How does Olympus get away with this?

[traveller]

http://www.dxomark.com/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/(appareil1)/909%7C0/(brand)/Olympus/(appareil2)/895%7C0/(brand2)/Canon/(appareil3)/793%7C0/(brand3)/Olympus

Take a look at the SNR 18% graph on the Measurements tab: neither Canon's nor Olympus' measured ISO values are anything near what the user believes they have set. DXO Mark's explanation is that:

"This difference stems from design choices, in particular the choice to keep some “headroom” to avoid saturation in the higher exposures to make it possible to recover from blown highlights." [http://www.dxomark.com/About/In-depth-measurements/Measurements/ISO-sensitivity]

The 70D reports ISO at about 1/4 stop over measured, but the EM5 reports it at a whole stop greater!

Presumably, both manufacturers deliberately underexpose the image and then leave instructions in the RAW file (or jpeg engine) to boost exposure by the requisite amount? This would allow extensive highlight recovery, but may also lead to greater shadow noise than necessary. Doesn't this simply mean that photographers who shoot RAW and know what the exposure histogram shows, have to factor even more ETTR into their calculations? If so, this makes the histogram even more misleading than it already is.

Has this trend been getting worse? Compare:

http://www.dxomark.com/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/(appareil1)/753%7C0/(brand)/Canon/(appareil2)/516%7C0/(brand2)/Canon

The old 1Ds is reporting ISO pretty accurately (at least at the lower values), whereas the 1D X is now at the seemingly standard for Canon, 1/4 stop overstated. Perhaps there is a technical reason why this is now the case that I don't understand. I believe that this makes it dangerous to rely on user reviews that make claims like "camera X produces very clean files up to ISO3200", as you now cannot compare this ISO rating (irony noted!) across brands.

 

[Don Haines]

I think that "electronic fudging" has become the standard for Olympus. It is one of the ways to save cost.... for instance, micro 4/3 lenses have a lot of distortion... Olympus relies on digital manipulation to get a pleasing image... it allows one to get passable results with inferior optics. I really like what they can do with their electronics, but optically, their lenses S__K compared to a good Canon lens.... That said, before they abandoned the 4/3 system, they had some NICE lenses... the 12-60 and the 50-200 were L-glass quality and they had several nice primes, but with micro 4/3 there are few quality lenses and most seem to be fairly slow lenses.

The pinnacle of electronic fudging may be the iPad... watch the "live view" as you pan the camera.... you can see the computer re-drawing the image to correct lens distortion as you move.... and exposure info is meaningless (they don't give it to you anyway)....

 

[dgatwood]

The pinnacle of electronic fudging may be the iPad... watch the "live view" as you pan the camera.... you can see the computer re-drawing the image to correct lens distortion as you move.... and exposure info is meaningless (they don't give it to you anyway)....

The f-stop (fixed) and ISO (adjustable) are there in the EXIF metadata. I wouldn't call either one meaningless, but yes, there's a huge amount of processing that happens to the image, so you can't really make assumptions that correlate noise levels with specific ISO settings, IIRC.

 

[old-pr-pix]

That said, before they abandoned the 4/3 system, they had some NICE lenses... the 12-60 and the 50-200 were L-glass quality and they had several nice primes, but with micro 4/3 there are few quality lenses and most seem to be fairly slow lenses.

Not to fire off a 4/3rds hate discussion, but... all those "NICE" lenses are still there and play well with the new E-M1. Plus, there are the 12mm, 17mm, 45mm, 75mm Oly primes. Not to mention the Leica 25mm, Panasonic X series and Sigma options. All better than average glass - some much better.

Most manufacturers have both less expensive kit lenses and better options. Just for fun, compare the Canon 100 mm f2 on a 70D (160mm eq.) vs. the Oly 75 mm f1.8 on a E-M5 (150 mm eq.) using DxO... the Oly scores 27 while the Canon is only 23. I know DxO scores are always suspect, and Olympus can be dunned for a lot of things, but Olympus does continue to deliver some excellent lenses.

