"The Nikon D3x offers the finest image quality in a DSLR the world has yet seen"

[Ray]

Thom will be publishing a review of the D3x, which he is working on now.  He has been using both the D3x and the D3 in a basketball gym (just like a "real" photographer would) and that's what he has based his preliminary evaluations on.  As for Thom and Iliah showing you their methodology; why don't you show us yours?  Only problem is that you have none to offer because you don't have either camera.

Tony,
Nor do you have both cameras. Having just one of them is not enough. But that doesn't stop you mouthing off.

I've already described my methodology in a previous post. I'll repeat it... "I do a few tests with the material available. I use my common sense, experience and what little knowledge I possess to make certain deductions and draw certain conclusions."

In situations like this where different experts who appear to be competent and experienced arrive at different conclusions, we have to get more tests and comparisons before we can be certain, and/or more details on what these experts are actually saying.

If you want to throw into the mix the variability of uncontrolled shooting conditions, camera shake with hand-held shots, moving subjects which cannot be photographed twice in the same position, unavoidable imprecision of focussing however slight, less than a perfect ETTR for the precise conditions which are likely to vary with each shot however slightly, then we are really into probability theory.

Even when the subject is stationary and I'm relying upon image stabilisation to give me a reasonably sharp result at a slow shutter speed, the sharpness of the results can vary noticeably with successive shots of the same scene at the same shutter speed. This is because camera shake can be slightly different with each shot. Take enough hand-held shots of the same scene at the same shutter speed and the best is likely to be significantly sharper than the worst.

I would concede the point that, if shooting conditions are such that it's not possible to get a sharp result with the D3X, then there's no resolution advantage of the D3X to be traded for lower noise. In those circumstances, the noise at high ISO will be in favour of the D3 statistically, whatever the print size.

Just as the resolution differences between the D3 and D3X will diminish as one stops down to diffraction limited f stops, the resolution differences will also diminish as one uses a slower and slower shutter speed. The resolution of the D3X will never be less than that of the D3 (statistically) as a result of a slow shutter speed, but it needs to be more.

[Ray]

For example, if you meter a scene at ISO 1600, F8, 1/100 in manual mode, and then drop the ISO to base ISO, shoot the scene, and then boost the exposure in a good RAW program like RPP, the noise of the A900 improves dramatically.

Well, that explains a lot and it's very surprising that anyone should produce a camera these days that behaves like that. Just so I've understood what you are saying, are you saying that an A900 shot which is underexposed 3 stops at ISO 200 actually has less noise than an ETTR shot at ISO 1600 employing the same F stop and shutter speed?

I used to complain that my first DSLR, the Canon D60, produced very little improvement in noise when comparing fully exposed images at ISO 800 with 3-stops-underexposed images at ISO 100. However, there was at least some improvement at ISO 800, most noticeable in the deep shadows. Two models later, with the 20D, the situation was radically different. High-ISO images were significantly cleaner across the entire tonal range, compared with underexposed shots at base ISO. That pleased me greatly.   .

[douglasf13]

Well, that explains a lot and it's very surprising that anyone should produce a camera these days that behaves like that. Just so I've understood what you are saying, are you saying that an A900 shot which is underexposed 3 stops at ISO 200 actually has less noise than an ETTR shot at ISO 1600 employing the same F stop and shutter speed?

I used to complain that my first DSLR, the Canon D60, produced very little improvement in noise when comparing fully exposed images at ISO 800 with 3-stops-underexposed images at ISO 100. However, there was at least some improvement at ISO 800, most noticeable in the deep shadows. Two models later, with the 20D, the situation was radically different. High-ISO images were significantly cleaner across the entire tonal range, compared with underexposed shots at base ISO. That pleased me greatly.   .

All cameras behave this way, more or less.  Iliah doesn't shoot over ISO 800 with the D3x. Once you find a camera's sweetspot, it is nearly always more affective to stay at that ISO and underexpose, then boost in RPP or equivalent. The secret is exposing for your RAW converter.  If your using ACR, then any IQ subtleties are thrown out the window, and there's really no point.  The a900s sweetspot happens to be 1/3 stop underexposure at ISO 100.  BTW,  ETTR is really only effective at base ISO.

