Fisheye lenses are pretty special when it comes to lenses. Their extreme distortion effect serves to be a novelty, which is sometimes overused, and can prove to be a handy feature to squeeze in extreme fields of view, or to ease 360-degree panorama photo-taking.
The number of fisheye options for DSLRs have generally be limited, and it was hard for many to justify spending real money on a real fisheye lens. In terms of options, the one I like most is the Samyang 8mm f/3.5 fisheye (~AU$380), available under a range of different rebadge names such as Bower, Pro-Optic, Rokinon, Opteka, Vivitar, Falcon, Polar, Walimex and Bell and Howwell. In fact, I liked it so much that I owned one for Canon and one for Nikon! The other that’s widely known is a Zenitar 16mm f/2.8 fisheye (~AU$260), which is more suited for full-frame users, and isn’t as nice when it comes to flare handling.
These options are both manual-type lenses, requiring manual focusing and aperture control, however, for fish-eye lenses with such a wide depth-of-field, the focusing is much less critical than it would otherwise seem. Both lenses cost just a fraction (under half) of what a Nikon 10.5mm f/2.8 DX fisheye (~AU$840) would cost.
Having a need to buy another fisheye lens, I went shopping and came across a Samyang lookalike which was 45% cheaper, at AU$210. There was no branding on the listing and the details were sparse, but the price was definitely attractive. It wouldn’t be entirely out of the question, as I got my Samyang lens for cheaper than AU$380 – the pricing is very much affected by the exchange rate, which, is quite poor at the moment. I decided to take the plunge in case it was indeed a rebadged Samyang.
It was a bit of excitement and a bit of a mystery to get the package and see what was inside. It was listed on eBay with no branding.
As it turns out, the lens is a Kelda branded lens, packed in a mostly generic white colour cardboard box.
The sides of the box lists some of the features of the lens – pretty much a generic spiel. One good feature is a removable petal lens hood – this was not a feature of the original Samyang CS (and was implemented in the CS II). The punctuation could do with some help. Curiously there is no mention of where the lens was made or assembled on the sides of the box, or even details of the company – you have to turn it over to find out more.
The capitalization is pretty bad, but it does clearly show the lens was Made in China, and Kelda Corporation is a US company (unexpectedly). Their URL, http://www.keldalens.com is actually not registered at the moment! How very reassuring.
The first thing you’re greeted by when opening the box is a pleather pouch for the lens. It’s a bit thin, but it feels good and it doesn’t smell like some of the other Chinese plastic pouches.
The lens itself is packed inside a moulded plastic shell in halves, similar to what my Samyang lens was packed in. The lens was placed in a plastic bag with a small bag of desiccant.
The lens features a front cap which clips onto the hood, which is released by squeezing the plastic ribbed portions. The petal hood comes off by rotating the hood until the symbols meet. A rear lens cap is also provided.
The lens features a red-coloured band around the ring, probably making a connection with the Canon L-series red colouring. Focusing is enacted by rotating the larger rubberized ring. The ring rotates with little resistance, and does not feel nicely damped, instead feeling a bit light and hollow. During focusing, the lens extends ever so slightly. The aperture is controlled by the metal ring in the rear, which has click stops.
The front element of the lens, as you would expect, is nicely curved but is also fairly dark coloured. There is evidence of an anti-reflection coating of some sort.
Without the hood attached, it should be possible to use the lens with a full frame sensor without any edges from the hood showing up. The downside is that you can’t fit the cap on, and thus, the front element is very vulnerable to damage or dirt.
The rear of the lens has a very very fine concentric circular patterning, which is different from any other lens I’ve seen. There are no electrical contacts, of course, nor mechanical linkages for aperture.
What’s a Kelda? Is it a Samyang?
As it turns out, some digging reveals that some people have noted the similarities in external design, but no one has been forthcoming with the answer.
Further digging shows that Kelda is NOT a Samyang lens. In fact, Kelda seems to be made by a company known as Kaili, as evidenced by the distributor’s website.
That indicates that the lens was very likely made by Changchun Kailii Optronics Co. Ltd. It’s a bit of a disappointment, but it seems as if their design was deliberately inspired by Samyang. I guess the performance is probably what matters most.
For avoidance of doubt, let me compare it with my Samyang lens, branded Rokinon, which I have had for a while. First thing, the Samyang lenses come in a thin suede cloth bag.
The lenses generally have a golden ring and clearly have Made in Korea inscribed in the lens. The fonts are much nicer as well. There is some deliberate similarities in terms of shape, size and even labelling, with both lenses known as the CS. The original CS does not have a removable hood however. There is, however, some major differences with the mechanical edge for the aperture, as well as the reversed direction of the focus ring. The number of clicks of the aperture ring is different as well, with an f/19 setting on the Kelda that is not on the Samyang.
The front element of the lens is a little more blue coloured, but the similarities in the retaining ring for the front element is seen as well.
One of the major similarities is also in the lens cap, but as it turns out, there are some differences as well – the shape of the central diamond and the patterning of the plastic are different. The patterning of the plastic seems to inherit the shape from the patterning of the metal body of the lens itself.
Samyang on Left, Kelda on Right
It’s also easy to see the colour differences between the front elements.
So, it’s not a Samyang, that much we know. But lets see how well it performs – that’s really the only thing that counts when it comes to lenses.
