Well, it might interest some...

[Artichoke]

Advertisement (gone after registration)

I question their methods

DXO does not support M8's DNG files

how did they do the conversions?

also look at the sensor scores from the Fujifilm S3 and S5 which use the identical sensor

these are quite different (color depth is 30/50 for one and 42/50 for the other)

clearly there is something amiss with DXO Mark

[Vieri]

...

 

 

also look at the sensor scores from the Fujifilm S3 and S5 which use the identical sensor

these are quite different (color depth is 30/50 for one and 42/50 for the other)

clearly there is something amiss with DXO Mark

 

Or you simply didn't consider that Fuji might have used the same sensor but optimizing its results. It happens all the time when a manufacturer uses the same sensor in a new generation of cameras; see Nikon and its use of the 12 MP sensor, or of the 6 MP sensor in subsequent generations of cameras. It sounds simply normal to me that the S5 offers better RAWs than the S3, as a matter of fact, (Leica aside) all manufacturers tend to do that: though, of course, I can see where your doubts come from, being we used to Leicas, where the 8.2 offers EXACTLY the same results as far as sensor outputs is concerned as the M8 for just a thousand bucks more...

[AlanG]

I question their methods

DXO does not support M8's DNG files

how did they do the conversions?...

 

They do not use DXO Optics for conversion or any other converter. They analyze the actual raw data.

 

"Only RAW-sensor performance evaluation allows comparison of the actual data delivered by the sensors, regardless of any image post-processing. Specifically, any demosaicing algorithm reduces noise by distinguishing between tiny details or patterns and noise; such a process is typically a design tradeoff made during algorithm tuning. Comparing only RAW data eliminates this possible source of bias when evaluating camera sensors."

[sean_reid]

They do not use DXO Optics for conversion or any other converter. They analyze the actual raw data.

 

"Only RAW-sensor performance evaluation allows comparison of the actual data delivered by the sensors, regardless of any image post-processing. Specifically, any demosaicing algorithm reduces noise by distinguishing between tiny details or patterns and noise; such a process is typically a design tradeoff made during algorithm tuning. Comparing only RAW data eliminates this possible source of bias when evaluating camera sensors."

 

When I first looked at their test information I was impressed that they were working that way. It does indeed remove some potential confounding variables.

 

Cheers,

 

Sean

[georg]

Thanks for the nice explanations, but somehow I have the feeling that we're not on the same page...

 

"Why are the dark areas with a low-contrast lens brighter? Where does the light come from?"

"From other, brighter parts of the image, obviously."

 

Exactly, that was my point! But this light doesn't belong here, it's false information!?

 

"Replace the lens by a lens that shows no contrast.

The whole image will be exposed in zone 5 with 18% noise."

 

Well, now you have perfect exposure on bright & dark areas but you also lost all the information, to my understanding, a low-contrast-lens doesn't "compress" the DR but rather covers the whole image with a "veil of light" (I hope that's the right translation).

 

You'll take a picture of a low contrast-scene (myst in the morning) and only the high-contrast lens can deliver the information while shooting a high-contrast-scene (sunny desert) you need the opposite performance (simply a "bad" lens, no lens-designer is aiming for) to deliver the information!? I simply don't come over that ;-)

 

When I find the time, I will make & post a comparison between my 50Asph and my '54 5cm summicron - let's see what happens.

[jaapv]

Georg, it is that contrast is seen as a separate function.It is not. In reality it is closely related to the frequency response of a lens and with that to its resolution. For that reason MTF curves tell us something about the quality of the lens, and those are based on contrast.

[sean_reid]

Advertisement (gone after registration)

When I find the time, I will make & post a comparison between my 50Asph and my '54 5cm summicron - let's see what happens.

 

That's a good idea. Try it on a bright (contrasty) outdoor subject and expose for the highlights. See what you think.

 

Cheers,

 

Sean

[sandymc]

Well, now you have perfect exposure on bright & dark areas but you also lost all the information, to my understanding, a low-contrast-lens doesn't "compress" the DR but rather covers the whole image with a "veil of light" (I hope that's the right translation)

 

Well, I'd argue that it does reduce DR - basically, what's happening is that light from the highlights is being redistributed into the dark portions. AKA, the highlights get darker, and the blacks go grey.......so DR is getting lower.

 

I'd also agree that information is being destroyed - this is not really behavior you want in a lens. But blown highlights also destroy information!

 

Sandy

[sean_reid]

Well, I'd argue that it does reduce DR - basically, what's happening is that light from the highlights is being redistributed into the dark portions. AKA, the highlights get darker, and the blacks go grey.......so DR is getting lower.

 

Sandy

 

Hi Sandy,

 

Yes, of course. Compression of the tonality means, by definition, that the DR delivered by the lens is reduced. That's the whole idea here - compressing the contrast range slightly before it reaches the sensor. And the point of the compression, of course, is to keep the shadow information above the noise floor while still holding the highlights.

