Performance in practice
Sigma’s own MTF chart for the extremes of the lens range tells a very clear story – the lens is ‘better’ at the long end. The red (10 cycles = coarse detail, high contrast) sagittal and meridional lines indicate really good performance with larger image structures at both 18mm and 250mm. The green (30 cycles = fine detail, low contrast) indicates how well fine detail like grass texture, skin, hair and similar things will be resolved. I find Sigma’s chart explanation slightly confusing, as I would normally expect this test to be performed with radial and tangential lines. What you actually see here with the green dotted lines in particular is a slightly different focus point applying to lines running diagonally one way, and those running the other. This is normally caused by residual coma and astigmatism, common off-axis aberrations. Since real subjects are three-dimensional the result is most visible on test charts, least in real life.
That over-simplified short throw focus scale hides an excellent macro range, to 1:3.4X. It beats the Sony design for close ups, but it’s not just a matter of distance and scale. The Sigma has some truly unexpected qualities, which overturn everything we have seen in superzooms before now. One of these is almost perfect straight-line geometry for the closest focus at 250mm.
This is a shot of an LCD screen (iMac 24 inch) and those blue lines are the rules on an InDesign page. This is minimum focus, wide open, 250mm. Frankly, it’s amazing. It is not critically sharp but it’s as even as you could possibly want across the field, meaning the focus field is flat. The geometry is nearly perfect. You could copy stamps with this lens (for stamp-copying purposes, as a professional stamp-copier!). But, of course, you don’t spend all your time focused approximately one hand-span away from the front element. That 45cm close focus is measured from the film plane. From the viewfinder eyepiece to the front of the lens hood is 30cm, leaving you just 15cm clear between gear and subject.
You can also shoot at a distance and get reasonably straight lines. This is 250mm from a long way off. The building, though old, is not subsiding in such a way that the lines are all curvy and our lens has corrected them. You will see a hint of pincushion on the short edges (further out in the field) but again almost perfect rendering 7mm out from the lens axis (the long edge). It’s not totally straight, and has a complex distortion pattern meaning that further into the frame pincushion reappears (see the top of the upper windows, compared to the string at the bottom).
But compared to most superzooms, it’s got very straight geometry indeed – no worse at 18mm than others, and considerable better at longer focal lengths.
It also has good bokeh (rendering of the out of focus image) compared to other lenses of this range (and, an addition to this article a few months on, this Scottish thistle shot has earned me a the price of a decent 12-year-old island malt):
Guess the working aperture? Wide open? No – f/9. Again, this is at 250mm – 1/160th at ISO 200 on the Alpha 700. I have used both the Sony and Tamron 18-250mms for close-ups, and the Sigma simply does it better. It has no field curvature issues, and whatever aperture shape it happens to offer, it produces a very good bokeh. I first encountered this testing the lens on the Canon EOS 500D. I got the first sample of the 18-250mm, before even dPreview had it (the same one as it happens, we had it for a few days before it was sent on to them) and I was asked not to publish results before dPreview or Amateur Photographer, so it had to be included in tests of the camera, not as a lens test.
On the smaller Canon 1.6X sensor this lens will naturally perform better – the 1.5X sensor, of course, presses it a bit harder for vignetting, disortion and sharpness to the corners. This is something which many reviewers forget about when they choose Canon APS-C bodies for reviewing lenses, they are not getting the full picture. Here is a Canon shot:
This was at f/8, when the cherry blossom was out – tells you how long ago I got the first sample! Then we waited for the Sony HSM version to come out, and we have been using it for over a month before finally reporting here on the conclusion. We made a mistake – the lens arrived one day before Shirley and I flew to Venice for a week’s fairly intensive stock shooting. I use a range of lenses, but Shirley prefers to use just the 18-250mm. We got the Sigma out, and she was put off by the size. It was also not going to fit in our single cabin bag (a rolling Kipling camera case sized exactly for airlines) as comfortably. She opted to take the Sony instead.
The week after, processing the Venice shots and testing the Sigma, it was clear that we should have taken it, not the Sony. Whenever the Sony was used at f/6.3 and the long end, results were likely to be too soft to pass quality standards for stock libraries. The Sigma, in contrast, is dead sharp in the centre at 250mm and f/6.3 – it does not need stopping down to f/8 like the Sony.
Here is the centre detail at f/6.3 from the building shot used above to show geometry, hand-held with OS at 1/400th, ISO 200, Alpha 700, autofocus, given +1 exposure boost in conversion as it was pretty dull:
Now you may say – that’s not impressive. Remember, this is extremely low contrast very fine detail, visible in the tiny cracks of the paintwork. It’s got normal capture sharpening (25 in ACR) and no noise reduction. The Alpha 700 in v4 firmware is fairly noisy even at ISO 200 as the dark shadow area of the window shows, and the one-stop exposure correction emphasises that. But it’s also very sharp. I suggest you view this image on a good LCD screen, and take 20 seconds or so to look steadily at the paintwork to the right side, let your eyes get accustomed to the contrast. You will see clearly resolved detail at the very limit of pixel sharpness.
Here is the extreme bottom left corner (actually cropped off the full image shown earlier, which was straightened up a slight amount to align the straight masonry string with the frame edge. You can see it is rougher, as you would expect, and fairly strong CA correction has been applied – +50 Red/Cyan, with -28 Blue/Yellow, and Defringe all Edges. The result of a straight process with no CA correction shows the exact extent of the strongest colour fringes given by the Sigma:
Note that the green fringe is very sharply defined. When CA is lateral (off-axis chromatic aberration, blue/red extreme wavelengths form smaller/larger images in the same focal plane) more than longitudinal (same wavelengths for same sized images, but focus in front of/behind the focal plane) a very clean post-process correction is possible.