Equal among firsts for high resolution

Here is an article which is mainly a test review of the Sigma 50mm f/1.4 EX DG HSM for Sony Alpha, and also Nikon, fit (both were tested). It also deals with the Alpha 900 and D3X, the two 24 megapixel full frame cameras used to test the lenses and two comparison 50mm f1.4s.

Alamy must be able to predict the future. One of the recent changes they made was to allow searches for images over 70MB size. Neatly, both the Sony Alpha 900 and Nikon D3X just fall at the last hurdle and come in about 500KB short!
Their 69.x megabyte, 24.x megapixel images – x because there is a tiny difference in final bytecount – are however the closest thing to medium format digital you can currently buy in a 35mm-compatible SLR system. The Alpha 900 is also the lowest cost current full-frame DSLR you can buy now that the Canon 5D mark 1 has been discontinued. Depending on where you live the 900, the Nikon D700 and the Canon 5D MkII jostle weekly for a price advantage but they are basically all at the same price-point.

Sigma’s new 50mm f/1.4 EX DG HSM for Sony – a lens on a massive scale even on the Alpha 900
This has caused a revival of interest in the 50mm standard lens, and in so-called ‘prime’ lenses generally. Prime is a misuse of an optical term, but you can’t avoid it now and it’s neater than writing ‘fixed focal length’ every time. We now talk casually about prime lenses opposed to zoom lenses, and that is not the correct term as many modern zooms are actually varifocals, meaning they need a focusing adjustment if you change the focal length, which a true zoom does not.
So, let’s just say prime for fixed focal length and zoom for anything with a variable focal length, and forget the pedantry. Almost any well-made prime, even of a simple construction, can give better contrast, resolution, even illumination and straight geometry than a zoom. When you move up to 24 megapixels and full frame, lenses like 24mm f/2.8, 35mm f/2, 50mm f/1.7, 85mm f/1.8 or 135mm f/2.8 get very interesting. They are small, light and focus accurately with a bright viewfinder image. To buy a zoom which matches their typical quality, at any given focal length, may cost three times as much or more.
Here, I will be showing at the end of this article links to an entire series of studio tests at full 24 megapixel size, which first allow comparison of the Nikon D3X and the Alpha 900 for this type of ISO 100 work, and secondly compare results at full aperture and f/8 from three very different lenses: the 1985 designed classic Minolta 50mm f/1.4 AF, the 2008 designed Nikkor AF-S 50mm f/1.4 G, and the 2008 radical rethink of what a 50mm f/1.4 should be, Sigma’s flasgship prime.
But first, a review of the Sigma – a lens intended to be a true ‘prime’, a first among equals. It’s certainly a first. No maker has ever built a 50mm standard with a 65mm diameter front element and a 77mm filter thread, weighing over 500g and costing almost £500.
The Sigma concept

Here is that 1985-ish Minolta next to the new Sigma. Despite appearances, they were indeed separated at birth. Both are basically the same type of Gauss design, the Sigma using just 8 elements in 6 groups. The difference is that in front of the iris, the design is progressively scaled up to improve illumination at the edges of the field, a critical issue with digital sensors.

This is what a light-table view of the old Minolta design looks like viewed at an angle equal to the corner of the frame, with the aperture held wide open.

This is what a similar view through the Sigma looks like. You can realise why from this outfield angle the Sigma is putting far more light on to the silicon, and indeed, when comparison photos are taken of a plain sky using both lenses wide open, the difference in field illumination is clear:

The Sigma is top, the Minolta bottom. During the period of testing this kit, I also had the Nikkor AF-S G, and it’s very similar to the Minolta, not the Sigma, in size and in full aperture vignetting.
This is not the end of the story. The Sigma is also made (like the Nikon) to have almost zero aperture related focus shift. This is a complex issue, which affects AF and manual focusing accuracy. Most AF sensor arrays will ‘see’ the lens at between f/2.8 (centre sensors on Alpha 900, for example) and f/7.1 (outer sensors on the 900) with f/5.6 being very typical (Nikon MultiCAM 3500, most Canon sensors except for central). Most focusing screens ‘see’ the lens around f/4.5, as do metering systems, which is why the vignetting issues don’t affect metering (the field is evenly illuminated by the time you stop down to f/4.5 with all such lenses).
With most lenses, including the old Minolta, the plane of exact focus shifts between f/1.4 and f/8. Cameras equipped with f/2.8 AF sensors will provide a better compromise, but full aperture may not be focused precisely as intended (stopping down masks the effect of shift, even when it happens). With the new lenses like the Sigma and Nikon, there is hardly any shift so they focus very accurately for shots all all apertures, both via AF and manually on-screen.