[Ray]

All cameras behave this way, more or less.  Iliah doesn't shoot over ISO 800 with the D3x. Once you find a camera's sweetspot, it is nearly always more affective to stay at that ISO and underexpose, then boost in RPP or equivalent. The secret is exposing for your RAW converter.  If your using ACR, then any IQ subtleties are thrown out the window, and there's really no point.  The a900s sweetspot happens to be 1/3 stop underexposure at ISO 100.  BTW,  ETTR is really only effective at base ISO.

Sorry! You are saying the opposite of what my eyes are telling me, at least with Canon DSLRs. I've made careful comparisons in relation to such issues and it's very clear to me that:

(1) ETTR is more important the higher the ISO and less important the lower the ISO.

(2) If noise is a concern, it's always better to aim for an ETTR at a higher ISO than use the same exposure at a lower ISO.

There might be a sweet spot where such differences are most noticeable, but as a general trend it works across the whole range of 'real' ISO settings. An ETTR at ISO 200 is better than a one-stop under-exposure at ISO 100. An ETTR at ISO 1600 is better than a one-stop underexposure at ISO 800 and also better than a 4-stop underexposure at ISO 100, with Canon cameras.

I understand the reason for this result is that at higher-than-base ISO, Canon amplify the analogue signal before it is subject to A/D conversion and all the other processes that might introduce noise. By amplifying the signal whilst it's still at the analogue stage, the total noise component of the signal is not reduced (in fact it's probably increased slightly) but the size of the signal in relation to the size of the noise is very much greater, therefore S/N ratio is much improved.

I'm surprised that Sony are not using such methods. Perhaps both Canon and Nikon have a patent on their own methods in this respect and Sony are unable to devise a different way of reducing noise without infringing on such patents.

[ErikKaffehr]

Hi,

There have been discussions on this forum about something called "unity gain". To my understanding once ISO is increased beyond "Unity Gain" ISO there is little or no utility of increasing preamplification before AD conversion.

From "Clarkvision.com"
"The Unity Gain ISO is the ISO of the camera where the A/D converter digitizes 1 electron to 1 data number (DN) in the digital image. Further, to scale all cameras to equivalent Unity Gain ISO, a 12-bit converter is assumed. Since 1 electron (1 converted photon) is the smallest quantum that makes sense to digitize, there is little point in increasing ISO above the Unity Gain ISO (small gains may be realized due to quantization effects, but as ISO is increased, dynamic range decreases)."

Best regards
Erik

Sorry! You are saying the opposite of what my eyes are telling me, at least with Canon DSLRs. I've made careful comparisons in relation to such issues and it's very clear to me that:

(1) ETTR is more important the higher the ISO and less important the lower the ISO.

(2) If noise is a concern, it's always better to aim for an ETTR at a higher ISO than use the same exposure at a lower ISO.

There might be a sweet spot where such differences are most noticeable, but as a general trend it works across the whole range of 'real' ISO settings. An ETTR at ISO 200 is better than a one-stop under-exposure at ISO 100. An ETTR at ISO 1600 is better than a one-stop underexposure at ISO 800 and also better than a 4-stop underexposure at ISO 100, with Canon cameras.

I understand the reason for this result is that at higher-than-base ISO, Canon amplify the analogue signal before it is subject to A/D conversion and all the other processes that might introduce noise. By amplifying the signal whilst it's still at the analogue stage, the total noise component of the signal is not reduced (in fact it's probably increased slightly) but the size of the signal in relation to the size of the noise is very much greater, therefore S/N ratio is much improved.

I'm surprised that Sony are not using such methods. Perhaps both Canon and Nikon have a patent on their own methods in this respect and Sony are unable to devise a different way of reducing noise without infringing on such patents.

[Ray]

Hi,

There have been discussions on this forum about something called "unity gain". To my understanding once ISO is increased beyond "Unity Gain" ISO there is little or no utility of increasing preamplification before AD conversion.

From "Clarkvision.com"
"The Unity Gain ISO is the ISO of the camera where the A/D converter digitizes 1 electron to 1 data number (DN) in the digital image. Further, to scale all cameras to equivalent Unity Gain ISO, a 12-bit converter is assumed. Since 1 electron (1 converted photon) is the smallest quantum that makes sense to digitize, there is little point in increasing ISO above the Unity Gain ISO (small gains may be realized due to quantization effects, but as ISO is increased, dynamic range decreases)."

Best regards
Erik

Erik,
My understanding is that Canon amplifies the initial analog signal before A/D conversion takes place, when an ISO higher than base is used. Such amplification also amplifies existing noise but ensures that further noise in the processing chain, such as Read Noise and noise from the A/D conversion, is smaller as a proportion of the total signal. Signal-to-noise ratio is thus better than it otherwise would be.

