Nice article on why to shoot RAW...

Started Oct 11, 2007 | Discussions
RomanJohnston
RomanJohnston Forum Pro • Posts: 18,827
Nice article on why to shoot RAW...

http://www.digitalphotopro.com/tech/exposing-for-raw.html

From one of my favorite magazines.

Worth the read.

Roman
--
'Miles to go before I sleep.'
--Robert Frost
http://www.pbase.com/romansphotos/

randyYork Regular Member • Posts: 179
...A day late...

I picked up this mag. yesterday specifically for this article. If i would have knows it would be online i would have saved myself $6.

It is a very good article, thanks for posting to let everyone else know..
--
Randy York
D200, 17-55, 70-200VR

Focal64 Regular Member • Posts: 127
funny that article Echoes ...

everything Thom and Julia have been preaching the past few ...

Ed_C Veteran Member • Posts: 3,808
just a shadow of it

This article is OK but IMO Julia has gone into a lot more detail and actually doesn't always agree with shooting to the right. It isn't just about not clipping highlights.

She went into a lot of interesting detail about optimizing ISO and how conventional wisdom about setting ISO and shooting to the right doesn't actually seem to be correct. Julia found that on her D2x bodies 320 was the optimal ISO (if my memory is correct) and, in fact, from a noise perspective that it was better to under expose and push in processing than to boost the ISO in camera and shoot to the right.

-- hide signature --

Ed C.

Kit Laughlin Regular Member • Posts: 255
Re: just a shadow of it

Would someone kindly give a link to Julia; Thom and multiple others I know, but not her.

I always push exposure on the D2Hs and D200 as far as possible to the right using the colour histograms as it is, but that article definitely is food for thought.
--
Kit Laughlin
http://www.pandf.com.au/ , http://www.bodypress.com.au

kielinski Junior Member • Posts: 33
Re: just a shadow of it

I'd really love to see where someone claims that 320 is the optimal ISO on a D2x.
If the argument holds water, I'd feel like shooting Tri-X all over again.

 kielinski's gear list:kielinski's gear list
Nikon D800 Fujifilm X-T1 Nikon D750 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 300mm f/4D ED-IF +12 more
Ed_C Veteran Member • Posts: 3,808
Re: just a shadow of it

Not sure where all of the references are but here are a couple of the threads:

I believe this was the first reference:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=24182147

Prompting the following
http://forums.dpreview.com/forums/read.asp?forum=1021&message=22330731
http://forums.dpreview.com/forums/read.asp?forum=1021&message=21905418

-- hide signature --

Ed C.

RomanJohnston
OP RomanJohnston Forum Pro • Posts: 18,827
From what I read.....

ISO 340 isnt optimal...just where you want to stop cranking up the iso because EV adjustments in RAW match or do better than shooting the higher ISO.

100 is still optimal.

Roman
--
'Miles to go before I sleep.'
--Robert Frost
http://www.pbase.com/romansphotos/

Bill Janes Senior Member • Posts: 1,848
Re: Absolutely Correct

RomanJohnston wrote:

ISO 340 isnt optimal...just where you want to stop cranking up the
iso because EV adjustments in RAW match or do better than shooting
the higher ISO.

100 is still optimal.

The main source of noise in digital images is due to shot noise, which is inherent in the collection of photons by the sensor and not a limitation of the sensor. Shot noise follows a Poison distribution and the standard deviation of the noise is equal to the number of photons collected. At base ISO, when you expose to the right, you collect the maximum number of photons and get the best signal to noise ratio.

When you expose above base ISO, the electron wells in the sensor are not filled, and the signal to noise ratio is determined mainly by the exposure (which determines the number of photons collected) and not by the ISO. The other main source of noise is read noise, which is noise introduced by transferring the collected electrons (or charge) from the wells of the pixels to the analog to digital converter (ADC) for conversion to a data number and the noise introduced by the ADC itself. Read noise is predominant in the deep shadows. It turns out that read noise is highest at base ISO and decreases ad ISO is increased. Therefore, if exposure is limited by shutter or aperture constraints, it is best to use a higher ISO to reduce read noise. You also get better quantization, since the full range of the ADC is used.

When 1 photo-electron corresponds to one data number (unity gain), then you have captured all the information, and increasing ISO further will not help. These considerations are explained in an article by Roger Clark on his web site. He gives the unity gain for various cameras. It is about ISO 1500 for the D50, 1050 for the D70, and 800 for the D200. He does not give the unity gain for the D2x, but I would expect it to be slightly less than for the D200.

If you expose at unity gain, rather than a higher ISO, then you can increase exposure in the raw converter when needed and get the same results as exposing at higher ISO, but you will have more highlight headroom as Roman points out.

http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/index.html

Bill Janes

Sam Jones Regular Member • Posts: 344
Re: Absolutely Correct

Bill Janes wrote:

RomanJohnston wrote:

ISO 340 isnt optimal...just where you want to stop cranking up the
iso because EV adjustments in RAW match or do better than shooting
the higher ISO.

100 is still optimal.

