The Alpha 900 as a high ISO body

Spread the love

Following the review on dPreview – more than anywhere else – Sony’s extremely poor JPEG engine with its associated wide radius chroma blur and strong luminance smoothing noise reduction has proved to be a dog well capable of biting its master. Definitely a dog, anyway. But this performance is not what the camera can really achieve. In fact it’s perfectly capable of delivering good high ISO shots in typical situations.

This is an ISO 3200 shot converted from raw (click for full size). ACR’s conversion is not perfect but overall levels of detail, saturation and noise are easily controlled.

Because the JPEG conversion and its accompanying NR process create artefacts, it pays to shoot raw.

This is a Nikon D3 ISO 3200 12 megapixel file processed from raw using ACR. Nikon 24-70mm f2.8 lens at 24mm, 1/124th at f/5.6, tungsten light. Zero sharpening, zero luminance NR, chroma NR 50, ACR 4.6.

This is an identical magnification section of an 11.2 megapixel export (scaled down in ACR) from the ISO 3200 Alpha 900 file, processed the same way. KM 17-35mm D lens at 30mm, 1/50th at f/5.6. Both images: click on image to open full size.

The problem is that while JPEG-only shooters report getting over 1,000 frames from a battery, RAW+JPEG shooters are getting a third of that. Taking typical low-light situations, many are the type of shots where JPEGs are preferred – parties, gatherings, domestic snaps, concerts, sports, clubs, theatres and so on. Many low light situations feature people or animals (if they don’t, you rarely need high ISO – a tripod will do). They are the situations where you want to take fifty pictures not five, and avoid the heavy duty work of processing 24 megapixel RAW.

This is a 6.3 megapixel reduction from a raw file at ISO 1600, with no luminance NR. The speed is necessary in daylight with a dark subject and the 500mm f/8 mirror lens. The noise is present, but in a much finer and tighter form than a typical 6-8 megapixel DSLR, accompanied by very good sharpness. Click for the full size image.

There is another reason why the Alpha 900 needs good high ISO. Full-frame, compared to APS-C on a direct basis, needs 1.5 stops down to get the same depth of field. You need f/11 at 24mm where f/6.3 at 16mm does fine on an Alpha 700. If you’re using flash, or having moving subjects, you may need higher ISO to do that – not more flash power or a longer shutter speed. And high ISO is needed in daylight for action shots, wildlife, and pets where focusing tracking however good still benefits from a bit of depth of field to crispen up tiny errors.

A 13-megapixel in-camera JPEG with normal NR and correct (generous) exposure. Click to see the full size.

How can the Alpha 900 compete? The first answer is easy enough – it is not just a 24 megapixel monster. For the JPEG shooter, it is also a full-frame 13 megapixel or 6.1 megapixel DSLR. Those situations which demand a JPEG most of all rarely need a file able to print out at 16 x 24 inches or fit neatly on to a 115 inch screen.

Another 13 megapixel in-camera JPEG, this time at 6400, displays plenty of chroma noise when you click it to open the full image – but it also has exceptional levels of detail, and the chroma noise is easily removed. Even without further processing a 13 x 19″ print from this is acceptable.

These are 100% sections from the edge and centre zones, also showing the quality of the 1985 Minolta 50mm f1.4 lens used here at f/3.2.

While there is no doubt at all that the Nikon 12-megapixel full frame CMOS sensor produces the best ISO 6400 (and up), the Alpha 900 results are not far from matching it at the 13 megapixel size and ISO 3200; they may even be slightly better at ISO 800 and 1600 given identical exposure conditions.

If RAW shooting is an option, even something as crude as Adobe’s Camera Raw converter (not well optimised to the .ARW files) will deliver results far better than the JPEGs, and reduced size output to 11.5 megapixels or less reveals a highly detailed image.

For those who want to see what the A900 can do, a better raw processor such as RawDeveloper (Mac OSX) is the answer:

Click this to visit the original as a full 24 megapixel conversion. This is what in-camera JPEGs should really be like!

The key points to securing a good low noise 6/13 megapixel in-camera JPEG with the Alpha 900 are:

  • Use ISO 3200 rather than 6400 if possible, as 6400 is an extended setting using ‘digital ISO boosting’ (this advice also applies to most cameras where ISO 6400 is the highest possible setting)
  • Do not underexpose, and don’t allow the auto metering to underexpose for you!
  • Set Low or Normal NR, never High.
  • Use Extra Fine JPEG not Fine, or RAW+JPEG (limited to Fine as its best option) if all you want is a JPEG.
  • Use Neutral Creative Style rather than Standard, and sRGB colour space (this is always less noisy than aRGB).
  • Use manual White Balance or take a Custom White Balance measurement and setting rather than relying on Auto White Balance in very warm tungsten, or mixed, lighting.*
  • Set +1 or +2 sharpness, as this changes the behaviour of the noise reduction to favour a tight crisp noise pattern rather than a blotchy soft one.
  • Use the open-shadows Zone setting (-1) with a corresponding cut in Brightness (-2) as this also changes the NR behaviour, while the two adjustments cancel out any change to midtone values.**

*It is even better to use a filter or filter pack to correct the light, as you would do with slide film, but this negates the benefit of high ISO. Use of an 80B conversion filter and Daylight WB will improve noise levels and dynamic range in Tungsten light, whatever ISO rating is used, by equalising the R/B peaks of the spectrum. This is a neglected aspect of digital technique which, with the arrival of extremely high resolution DSLRs such as the Alpha 900, is worth reviving for the highest quality architectural interiors and tungsten studio work.

