Exposure basics, lesson two point one (& ISO)

Started Mar 19, 2013 | Discussions
OP texinwien Veteran Member • Posts: 3,326
Re: Of course, i already told you this

... you were just a ball of hubris-fuelled ad hominem. and appeals to (false) authority. I tried, Jack.

I myself see Jack as an intelligent, knowledgeable, and a reasonable fellow to communicate with. There are times when I do find that I may see some technical things somewhat differently than him, but have never known him to in my view be anything of the sort of what you describe above.

None of us are always and invariably immaculate in our reasoning:

http://forums.dpreview.com/forums/post/41980728

Even when we do learn and progress we do not become invincible authorities.

Vigorous technical debate is one thing. I rarely make comments surrounding tone of communication of others, but find such distinct, ongoing personalizations both unfortunate and in very bad taste.

In my very first posts to Mr. Hogan, I was both polite and clear that I did not think less of him for making the mistakes in thinking that he had made. I explained that I understood that the problem was complex, that it was easy to make a simple error in logic, and that I had also been mistaken about these very topics some few months ago. I told him I thought he was 100% right in 99.9% of the argument, but that the last 0.1% was ruining his conclusion.

I approached jack because he struck me as being much more intelligent than the average schlub and told him so, up front. Had he not replied with full on personal attacks and implications that I would be unable to comprehend the explanations of his correctness, and, on those grounds, refused to even justify his positions, well, I'm no saint, and may not respond to this kind of hubris too well, especially when I start out trying to be helpful.

So, there you have it. He's wrong. I would have gladly continued the discussion on a purely friendly level, but if he wishes to make it personal, well, I'm going to take a little pleasure in watching him get his well-deserved comeuppance. Not as zen as I'd like. Perhaps I'll have to do another round as a snail.

Jack sometimes has a bit of a flourish about him, but in my communciations with him I have found him to be overall a rather gentle and reserved fellow who is not at all prone to making personal and mean spirited attacks. Have been following your recent interchanges. While I confess I have not been immersing myself deeply in the technical details of the various averments, I have read the surrounding text, and am familiar with the tone of communications.

Since you are making such charges, could you please provide specific quoted examples of statements made by Jack that you feel rise to the level of "ad hominem" attacks upon you as a person? Thanks.

Read my post to him, linked above. Totally focused on the issue. Totally polite, although he'd already made snide comments elsewhere, suggesting that I had run out of arguments and was trying to 'shoot the messenger' - no idea why.

Then read his reply to the comment of mine linked above. Sorry, but I was being nothing other than polite and patient, and was sincerely hoping to help him post his mental block.

Ah, when I know I'm right, and when I'm trying to politely explain what is right to someone who has made an error in his logic, and his response is that he thinks I wouldn't be able to understand the explanation anyway (accompanied by a BS 'joke' defense of his wrongheaded ideas), yeah, not yet zen enough to just turn the other cheek yet. Working on it, but still have a ways to go.

I did not ask for your personal impressions surrounding statements made by Jack (which I think that you have made evident), or for your own personal impressions of the tone of your statements.

I did ask you to cite:

... specific quoted examples of statements made by Jack that you feel rise to the level of "ad hominem" attacks upon you as a person.

They're spread across a number of his replies, so I suggest you have a look at the original thread.

Here an early post from me to jack (linked for the second time in this thread)

And here the beginning of his reply (quoted and linked, for your convenience):

 texinwien's gear list:texinwien's gear list
Panasonic Lumix DMC-GM5 Olympus E-M5 II Olympus 12-40mm F2.8 OnePlus One Canon EOS 300D +20 more
Detail Man
Detail Man Forum Pro • Posts: 17,026
Re: Of course, i already told you this

... you were just a ball of hubris-fuelled ad hominem. and appeals to (false) authority. I tried, Jack.

I myself see Jack as an intelligent, knowledgeable, and a reasonable fellow to communicate with. There are times when I do find that I may see some technical things somewhat differently than him, but have never known him to in my view be anything of the sort of what you describe above.

None of us are always and invariably immaculate in our reasoning:

http://forums.dpreview.com/forums/post/41980728

Even when we do learn and progress we do not become invincible authorities.

Vigorous technical debate is one thing. I rarely make comments surrounding tone of communication of others, but find such distinct, ongoing personalizations both unfortunate and in very bad taste.