If you expose at unity gain, rather than a higher ISO, then you can
increase exposure in the raw converter when needed and get the same
results as exposing at higher ISO, but you will have more highlight
headroom as Roman points out.

If you expose at unity gain, rather than a higher ISO, then you can
increase exposure in the raw converter when needed and get the same
results as exposing at higher ISO, but you will have more highlight
headroom as Roman points out.

http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/index.html

Bill Janes

Bill,

In layman's terms lets say for examples sake that at 200 ISO a full aperture of F1.4 requires a shutter speed of 1/8 to expose an important highlight reading somewhere between 250 - 255.

I wish to increase the shutter speed to 1/30 for less chance of camera shake.

1. Should I increase the ISO two stops to 800 ISO and set the shutter speed to 1/30?

or

2. Should I expose at 200 ISO F1.4 at 1/8 sec, thus underexposing by 2 stops, and increase the exposure in ACR by two stops?

Working out manual exposure techniques and adjusting an incident lightmeter/spot meter to each sensor from GM ColorChecker tets, whether very close to 255 (within 1/10th of a stop for normal), or routinely over at +1/3 - +2/3 and recovering highlight data is not an issue for me but sensor science (unity gians, etc) goes over the top of my head as a photographer.

I have heard arguments for underexposing and making large gamma moves in RAW converters but I see ugly noise in the shadows (ACR V4) although the tonal range is less damaged.

I do understand that as the ISO is boosted the tonal range is squeezed and the highlights are affected - less headroom.

Is it to be method 1 or 2 above?

Sam Jones Regular Member • Posts: 344
Re: Absolutely Correct

Error

Sam Jones Regular Member • Posts: 344
Re: Absolutely Correct

Ooops!!!

I meant to say - 2. Should I expose at 200 ISO F1.4 at 1/30 sec, thus underexposing by 2 stops, and increase the exposure in ACR by two stops?

Bill Janes Senior Member • Posts: 1,848
Re: My Advice

You did not say what camera you were using, but let us assume the D200 since Roger has posted data for that camera. Unity gain is at ISO 800, so it doesn't make much sense to use ISOs above 800, but up to that value I would increase the ISO to get the best exposure according to the meter. Above 800, your headroom decreases and you don't really gain anything.

In your case, I would use ISO 800. When you increase exposure in the raw converter, you are not adjusting gamma, which affects the midtones, but rather lifting up the entire exposure in a linear manner. For example, at +1 exposure, every value doubles just as you had changed exposure by 1 stop.
--
Bill

Kim Letkeman
Kim Letkeman Forum Pro • Posts: 33,435
Re: My Advice

So, to summarize:

One could stop using the normal camera controls after unity gain has been achieved and start using underexposure to get increases in shutter speed. There is no real difference between the two techniques, except of course for the inconvenience of switching metering techniques in the middle of the ISO range (which I think is a big inconvenience and will lead to more bad exposures than simply using high ISO boost in the cam, but that's another debate.)

Have there been any decent tests that show the difference in the final image between the two techniques? That is, if we boost by two stops using either of:

1) analog exposure boost at the sensor using high ISO, which allows readout noise to remain in the appropriate proportion to data coming off the sensor and which allows the storage of highlights to use the appropriately allocated bits (i.e. the highest half the bits as expected) and the 6th stop to have 64 levels in which to be stored; versus

2) digital exposure boost in the RAW converter, which stretches out the incoming data (lets say it is 2 stops of boots, so we are taking the bottom 1/4 of the linear data and stretching it to fill the entire space). This quadruples the readout noise in proportion to the inherent high ISO boost noise, so maybe it is a wash. This also means that the 6th stop, instead of having 64 levels in it from a proper exposure, has only 16 levels in it from a 2 stop underexposure. The highest stop has, instead of the expected 2048 levels, only 512 levels.

I've seen attempted tests the last time this debate raged on here (and it raged ) ... but none of them controlled any of the key variables. Many have speculated that the camera is simply applying digital boost after a certain point, but I find that hard to believe. Each well can be amplified before being read out, so why bother boosting digitally when that can easily be done later and does not maintain an appropriate s/n ratio for readout?

Do you know of any such tests?

-- hide signature --
 Kim Letkeman's gear list:Kim Letkeman's gear list
Nikon Coolpix 990 Fujifilm FinePix F770EXR Nikon D600 Nikon D7200 Panasonic Lumix DMC-GX85 +17 more
Sam Jones Regular Member • Posts: 344
Re: My Advice

Bill Janes wrote:

You did not say what camera you were using -

I am using a 1D MKIII.

In your case, I would use ISO 800. When you increase exposure in the
raw converter, you are not adjusting gamma, which affects the
midtones, but rather lifting up the entire exposure in a linear
manner.

Bad use of terminology on my behalf as gamma is applied during demosaicing and not when the exposure is adjusted in ACR - as you say the exposure control in ACR lifts the entire exposure in a linear manner.

Thank you for clarifying the ISO v underexpose and exposure increase in the RAW converter.

Bill Janes Senior Member • Posts: 1,848
Re: Astronomers are experts in low light photography

and you might look at Roger Clark's web site. The reduced number of levels resulting from using less than the full scale of the ADC is often not a major problem, as the lower levels are obliterated by noise with high ISO.