 

And, of course, I don't think anyone is arguing that moderate contrast lenses are "better" in any broad sense. They just provide certain pros and certain cons. For some kinds of lighting, the mixture of pros and cons they offer is appealing to some.

 

For photographers reading this who've never done much darkroom to work it is important to keep in mind that a higher contrast negative can limit the range of tonality (in the print) that shows detail. A longer scale, lower contrast, negative gives one more room in printing because the detail is held from bright to dark. This relationship has been known for decades (see Ansel Adam's "The Negative" for example) and every good exhibition printer knows it (or should). The Zone system is, in part, predicated on this understanding.

 

What lower contrast lenses can do, in certain lighting, is to essentially give one a lower contrast "negative" (RAW file). And digital adds a noise floor component that is different from film. In film one can't recover thin shadows or (after a point) blocked-up highlights. In digital we have over-exposed highlights and shadow detail at the noise floor level to contend with.

 

By all means, try this and see.

 

Cheers,

 

Sean

[sandymc]

Sean,

 

Yes. And I'd admit to being a big fan of the zone system, as unfashionable as it might be today.

 

The thing is, as undesirable as a lens "losing" information might be, as photographers we only have three choices when the scene DR exceeds the sensor's ability to capture it:

 

1. Blow the highlights

 

2. Lose detail in the shadows

 

3. Or use a "low contrast" lens, and accept that we are losing something in the way of detail in the process.....

 

As I mentioned to you elsewhere, sometimes (3) is the best option.

 

Sandy

[sean_reid]

Sean,

 

Yes. And I'd admit to being a big fan of the zone system, as unfashionable as it might be today.

 

The thing is, as undesirable as a lens "losing" information might be, as photographers we only have three choices when the scene DR exceeds the sensor's ability to capture it:

 

1. Blow the highlights

 

2. Lose detail in the shadows

 

3. Or use a "low contrast" lens, and accept that we are losing something in the way of detail in the process.....

 

As I mentioned to you elsewhere, sometimes (3) is the best option.

 

Sandy

 

Hi Sandy,

 

Yes, those are indeed the only options (substituting "lower contrast" for "low contrast). So, we basically see this quite similarly.

 

Cheers,

 

Sean

[jrc]

And then an argument might be the Leica glass, but try the D700 with Zeiss Glass.....

 

I have a D3 and I bought the Zeiss ZF 25 f2.8 just to see what the talk was all about, and to tell you the truth, it didn't exactly bowl me over. Nice lens, good machinery, etc., but I've got the Nikon 14-24 and to my eye, it's just as good. Sean Reid has tested some ZF lenses and he likes them, but in the one test I remember, which I think was at 50mm, he didn't find a screaming difference in quality between the Zeiss and a comparable Nikon, except that the Nikon was AF...

 

JC

[henning]

I'm coming rather late to this discussion. I had a look at the DXO site a couple of weeks ago, and was both impressed by the effort and wondering why my subjective experience wasn't really substantiated.

 

I have M8's, and have used mostly Canon DSLR's. I have some limited experience with Nikons and various other DSLR's and compact cameras, and I also have a G10.

 

Particularly the latter caused me to consider how the data was acquired, because clearly the lens used was the lens that comes built in. It appears to be a decent lens, but clearly is not a lens that matches Leica or Canon L primes. How is that fact separated from the acquired data? In any case, the data acquired from the sensors for the interchangeable lens cameras also depends on the lenses, even though an effort can obviously be made to reduce the effects of specific lenses. But in this scenario only cameras using the same lens can truly be compared, especially in cases of relatively similar performance.

 

My experience reflects that of others, namely that the M8 files are overachievers. This might have to do with my slightly greater familiarity with rangefinders and constant use of Leicas for over 45 years compared with 40 years of SLR usage of various brands. Probably for that reason I tend to produce more pictures that I like with rangefinders than with SLR's. This becomes self fulfilling since I use them more for that reason.

 

In any case, in spite of having many top rated Canon primes, files from the 5D are generally not as good as the M8 files, unless ISO's over 800 are required.

 

With the 5D I tend to increase the ISO in low light; with the M8 I shoot wide open. With this choice I still generally use the M8 as the the files from the M8/Noctilux at 640 are still more appealing to me than 5D files with the 50/1.4 at 1600.

 

As far as dynamic range is concerned, I'm surprised that the M8 didn't do as well as many other cameras, but I haven't really tested that. I have some old lenses, including an Elmar 50 from an early Standard, that are very low contrast, but most of my lenses are quite new. I generally don't like files that have the dynamic range compressed due to very low contrast and flarey lenses. Since I used to shoot Kodachrome a lot, I'm quite at home with short dynamic range results and can generally live with what the M8 delivers. Until a full frame M with better high ISO performance comes along, the M8 is just fine. I'll keep using it as my main camera for personal use, and otherwise, I'll be getting a 5DII shortly, maybe get the 18 and/or 21 Zeiss lenses, and hope Canon keeps on upgrading cameras and lenses.

 

I still would like to know more specifics on how DXO tests are conducted, and to what extent factors other than the sensor and firmware are involved.

 

Henning