In my very first posts to Mr. Hogan, I was both polite and clear that I did not think less of him for making the mistakes in thinking that he had made. I explained that I understood that the problem was complex, that it was easy to make a simple error in logic, and that I had also been mistaken about these very topics some few months ago. I told him I thought he was 100% right in 99.9% of the argument, but that the last 0.1% was ruining his conclusion.

I approached jack because he struck me as being much more intelligent than the average schlub and told him so, up front. Had he not replied with full on personal attacks and implications that I would be unable to comprehend the explanations of his correctness, and, on those grounds, refused to even justify his positions, well, I'm no saint, and may not respond to this kind of hubris too well, especially when I start out trying to be helpful.

So, there you have it. He's wrong. I would have gladly continued the discussion on a purely friendly level, but if he wishes to make it personal, well, I'm going to take a little pleasure in watching him get his well-deserved comeuppance. Not as zen as I'd like. Perhaps I'll have to do another round as a snail.

Jack sometimes has a bit of a flourish about him, but in my communciations with him I have found him to be overall a rather gentle and reserved fellow who is not at all prone to making personal and mean spirited attacks. Have been following your recent interchanges. While I confess I have not been immersing myself deeply in the technical details of the various averments, I have read the surrounding text, and am familiar with the tone of communications.

Since you are making such charges, could you please provide specific quoted examples of statements made by Jack that you feel rise to the level of "ad hominem" attacks upon you as a person? Thanks.

Read my post to him, linked above. Totally focused on the issue. Totally polite, although he'd already made snide comments elsewhere, suggesting that I had run out of arguments and was trying to 'shoot the messenger' - no idea why.

Then read his reply to the comment of mine linked above. Sorry, but I was being nothing other than polite and patient, and was sincerely hoping to help him post his mental block.

Ah, when I know I'm right, and when I'm trying to politely explain what is right to someone who has made an error in his logic, and his response is that he thinks I wouldn't be able to understand the explanation anyway (accompanied by a BS 'joke' defense of his wrongheaded ideas), yeah, not yet zen enough to just turn the other cheek yet. Working on it, but still have a ways to go.

I did not ask for your personal impressions surrounding statements made by Jack (which I think that you have made evident), or for your own personal impressions of the tone of your statements.

I did ask you to cite:

... specific quoted examples of statements made by Jack that you feel rise to the level of "ad hominem" attacks upon you as a person.

They're spread across a number of his replies, so I suggest you have a look at the original thread.

Here an early post from me to jack (linked for the second time in this thread)

I do not find a single statement by Jack quoted in that post that seems in any way "ad hominem".

And here the beginning of his reply (quoted and linked, for your convenience):

That is just so cruel and heartless of Jack to even imply that he might be "wasting his time". I must say that I have seen far more severe statements posted on these forums in my time. Is that all ?

Why would it not "make much sense" for you to present statements substantiating your allegations ? When people make technical claims of fact, it is reasonable to ask them to directly present the specific bases of their claims. While judgments of "ad hominem" statements may be subjective in nature, why would it not "make much sense" for you to specifically identify such statements ?

Steen Bay Veteran Member • Posts: 7,418
Re: It's about useful comparisons - not which is better

How you can check, without taking DPR's word for it, whether two cameras actually got the same signal (same amount of light per sensor area unit) at the same camera ISO setting is explained here:

http://forums.dpreview.com/forums/post/51123359

As you can see, the criterion of equal signal (same amount of light per sensor area unit) at the same camera ISO setting (I took ISO 200 as an example) was met for the E-M5 and the G3 within an error margin of 1/8 EV. Given that such is the case, what is wrong with comparing the sample images at the same camera ISO setting?

I am trying to follow, but the inconsistencies are starting to pile up In your link above it appears that you have confirmed DxO's relative saturation ISOs within 1/8th of a stop. Please read on for the answer to your question.

It follows logically from the above, that if we compare the studio scene samples with the E-M5 set to ISO 3200 and the G3 to ISO 1600, as you propose, the G3 has received nearly twice the signal of the E-M5. So that comparison certainly does not meet your "same signal" requirement. Given that such is the case, what would be the reason to prefer it?

