400 , 800 ISO, there was no difference really between 200 and 400, but 800 had more grain but was sharper, due to doubling shutter speed from 400.
02-14-2008, 06:16 PM   #4
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
Original Poster
QuoteOriginally posted by Lowell Goudge Quote
You can't change ISO and not change either apature or shutter speed, because you need to modify the amount of light when you change ISO.
This is not what was meant. Setting all of aperture, speed and ISO manually, you of course can do such a thing.

The question was about the difference of, e.g., exposing right at ISO 800 and underexposing 1 stop at ISO 400 (and then "push up" in photoshop afterwards).

02-14-2008, 06:38 PM   #5
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
Original Poster
QuoteOriginally posted by AVANT Quote
To obtain higher sensitivities from the sensors, more power is sent to it (amplifying the signal). When you amplify a signal, interference is introduced and the A/D converters interprets this as slightly different colors. These random color inaccuracies are what we call noise
Thank you for the interesting reply.

As always, it answered one question and created two new ones with me

One question is whether there is a difference to be expected by amplyífying a signal by a factor of two, electronically, then digitize, or first digitize, then multiply the number by 2.

Another thing which puzzles me is that a sensor cell is supposed to build up charge before it is read out (reading out a cell is decharging it and measuring the electric current at which it decharges -- at least I believe this to be true for CCD). So how can a lower ISO setting prevent the cell to become full?

The point is that, if you are right AVANT, a simple amplification of the current during read-out would not increase the cell's sensitivity and THEORY #1 would still apply.

On the other hand, it could be that higher ISO leads the cell fill up with electric charge more quickly at the price of higher thermodynamic noise. This would be THEORY #2 then.

Still don't know...


A side note on what was written in the other thread... Current sensors measure light as an analog feature. But not because light is analog but because of practical reasons. The photo-electric effect explained by Einstein and used in sensors causes an electron to be separated when a photon hits which adds up to a recombination current which therefore, takes discrete values. Photons could be counted. This is so because of the quantum nature of light, a photon either separates an electron from its atom, or not. 1 or 0. Think of a photon kicking out an electron. And all kicked out electrons combine to flood, the electric current, which eventually gets measured and digitized.

This is why I wonder where the hell anybody could mess up with the sensitivity aka ISO.
02-14-2008, 07:45 PM   #6
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,434
QuoteOriginally posted by falconeye Quote
This question has been raised in another thread https://www.pentaxforums.com/forums/177023-post72.html and I would be interested in the reponse:

At same aperture/shutter speed, does ISO have impact on IQ?

THEORY #1:
The exact same amount of light is captured, the image imformation only is multiplied by a constant factor which can be undone in photoshop. So, same quality.

THEORY #2:
ISO somehow primes the sensor cells (i.e., makes them more or less sensitive to build up charge with light, maybe by changing a resistor). So, different quality.

Somebody out there knows what ISO is at the electronical level?
Depends on the camera and on the iso of a particular model. There were plenty of discussions at the Canon forum because apparently in-between iso's (ie 250 ect) were mathmatically derived as opposed to amp gain. I also have a source that stated (from examining the RAW data of the K10) that it is also strictly "mathmatical"(ALL iso's). I will try to verify this.
And this quote gives a fairly good summary of parts of this mixed bag of AMP gain and post digital math..:
The Sigma’s do use post processing gain to handle ISO changes. The image data in the RAW file is stored at baseline ISO. A few other cameras (mostly older models) use this method as well. Most other cameras apply ISO gain prior to writing the data to their RAW files either at the hardware amplifier level or with the internal camera processor. My guess is that most probably use hardware amplification for capture performance reasons to reduce the amount of post capture processing as much as possible.

The new Canon highlight tone priority mode is sort of similar where the RAW data is stored basically a full stop underexposed and then gain and/or a custom camera tone curve is applied later in software post processing. My understanding is that Canon software uses different camera tone curves for images in this mode. Basically if you blow the exposure at capture time you’ve got an additional full stop of highlight data to work with. Of course if you totally blow the exposure say by 2 stops you’re still out of luck and if you hit the exposure right on you’re basically tossing a full stop of dynamic range in the shadows. The new 14 bit files should make up for this.