Dynamic range usually decreases with increasing ISO in all cameras. For example, DXOmark shows the 5D2 at the measured ISO of 1600 having a DR of about 9.5EV whereas at ISO 100 it's about 11.16EV, approximately 1.6 stops greater. However, a full ETTR at ISO 100 would require 4 stops more exposure than a full ETTR at ISO 1600. Why is the difference in DR between ISO 100 and ISO 1600 only 1.6EV and not 4EV?

By the way, the 4 stop difference in exposure with the A900 does not translate to a 4 stop difference in DR either. But it's almost a 3 stop difference between the measured ISOs of 119 and 1904.

However, using the camera's ISO of 1600 and 100 for the 5D2 (instead of the values on the horizontal axis) we get DR values of 11.16EV at ISO 100 (measured as ISO 73) and 10.05EV at the camera's ISO of 1600 (measured as ISO 1093). The difference in DR is hardly more than 1EV. Applying the same method of comparison with the A900, the measured ISOs are 119 and 1264 and the DR figures are 11.5EV and 8.46EV respectively, which represents a full 3 stops difference.

The DR graph for the A900 is a straight line, all the way from the camera's nominated value of ISO 100 to ISO 6400. However, the DR graph for the 5D2 has a more gently slope down to ISO 1600, after which it becomes a straight line and for each increase in ISO of one stop beyond ISO 1600, DR is reduced by almost the same amount.

It would therefore seem from this analysis that the 5D2's DR performance does not vary consistently across the entire ISO range. Up to ISO 1600 there's minimal loss of DR. Beyond ISO 1600 there's a more dramatic loss of DR. The A900 seems to show a more consistent and dramatic loss of DR across the entire ISO range, moving from base ISO to maximum ISO. However, it does appear to be less than a 1EV difference in DR for each 1 stop of ISO increase up to about ISO 200, and slightly less than 1EV between ISO 200 and 400. Beyond ISO 400, for each doubling in ISO there's a corresponding loss of 1EV of DR.

These graphs would suggest that little point is served using ISOs higher than 400 with the A900. You might as well underexpose at a lower ISO.

[douglasf13]

Ray, the D3x also uses column adcs on the chip, like A900. The a900 and D3x have the lowest read noise of base ISO of any of the new fullframes. There are lots of tradeoffs. CFA makes a big difference.

[ErikKaffehr]

Ray,

I agree with you on this. The reasoning that Roger Clark does is that noise will decrease with increasing ISO when compared with "pushing" the raw data in conversion at low ISO, up to a certain level which used to be around 800 ISO (or so) increasing ISO beyond that amplifies noise and signal the same. So ISO up "unity gain ISO" is "is real ISO" and beyond that is "fake ISO". The 800 ISO figure is not valid any longer, because 14 bit ADCs redefine the playing field (causing Unity gains ISO to go down) and also because of cameras with big sensels (like the Nikon D3 and the original 5D) which have very high Unity Gain ISO.

I had a short discussion about this with Guillermo Glujik on this forum and I think that he both explained what was going on and demonstrated the concept experimentally.

At this stage we now very little about the Nikon D3X, especially we don't know how much of the advantage that some thrustworthy testers report comes from sensor, camera electronics and "raw" processing. DxO-mark analyses "raw" data and that should offer some insight in what the camera delivers to "raw conversion".

It seems that the sensor itself is the same on the A900 and the D3X but there may be differences in color filter array (which DxO also covers, BTW), AA-filtering and signal processing. Regarding signal processing the A900 has column AD converters, but they are 12 bit AFAIK. For 12 bit conversion Nikon would need off chip AD-conversion and preamps, essentially giving a fully new signal processing pipeline.

From the data at DxO-mark it seems that there is something like 1 eV advantage of the Canon 5DII sensor over the A900 sensor, except for DR where DR drops of much faster on the Alpha than on the 5DII. DxO did not publish data for the Alpha 700 which shares sensor technology with the Nikon D300, unfortunately.

Regarding DxO-mark I don't really care about the DxO-mark itself. For one thing trying to turn a lot of data into a single figure of merit is not a smart thing in my view. The other issue is that DxO clearly states that it takes about 5 DxO mark units for any visible difference. So it essentially says that a camera with DxO-mark 83.9 performs the same as a camera with DxO mark of 82, 83, 84, 85, 86 and so on.