Perhaps the reason to prefer it is that your assumption that the G3 received twice the signal appears incorrect. Warning, this is about to get very tedious for 95% of readers. The other 5% may actually enjoy it

In the previous paragaph you have confirmed the relative saturation ISOs, but that does not mean that you can use mean raw values directly to estimate the signal (alias exposure, alias the number of photons that hit the sensors). To do that you need good ol' SNR, around a mean of such a level that the vast majority of the N is shot noise. At ISO 3200 and 1600 respectively for our two little cameras that means at around 25-50% of full scale. We find just such a compliant little area in DPR's RAW captures under the letter A of the Kodak Gray Scale. We find that the EM5 at in-camera ISO 3200 has an SNR of 19.39 and the G3 at ISO 1600 has an SNR of 21.04. Squaring these values will give us the number of electrons output by each sensor. That is 376 and 443. Taking into consideration the G3's 11% lower Absolute Quantum Efficiency we can calculate that for the two captures in question the G3 received 0.41 stops more photons,/exposure/signal than the EM5.

So in DPR's test, the one I referred to with the images earlier, the G3 at in-camera ISO 1600 received 0.41 stops more light than the EM5 at ISO 3200, not twice the signal or 1 stop as you suggest. This is fully explained by its 0.32 stop slower shutter speed (1/640 vs 1/800 and possible inaccuracies in the system), as mentioned in my post with the pictures. Since the lens appears to be the same, both cameras were illuminated by the exact same light, within +0.1 of a dB, good job DPR.

All this to say that when these two captures were taken the cameras were receiving virtually the same light, meeting the "same signal, exposure, number of incident photons requirement", within the allowed 1/3 of a stop difference fully accounted for by the different shutter speeds.

So both cameras, the EM5 at ISO3200 and the G3 at ISO1600, were illuminated by the same light, were subject to the same exposure, and would reach saturation at the same absolute luminous exposure, with the usual 1/3 of a stop proviso. A very good match indeed, and imho the most useful of the ones available for the G3 to compare SNR performance with the EM5 at ISO 3200.

Cheers,
Jack

Well, the G3 would win in such a comparison, would that be more fair/useful? (according to DxO the E-M5 has app. 1 dB better SNR 18% and 0.5 Ev better DR at high ISOs).

Great Bustard Forum Pro • Posts: 42,852
Re: Pushing ISO in order to ETTR
1

Certainly. Remember that my entire post is written under the assumption that exposure can't be increased (as implied by the low-light scenario described by Jack Hogan). If exposure can be increased without negative side-effects, it is of course better to stay at ISO 200 and increase exposure to reach the ETTR criterion than to keep exposure unchanged and go to ISO 400. However, if exposure cannot be increased, it is better to go to ISO 400 and ETTR rather than to remain at ISO 200 and stay one EV short of ETTR and that was the point I was making here.

True from ISO 200 to ISO 400, as the improvement in DR is greater than the increase in noise. Not true from ISO 1600 to ISO 3200, as the noise increase is now greater than the DR increase. "Pushing" ISO in order to ETTR only works in the linear noise region of the sensor -- up to ISO 400 or so for CMOS.

Not quite. Increasing ISO so as to ETTR while keeping exposure constant is an advantage in that region where the relationship between camera ISO and DR is nonlinear, i.e., the region in which read noise (as measured in electrons) keeps falling with increasing camera ISO. On the E-M5 it falls up to about ISO 1600 although the main read-noise reduction occurs between ISO 200 and ISO 400.

I thought the EM5 had relatively constant read noise from ISO 800 on up:

http://www.sensorgen.info/OlympusOM-D_E-M5.html

By the way, probably better to refer to "ETTR via higher ISO" as "BTTR" (brightness to the right) or "GTTR" (gain to the right), as the exposure is unaffected.

Chris Noble
Chris Noble Veteran Member • Posts: 3,059
ETTR vs. GTTR
 

By the way, probably better to refer to "ETTR via higher ISO" as "BTTR" (brightness to the right) or "GTTR" (gain to the right), as the exposure is unaffected.

I like it! Not just the Read noise; also analog gain noise?

So "ETTR up to ISO 400"? 800?

 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
richarddd
richarddd Veteran Member • Posts: 3,297
Re: Pushing ISO in order to ETTR

Not quite. Increasing ISO so as to ETTR while keeping exposure constant is an advantage in that region where the relationship between camera ISO and DR is nonlinear, i.e., the region in which read noise (as measured in electrons) keeps falling with increasing camera ISO. On the E-M5 it falls up to about ISO 1600 although the main read-noise reduction occurs between ISO 200 and ISO 400.