 

[Don Haines]

That said, before they abandoned the 4/3 system, they had some NICE lenses... the 12-60 and the 50-200 were L-glass quality and they had several nice primes, but with micro 4/3 there are few quality lenses and most seem to be fairly slow lenses.

Not to fire off a 4/3rds hate discussion, but... all those "NICE" lenses are still there and play well with the new E-M1. Plus, there are the 12mm, 17mm, 45mm, 75mm Oly primes. Not to mention the Leica 25mm, Panasonic X series and Sigma options. All better than average glass - some much better.

Most manufacturers have both less expensive kit lenses and better options. Just for fun, compare the Canon 100 mm f2 on a 70D (160mm eq.) vs. the Oly 75 mm f1.8 on a E-M5 (150 mm eq.) using DxO... the Oly scores 27 while the Canon is only 23. I know DxO scores are always suspect, and Olympus can be dunned for a lot of things, but Olympus does continue to deliver some excellent lenses.

I have some of that nice glass..... And it's weather sealed too.....

 

[Mt Spokane Photography]

The cameras all expose accurately! Your images will not be under exposed.

DXO is confusing the issue with their measurements, which lead some to think there images are not correctly exposed. You can find DXO's explanation saying this buried in their test methods, but the bottom line is that the ISO's are accurate and DXO is measuring the sensor and not accuracy of the camera at each ISO setting. They are known for inventing their own tests and then claiming they mean something while confusing people in the process.

DPR, for example actually checked the 70D camera and found it to be within 1/6 stop.
http://www.dpreview.com/reviews/canon-eos-70d/15

 

[traveller]

The cameras all expose accurately! Your images will not be under exposed.

DXO is confusing the issue with their measurements, which lead some to think there images are not correctly exposed. You can find DXO's explanation saying this buried in their test methods, but the bottom line is that the ISO's are accurate and DXO is measuring the sensor and not accuracy of the camera at each ISO setting. They are known for inventing their own tests and then claiming they mean something while confusing people in the process.

DPR, for example actually checked the 70D camera and found it to be within 1/6 stop.
http://www.dpreview.com/reviews/canon-eos-70d/15

I wasn't trying to claim that actual exposure was off by this amount, but I take your point: if the exposure is correct and the SNR is acceptable, what the heck does it matter if DXO Mark's measured ISO is different to what the camera reports. Re-reading DXO's tests, I think that the relevance of the DXO method based upon measuring the exposure required to reach sensor saturation, is to allow us to understand what decisions the manufacturer has made about how the sensor should handle the signal.

Given that actual sensitivity of the sensor to light is fixed, I assume that what they're measuring is the amount of 'hardware' amplification that the signal is given. I have seen it written (I am no expert) that all cameras use hardware amplification up to a certain level and then start to boost the signal using 'software' methods. Thus the difference between measured and manufacturers' ISO ratings must be the amount of 'software' amplification after the signal has been converted from analogue to digital. I would assume that software amplification would generate greater noise than hardware amplification (or why bother with the latter?), but again this is beyond my level of expertise. I'm guessing that under amplifying the signal from the sensor allows Olympus to retain more of the highlight data, perhaps at the expense of shadow noise. If Olympus are using one of Sony's excellent current generation sensors, read noise might not be a problem. This would therefore enable them to make the output from their cameras more forgiving in the highlights.

As I have stated previously, this is all speculation on my part as I am no expert in this field. Perhaps you are right when you criticize DXO Mark for presenting this data without more extensive explanation of its implications, thus opening the door for DXO data to be used as fuel for unwarranted claims. Perhaps it is us (well, me in this instance) that are the real problem; a little knowledge is truly a dangerous thing!

 

[Don Haines]

The cameras all expose accurately! Your images will not be under exposed.

DXO is confusing the issue with their measurements, which lead some to think there images are not correctly exposed. You can find DXO's explanation saying this buried in their test methods, but the bottom line is that the ISO's are accurate and DXO is measuring the sensor and not accuracy of the camera at each ISO setting. They are known for inventing their own tests and then claiming they mean something while confusing people in the process.