One other nice thing with DxO is that they measure many cameras and sensors under well defined lab conditions. Very few reviewers are in a similar situation. Add to this that the DxO folks are image scientists and the developers of an advanced "raw" converter pipeline. No doubt that Phase One, Adobe, Bibble or Iridient would be able to supply similar data, DxO decided to share this information doing a great service to anyone interested in sensor technology.

Best regards
Erik

Erik,
My understanding is that Canon amplifies the initial analog signal before A/D conversion takes place, when an ISO higher than base is used. Such amplification also amplifies existing noise but ensures that further noise in the processing chain, such as Read Noise and noise from the A/D conversion, is smaller as a proportion of the total signal. Signal-to-noise ratio is thus better than it otherwise would be.

Dynamic range usually decreases with increasing ISO in all cameras. For example, DXOmark shows the 5D2 at the camera's ISO of 1600 having a DR of about 9.5EV whereas at ISO 100 it's about 11.16EV, approximately 1.6 stops greater. However, a full ETTR at ISO 100 would require 4 stops more exposure than a full ETTR at ISO 1600. Why is the difference in DR between ISO 100 and ISO 1600 only 1.6EV and not 4EV?

By the way, the 4 stop difference in exposure with the A900 does not translate to a 4 stop difference in DR either. But it's almost a 3 stop difference between the measured ISOs of 119 and 1904.

[ErikKaffehr]

Hi,

Are you sure? On the D700 Nikon can use either 12 bit readout from column ADCs or 14 bit readouts from a set of readout channels. I'm pretty sure column ADCs are only 12 bits!

Have you any reference for your statement that: "The a900 and D3x have the lowest read noise of base ISO of any of the new fullframes." ? I have not found any information on this, but I got the impression that this is an are where Canon used to excel.

The CFA probably matters, and different vendors may opt for different CFAs to be able to achieve specific colors. There is also some info on CFA implementation on the DxO-mark site, under the "Color Response" tab. It's missing from the "camera comparison pages".

Best regards
Erik

Ray, the D3x also uses column adcs on the chip, like A900. The a900 and D3x have the lowest read noise of base ISO of any of the new fullframes. There are lots of tradeoffs. CFA makes a big difference.

[Slough]

Hi,

My impression is that nowdays the frequency at which MTF of 50 percent is achieved is regarded a better measure of image quality than pure resolution. This does not contradict your findings in any way, but can explain if we perceive digital images to be sharper than film even if digital does not resolve higher.

I think you are right. Basically the micro-contrast of the digital image is far higher. Although the two are providing about the same level of detail, the film image has so much noise that the IQ is degraded significantly. So, from an aesthetic point of view, the film image is significantly inferior. However, were I to pass the film image through some form of grain removing software, the two might be rather close. And I will try that if I can find some software without having to pay an arm and a leg for it.

This incidentally is probably why some people still claim that a camera such as the D2x does not outperform 35mm slide film, despite the fact that we all know that in practice it does.

Another aspect is that sharpening is always a part of the digital workflow. So we need to compare optimally sharpened images fro each medium. What is optimal sharpening is also a matter of perception.

Best regards
Erik

True, though in the tests I carried out, the sharpening is not an issue as the detail provided by the DSLR is limited by the number of pixels.

[Ray]

Regarding DxO-mark I don't really care about the DxO-mark itself. For one thing trying to turn a lot of data into a single figure of merit is not a smart thing in my view. The other issue is that DxO clearly states that it takes about 5 DxO mark units for any visible difference. So it essentially says that a camera with DxO-mark 83.9 performs the same as a camera with DxO mark of 82, 83, 84, 85, 86 and so on.

One other nice thing with DxO is that they measure many cameras and sensors under well defined lab conditions. Very few reviewers are in a similar situation. Add to this that the DxO folks are image scientists and the developers of an advanced "raw" converter pipeline. No doubt that Phase One, Adobe, Bibble or Iridient would be able to supply similar data, DxO decided to share this information doing a great service to anyone interested in sensor technology.

Best regards
Erik

Erik,
Nor do I, but the graphs are meaningful and, as you mention, the lab conditions and methodology are consistent which is essential for accurate comparisons. I also like the way one can compare three different cameras on the same screen and also get an idea of how 8x12 prints would compare, with respect to DR and noise. However, I would prefer it if DXO were to provide another option showing results in respect of a larger print, say 24x36".