Here’s a shot at f/2 on the Sigma, 1/30th hand held at ISO 400 on the Alpha 900. One stop down is always a good idea, even with the excellent performance of this lens. The pub sign shows how well it focuses and how sharp it is:

Testing the Sigma at full aperture on a distant subject with the Alpha 900, here are some results:

This is the centre of the frame – trust me, at 24 megapixels 100% and without any sharpening, this is state of the art for a 50mm f/1.4 used wide open. The corners of the frame, all photographed in turn right into the extreme corner as the 100% finder of the Alpha 900 allows this, show both the slight darkening and the slight softening present:

Click the image for a 100% view of this composite.
The Sigma 50mm f/1.4 is a great lens to use for its bokeh – which is best between f/1.4 and f/2.8, here shown at f/2.2:

The iris is nine-blade, which keeps things pretty tidy even when stopped down, but the design itself does produce an amount of colour bokeh and can produce some colour fringes, so processing carefully from raw files is the best way to handle its output.

Finally, this expensive lens comes with a good petal hood and a case. I will not show examples here, but be warned of one phenomenon – the view through the lens at f/1.4 when the sun is just out of the frame can be perfect, no flare. Then your final picture at f/8 is totally obscured by flare! You must check each shot in these conditions, as the stopped-down aperture blades clearly create the flare which is entirely absent in a wide open lens for viewing.

Do you need this? Would the current latest Sony – a revision of the Minolta design – be better? Well, I am not rushing to buy the Sigma. My tiny Minolta is good enough – soft at f/1.4, but sharper with every 1/3 or 1/2 stop step it is closed down, fantastic by f/2.8, and although it seems prone to focus errors on the Alpha 700 it’s proving accurate on the 900.
But if I wanted to use f/1.4 for a high proportion of shots, I would get the Sigma. It has HSM focusing and it’s every bit as good as Sony SSM, working perfectly with smooth manual focus over-ride and very fine adjustments in virtual silence. It is also built like a tank and the two (new) samples I tested on Nikon and Sony were equally good.
The studio tests
On to the final part, the studio tests of the various 50mm lenses and of the two 24 megapixel bodies. What I show here is a fraction of the many tests done, and raw processing exports produced using a whole range of raw converters and different parameters. The ‘native’ results of the Nikon and Sony look so different you would believe they were different sensors, but by careful alignment of the default settings in Adobe Camera Raw 5.3 it was possible to see they are clearly identical or very similar. Nikon treats exposure calibration differently, and I believe they pay Sony for the best test-selected sensors from the production. I don’t think they add a magic pizza topping. Everything in the difference between the A900 and D3X, including the price, points to selection of the absolute pick of the sensor crop after fabrication.
That is perfectly possible, it happens already in most fields of similar production, every sensor has to be tested anyway (each one has a different calibration table) and with the D3X, you get a premium quality imager for £6,000 where Sony give you a standard quality for £2,000.
Here is the basic set-up used for the test shots, and the first shot, on the Sigma at f/8 using the Sony:

To view the full size image (‘original’ size) go to:
http://www.pbase.com/davidkilpatrick/image/109053091
It is very large – over 15MB of Level 12 JPEG data – so be prepared for a long wait unless you have fast broadband. The image can be viewed or downloaded. Full caption details are provided on pBase.
This is the first image in a series of ten. They include two examples where the Sony and Nikon are set to totally flat, linear output to show with accuracy the dynamic range (these will look dull and desaturated to the eye, but are a bit closer to the Bayer value contrasts than a normal raw conversion – sure, nothing like as dark and dull as a literal Bayer readout looks!).
They also include full aperture results from all four lenses tested (two Sigma 50s, Minolta and Nikkor), and f/8 results as above. The small differences in framing you will see are partly due to the true focal length of the lenses varying – the Sigma is around 48.5mm, the Nikon is around 52mm – and their different methods of focusing, internal or whole lens. Also, the D3X is a different shape of body and the tripod column had to be raised to place the A900 in the same position. Differences in focus are down to the way the AF worked and how the locked manual focus held. In all cases the intended focus point was the label/wool right hand edge of the pastel multicoloured ball of wool.
To view the sequence, just click NEXT after each one. Tip: to compare images, after they have been cached/loaded, you can click PREVIOUS/NEXT and your browser will flip between them rapidly.
The Sigma lens, though it focuses down to 45cm, is not at its best with close-up subjects and the other two conventional 50mms render fine detail rather better across the frame for this setup. Also, when focused close enough for test charts, the Sigma distortion appears to be around -2% but when used at infinity it’s very low – more like -1%, not worth bothering with correction. Here is an uncorrected typical medium close shot:

I draw no conclusions about the cameras, or lenses, and will let you decide. But as someone who shoots 90% of everything at ISO 100-200 (200 is a bad light setting, 400 is dusk!) I’m in no way unhappy with the Alpha 900. It has, also, so many overriding user-interface, ergonomic, size, weight and cost advantages that every time I process a batch of shots from it and remember the cost of the D3X I can’t help sitting there with a big smile on my face!
– David Kilpatrick FBIPP Hon FMPA