Tamron 10-24mm f/3.5-4.5

Tamron’s new ultra wide angle zoom for APS-C/DX is getting a bit of a blasting from reviewers. Now, when I see this happen, I get curious. Lens testing is often badly designed for such zooms, involving test chart targets at distances which are extremely close and result in very bad figures caused mainly by a strong curvature of field (dished, ‘cap’ shape relative to the camera) when gets worse in effect the closer you focus.

So I obtained a Tamron 10-24mm to test. The only problem is that Tamron in the UK (Intro2020) has a new policy that all review lenses will only be available in Nikon mount. I found their 70-200mm f/2.8 so hesitant in focusing on the D3X that I did manage to obtain a sample in Canon mount, which I tested on the 50D. But this was a special case because I felt the lens simply did not agree well with the Nikon AF system. Indeed, it proved far more positive and faster on the Canon.
It’s very hard to request a loan body to test a rival maker’s independent lens (indeed, very presumptuous as well!). Therefore, in February I purchased a Nikon-fit body of the most unlikely type – a vintage 2001 Kodak DSC760C, only 6 megapixels but probably the bext 6 megapixels around, with no anti-aliasing filter and a 1.3X factor. I also had a Fuji FinePix S5 Pro to hand, but there’s no way the Fuji could compare with the older Kodak for checking lens performance – it’s just too soft in rendering.
The Tamron did not AF on the Kodak, despite the old DSLR being supposedly compatible with its in-lens motor system. It did on the Fuji. When testing it using MF with focus confirmation on the Kodak, I discovered that a huge range of distances were ‘in focus’ according to the AF sensor.
So, what is the lens really like?
Aim it at a wall, at 10mm, and you will get to see exactly why it gets bad reviews. This ‘test chart’ would be two by three metres, and most tests are done even closer. Full aperture shows how those corners just disappear from the MTF chart:

With my sample pictures here, you can click the image to open a larger version. But even at this small size you can see the sudden death of illumination on the S5 Pro, combined with serious loss of flat-plane sharpness. This lens seems to be made for those mini-sensors you find in Canon cameras, 1.6X instead of 1.5X like the big boys over at Sony and Nikon (and Pentax and Samsung). I’ve marked the exact Canon sensor coverage – you can see how much you lose of the 10mm view, and how the soft corners are eliminated. Tamron should have issued this lens in Canon fit only for testing, they would have got much better reviews!
But, still working wide open, still working at 10mm, a subject with a very different distribution of depth in focus creates a slightly more favourable result:

The cap-shape curved plane of the lens focus has suited the way the pillars are closer than the distant aisle. You can also get some idea from this of the distortion (barrel, simple) at 10mm. It’s certainly there, but it is also very simple to correct, unlike the complex moustache-wave type distortions often found in ultrawide zooms. Fuji’s extended dynamic range helps with my favourite difficult subject, one which allows me to get nice looking pictures even in the depth of a poor winter season. Click this and you get a full sizer, albeit rather heavily compressed.
What I found in practice was that angled perspectives, with the ground close to the lens and receding into the distance, favoured extra sharpness with the Tamron’s curved field in the foreground but could produce some bad blur and aberration in the corresponding distant opposite corners. It’s a ) shaped focus field, for sure, and if you learn how to use it, it’s not so bad.
Fitted to the 1.3X Kodak camera I discovered it almost covered the large sensor size. The superb colours given by the Kodak were a pleasure to work with, as was the non-anti-aliased image. I picked up two of these cameras (one high speed 720X and one ‘high res’ 760C) for a modest outlay thinking they would be of little use, but the 760 proves very usable indeed:

This file needed extra compression to get down to my 2MB attachment limit! It’s so detailed. It is a crop, mainly cutting off the long edges, down from the 1.3X sensor but you are actually seeing more height than a 1.5X sensor – so it goes beyond the official coverage of the lens. At f/11, the curved field has helped bring the foreground into better focus, and left the branches at the top slightly rough. Again, like most of my shots, it’s at 10mm – which is frankly the main reason why I ever buy such a lens. I do not need the range from 16mm to 24mm, all my other ‘standard zoom’ lenses for this format cover from 16mm to 18mm at their wide end.
This is the dilemma. Tamron has gone for specification, not utility. The lens has fewer elements in fewer groups than the 11-18mm f/4.5-5.6 but it’s almost a stop faster and it is a 2.4X zoom not a 1.63X zoom, and it goes wider at the wide end. It is barely any larger, same 77mm thread, even takes exactly the same lens hood – physically it is nearly identical in size and weight to the 17-35mm f/2.8-3.5 full frame design.
In the process, they have definitely lost quality in terms of even coverage and illumination. The 11-18mm is probably not quite as sharp centrally, but beats the 10-24mm in every other respect. Maybe the simple distortion of the 10-24mm at 10mm is a superior feature. Here, you can see it clearly again:

