Dyn Range

Great Bustard said:

The comprehension fail is all yours. I never mentioned ISO at all --- you did.

Click to expand...

http://forums.dpreview.com/forums/read.asp?forum=1022&message=38981110

Great Bustard said:

And you made an unsubstantiated claim. I'm waiting to see that backed up.

Click to expand...

http://forums.dpreview.com/forums/read.asp?forum=1022&message=38981359

Great Bustard said:

And, as for DR, you're just plain wrong.

Click to expand...

DxO claims:

• E-5: 10.47 EV DR at ISO 200
• E-3: 10.49 EV DR at ISO 100.

User reports:

• E-5 has more DR than the E-3.

My view is that those using both bodies are likely to be able to determine the respective DR capability, thus DxO's methodology is flawed.

kieth kiesel said:

Well, I guess it's time to experiment.

Click to expand...

It would be interesting to know whether those people using 1/3 stop ISO settings would be better served to stick with 1 EV steps and under (or over) expose, then bring the brightness up (or down) in post.

boggis the cat said:

Great Bustard said:

The comprehension fail is all yours. I never mentioned ISO at all --- you did.

Click to expand...

http://forums.dpreview.com/forums/read.asp?forum=1022&message=38981110

boggis the cat said:

And you made an unsubstantiated claim. I'm waiting to see that backed up.

Click to expand...

http://forums.dpreview.com/forums/read.asp?forum=1022&message=38981359

boggis the cat said:

And, as for DR, you're just plain wrong.

Click to expand...

DxO claims:

• E-5: 10.47 EV DR at ISO 200
• E-3: 10.49 EV DR at ISO 100.

User reports:

• E-5 has more DR than the E-3.

Click to expand...

Guess which one I place more stock in? Not to put too fine a point on it, but I can link to some pretty lame pics which even "knowledgable" people use to extol the virtues of a lens, and these are the same people who talk about DR.

boggis the cat said:

My view is that those using both bodies are likely to be able to determine the respective DR capability, thus DxO's methodology is flawed.

Click to expand...

My view is that most people (guess who?) don't even know what DR is. For those people, this is an excellent start:

http://www.dpreview.com/learn/?/Guides/The_art_of_HDR_Photography_part_1_01.htm

boggis the cat said:

kieth kiesel said:

Well, I guess it's time to experiment.

Click to expand...

It would be interesting to know whether those people using 1/3 stop ISO settings would be better served to stick with 1 EV steps and under (or over) expose, then bring the brightness up (or down) in post.

Click to expand...

For ISOless sensors (sensors where the read noise does not depend on ISO), it makes no difference in terms of noise, but using a higher ISO may clip highlights, so it's safer to use base ISO and push in the conversion.

For non-ISOless sensors, the higher ISOs almost invariably have less noise, so it's better (in terms of noise) to use the higher ISO and pull down (although this may clip highlights).

Let's take the E5 as an example:

http://www.sensorgen.info/OlympusE-5.html

At ISO 200, the read noise is 15.4 electrons, whereas at ISO 400 the read noise is 11.2 electrons (about a 0.5 stop difference). So, it is better to use ISO 400 and pull down rather than use a lower ISO (1/3 stop increment or not), although this may result in more blown highlights for a given f-ratio and shutter speed.

For the E3:

http://www.sensorgen.info/OlympusE-3.html

it's an even better idea to use the higher ISO (in terms of noise) since the read noise differential between ISO 200 and ISO 400 is even greater (22.1 electrons vs 13.0 electrons -- about a 1 stop difference).

Of course, the read noise differential in stops doesn't reflect the total noise differential, since read noise is only really noticable in the shadows of a photo, as the rest of the photo is dominated by photon noise, which is unrelated to read noise. However, for those pushing the shadows of a photo, it's an important consideration.

boggis the cat said:

kieth kiesel said:

Well, I guess it's time to experiment.

Click to expand...

It would be interesting to know whether those people using 1/3 stop ISO settings would be better served to stick with 1 EV steps and under (or over) expose, then bring the brightness up (or down) in post.

Click to expand...

thats been done here already for both E3 and E5

--
Riley

any similarity to persons living or dead is coincidental and unintended

no text

Great Bustard said:

boggis the cat said:

kieth kiesel said:

Well, I guess it's time to experiment.

Click to expand...