Many RAW converters will apply automatic noise reduction based on the stored ISO value so one negative to shooting at ISO 100 and then using digital exposure compensation (gain) later in post processing is that you’d lose this automatic noise reduction.

Not sure of this source but from other readings it seems quite credible and likey. One thing is Canon and now Nikon do do a fair bit of on-chip NR at high iso's to make the FF as clean as it looks... but it is not a "pure capture".......
Nothing is as black and white as it seems.........
What do we mean by over-exposure in this digital age? | photostream
From the same thread:
I went off and re-read some of the Canon technical papers. I’d presume most manufacturers are similar.
ISO is a mix of straight up amplification at the sensor chip plus some magic in the A/D conversion to tweak noise etc.
The HTP, looking at various samples & reading views of those examining files, is a mix of ISO amplification and an adjusted tone curve to roll-off the highlights. Not quite RAW any more but reducing the risk of blown highlights.

When Pentax said they wanted to keep things "film like" I think they also meant to leave the processing to us....
Note: Apologies for this collection of random thoughts..........

Last edited by jeffkrol; 02-14-2008 at 07:54 PM.
02-14-2008, 10:10 PM   #7
Veteran Member
distudio's Avatar

Join Date: Jul 2007
Location: Sydney
Photos: Albums
Posts: 440
QuoteOriginally posted by falconeye Quote
One question is whether there is a difference to be expected by amplyífying a signal by a factor of two, electronically, then digitize, or first digitize, then multiply the number by 2.
It depends on the number of discrete steps vs sensitivity vs linearity of the analogue to digital converter employed. The New K20D sensor employs a programmable gain amplifier (PGA) before the integrated analogue to digital converter (ADC). I assume it's been implemented this way to create the best match between the analogue output of the photocell and the following ADC for the expected levels to be encountered through it's design range. They may also employ digital amplification for high ISOs too. It's a matter of optimizing system signal to noise ratios.

QuoteOriginally posted by falconeye Quote
Another thing which puzzles me is that a sensor cell is supposed to build up charge before it is read out (reading out a cell is decharging it and measuring the electric current at which it decharges -- at least I believe this to be true for CCD). So how can a lower ISO setting prevent the cell to become full?
The sensor has an effective native ISO based loosely on the relationship between the noise floor and the photocell saturation, any deviation from this ideal native ISO should result in a some loss of overall exposure latitude. So it's not that lower ISOs give the sensor more capacity but that setting an ISO higher than the native ISO prevents the sensor from ever being saturated given a normal exposure at the higher effective ISO.

QuoteOriginally posted by falconeye Quote
On the other hand, it could be that higher ISO leads the cell fill up with electric charge more quickly at the price of higher thermodynamic noise. This would be THEORY #2 then.
Yes it fills more quickly as it's been deemed to be diminished in capacity at any ISO higher then the native ISO. Noise is worse because of the amplification factor (digital or analogue) that's required to simulate the higher ISO also raises the level of all the noise components.

Last edited by distudio; 02-15-2008 at 04:37 AM.
02-15-2008, 07:26 AM   #8
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,434
Base iso or sensor iso in general

I am going to assume that "base iso" is equal to the point where dynamic range of the sensor is at max...
Kodak's way(s) of measuring base iso for anyone interested though:
http://www.kodak.com/ezpres/business/ccd/global/plugins/acrobat/en/supportdo...asurements.pdf

There are two basic types of ISO: saturation-based and noise-based. The saturation-based ISO is also
referred to as the ‘base ISO.’.......................To measure the scene luminance, the lighting is varied until the camera outputs an average value
equal to 18/106 of full scale when imaging the 18% reflecting surface (edit note: 18% is somewhat arbitrary, see rest of paper). During this adjustment the
exposure time is held constant. Full scale is either where the imager saturates or where the camera’s
ADC reaches its maximum value, whichever happens first.
The use of the number 106 is not arbitrary—it means that we want to accommodate a reflectance of
up to 106% for image highlights. Very demanding applications might need more headroom for
highlights, and so might use a higher number when calculating base ISO.
The scene luminance is measured with a light meter in cd/m2 and entered into the equation with the f
number and integration time. For a color imaging system, the ISO measurement is performed on the
highest speed channel--green for an RGB imager or yellow for a CMY image sensor.
This measurement depends on just about everything—lenses, IR cut filters, camera gain and offset
settings, mapping of the output voltage to the ADC, just to name a few. So take care that the camera
system is set up exactly as it will be used in the end application.
Kodak’s KAF-16801CE, a 16 megapixel full-frame CCD image sensor with 9 micron pixels, a Red-
Green-Blue color filter array, and no microlenses has a base ISO just over 100 assuming a
demanding 170% reflectance tolerance for highlights. This ISO is more than adequate for the studio
lighting conditions under which this sensor is typically used. Kodak’s KAC-1310, a megapixel
CMOS image sensor with 6 micron square pixels, a Cyan-Yellow-Magenta color filter array, and
microlenses has a base ISO of 180 assuming 106% reflectance. This is more appropriate for the
lower light conditions often encountered by the snap shot photographer.
Reference:
International Organization for Standardization standard 12232, Photography – Electronic Still