[douglasf13]

Erik, the D3x 14bit mode is a cludge just like the D300, and Im pretty convinced by the evidence that points to it not being a true 14bit mode like on the D3, but rather a resampling of the 12bit column ADCs. This extra sampling does result in less shadow noise, but that isn't so much a by-product of 14bits, but rather a benefit from a slow resample. There is no evidence showing a seperate analogue output from the D3x sensor; and Nikon themselves claim the chip has onchip ADCs.  

There specific benefits to each of these cameras, and there is a reason that guys like Iliah own and use he D3, D3x, and A900 (and btw, not the 5dii.)

[jjj]

Just a wild thought here, did it ever occur to anyone just to take some real world photos [not test charts] with some cameras, develop them to their best and and then simply look at some prints to see which one looks best?
In Hifi you do blind listening to determine which kit sounds best - also gets rid of fanboyism bias as well. For all the talk about best s/n, least distortion, most power etc, in reality these absolute figures told you squat about how good stuff sounded or even how loud it was. A 20W amp could be louder than a [claimed] 80w amp for example.

I did see some real world shots recently comparing the 5DII's noise with a Nikon and what was at first very obvious, is that all the Canon shots were underexposed - which isn't going to help when comparing noise.

[ErikKaffehr]

Point taken! Appreciate your comment!

Best regards
Erik

Just a wild thought here, did it ever occur to anyone just to take some real world photos [not test charts] with some cameras, develop them to their best and and then simply look at some prints to see which one looks best?
In Hifi you do blind listening to determine which kit sounds best - also gets rid of fanboyism bias as well. For all the talk about best s/n, least distortion, most power etc, in reality these absolute figures told you squat about how good stuff sounded or even how loud it was. A 20W amp could be louder than a [claimed] 80w amp for example.

I did see some real world shots recently comparing the 5DII's noise with a Nikon and what was at first very obvious, is that all the Canon shots were underexposed - which isn't going to help when comparing noise.

[DarkPenguin]

I thought this was an interesting point ...

http://forums.dpreview.com/forums/read.asp...essage=30639804

anyone noticed polarizing oddities with their cameras?

[Tony Beach]

It got so rude and personal, this will be my last post in the thread since it does not worth my time to say anything more.

"If someone tells you D3x does not record any more details than the D3 at 1600 ISO, what it tells is only that you were listening to an incompetent tester, even he or she actually has the means to own the equipment."

Calling others incompetent is rude in my book.

Just as I said, there are hundreds of situations in which the D3 makes a better camera in the world, such as handhold, and the rude party is admitting he is talking about handhold.

Actually it is only one variable, and the shot I showed was done with a tripod.  Another variable that you completely ignore is spectral properties differences between the two cameras and what happens when those are compensated for by conversion; then you impugn the competence of those that are aware of these issues because you saw a comparison made under studio lighting.

Oranges to apples, to say the least.  I never found a tripod setup less than 20lb in weight is adequate, and I drill holes through the center column to hang my 35+ pound backpack before I even consider a shot.  Apparently some handhold people don't understand how capable landscape photographers work, and simply categorize anything not relevant to their style "nonsesnss", as their corner is the whole world.  More than enough said.

Galen Rowell was a hand hold person, he couldn't hold on to 55 lbs while hanging from a rock face and taking the shot you missed because you stayed on the ground.  Thom Hogan is his most notable protege; and he points out that the farther you get from the main road, the fewer big rigs you see.  You think pretty highly of yourself when you refer to such people as not being "capable landscape photographers".

[Dan Wells]