I thought the EM5 had relatively constant read noise from ISO 800 on up:

http://www.sensorgen.info/OlympusOM-D_E-M5.html

By the way, probably better to refer to "ETTR via higher ISO" as "BTTR" (brightness to the right) or "GTTR" (gain to the right), as the exposure is unaffected.

Those who insist that people use the official definition of exposure might look a bit odd using their own versions of the well accepted ETTR

 richarddd's gear list:richarddd's gear list
Sony RX100 III Olympus OM-D E-M5 Panasonic Lumix G Vario 7-14mm F4 ASPH Olympus M.Zuiko Digital ED 40-150mm 1:4-5.6 R Olympus M.Zuiko Digital 45mm F1.8 +4 more
Great Bustard Forum Pro • Posts: 42,852
Re: ETTR vs. GTTR

By the way, probably better to refer to "ETTR via higher ISO" as "BTTR" (brightness to the right) or "GTTR" (gain to the right), as the exposure is unaffected.

I like it! Not just the Read noise; also analog gain noise?

The read noise is simply all the sources of noise from the sensor and supporting hardware lumped into a single figure.  Thus, Total Noise = Photon Noise + Read Noise, where "+" represents the quadrature sum:  P + R = sqrt (P² + R²).

So "ETTR up to ISO 400"? 800?

ETTR is always best when possible, as it increases the actual exposure which lowers the photon noise, and this has a more dramatic effect on lowering the overall noise than using a higher ISO to lower the read noise.

Detail Man
Detail Man Forum Pro • Posts: 17,026
Re: G3 vs EM5 - Estimates of Ratios of Light Exposure
2

How you can check, without taking DPR's word for it, whether two cameras actually got the same signal (same amount of light per sensor area unit) at the same camera ISO setting is explained here:

http://forums.dpreview.com/forums/post/51123359

As you can see, the criterion of equal signal (same amount of light per sensor area unit) at the same camera ISO setting (I took ISO 200 as an example) was met for the E-M5 and the G3 within an error margin of 1/8 EV. Given that such is the case, what is wrong with comparing the sample images at the same camera ISO setting?

I am trying to follow, but the inconsistencies are starting to pile up In your link above it appears that you have confirmed DxO's relative saturation ISOs within 1/8th of a stop. Please read on for the answer to your question.

It follows logically from the above, that if we compare the studio scene samples with the E-M5 set to ISO 3200 and the G3 to ISO 1600, as you propose, the G3 has received nearly twice the signal of the E-M5. So that comparison certainly does not meet your "same signal" requirement. Given that such is the case, what would be the reason to prefer it?

Perhaps the reason to prefer it is that your assumption that the G3 received twice the signal appears incorrect. Warning, this is about to get very tedious for 95% of readers. The other 5% may actually enjoy it

In the previous paragaph you have confirmed the relative saturation ISOs, but that does not mean that you can use mean raw values directly to estimate the signal (alias exposure, alias the number of photons that hit the sensors). To do that you need good ol' SNR, around a mean of such a level that the vast majority of the N is shot noise. At ISO 3200 and 1600 respectively for our two little cameras that means at around 25-50% of full scale. We find just such a compliant little area in DPR's RAW captures under the letter A of the Kodak Gray Scale. We find that the EM5 at in-camera ISO 3200 has an SNR of 19.39 and the G3 at ISO 1600 has an SNR of 21.04. Squaring these values will give us the number of electrons output by each sensor. That is 376 and 443. Taking into consideration the G3's 11% lower Absolute Quantum Efficiency we can calculate that for the two captures in question the G3 received 0.41 stops more photons,/exposure/signal than the EM5.

So in DPR's test, the one I referred to with the images earlier, the G3 at in-camera ISO 1600 received 0.41 stops more light than the EM5 at ISO 3200, not twice the signal or 1 stop as you suggest. This is fully explained by its 0.32 stop slower shutter speed (1/640 vs 1/800 and possible inaccuracies in the system), as mentioned in my post with the pictures. Since the lens appears to be the same, both cameras were illuminated by the exact same light, within +0.1 of a dB, good job DPR.

All this to say that when these two captures were taken the cameras were receiving virtually the same light, meeting the "same signal, exposure, number of incident photons requirement", within the allowed 1/3 of a stop difference fully accounted for by the different shutter speeds.

So both cameras, the EM5 at ISO3200 and the G3 at ISO1600, were illuminated by the same light, were subject to the same exposure, and would reach saturation at the same absolute luminous exposure, with the usual 1/3 of a stop proviso. A very good match indeed, and imho the most useful of the ones available for the G3 to compare SNR performance with the EM5 at ISO 3200.