Bill Janes

Bill Janes Senior Member • Posts: 1,848
Re: My Advice

Sam Jones wrote:

Bill Janes wrote:

You did not say what camera you were using -

I am using a 1D MKIII.

I have not seen any test data for the 1D MIII, but with the 14 bit ADC of the MKIII unity gain would occur at a lower ISO, making it less necessary to fiddle with ISO. With a 16 bit ADC, unity gain would occur near base ISO and you could do everything with the raw converter.

-- hide signature --

Bill

mlmusto Senior Member • Posts: 2,109
Re: My Advice

what a great technical discussion. this confirms my lay nontechnical experience w/ my D70S -- ISO above 1000 yield muddy, noisy results -- - shoot indoor club volleyball in very poor lighting where i need 1/400 minimum. have been dialing in EV adjustment (-) to get me to 1/400 but at ISO 1600 and unhappy w/ results.

till the d300 arrives, guess that means max out my ISO at 1000 and see if i can push it in RAW converter. wish i could afford d3 and my troubles would be no more but that is not in the cards.

thanks for the theoretical insight with practical application, all!

 mlmusto's gear list:mlmusto's gear list
Nikon AF-S 105mm F1.4E ED Fujifilm X100S Nikon D700 Nikon D800 Nikon Z6 +1 more
cycle61 Senior Member • Posts: 1,550
Re: Unity gain

Kim Letkeman wrote:

So, to summarize:

One could stop using the normal camera controls after unity gain has
been achieved and start using underexposure to get increases in
shutter speed. There is no real difference between the two
techniques, except of course for the inconvenience of switching
metering techniques in the middle of the ISO range (which I think is
a big inconvenience and will lead to more bad exposures than simply
using high ISO boost in the cam, but that's another debate.)

That's how I understand it also. I don't think it requires you to switch metering techniques, just be aware of where "unity gain" is for your camera and don't go above that. I frequently shoot using manual mode with auto ISO set to max out at 800. I pick shutter speed and aperture as needed, and keep an eye on the ISO indicated in the VF.

Have there been any decent tests that show the difference in the
final image between the two techniques? That is, if we boost by two
stops using either of:

1) analog exposure boost at the sensor using high ISO, which allows
readout noise to remain in the appropriate proportion to data coming
off the sensor and which allows the storage of highlights to use the
appropriately allocated bits (i.e. the highest half the bits as
expected) and the 6th stop to have 64 levels in which to be stored;
versus

2) digital exposure boost in the RAW converter, which stretches out
the incoming data (lets say it is 2 stops of boots, so we are taking
the bottom 1/4 of the linear data and stretching it to fill the
entire space). This quadruples the readout noise in proportion to the
inherent high ISO boost noise, so maybe it is a wash. This also means
that the 6th stop, instead of having 64 levels in it from a proper
exposure, has only 16 levels in it from a 2 stop underexposure. The
highest stop has, instead of the expected 2048 levels, only 512
levels.

Once past unity gain, you have more potential digital output levels than you have electrons to count, so analog amplification is no longer advantageous.

I've seen attempted tests the last time this debate raged on here
(and it raged ) ... but none of them controlled any of the key
variables. Many have speculated that the camera is simply applying
digital boost after a certain point, but I find that hard to believe.
Each well can be amplified before being read out, so why bother
boosting digitally when that can easily be done later and does not
maintain an appropriate s/n ratio for readout?

Do you know of any such tests?

Maybe my test was on of the ones that failed to control enough variables, but I think I did a decent job....
http://forums.dpreview.com/forums/read.asp?forum=1021&message=24058189

The ISO 100 shot looked the worst, until Julia Borg schooled me on the evils of ACR and kindly re-processed the files with RawMagick.
Here are the shots re-processed:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=24064373
The images are in order from ISO 800 to 100.

Rendered at all defaults through RAWMagick, all noise reduction set to
off.

-- hide signature --

Links to fullsize jpegs

http://www.pochtar.com/jb/NDavis/_LND6075.jpg
http://www.pochtar.com/jb/NDavis/_LND6076.jpg
http://www.pochtar.com/jb/NDavis/_LND6077.jpg
http://www.pochtar.com/jb/NDavis/_LND6078.jpg

I was personally amazed at how close a grossly underexposed ISO 100 shot can be compared to a "correctly" exposed ISO 800, if you're careful with the RAW processing. It seems that for values over ISO 800, you're better of with underexposure and restoration by software. This also, as mentioned before, practically eliminates any risk of losing highlights in higher ISO shots.

--
-Nick Davis
I have a full frame Nikon already......an F4s!
Please feel free to critique anything I post. I'm here to learn.

A humble reflection of the beauty in the world around us:
http://nickdavis.zenfolio.com/

Iliah Borg Forum Pro • Posts: 25,779
Re: Absolutely Correct

Dear Bill,

Do you have any links to actual tests confirming unity gain values?

-- hide signature --

no text

Keyboard shortcuts:
FForum MMy threads