The comparisons I've made that find a D3x close if not equal to 6x9 cm film on detail extraction , and noticeably superior on overall image quality, use a Nikon Super Coolscan 9000 to scan the film (all of my film vs. digital subjective numbers are based on the Super Coolscan 5000 and 9000, the scanners with which I have the most experience) . That's actually a very fair comparison, because any scanner better than the Super Coolscan would cause the 6x9 outfit to significantly exceed the cost of the D3x (without factoring in the cost of film and processing). A better scanner could not extract much more detail from the Velvia, because the Super Coolscan is already scanning film grain - it could, however, potentially extract more dynamic range, especially in the shadows, bringing the overall image quality closer.
       I am confident in saying that the only film setup that can produce an image with overall quality superior to the D3x (and potentially the 20+ mp Canons and/or the Alpha 900, about none of which I claim knowledge) correctly handled (ISO 100, good lens, tripod) without significantly exceeding the cost of the D3x (therefore no drum scanners or $50,000 EverSmart flatbeds) would rely upon 4x5 or larger sheet film or possibly slow, exotic 120 black and white film (Tech Pan, etc...).
         It is certainly possible to beat the detail extraction of ANY digital camera, including the 50+ MP medium format digital backs, by the simple expedient of using a large enough piece of film. Right now, it seems to me that the best digital sensors are roughly equivalent to a piece of film 4x their ares (FF35 digital ~6x9 cm film, 645 digital ~4x5 film), and that various technical limits (diffraction, quantum-related noise) will hold them not too far above that (I would have a hard time seeing pure detail extraction ever exceeding a piece of film 6x the area of the sensor), although perhaps with improved image quality in other areas such as dynamic range or tonal scale. If this is even roughly correct, it will take an image sensor somewhere on the scale of 4x5 inches to equal the quality possible from 8x10 film, and around 5x7 inches to equal 11x14 film. While the cost of silicon of a given capability goes down rapidly with time, the cost of a given AREA of silicon does not. A 4x5 inch single-shot image sensor will remain prohibitively expensive to anyone who is neither Jim Jannard (the eccentric billionaire founder of Oakley sunglasses and the RED movie camera company, who owns 1500 cameras) nor employed by the "Bureau of Public Roads" (the famed label on the CIA's highway exit) in Langley, VA. An 11x14 Canham field camera will remain forever capable of capturing an image that no digital sensor in general circulation can touch without a lot of stitching (of course, an 11x14 camera is so tricky to set up and use that I'm not AT ALL sure that a 20-shot stitch with a D3x isn't an easier way to that image quality).

                                                                       -Dan

[Bill Caulfeild-Browne]

Sorry! You are saying the opposite of what my eyes are telling me, at least with Canon DSLRs. I've made careful comparisons in relation to such issues and it's very clear to me that:

(1) ETTR is more important the higher the ISO and less important the lower the ISO.

(2) If noise is a concern, it's always better to aim for an ETTR at a higher ISO than use the same exposure at a lower ISO.

There might be a sweet spot where such differences are most noticeable, but as a general trend it works across the whole range of 'real' ISO settings. An ETTR at ISO 200 is better than a one-stop under-exposure at ISO 100. An ETTR at ISO 1600 is better than a one-stop underexposure at ISO 800 and also better than a 4-stop underexposure at ISO 100, with Canon cameras.

I understand the reason for this result is that at higher-than-base ISO, Canon amplify the analogue signal before it is subject to A/D conversion and all the other processes that might introduce noise. By amplifying the signal whilst it's still at the analogue stage, the total noise component of the signal is not reduced (in fact it's probably increased slightly) but the size of the signal in relation to the size of the noise is very much greater, therefore S/N ratio is much improved.

I'm surprised that Sony are not using such methods. Perhaps both Canon and Nikon have a patent on their own methods in this respect and Sony are unable to devise a different way of reducing noise without infringing on such patents.

Ray, based on tests with my A900 (and my Phase back, for that matter) you are absolutely correct. Whatever ISO you're using, flood the pixels with as much light as you can w/out serious clipping and then dial-in the right density in post processing. (Thomas Knoll expounded on that several years ago somewhere in an article MR wrote on this site.)
Bill

[Slough]

The cost of a sensor of a given size is slowly going down, in part due to economies of scale, and in part due to improved manufacturing techniques. A large part of the cost arises from the high defect rate, so addressing that could easily halve the production cost per unit. Also, we are seeing displays being made using printing processes, whereby the transistors are printed on a substrate. I do not know enough about the physics to know if it would be possible to print a sensor. It might be that the incident photon to recorded electron ratio would be low, so ISO performance might not be good. But then again, a 4"x5" digital camera with a D3 equivalent pixel density restricted to ISO 100 might be a pixel peepers dream. As to whether or not a lens could be manufactured to illuminate such a large image circle and make use of such a high pixel density, who knows.

[douglasf13]

Bill, using ETTR is fine by me, and I use it frequently. However, Ettr at an ISO above your sensor sweetspot does not gain anything. Even if you meter for ETTR, you're still better off holding those shutter and fstop settings and then cranking down the ISO that happens to be your camera's sweetspot, and that would be ISO 100 on the A900. After that, boost exposure in raw with highlight compression.    What raw converter are you using?


Ps. This all about eeking out best results. Shooting regularly, the A900 is competitive through ISO 800 with the other makes.