Last edited by jeffkrol; 02-15-2008 at 07:59 AM.
02-15-2008, 09:22 AM   #9
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
Original Poster
Thanks for all the interesting replies.

In summary it looks as if it is THEORY #1 rather than THEORY #2. But implementation tweaks in various cameras may still bring out significant differences not to expect from THEORY #1 alone.

As soon as I get my K20D, I will do some shoots of a gray card and compare histograms to find out.
02-15-2008, 11:01 AM   #10
Veteran Member
Gooshin's Avatar

Join Date: Sep 2007
Location: Toronto, the one in Canada.
Posts: 5,611
remember from that same thread, my picture of an ISO 200 image blown up to ISO 3200, vs a normal ISO 3200 image



if all this holds true, the blown up iso 200 image in Lightroom should have been identical to the native iso 3200 shot, except it wasnt, not by a long shot.

so, either Lightroom sucks, or setting the ISO does matter.
02-15-2008, 01:49 PM   #11
Veteran Member
distudio's Avatar

Join Date: Jul 2007
Location: Sydney
Photos: Albums
Posts: 440
QuoteOriginally posted by Gooshin Quote
remember from that same thread, my picture of an ISO 200 image blown up to ISO 3200, vs a normal ISO 3200 image



if all this holds true, the blown up iso 200 image in Lightroom should have been identical to the native iso 3200 shot, except it wasnt, not by a long shot.

so, either Lightroom sucks, or setting the ISO does matter.
Both I assume, the fact is that in Lightroom you can only apply digital gain at the end of the signal path, for various reasons this may be the poorest place to apply gain to minimize noise and maximize latitude for the given ISO.
02-15-2008, 05:18 PM   #12
Pentaxian
Lowell Goudge's Avatar

Join Date: Jan 2007
Location: Toronto
Photos: Gallery | Albums
Posts: 15,312
QuoteOriginally posted by Gooshin Quote
remember from that same thread, my picture of an ISO 200 image blown up to ISO 3200, vs a normal ISO 3200 image



if all this holds true, the blown up iso 200 image in Lightroom should have been identical to the native iso 3200 shot, except it wasnt, not by a long shot.

so, either Lightroom sucks, or setting the ISO does matter.
I think you will find that setting ISO matters because the fundamental difference between RAW and JPEG is the image color depth.

JPEG is 8 bit RAW is 12 bit.

Both formats still compress the color and exposure data off the median band (i.e. below 25 greyscale and above 230 greyscale) so when you try to "push" in any photo editor, you are still trying to pull an image out of the mud, it's just in RAW the mud is a little finer, and you can get some detail that was lost in the compression from 12 bits to 8 bits, BUT when you amplify the data in the sensor, (increase the ISO) you move the whole image up through the range, and have much better definition of the mid values because there was more data available to begin with,
02-15-2008, 07:50 PM   #13
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,434
Quick test....

QuoteOriginally posted by Gooshin Quote
remember from that same thread, my picture of an ISO 200 image blown up to ISO 3200, vs a normal ISO 3200 image



if all this holds true, the blown up iso 200 image in Lightroom should have been identical to the native iso 3200 shot, except it wasnt, not by a long shot.

so, either Lightroom sucks, or setting the ISO does matter.
Well maybe you are being a bit extreme:
Here's 2 images
One at 200, pushed 3 stops (per RSE)
One at 1600 "normal". Small -ev adj just to make things a bit more equal looking..
To be completely honest, the 1600 looked a little better because the sensor (which was swapped out) on my D now has some horiz. pattern noise. The 1600 image came out much flatter than the way underexposed 200 onr though.....