It would be interesting to know whether those people using 1/3 stop ISO settings would be better served to stick with 1 EV steps and under (or over) expose, then bring the brightness up (or down) in post.

Click to expand...

For ISOless sensors (sensors where the read noise does not depend on ISO), it makes no difference in terms of noise, but using a higher ISO may clip highlights, so it's safer to use base ISO and push in the conversion.

For non-ISOless sensors, the higher ISOs almost invariably have less noise, so it's better (in terms of noise) to use the higher ISO and pull down (although this may clip highlights).

Let's take the E5 as an example:

http://www.sensorgen.info/OlympusE-5.html

At ISO 200, the read noise is 15.4 electrons, whereas at ISO 400 the read noise is 11.2 electrons (about a 0.5 stop difference). So, it is better to use ISO 400 and pull down rather than use a lower ISO (1/3 stop increment or not), although this may result in more blown highlights for a given f-ratio and shutter speed.

For the E3:

http://www.sensorgen.info/OlympusE-3.html

it's an even better idea to use the higher ISO (in terms of noise) since the read noise differential between ISO 200 and ISO 400 is even greater (22.1 electrons vs 13.0 electrons -- about a 1 stop difference).

Click to expand...

Seems a bit strange though, that the read noise is higher at iso200 than at iso100.

Great Bustard said:

Of course, the read noise differential in stops doesn't reflect the total noise differential, since read noise is only really noticable in the shadows of a photo, as the rest of the photo is dominated by photon noise, which is unrelated to read noise. However, for those pushing the shadows of a photo, it's an important consideration.

Click to expand...

boggis the cat said:

no text

Click to expand...

check it out
http://www.flickr.com/groups/olympusesystem/discuss/72157626252480834/

http://forums.steves-digicams.com/olympus-dslr/185525-1-3-iso-noise-examples.html

--
Riley

any similarity to persons living or dead is coincidental and unintended

Rriley said:

check it out
http://www.flickr.com/groups/olympusesystem/discuss/72157626252480834/

http://forums.steves-digicams.com/olympus-dslr/185525-1-3-iso-noise-examples.html

Click to expand...

So the basic message is: use 1 stop ISO steps in preference, or 1 stop +1/3, but skip the lower ISO +2/3 stop increments -- except ISO 1250, for the E-3. Such simplicity... Pretty much in accord with what E-3 owners seem to have found (e.g. ISO 1600 ideally, or 2000 at a push).

The question was really whether, for example, over-exposing ISO 1600 by 1/3 stop then correcting in post would yield a better result than using the "native" ISO 2000 setting. It seems that it would, but it hasn't been specifically tested.

(There may be a difference in correct exposures at ISO 1600 and 2000, where the 2000 is "faked" in-camera, compared to deliberate over-exposure at ISO 1600 then pulling it back in post.)

boggis the cat said:

Rriley said:

check it out
http://www.flickr.com/groups/olympusesystem/discuss/72157626252480834/

http://forums.steves-digicams.com/olympus-dslr/185525-1-3-iso-noise-examples.html

Click to expand...

Click to expand...

boggis the cat said:

So the basic message is: use 1 stop ISO steps in preference, or 1 stop +1/3, but skip the lower ISO +2/3 stop increments -- except ISO 1250, for the E-3. Such simplicity... Pretty much in accord with what E-3 owners seem to have found (e.g. ISO 1600 ideally, or 2000 at a push).

Click to expand...

yep 100 200 400 800 1250 2000, but of all things avoid 320

boggis the cat said:

The question was really whether, for example, over-exposing ISO 1600 by 1/3 stop then correcting in post would yield a better result than using the "native" ISO 2000 setting. It seems that it would, but it hasn't been specifically tested.

(There may be a difference in correct exposures at ISO 1600 and 2000, where the 2000 is "faked" in-camera, compared to deliberate over-exposure at ISO 1600 then pulling it back in post.)

Click to expand...

well Andy did specifically say E3 was tested using SOS and from memory the exposures balance, so ...

that said, it isnt exactly territory i engage, so Im not the person to ask. However I think Greg has been here before ...

--
Riley

any similarity to persons living or dead is coincidental and unintended

I am bring this up again probably for no good reason.

But for interested the DXO DR numbers are calculated this way:

(sensor saturation at the ISO selected - bias offset) divided by the Read noise at this ISO level. Thats it!

