Image quality
The final question about the Alpha 580 (more so than the 560) is whether or not it has identical image quality to the A55, or to rival makes using a Sony-derived sensor of the same basic type – the Nikon D7000 and Pentax K-5. I can make some answer on the A55 front, I can’t comment on the other two makes. When the A580 first arrived, no raw conversion was available through Adobe (Lightroom and ACR) which is my standard workflow. Using a batch processor, the identity of my first A580 files was changed so that Adobe products saw them as A55 raws. Later, an update was issued. Comparing the results, they are identical which probably means Adobe has done much the same thing – simply updated the tables in the converter to treat the A580 the same as the A55.
The strong contrast curve in Adobe Camera Raw 6.3 restored this ISO 200 shot to a crisp look
This leaves the A580 raw images looking slightly softer than the A55. The light-loss in the A55 is not compensated for by just lifting the gain on the sensor, that much is clear from other specifications (including the metering EV range). It’s achieved using a different conversion curve after the data leaves the A to D converters which are part of the sensor itself. ISO speed in digital cameras is not calculated the same way this it is for film, even if DxO in their tests use a noise threshold method which is not far removed from ‘0.1D above fog’. In real life it’s assigned on the basis of midtone density, the light value and exposure settings which produce 128/128/128 RGB from a 12% reflectance grey target.
The conversion curve of the A55 would seem to compensate for the 2/3rds of a stop or so light loss from the Translucent mirror, leaving the deep shadows without much lift but increasing brightness and contrast in the midtones while preserving highlight detail. This is confirmed by raw conversions which have sudden descent into black at the shadow end, and the type of slightly enhanced colour saturation and sharpness given by a steeper midtone section of the characteristic curve. This can also create higher resolution readings, as these are based on contrast.
A580 JPEGs are processed in camera to produce as much the same look as other Sony models as possible. Lens choice is likely to have more real effect on results than body choice, though, especially at such a high sensor resolution. With the A580, our Sigma 18-250mm proves more the good enough at short to medium settings but begins to run out of resolution above 150mm and looks soft by 250mm. The 16-80mm CZ Sony lens really comes into its own, as it does on the A55. One most noticeable thing, shooting with several cameras and lenses, is how much higher the contrast is from the CZ at 16mm on A55 or A580 compared to the contrast of the E-mount 16mm pancake lens on NEX.
The A580 appears to be handled by some raw converters almost as if it was an A55. It doesn’t have the same contrast boost in camera and therefore the results look just a touch flatter and more neutral, and not as ‘crisp’. I speculated in the A55 report that the configuration of the AF module with the simple SLT design might be responsible for improved focus accuracy. Using Phase Detect AF in Focus Check LV mode on the A580, then magnifying the result, tells me the new 15-sensor AF module may be just as much a reason. The A580, like the A55, has a high ‘spot on AF’ success rate with a wide range of lenses.
My conclusion is that the A580 is not ‘less sharp’ than the A55, it just has a different tone curve when processed identically from raw and shows less detail contrast. That can be corrected by changing the raw processing parameters – adjusting Brightness and Contrast, and the selected tone curve, in ACR/Lightroom. The JPEG engine is pretty crude (come back ‘Extra Fine’ KM quality please!) and any reviews judging the camera on JPEGs alone are incomplete.
When I can shoot in extremely low light, hand holding 1/8th at f/9 with the 16-80mm at 16mm, and get great colours and no noise at ISO 1600, I find it hard to fault the performance of the Alpha system (SSS) or the 580. Click the image for our subscriber-only link page, to download and view the original full size JPEG conversion
As for the high ISO issue, Sony’s policy of retaining as much colour information as possible for settings over ISO 1600 contrasts with Nikon’s deliberate reduction of the colour saturation and discrimination. Nikon perhaps has a better understanding of how human vision works, with the monochrome-sending rods taking over from the colour-sensing cones in low light. We perceive dusk and night colours as desaturated, so there’s no need for a camera to try to capture high levels of colour at higher ISOs.
The full picture at ISO 3200 – auto exposure with the 580 has proved very accurate in unusual conditions. 1/80th at f/4 with the 28mm f/2 Minolta RS lens, hand held with stabilisation
Look at this in-camera JPEG and you would perhaps agree that ISO 3200 quality misses the mark. But Sony spend exactly $5 a year on JPEG quality R&D – or so their track record so far suggests. Most likely, they are using fast dirty compression to enable fast fast shooting
Conversion using ACR 6.3 tells a different story, using default sharpness and Luminance NR 25/50detail plus Chroma NR 25/50detail
Sony does not agree. Colour values are not reduced, and that produces more visible colour noise. What you lose in low light is gained in bright conditions when high ISOs are used for action photography, with the Sony image typically looking normal and Nikon’s results muted. Unless Sony follows Nikon’s example, high ISO shots will always look more noisy and coarser than Nikon because retaining colour values means using a larger sample of RGGB sensel values to create each RGB image pixel.
I am happy with either approach, partly because I hardly ever need speeds above ISO 1600 anyway. And, at 1600 or under using Auto ISO, the 580 creates such fine grain and high sharpness I can use it freely for any type of work including stock library submissions. My ISO 1600 shots from the 580 look as good as ISO 200 shots from five years ago.