Hi Jack,

I have not been following the technical particulars of this thread closely (having burnt out on comparing the EM5 to the G-Series cameras in all too many previous threads). The gist of this seems to be the ratio of light that the G3 and EM5 have received in the DPR SCT RAWs using the same lens and the same F-Number (F=6.3). So I finally downloaded those RAWs for both cameras recorded at ISO=1600 and ISO=3200, and used ExifTool and RawDigger to analyze them. I averaged the two Green RAW channels and added them to the Red and Blue RAW channels.

I will use EV (powers of 2) to state what I came up with. Calculated to more places than shown.

The EM5 (due to higher QE) transduces 0.236 EV more electrons for the same photon illumination.

The G3 shots (only) have a tiny tad of RAW Green channel clipping (in the specular reflection from the infamous little shiny fake gem near the center of the image-frame). I have ignored that.

ISO=1600

(For the G3 relative to the EM5) at ISO=1600, the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is -0.356 EV, but when receiving 0.678 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received seems to be around:

- 0.356 + 0.678 - 0.236 = + 0.0864 EV (6.168% more light)

ISO=3200

(For the G3 relative to the EM5) at ISO=3200, the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is -0.347 EV, but when receiving 0.700 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received seems to be around:

- 0.347 + 0.700 - 0.236 = + 0.1173 EV (8.468% more light)

G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

Perhaps I flubbed up somewhere in my reasoning - but my results do seem to differ from yours.

DM ...

Detail Man
Detail Man Forum Pro • Posts: 17,026
Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
1
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

Martin.au
Martin.au Forum Pro • Posts: 13,250
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

Silly question, but could we also determine the accuracy of Jack's claim by taking two photos with with a G3 and an OM-D using the same settings, except for ISO. Processing them using the same method as DPR, and, well, looking at them. 1 stop is pretty obvious.

 Martin.au's gear list:Martin.au's gear list
Olympus OM-D E-M5 Olympus E-M1 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G Fisheye 8mm F3.5 Olympus M.Zuiko Digital ED 12-50mm 1:3.5-6.3 EZ +7 more
Jack Hogan Veteran Member • Posts: 6,982
Re: It's about useful comparisons - not which is better

How you can check, without taking DPR's word for it, whether two cameras actually got the same signal (same amount of light per sensor area unit) at the same camera ISO setting is explained here:

http://forums.dpreview.com/forums/post/51123359

As you can see, the criterion of equal signal (same amount of light per sensor area unit) at the same camera ISO setting (I took ISO 200 as an example) was met for the E-M5 and the G3 within an error margin of 1/8 EV. Given that such is the case, what is wrong with comparing the sample images at the same camera ISO setting?

I am trying to follow, but the inconsistencies are starting to pile up In your link above it appears that you have confirmed DxO's relative saturation ISOs within 1/8th of a stop. Please read on for the answer to your question.

It follows logically from the above, that if we compare the studio scene samples with the E-M5 set to ISO 3200 and the G3 to ISO 1600, as you propose, the G3 has received nearly twice the signal of the E-M5. So that comparison certainly does not meet your "same signal" requirement. Given that such is the case, what would be the reason to prefer it?

Perhaps the reason to prefer it is that your assumption that the G3 received twice the signal appears incorrect. Warning, this is about to get very tedious for 95% of readers. The other 5% may actually enjoy it

In the previous paragaph you have confirmed the relative saturation ISOs, but that does not mean that you can use mean raw values directly to estimate the signal (alias exposure, alias the number of photons that hit the sensors). To do that you need good ol' SNR, around a mean of such a level that the vast majority of the N is shot noise. At ISO 3200 and 1600 respectively for our two little cameras that means at around 25-50% of full scale. We find just such a compliant little area in DPR's RAW captures under the letter A of the Kodak Gray Scale. We find that the EM5 at in-camera ISO 3200 has an SNR of 19.39 and the G3 at ISO 1600 has an SNR of 21.04. Squaring these values will give us the number of electrons output by each sensor. That is 376 and 443. Taking into consideration the G3's 11% lower Absolute Quantum Efficiency we can calculate that for the two captures in question the G3 received 0.41 stops more photons,/exposure/signal than the EM5.