Read this if interested:
http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p2.html#readandshot

The read noise is the lowest detected number of photos.

Since for a 14bit ADC the max number of saturation is 16383, for 12-bit ADC the max is 4096.

The number listed for saturation of E3 at 200 iso is 13411 and at 400 iso is 7510. with the read noise numbers at these iso levels of 22.1 and 13 respectfully.

For the E5 at 200 iso the saturation is 18001 and at 400 iso is 9073, with the read noise numbers at 15.4 and 11.3 respectfully.

The claim DR numbers are E3 200: 9.2 E5 200 10.2 E3 400:9.2 E5 400 9.7.

So what does all this mean. Really not much. The unit used to determine DR at DXO is completely useless. Its is the max number of electrons (clip level) to the lowest detected or noise level. If you get the noise level lower you DR improves.

This has nothing to do with color, what color clips first, what shades of color you can see. only pure electrons. Also these numbers are less useful considering the 12bit cut off.

But if you need a number then DXO says that the DR of the E5 is 1 stop better than E3 at ISO 200. But this could because the read nosie for the E3 is unusually high at ISO 200, at ISO 100 it was 16.9. Since read noise drops as you increase ISO it would be expected that the noise would be around 14 at ISO 200 and thus the E3 would have better "DXO DR" than the E5.

dbateman said:

I am bring this up again probably for no good reason.

But for interested the DXO DR numbers are calculated this way:

(sensor saturation at the ISO selected - bias offset) divided by the Read noise at this ISO level. Thats it!

Click to expand...

I think, but am not sure, that they normalize these figures for 8 MP. For example, consider two sensors with the same DR, but one has more pixels than the other. When normalized to the same display dimensions, the photo made with more pixels will have more DR.

dbateman said:

Read this if interested:
http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p2.html#readandshot

Click to expand...

That's one of the best articles on the subject.

dbateman said:

The read noise is the lowest detected number of photos.

Click to expand...

No. The read noise is the additional noise added by the sensor and supporting hardware.

dbateman said:

Since for a 14bit ADC the max number of saturation is 16383, for 12-bit ADC the max is 4096.

Click to expand...

No. One bit does not have to represent one electron.

dbateman said:

The number listed for saturation of E3 at 200 iso is 13411 and at 400 iso is 7510. with the read noise numbers at these iso levels of 22.1 and 13 respectfully.

Click to expand...

Yes, and the DR is the number of stops from the read noise to the saturation. So, at ISO 200, that gives us a DR of log2 (13411 / 22.1) = 9.2 stops, and the DR at ISO 400 is log2 (7510 / 13) = 9.2 stops.

dbateman said:

For the E5 at 200 iso the saturation is 18001 and at 400 iso is 9073, with the read noise numbers at 15.4 and 11.3 respectfully.

The claim DR numbers are E3 200: 9.2 E5 200 10.2 E3 400:9.2 E5 400 9.7.

Click to expand...

Not "claim", but measured and calculated.

dbateman said:

So what does all this mean. Really not much. The unit used to determine DR at DXO is completely useless. Its is the max number of electrons (clip level) to the lowest detected or noise level. If you get the noise level lower you DR improves.

Click to expand...

So dynamic range doesn't "mean much"? That's an odd thing to say. DR measures the number of stops where a pixel can record detail. Some people, such as myself, feel a more meaningful measure of DR is to use the 100% NSR instead of the read noise for the "noise floor". And, most certainly, it makes more sense to speak of DR / area of the photo rather than DR / pixel.

dbateman said:

This has nothing to do with color, what color clips first, what shades of color you can see. only pure electrons. Also these numbers are less useful considering the 12bit cut off.

Click to expand...

DxOMark, in case you didn't notice, does give the bit depth for color.

dbateman said:

But if you need a number then DXO says that the DR of the E5 is 1 stop better than E3 at ISO 200. But this could because the read nosie for the E3 is unusually high at ISO 200, at ISO 100 it was 16.9. Since read noise drops as you increase ISO it would be expected that the noise would be around 14 at ISO 200 and thus the E3 would have better "DXO DR" than the E5.

Click to expand...

Numbers are useful when you know what the measure and how that measurement corresponds to the visual properties of the final photo. Otherwise, the numbers are just numbers.