I set up a test to see how the two uses of a similar sensor type compared at the highlight end, around native ISO. The Alpha 900 was able to focus accurately by AF on the side of the silver urn bowl which shows a slight bright edge against the paper. The D3X was completely unable to focus on the same spot by any AF method, and had to be manually focused; unfortunately the accuracy of manual focusing with standard screens in cameras like this is zero, and the Nikon screen provides very few visible engraving or marks to help adjust the dioptre correctly for viewing. So the Nikon focus point is different. This doesn't affect the aspect I wanted to check.
Raw file exposures were taken by tungsten light (this is not ideal, and automatically cuts highlight headroom by over one stop - approximately 1.5 stops of dynamic range is required for the camera to use tungsten WB in place of its presumed native daylight WB). The D3X's Tungsten preset is not as accurate as the A900, and produces a much warmer result. This makes the comparison difficult by eye and I have therefore used one extra control when processing the raw - I have set the WB based on the white paper, so that it matches the A900.
RAW Developer uses two very different curves to convert D3X and A900 files:
While the A900 benefits (as you will see below) from using Nikon curve instead of its own, the curve Iridient have written for the A900 is too dark and makes the files need exposure correction - it makes the Nikon appear to be almost twice the true ISO value compared to the A900. For this test, I have processed a set of raw files taken at 1/3rd stop intervals all the way from 2 seconds to 1/8th, using the Nikon curve. The Nikon was set to ISO 100 (claimed to be 'native') and the Sony was set to ISO 160 (while 200 is said to be 'native'). The reason for this is that several technical reports indicate the sensor may have a real native sensitivity best expressed around ISO 125-160. Doing this gives the Nikon the benefit of 1/3rd stop versus the A900, and compensates for the slight tungsten value adjustment in raw processing.
Here are the two results, Sony to the left, Nikon to the right (or below depending on how you are viewing this):
The value in the bright rim of the silver next to the paper is around 242-244 RGB in both cases, and the highlights clip in the brighter reflections to 0-0-0 in both cases. The actual exposure needed by the Nikon was 0.5s at f/11, and the Sony 0.4s at f/11. This is within 1/3rd of a stop of the ISO settings, but it is based on values around 235-245. If 255 clipping is used for comparison, both 0.4s frames are a better match.
What you should be able to see is that the Sony image is much brighter as you get down into the shadow tones. I leave you to work out the benefits of this, and why the Nikon image is more contrasty and darker for the same highlight detail recording (hint: try moving the left hand slider of any photo to the right in Photoshop levels, and you will see a similar darkening).
Please note that the Iridient curves shown are not the same as post-conversion curves in ACR (or indeed in RAW Developer - there is a complete additional Curves window for applying these). These are camera input calibration curves, designed to correct for the way each different camera raw format records brightness values.
Sorry for the rather obscure nature of this stuff. I do not plan to write a blog post about this yet, because I don't know enough about it yet. Still learning. RAW Developer has been a useful tool (so has Raw Photo Processor) but two or three other raw processors need to go live with D3X conversion. I want to look at as many real-life shots in different converters as possible before making an conclusions. Just let's say I am very happy with some aspects of the A900 image handling at low ISOs, and as things stand right now, the strategy seems to have been to tune performance for low ISOs with maximum recovery of both highlight and shadow detail along with colour. Nikon has been much more concerned about how high ISO performance is perceived, and they seem to have sacrificed ultimate low-ISO flexibility to get better high ISO results.
It would be understandable for a sensor maker (Sony) to 'go by the book' and set everything up perfectly around the native base ISO - and typical of a camera maker, like Nikon, to deviate from specifications to squeeze something more for real photographic purposes, by experimenting with different reading and processing strategies.