So in DPR's test, the one I referred to with the images earlier, the G3 at in-camera ISO 1600 received 0.41 stops more light than the EM5 at ISO 3200, not twice the signal or 1 stop as you suggest. This is fully explained by its 0.32 stop slower shutter speed (1/640 vs 1/800 and possible inaccuracies in the system), as mentioned in my post with the pictures. Since the lens appears to be the same, both cameras were illuminated by the exact same light, within +0.1 of a dB, good job DPR.

All this to say that when these two captures were taken the cameras were receiving virtually the same light, meeting the "same signal, exposure, number of incident photons requirement", within the allowed 1/3 of a stop difference fully accounted for by the different shutter speeds.

So both cameras, the EM5 at ISO3200 and the G3 at ISO1600, were illuminated by the same light, were subject to the same exposure, and would reach saturation at the same absolute luminous exposure, with the usual 1/3 of a stop proviso. A very good match indeed, and imho the most useful of the ones available for the G3 to compare SNR performance with the EM5 at ISO 3200.

Cheers,
Jack

Well, the G3 would win in such a comparison, would that be more fair/useful? (according to DxO the E-M5 has app. 1 dB better SNR 18% and 0.5 Ev better DR at high ISOs).

Yes, we are not talking about which camera is better (we know the EM5 is better per DxO), we are talking about how to compare the two cameras in a manner that will produce useful data to base a buying decicion on.  The point is that, because manufacturers are not consistent in their ISO labelling, comparing two cameras at the same ISO is not always the best choice, especially in cameras like the EM5 that deviate substantially from past common practice.

I didn't choose the EM5-G3 face to face at high ISO, and I donìt particularly care.  It has just proven to be a good example of why more consistent ISOs (perhaps splitting them into a 'gain' and an 'average brightness' component) would be more meaningful for useful comparisons.

Jack

Detail Man
Detail Man Forum Pro • Posts: 17,026
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

Silly question, but could we also determine the accuracy of Jack's claim by taking two photos with with a G3 and an OM-D using the same settings, except for ISO.

Since the above analysis was performed on two RAW images recorded by each camera of the same scene with the ISO settings stated, the (perhaps seemingly obvious ?) answer would be "yes".

Processing them using the same method as DPR, and, well, looking at them. 1 stop is pretty obvious.

DPR provided unprocessed RAWs. In-camera JPG processing would muddy the measurement waters.

Jack Hogan Veteran Member • Posts: 6,982
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

DM, I don't see any squaring of SNRs You can't read off mean values and relate them to electrons/photons/exposure because you do not know each cameras' gain.The only way to get back to relative electrons/photons/exposure is through SNR when most of the noise is shot, which is what I did.

Also, I calculated the difference in QE at these ISOs to be less than 1/6 of a stop, which correlates relatively well with Sensorgen's which you used. On the other hand, I believe mine is more precise because I assume a blackbody radiator at 5000k as the light source, while Bob assumes a green laser

How about a peer review of my post where I show that the lighting under which both the G3's ISO 1600 and EM5's ISO 3200 captures was for all practical intents and purposes identical.above, Bob?

Jack

bobn2
bobn2 Forum Pro • Posts: 60,932
Re: Pushing ISO in order to ETTR
1

Not quite. Increasing ISO so as to ETTR while keeping exposure constant is an advantage in that region where the relationship between camera ISO and DR is nonlinear, i.e., the region in which read noise (as measured in electrons) keeps falling with increasing camera ISO. On the E-M5 it falls up to about ISO 1600 although the main read-noise reduction occurs between ISO 200 and ISO 400.

I thought the EM5 had relatively constant read noise from ISO 800 on up:

http://www.sensorgen.info/OlympusOM-D_E-M5.html

By the way, probably better to refer to "ETTR via higher ISO" as "BTTR" (brightness to the right) or "GTTR" (gain to the right), as the exposure is unaffected.

Those who insist that people use the official definition of exposure might look a bit odd using their own versions of the well accepted ETTR

I wouldn't call ETTR 'well accepted'. Talking to its proponents, you find most have their own individual definition. In fact the original motivation for ETTR (giving more raw levels) was completely bogus. The effect of ETTR is due to increasing exposure at base ISO or minimising read noise at higher ISO's. The idea that it was down to anything else is probably due to not seeing the obvious for lack of knowledge about what the effects of raising exposure are. In any case, ETTR throws the Petersen 'exposure triangle' out of the window, so its not a strong starting point ofr proponents of that.