02-16-2008, 10:29 AM   #14
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
Original Poster
QuoteOriginally posted by Gooshin Quote
remember from that same thread, my picture of an ISO 200 image blown up to ISO 3200, vs a normal ISO 3200 image
I do remember. On the other hand, your ISO 800 vs 1600 suggested the opposite.

At ISO200 +4EV, maybe LightRoom failed.

It would be interesting to check the RAW histograms -- with THEORY #1 valid, both histograms should look identical, one only shifted left by 4EV (of course, if this left shifting caused clipping, you cannot compare anymore).

Maybe, this is the only thing the ISO setting really does: prevents zero light to be recorded aka quantization noise (-4EV means that 2^4-1 or 15 (@ISO3200) becomes 0 (@ISO200)).



I'll say that the conclusion in this thread so far, is that compared to ISO200, ISO400 and ISO800 don't have an impact on IQ. At ISO1600, there may be a minor improvement, at ISO3200 the improvement may be significant for areas even dark at ISO3200. Of course, all this applies to RAW only.
02-16-2008, 11:22 AM   #15
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,434
QuoteOriginally posted by falconeye Quote
I do remember. On the other hand, your ISO 800 vs 1600 suggested the opposite.

At ISO200 +4EV, maybe LightRoom failed.

It would be interesting to check the RAW histograms -- with THEORY #1 valid, both histograms should look identical, one only shifted left by 4EV (of course, if this left shifting caused clipping, you cannot compare anymore).

Maybe, this is the only thing the ISO setting really does: prevents zero light to be recorded aka quantization noise (-4EV means that 2^4-1 or 15 (@ISO3200) becomes 0 (@ISO200)).



I'll say that the conclusion in this thread so far, is that compared to ISO200, ISO400 and ISO800 don't have an impact on IQ. At ISO1600, there may be a minor improvement, at ISO3200 the improvement may be significant for areas even dark at ISO3200. Of course, all this applies to RAW only.
Well you practically never see a RAW histogram.... I wouldn't make a lot of visual sense.
Remember the original data is just voltage numbers assigned to a pixel that had a color filter attached to it.
Bayer Patterns and Raw Image File FAQ
BUT you can use things like dcraw or this: See the "statistic" gif's. Not a real image mind you, just a plot of the color point in RAW....i
Re: OK, solved.sort of: Pentax SLR Talk Forum: Digital Photography Review
all your work has been done for you:
If you analyse the identical shots at different ISO's (ie. tripod, same scene and lighting, same exposure compensating ISO change with shutter speed), you will find that the ISO 200 image has a gain that is about 0.5 stops higher than any of the other ISO's, that neither ISO 200 nor ISO 400 have any gaps, that ISO 800 begins to have gaps due to an internal scaling done before writing to the raw file (looks to skip about just over 10th code, which indicates it is scales by a little less than 10%), the ISO 1600 image is just a left bit by one shift of the ISO 800 image and an ISO 3200 image would just be one more left shift. This means that there every other code is dropped at ISO 1600 plus another few due to the scaling, and it keeps only every fourth code at ISO 3200 plus loses a few more due to the scaling.
Re: OK, solved.sort of: Pentax SLR Talk Forum: Digital Photography Review
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
camera, dslr, impact, iso, light, photography, quality, theory
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Quick Guide to Shutter Speed/Aperture/ISO Copyright Troubleshooting and Beginner Help 3 10-24-2010 12:06 PM
Shutter Speed and Aperture Comparison indy Pentax SLR Lens Discussion 6 05-28-2010 06:25 AM
How do you where to start with Aperture and SHutter speed?? hockmasm Pentax DSLR Discussion 22 05-04-2010 12:57 PM
Exposure: Shutter Speed, Aperture, ISO ltdstar Photography Articles 0 06-05-2009 06:27 PM
Please help a K10D n00b with aperture and shutter speed PaddyB Photographic Technique 11 09-25-2008 06:46 AM



All times are GMT -7. The time now is 03:26 PM. | See also: NikonForums.com, part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top