Sharpness vs Aperture Comparison
This will be a shootout between the Samyang 8mm f/3.5 CS and the Kelda 8mm f/3.5 CS. Both lenses were used with the Nikon D3200 DX body, which actually has a very demanding sensor. It is capable of 24.2MP in a DX format, thus being able to image a much higher resolution than even a 36MP full frame sensor. It is diffraction limited at f/5.6, so a lens which isn’t capable of maximal sharpness at f/5.6 will never be able to make the most of the sensor (and thus its flaws will be evident). A single test scene was used with a large variety of elements at different focal lengths. Focus was adjusted by hand using Live View at maximum magnification, which was at infinity for both lenses. Shots were taken at ISO200 (due to live view limitation), one at each aperture click stop, with very rough compensation for brightness made by adjusting exposure time (1/160s to 1/4000s). A series of 200x200px 100% crops were taken from the RAWs for comparison. All RAWs were processed through the same workflow without any deliberate enhancement.
The test scene was the image above, with six areas concentrated on for the comparison, covering different areas from the centre and different focal distances. This would compensate for any mis-focusing which would show up as lenses trading blows depending on where the crop was taken from. Output was compressed at maximum JPG quality (12) in Adobe Photoshop. Click for full size comparison sets.
A close look shows that the Kelda lens seems quite soft and blurry compared to the Samyang. The Samyang has a lot of chromatic aberration at f/3.5 but sharpens very nicely at f/5.6 and slowly reduces to softness due to diffraction. The Kelda seems to reach sharpness only by f/8 to f/9.5 where it holds on and doesn’t quite soften as badly at f/22 as the Samyang does. That’s too late to make the most of the 24.2MP DX sensor but is probably more suitable for a 10MP DX sensor.
Looking at the nuts and rivets on the pipe on the right is a very detail-laden area that clearly exemplifies the differences between the lenses. Again, the Samyang shows chromatic abberation at f/3.5 that makes the image a bit less clear, but f/5.6 is getting pretty much sharp as you can expect, and starts to smooth off due to diffraction by f/11 onwards. The Kelda never quite reaches the same level of sharpness, and it seems to peak at f/9.5 to f/14. It doesn’t get so soft as the Samyang at f/22, but that’s hardly really relevant.
A close look at a leaking joint with rust shows the same sort of pattern, where the Kelda sharpens after f/9.5. Aside from the Samyang’s f/3.5 result, it seems the Samyang lens’ worst result matches or betters the Kelda’s best result.
The Kelda seems to sharpen at f/8 based on the needles, although it’s harder to see the differences in this sample. Again Samyang wins almost hands-down.
The same trends are seen when you look at the detail on the insulator – but the Kelda doesn’t soften as badly at f/22 as the Samyang. At f/3.5, Samyang has a large amount of chromatic aberration.
The sign nearer the edge seems to show the Samyang has less sharpness than expected, with green-purple CA at f/3.5, and sharpness peaking at f/8 for the Samyang. The Kelda has a blue-yellow CA at f/3.5, and sharpens up at f/9.5 onwards.
On the whole, the Samyang wins the sharpness battle, hands down. While the Samyang sharpens nicely at f/5.6 most of the time, the Kelda doesn’t sharpen until f/9.5 and it doesn’t even match the Samyang lens. The Samyang is a much better lens for use with demanding sensors. Given how little the fisheye might be used by some, the Kelda compromise may be acceptable when producing outputs of 10MP or less, or for filming video although the clicky aperture might not be preferred.
Full Sample Images
These sample images were shot by hand using the same gear, this time, in ISO100 through the viewfinder, to test the lenses side by side for different conditions. The same exposure settings as well as aperture and focus distance were used for both lenses. Full size shots available by clicking on thumbnails. Crops emphasizing the differences are presented.
Please ignore the slight vignetting in the bottom right corner of Kelda shots – this was due to a slightly mis-aligned removable hood – probably one good reason not to have a removable hood after all!
The detail in the braiding of the ropes is subtly lost in the case of the Kelda. Even with a large set of focus bracketed shots, with the softness of the lens, it was hard to find any sharpness more than this.
It’s clear that the bark patterning on the trees is sharper with the Samyang, but the detail on the Kelda is not entirely objectionably poor.
Both lenses did not seem to have any big issues with having sun directly entering the lens, which is good, because it’s quite a likely circumstance with a fisheye. The coating seems to have a slight impact on the final white balance and colouration, with the Kelda being slightly more red, and the Samyang slightly more green.
Park Bench Focus
Focusing with the Kelda even with live view on full magnification was a challenge. Despite my best efforts, I could never get sharpness in the park benches, but I had no such issue with the Samyang. I think this is the best the lens can give for “intermediate” focus between infinity and 30cm.
Underneath a Tree
Again, the same sort of results are shown – the Kelda is somewhat soft at the pixel level, although overall, it’s probably fine for video and 10MP image print level.
For many, justifying the cost of a fisheye lens is hard, given the speciality of it. That being said, the Samyang has always represented great value in the fisheye department, but don’t be fooled by this look-alike.
It seems as if this particular lens was made deliberately similar to the Samyang, to inspire confusion, but the lens itself is definitely not a Samyang lens. In terms of performance, it doesn’t even live up to Samyang’s level of performance.
The lens is soft in comparison, and doesn’t come up to sharpness until about f/9.5 meaning that its resolution is more suited to the 10MP DX sensor level, and is not able to make the best of a 24MP DX sensor.
It also was fairly difficult to focus as it wasn’t clear when it was focused or not when it’s mixed in with the lens’ softness. It probably reflects the lower quality glass or design, which explains the lower price.
It wasn’t entirely a loss, with the flare and chromatic aberration performance being on par with the Samyang, and it would be an adequate lens if high-definition DSLR video is your thing. The build quality feels slightly less sturdy, with a noticeable extension of the front element in focusing and a not-well damped/smooth zoom ring.
I suppose this lens has a market where you want the cheapest proper fisheye lens for a DSLR, or where you don’t need to extract every pixel out of a sensor (e.g. for taking print-quality or web-quality photographs or filming video). It is definitely affordable. But the Samyang which is almost twice the cost is significantly sharper, for those who care.