-- hide signature --

Bob

Jack Hogan Veteran Member • Posts: 6,982
Re: Pushing ISO in order to ETTR

Certainly. Remember that my entire post is written under the assumption that exposure can't be increased (as implied by the low-light scenario described by Jack Hogan). If exposure can be increased without negative side-effects, it is of course better to stay at ISO 200 and increase exposure to reach the ETTR criterion than to keep exposure unchanged and go to ISO 400. However, if exposure cannot be increased, it is better to go to ISO 400 and ETTR rather than to remain at ISO 200 and stay one EV short of ETTR and that was the point I was making here.

True from ISO 200 to ISO 400, as the improvement in DR is greater than the increase in noise. Not true from ISO 1600 to ISO 3200, as the noise increase is now greater than the DR increase. "Pushing" ISO in order to ETTR only works in the linear noise region of the sensor -- up to ISO 400 or so for CMOS.

Not quite. Increasing ISO so as to ETTR while keeping exposure constant is an advantage in that region where the relationship between camera ISO and DR is nonlinear, i.e., the region in which read noise (as measured in electrons) keeps falling with increasing camera ISO. On the E-M5 it falls up to about ISO 1600 although the main read-noise reduction occurs between ISO 200 and ISO 400.

I thought the EM5 had relatively constant read noise from ISO 800 on up:

Yes GB, the issue in this comparison is that the G3's is not from 1600 to 800. Which is the wrong assumption I made (I am not an MFT guy) a few posts up and why AndersW had (half ;-)) a point in that case.

Detail Man
Detail Man Forum Pro • Posts: 17,026
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

DM, I don't see any squaring of SNRs You can't read off mean values and relate them to electrons/photons/exposure because you do not know each cameras' gain.The only way to get back to relative electrons/photons/exposure is through SNR when most of the noise is shot, which is what I did.

That's interesting, Jack. I do follow your procedure in the quoted text below. So, you are saying that since the ratio between each camera's ADU/Photon gain may be different, then simply dividing the ratio of Average RAW levels by the ratio of DxOMarks "Saturation ISOs" is not sufficient. That seems worthy of consideration. I wonder what Anders thinks about that. Will be interesting to hear from him.

Also, I calculated the difference in QE at these ISOs to be less than 1/6 of a stop, which correlates relatively well with Sensorgen's which you used. On the other hand, I believe mine is more precise because I assume a blackbody radiator at 5000k as the light source, while Bob assumes a green laser

How did you calculate the QE then ? Seems like it would be a factor in the ADU/Photon gain ? The SNR varies by the square-root of QE multiplied together with the number of Photons. How would one be able to separate-out the QE from the number of Photons ? Your stated 11% (below) is quite a bit less than the 17.78% difference that Bob's Senorgen data indicates. How do you relate it to a black-body radiator (as opposed to a monochromatic Green wavelength) ?

QUOTED FROM JACK'S PREVIOUS POST:

... you have confirmed the relative saturation ISOs, but that does not mean that you can use mean raw values directly to estimate the signal (alias exposure, alias the number of photons that hit the sensors). To do that you need good ol' SNR, around a mean of such a level that the vast majority of the N is shot noise. At ISO 3200 and 1600 respectively for our two little cameras that means at around 25-50% of full scale. We find just such a compliant little area in DPR's RAW captures under the letter A of the Kodak Gray Scale. We find that the EM5 at in-camera ISO 3200 has an SNR of 19.39 and the G3 at ISO 1600 has an SNR of 21.04. Squaring these values will give us the number of electrons output by each sensor. That is 376 and 443. Taking into consideration the G3's 11% lower Absolute Quantum Efficiency we can calculate that for the two captures in question the G3 received 0.41 stops more photons,/exposure/signal than the EM5.

So in DPR's test, the one I referred to with the images earlier, the G3 at in-camera ISO 1600 received 0.41 stops more light than the EM5 at ISO 3200, not twice the signal or 1 stop as you suggest. This is fully explained by its 0.32 stop slower shutter speed (1/640 vs 1/800 and possible inaccuracies in the system), as mentioned in my post with the pictures. Since the lens appears to be the same, both cameras were illuminated by the exact same light, within +0.1 of a dB, good job DPR.

All this to say that when these two captures were taken the cameras were receiving virtually the same light, meeting the "same signal, exposure, number of incident photons requirement", within the allowed 1/3 of a stop difference fully accounted for by the different shutter speeds.