**This was suggested on dPreview forums by ‘Agorabasta’ and followed up on photoclubalpha forums. Similar results are obtained using combinations of low contrast, basic DRO (which applies a similar lightening gamma curve before raw conversion) and/or exposure compensation. It is also possible to improve ACR raw conversions by using the Linear conversion curve, and enhancing contrast later. Agorabasta’s suggestion has the benefit that it can be applied to one Creative Style, such as Neutral, and remembered for quick access.

The result will retain maximum detail and colour saturation, but may have some tight, strong chroma noise remaining which can be removed in application which allows chroma noise reduction. This noise will not be apparent at all in many areas of medium tone and detail, only in shadows, and if you can live with it the colour rendering will be much richer.

Here is a Nikon D3 ISO 6400 12 megapixel in-camera JPEG. Again, clicking it will open the full size original (resaved in sRGB for pBase – shot as AdobeRGB). Chroma noise is almost non-existent but the actual noise size is not so different from the Alpha 900 at 13 megapixels:

The benchmark against which the Alpha 900 should be compared is, realistically, the Nikon D300 or the Alpha 700 rather than the Nikon D3/D700. It has a similar pixel pitch (sensel size) and a similar ISO 3200 ‘safe limit’ with 6400 boost. When the JPEG is Medium size it also offers a similar final file size. It would be remarkable if the Nikon D3/D700 with double the sensel area did not manage to achieve at least double the ISO at identical quality, even if Sony improved their noise handling to match Nikon outright.

Comments made about the ‘high density’ of the Alpha 900 sensor are misplaced. It is only equivalent to an 11 megapixel APS-C (DT/DX) sensor and barely the density of Canon 10 megapixel 1.6X format silicon. Where it differs is in the readout speed and volume, before the twin BIONZ processor stage – and even then, each BIONZ though identical to the A700 chip is processing twice as much data, they do not divide the load between them, they take it in turns on a ‘first free, first served’ basis. This may account for the Alpha 900 (v1 firmware) being no match for the Alpha 700 (v4 firmware) in noise handling and JPEG conversion. The fact that enabling NR slows down sequence shooting confirms the workload involved.

Here, in good lighting and well exposed, is a final ISO 6400 image from the Alpha 900 resized to 12 megapixels:

And here is a 100% clip from this image:

The full 24 megapixel image can also be viewed and downloaded:

I have printed this to A2 bleed on the Epson 3880 and shown it to a group of MPA/BIPP professional photographers locally, including several using Nikon D3; it was felt that the Nikon would be cleaner, but at half the size only. Their general reactions to the Alpha 900 were not as negative as many reviews and the main comments were how small and light the body was for 24 megapixels and how good the viewfinder was… but what a stupid non-standard hot shoe!

– David Kilpatrick FBIPP Hon FMPA


  • Mike Johnston of “The Online Photographer” has just added his very positive observations about the high-iso qualities of the A900 in his blog entry The Sony A900 and the Nikon D3. Definitely worth reading for anybody considering the purchase of the A900.

  • I hope DxO Optic Pro adds the A900 soon. I would like to see how it handles the A900. 5.3 is doing a very good job with the A350 images.

    So for now, the A900 need to be used in RAW. Maybe in four software releases they will fix it.

    I love the Zeiss lenses and the camera bodies. I do not like Sony’s handling of the data. DxO Optic Pro is working very very well, but makes a much slower work flow.

  • I have added a RawDeveloper conversion of the 6400 street shot which shows exactly this comparison – RD knocks ACR sideways for handling of the noise-grain structure though I suspect ACR retains sharper edges.

    Also, I have changed all the direct links to full size files so that they now go to the pBase page, instead of opening the original size file directly. pBase went down and would not deliver any of these files within about 12 hours of the article being posted. This change should lower the ‘load’ on my pBase pages.


  • It’s your comment on the ACR plug-in that has struck a cord with me, because sitting on my iMac at the moment I have Aperture 2.1.2 together with LightZone 3.6 and Photoshop Elements 6 (with ACR 5.1 just installed). I have a trial version of the Nik Software package.

    Digital photography is expensive. Your set-up costs are at least a new SLR, a couple of new lenses and a new photo printer. And this does not include all the hours of research and cost benefit analysis.

    Are we now at a stage when camera and lens reviews are going to have to come with pages of examples and graphs of performance data with each of half a dozen different raw developers or graphic programs ?

    For example

    Adobe Camera Raw / DNG with Photoshop Elements or Lightroom

    Phase One LE ( consumer ) or Pro ( pro models )

    Bibble Lite or Bibble Pro


    Apple Aperture

  • Obvious I will come back wen images will be available.

    But just in case I wonder if you will take ISO1600-ISO6400 from A900 and D3 RAW and then do a strong noise reduction in ACR (luminance 50-100%) for sure you will get similarly clean images from both cameras and resolution will be still better for A900. Thus up till ISO6400 the noise/signal ratio for both cameras is the same.

  • Interesting, as usual.
    I’m still undecided re. DR+ setting…

    Would love to see the images though, but unfortunately as often is the case (or perhaps the rule) pbase is having problems.

Leave a Reply