p.2 #2 · WTH? Really? Any Pentax Users Can Verify?
you are assuming that the tweaking of the curves and other processing inside the camera past the sensor is the same. that and that Samsung sensors in general are known for their noise. Sony sensors used in these cameras are much less noisy at the same resolution.
Herb...
Bifurcator wrote:
Does this look right to you?
Is there really that much more DR in the entry level K-x than there is in the Semi-Pro K-7?
p.2 #3 · WTH? Really? Any Pentax Users Can Verify?
Well, to me it is - mostly.
I keep looking at newer and newer sensors from release to release and a few of the test sites like DXO and DPR say and show how much improved they are at noise suppression. Yet when an honest user shows up and posts a RAW or when I go to the shops and record a few of my own, it turns out to be mostly bullshit and the chip is delivering RAW files with the amount and size of noise exactly as it should for the chip's size and density. There have been a few exceptions where the sensor's supporting circuitry and components (on-chip and off) were substandard like with early Panasonic µ4/3's but other than that nope. The latest one and the one with perhaps the biggest BS story to it is the Nex 5n. I need to go record a few for myself but the RAW images I've received from 5n owners show DPR at least to be cheating or fudging somehow.
DR on the other hand I don't think I understand. I mean I know what it is and I can see the results of it in tones and levels but what's responsible for it exactly eludes my full understanding. Sensor size is related, I think the size and shape of the photo-cites is related as well but I'm not exactly sure of the dynamics there. Etceteras...
So yeah, I am expecting and assuming some things. So far I haven't read or seen any reason not to.
p.2 #4 · WTH? Really? Any Pentax Users Can Verify?
The tech stuff makes my head spin. The only thing I pay attention to is whether or not the camera is able to achieve the images I want. Having owned both the K-x and K-7 at one point, I'll vouch for the K-x's superior high ISO capability. It was my "go to" camera for low light situations. As for features, I definitely liked the K-7 more in terms of ergonomics, manual controls, and pentaprism (vs. K-x's pentamirror). I've since sold both and moved on to the K-5, which to me combine the best of both cameras.
One thing about the K-x that drove me crazy was how it had no AF illuminated points in the viewfinder. With such a dark VF, focusing in low-light was rather difficult, even with center point focusing.
p.2 #5 · WTH? Really? Any Pentax Users Can Verify?
Bif;
There has been a tendency for manufacturers to not expose as high as they can on newer cameras, but to leave a more generous highlight headroom. This is only noticeable in raw.
To get the most out of a new camera, you often have to adjust exposure by +2/3. But then you lose the flexibility of the file, you get very sensitive to less than perfect exposures.
You can also notice this if you look at the absolute exposures. A newer camera will often chose a shorter exposure time at the same ISO / aperture - this results in less actual light reaching the sensor - and this of course affects the amount of noise.
If you want to compare cameras, there's really only one way to do it... If you compare cameras of the same brand (like here, Pentax-Pentax) it's really easy...
Shoot the first one in "A" mode at a certain ISO. Note the settings. Put that lens on the other camera, and in manual mode shoot the SAME absolute exposure and ISO. Then compare.
You don't think you can stretch the budget to a K-r? It's newer, so you might not find it at the same bargain prices, but over here, the price differential isn't that big. 100 bucks, maybe.
That's really a more competent camera BODY than the K-x, with several quite nice updates, and a more "quality" feel to it.
but does it really make that much difference? Almost two whole stops? Or is this one of those areas where DXO it producing a data-fart?
Yes, and the explanation is very simple...The law of quantum physics!
Given a fixed sensor size, the more pixels you try to cram onto it, the smaller the photosites become and therefore the less sensitive and noisier they become. Its unavoidable.
The K-x not only has slightly larger sensor than the K-7 but it has lower resolution than the K-7 too which of course means that it has considerably larger photosites than the K-7, and therefore its photosites are less noisy and more sensitive, increasing its dynamic range too.
In other words, you can have great high ISO performance and dynamic range, but only by compromising by having a having lower resolution.
The 21.1mp Canon EOS 1Ds MkIII and the 18mp Canon EOS 1Dx are perfect examples of this "law of deminishing returns"...Both are large pro bodies with full frame sensors but the one with higher resolution can only goto ISO3200 (as an option) whereas the lower resolution one can goto ISO51200 (and even higher with boost!)....Thats means that even though the 1Ds MkIII seems to only have a modest increase in resolution compared to the 1Dx, it makes the 1Ds MkIII at least 4 stops less sensitive!
So 2 stops between K-x and K-7 is certainly possible.
p.2 #7 · WTH? Really? Any Pentax Users Can Verify?