So both cameras, the EM5 at ISO3200 and the G3 at ISO1600, were illuminated by the same light, were subject to the same exposure, and would reach saturation at the same absolute luminous exposure, with the usual 1/3 of a stop proviso. A very good match indeed, and imho the most useful of the ones available for the G3 to compare SNR performance with the EM5 at ISO 3200.

bobn2
bobn2 Forum Pro • Posts: 60,932
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

DM, I don't see any squaring of SNRs You can't read off mean values and relate them to electrons/photons/exposure because you do not know each cameras' gain.The only way to get back to relative electrons/photons/exposure is through SNR when most of the noise is shot, which is what I did.

Also, I calculated the difference in QE at these ISOs to be less than 1/6 of a stop, which correlates relatively well with Sensorgen's which you used. On the other hand, I believe mine is more precise because I assume a blackbody radiator at 5000k as the light source, while Bob assumes a green laser

How about a peer review of my post where I show that the lighting under which both the G3's ISO 1600 and EM5's ISO 3200 captures was for all practical intents and purposes identical.above, Bob?

Hi Jack,

Don't have the time for more than superficial posts (ones I fit in between real work) right now, but I will get onto it. I think the abuse you're taking for just putting your point of view here is quite unjustified, whether you're wrong or right (not to include DM as an abuser, since he has also pointed out this).

-- hide signature --

Bob

Jack Hogan Veteran Member • Posts: 6,982
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
I wonder what Anders thinks about that. Will be interesting to hear from him.

I am waiting with garlic breath

Also, I calculated the difference in QE at these ISOs to be less than 1/6 of a stop, which correlates relatively well with Sensorgen's which you used. On the other hand, I believe mine is more precise because I assume a blackbody radiator at 5000k as the light source, while Bob assumes a green laser

How did you calculate the QE then ? Seems like it would be a factor in the ADU/Photon gain ? The SNR varies by the square-root of QE multiplied together with the number of Photons. How would one be able to separate-out the QE from the number of Photons ? Your stated 11% (below) is quite a bit less than the 17.78% difference that Bob's Senorgen data indicates. How do you relate it to a black-body radiator (as opposed to a monochromatic Green wavelength) ?

Perhaps it would be worth a separate thread in the Open Forum. Here is a sneak preview with the D5200 as the subject

Jack

PS WTH is going on with the editor?

Martin.au
Martin.au Forum Pro • Posts: 13,250
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure
G3 at ISO=1600 and EM5 at ISO=3200

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving -0.322 EV less light (due to it's higher Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

SHOULD READ:

(For the G3 relative to the EM5), the ratio of the Average RAW channel levels divided by the ratio of the DxOMark "Saturation ISO" ratings is 1.494 EV, but when receiving 0.322 EV more light (due to it's lower Shutter Speed), and transducing -0.236 EV less electrons due to it's lower QE. So, the G3/EM5 ratio of the amount of light received in that case seems to be around:

+ 1.49 - 0.322 - 0.236 = + 0.9358 EV (91.289% more light)

DM, I don't see any squaring of SNRs You can't read off mean values and relate them to electrons/photons/exposure because you do not know each cameras' gain.The only way to get back to relative electrons/photons/exposure is through SNR when most of the noise is shot, which is what I did.

Also, I calculated the difference in QE at these ISOs to be less than 1/6 of a stop, which correlates relatively well with Sensorgen's which you used. On the other hand, I believe mine is more precise because I assume a blackbody radiator at 5000k as the light source, while Bob assumes a green laser

How about a peer review of my post where I show that the lighting under which both the G3's ISO 1600 and EM5's ISO 3200 captures was for all practical intents and purposes identical.above, Bob?

Jack

Are you able to do this calculation for the E520, based on its old studio scene image?

http://www.dpreview.com/reviews/olympuse520/26

Then, if there's any issue with calculations, we at least have two cameras that can provide real world data.

 Martin.au's gear list:Martin.au's gear list
Olympus OM-D E-M5 Olympus E-M1 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G Fisheye 8mm F3.5 Olympus M.Zuiko Digital ED 12-50mm 1:3.5-6.3 EZ +7 more
Jack Hogan Veteran Member • Posts: 6,982
Re: Correction: G3 vs EM5 - Estimates of Ratios of Light Exposure

Are you able to do this calculation for the E520, based on its old studio scene image?

Afraid not, you need the raw data.

Keyboard shortcuts:
FForum MMy threads