This subject also helped me work out the CA settings – minus 30 Red/Cyan and +30 Yellow/Blue in ACR/Lightroom at 10mm.
The lens is unusual in focusing extremely close (the focus field then resembles a bowl in front of the lens!) and in having a huge exit pupil and rear element size:

I’ve seen 50mm f/2 lenses with smaller rear glass than this design!
Where Tamron has misjudged is in increasing focal length range, rather than aperture. A maximum of f/3.5 is not much use as it does not actually brighten focusing screen views much, and does not enable f/2.8 AF sensors to function. Tokina’s 11-16mm f/2.8 constant aperture zoom, which I have tried on Nikon D90 with excellent results, focuses with far greater accuracy because (depending on the camera body) it switches on the double-precision f/2.8 sensor/s. This does not apply to Nikon as none of their DSLRs incorporate f/2.8 capable sensors, all are f/5.6 limited. Canon use f/5.6 sensors with the addition of f/2.8 capable sensors in some bodies, Sony use f/7.1 sensors with central f/2.8 sensors in the A700 and A900.
Just to show you, for the sake of it, how much depth of field a 10mm lens has at f/11 (which this lens demands as a working aperture, if you want to lose the soft dark corners) here is a series of shots taken using the Kodak DCS760C, all of which fell within the range of ‘focus confirmed’ by the camera:

This is at ‘beyond infinity’ – note the moirés on the clock, caused by having no AA filter. It is a 100% clip from the six megapixel file, not bad for 2001!

Here’s 3 feet/1 metre (the focus scale is not very precise!).

This is 2 feet. I mislabelled these pix because on seeing them, I was sure the distances were metres – surely the depth of field could not be THAT great, given the common focusing errors with ultrawides? But no – I had been adjusting the setting in FEET not metres…

We are now at 1.5ft and the focus confirm light (on a Nikon F5 based body) is still saying all is well, and that is at an effective f/5.6, let alone the f/11 actual shooting aperture. Nikon bodies do not generally have f/2.8 sensors and use f/5.6 for all AF points.

Finally, with the lens focused on 1 foot, we get the AF/manual focus confirm light to go off! Now I had worse focus than this on my Dynax 7D with my first 11-18mm Tamron, which gives me some idea just how badly calibrated the camera was before I adjusted its focus screws.
Message – if your ultrawide or wide shots do not look sharp at f/11, your camera does have an FF/BF error physically. I would never have thought I could focus all the way from infinity+ to 18 inches and still get acceptably sharp results at this resolution (which is more like 4 megapixels than 6 relative to the 1.5X format).
For reference, also showing the entire field of view on the 1.3X sensor with the petal lens hood in place, here is the overall view. I cleaned the Kodak’s sensor after taking these shots:

What conclusion, if any, do I have about the Tamron 10-24mm? First, I have a Sony 11-18mm and as far as this test goes, I’m sure it is better and I would not swap it. Secondly, f/3.5-4.5 is no benefit for AF and not much benefit for viewing – but f/2.8 would have been a real advance. Thirdly, my own uses for such lenses are more technical than creative, to record very wide architectural shots and interiors, therefore sharpness corner to corner counts most of all.
The 10-24mm Tamron is almost a photojournalistic lens. Where the subject is centred, and the corners do not matter, it produces an image like an old Summar f/2 or a 1960s Nikkor 28mm f/3.5 wide open. It has a look which could be exploited.
Should Sony issue a version? At their peril! There are reasons why this lens has been reviewed unfavourably. In fact it is very sharp centrally, has an amazing range, great close focus, simple distortions in place of dirty ones and would be wonderful on the Olympus 4/3rds format. It even manages to nearly cover 1.3X when stopped well down, to f/11 or f/16.
My recommendation must be to prefer the Sony or Tamron 11-18mm f/4.5-5.6 despite the loss of maximum aperture. At least the wide open setting is fully usable on these lenses, and by f/11, they are in a different class to the 10-24mm technically.
Brave attempt, but not what was needed.
– David Kilpatrick FBIPP Hon FMPA