Sorry, but the Nikon D3x has twice the amount of pixels as the D3s, and the D3x has a very real 1.5Ev more of DR at base ISO.
Quantum physics state that the noise is always sqrt(e-) in the light measured, it doesn't matter if you measure 10,000 photons or 1,000 photons. Smaller pixels do sometimes - if constructed badly - have less efficiency, but that is not generally true. The D7000 for instance has 25% smaller pixels than the D300s, and is almost 50% more efficient.
The recipe for DR and pictorial noise is sensor area x sensor efficiency (and then subtract for electronic noise). Within reasonable sizes, pixel size doesn't even register as a parameter. What pixel size does is to alter the requirements on the interpolation engine, a thing most software developers are still sadly lagging behind on.
The Canon G12 is another very good example. It has 2x2 = 4µm² pixels, and it's more than twice as effective as the more than the fifteen times bigger pixels in the Canon 5Dmk1 (8x8 = 64µm²). It also has a lower noise sum per mm² of sensor area. An 5Dmk1 camera sensor built with G12 pixel technology would be about twice as good as the original (ISO6400 would be just as good as ISO3200 on the original). And it would be 216MP.
p.2 #8 · WTH? Really? Any Pentax Users Can Verify?
Well, you don't have to TRUST DPreview to use their raw files... :-) They do have the K-7 included in their comparator database
Use the download marker in the comparison window (your top screencap) and check them out yourself. Mind the EXIF data - check that exposure was the same. If they aren't give the camera with less absolute exposure (shutter speed / aperture) some mental note handicap when looking at the results.
In the linked shots, the K7 is given 1/3Ev more light.
I just like the K-r for the slight increase in body "feel". Even though the K-x is pretty good by entry-level standards, the K-r is better. The better viewfinder and LCD helps too. But I doubt you'll see the almost not measureable difference between the images they produce.
p.2 #9 · WTH? Really? Any Pentax Users Can Verify?
theSuede wrote:
Sorry, but the Nikon D3x has twice the amount of pixels as the D3s, and the D3x has a very real 1.5Ev more of DR at base ISO.
But the D3x can only goto ISO1600 (ISO6400 with boost) but the lower resolution D3s can goto ISO12800 (ISO102400 with boost).
theSuede wrote:
Quantum physics state that the noise is always sqrt(e-) in the light measured, it doesn't matter if you measure 10,000 photons or 1,000 photons. Smaller pixels do sometimes - if constructed badly - have less efficiency, but that is not generally true. The D7000 for instance has 25% smaller pixels than the D300s, and is almost 50% more efficient.
Very true but I'm not sure that the 1.9 stops difference in DR between them equals 50% more efficient?
The D7000 may not seem to strictly obey the LOP, but lets not forget its also about 1 year newer technology than the D300s, so significant sensor improvements could have been implemented in that time.
theSuede wrote:
The Canon G12 is another very good example. It has 2x2 = 4µm² pixels, and it's more than twice as effective as the more than the fifteen times bigger pixels in the Canon 5Dmk1 (8x8 = 64µm²). It also has a lower noise sum per mm² of sensor area. An 5Dmk1 camera sensor built with G12 pixel technology would be about twice as good as the original (ISO6400 would be just as good as ISO3200 on the original). And it would be 216MP.
Thats easy to understand if you factor in that the older technology 5D might only have 3 Transistors per photosite and the the newer technology G12 may have 4 Transistors per photosite...
p.2 #10 · WTH? Really? Any Pentax Users Can Verify?
theSuede wrote:
Well, you don't have to TRUST DPreview to use their raw files... :-) They do have the K-7 included in their comparator database
No but maybe the camera they were given? And RAW files ARE editable. All files are. The Nex 5n RAW files have changed since they first went up BTW... I dunno if that means anythng or not. They are different too. Now they are normal looking whereas before they virtually noiseless at 3200. <shrug>
Use the download marker in the comparison window (your top screencap) and check them out yourself. Mind the EXIF data - check that exposure was the same. If they aren't give the camera with less absolute exposure (shutter speed / aperture) some mental note handicap when looking at the results.
In the linked shots, the K7 is given 1/3Ev more light.
I just like the K-r for the slight increase in body "feel". Even though the K-x is pretty good by entry-level standards, the K-r is better. The better viewfinder and LCD helps too. But I doubt you'll see the almost not measureable difference between the images they produce....Show more →
That was my take too. Thanks for putting this together BTW, very cool!
Or is this one of those areas where DXO it producing a data-fart?
The Nex 5n RAW files have changed since they first went up BTW... I dunno if that means anythng or not. They are different too. Now they are normal looking whereas before they virtually noiseless at 3200. <shrug>