DPR, for example actually checked the 70D camera and found it to be within 1/6 stop.
http://www.dpreview.com/reviews/canon-eos-70d/15

I wasn't trying to claim that actual exposure was off by this amount, but I take your point: if the exposure is correct and the SNR is acceptable, what the heck does it matter if DXO Mark's measured ISO is different to what the camera reports. Re-reading DXO's tests, I think that the relevance of the DXO method based upon measuring the exposure required to reach sensor saturation, is to allow us to understand what decisions the manufacturer has made about how the sensor should handle the signal.

Given that actual sensitivity of the sensor to light is fixed, I assume that what they're measuring is the amount of 'hardware' amplification that the signal is given. I have seen it written (I am no expert) that all cameras use hardware amplification up to a certain level and then start to boost the signal using 'software' methods. Thus the difference between measured and manufacturers' ISO ratings must be the amount of 'software' amplification after the signal has been converted from analogue to digital. I would assume that software amplification would generate greater noise than hardware amplification (or why bother with the latter?), but again this is beyond my level of expertise. I'm guessing that under amplifying the signal from the sensor allows Olympus to retain more of the highlight data, perhaps at the expense of shadow noise. If Olympus are using one of Sony's excellent current generation sensors, read noise might not be a problem. This would therefore enable them to make the output from their cameras more forgiving in the highlights.

As I have stated previously, this is all speculation on my part as I am no expert in this field. Perhaps you are right when you criticize DXO Mark for presenting this data without more extensive explanation of its implications, thus opening the door for DXO data to be used as fuel for unwarranted claims. Perhaps it is us (well, me in this instance) that are the real problem; a little knowledge is truly a dangerous thing!

It does get interesting when you try to compare two cameras... You set them both at ISO1600 and look at the noise.... camera A appears to be cleaner than B... then you realize that the ISO's are different and you are comparing 800 to 1600....

It looks to me like DXO Mark has come up with a test that proves you can not compare apples to oranges

 

[traveller]

It does get interesting when you try to compare two cameras... You set them both at ISO1600 and look at the noise.... camera A appears to be cleaner than B... then you realize that the ISO's are different and you are comparing 800 to 1600....

It looks to me like DXO Mark has come up with a test that proves you can not compare apples to oranges

Yes, I thought that, but I reconsidered after reading Mt Spokane Photography's comments. The way I see it, if camera A reports ISO1600 and shows a SNR of say 26dB and camera B reports the same ISO and SNR, what does it matter if the sensor output from camera A is only at ISO800 and has been brightened in software? To the photographer, the output looks the same. It is not as if you could brighten the image from camera B by a stop and see less noise, as the SNR is the same.

Unless someone can think of a reason why the opposite is true...?

 

[Don Haines]

It does get interesting when you try to compare two cameras... You set them both at ISO1600 and look at the noise.... camera A appears to be cleaner than B... then you realize that the ISO's are different and you are comparing 800 to 1600....

It looks to me like DXO Mark has come up with a test that proves you can not compare apples to oranges

Yes, I thought that, but I reconsidered after reading Mt Spokane Photography's comments. The way I see it, if camera A reports ISO1600 and shows a SNR of say 26dB and camera B reports the same ISO and SNR, what does it matter if the sensor output from camera A is only at ISO800 and has been brightened in software? To the photographer, the output looks the same. It is not as if you could brighten the image from camera B by a stop and see less noise, as the SNR is the same.

Unless someone can think of a reason why the opposite is true...?

I think I agree with you... To me, this whole thing speaks to how difficult it is to get a comparison between two different brands.....

 

[Kit.]

The way I see it, if camera A reports ISO1600 and shows a SNR of say 26dB and camera B reports the same ISO and SNR, what does it matter if the sensor output from camera A is only at ISO800 and has been brightened in software? To the photographer, the output looks the same.

To the photographer, the output from camera A should look better, because of (potentially) better highlight control.