Rokinon Aspherical IF MC 8mm f/3.5 Fisheye

This is the lens version that has an “auto exposure” chip to communicate to the camera, plus a built-in lens hood. The lens is purely manual focus, but auto-focus on a lens this wide would be a bit superfluous anyway (f/5.6 with the scale at about a meter is sharp to infinity). It uses a very sophisticated optical projection formula called “stereographic”, which is extraordinary in a lens of this price.

The lens comes in many flavors, including Canon, Sony, Pentax, and Nikon. For some inexplicable reason, the Samyang company calls this lens Rokinon. Even the internal EXIF data for this lens says “Samyang”.

Being a fisheye, this lens is very specialized and is an acquired taste (I just cannot resist puns). If you have something like Photoshop or Adobe Elements, then you can use something called “spherize” to largely straighten the rather extreme curved lines without losing the 180-degree field of view. More on that later. Other straightening methods, such as “warp” are also mentioned later, but those techniques will need trimming of the edges, and you won’t end up capturing 180 degrees.

If you stick with landscapes and avoid horizon lines, people might not even guess that the pictures were done with a fisheye. See examples below. No single lens can do it all; use the right tool for the right job.

First off, promise me that you won’t go taking any pictures of females with this lens. Paraphrasing Hagrid, “they won’t thank you for that”.

On the other hand, your rooms will look so huge they’d make inhabitants of Buckingham Palace jealous.

You cannot use filters with this lens, so be ready for a bit of a challenge to clean it; the front element is an extreme spherical shape.

This is a DX lens; you’ll get a circular image if you try it on an FX sensor camera.

Surprisingly, the lens is metal and glass; for its price, you’d expect to see lots of plastic. The lens hood is plastic, though, but not flimsy in any way. The focus ring is rubberized like you’d expect, and has very smooth operation.

The lens also has a proper metal lens mount, but no rubber seal. I wouldn’t go out in heavy rain with this baby, but then again I wouldn’t go out in the rain with hardly any of my equipment.

The resolution this lens can produce is quite amazing. Just don’t expect that resolution in the corners with the lens wide open.

Something I noticed that isn’t quite so nice is that the aperture isn’t entirely accurate. You can get some variation in exposure as you switch to different (small) apertures. The physical area for a given aperture on this lens is incredibly small, given its focal length and is probably very hard to build the mechanics to achieve it precisely. Just be aware of this characteristic.

There is also a fairly large amount of chromatic aberration that can be a challenge to control in post processing. The most difficult is the “axial” type; I use Capture NX2 to minimize it.

I’d recommend you try out Live View at high magnification to get a feel for how the distance scale on the lens relates to actual focus. My lens has perfect infinity focus when the lens is at the “infinity” hard stop, but at nearer distances the scale differs a bit with reality. This becomes a near “don’t care” when you stop the lens down.

What you don’t get: vibration reduction. Not such a big deal on short focal lengths, however.

Yes, you’ll get internal reflections to show up when you point it at the sun; they’re generally a pretty small part of the picture, though. It can be difficult to avoid getting the sun in the shot, given the extreme field of view.

I didn’t notice any objectional vignetting, which surprised me. Again, take a look at the sample photos at the bottom to see for yourself. Use your favorite image editor to deal with any lateral chromatic aberrations.

This is a good video lens, if you don’t get too close to people or pan around too much. If you do pan around too much, your audience will be barfing in the aisles.

Although the lens has a computer chip to communicate to the camera, distance information is not properly transmitted and is therefore incorrect in the EXIF data.

Rokinon (Samyang) 8mm with fixed lens hood. Looks like 2001 HAL.

Distortion Analysis

Distortion of the resolution target. Only 18 inches from sensor plane.

Same 8mm photo after software distortion correction “spherize -52”.

An interesting feature in Photoshop and Elements is “spherize”. It can counteract the distortion to a large extent, although it makes some dog-legs near the frame edges. It maintains the full 180 degree field of view. Other software with more mainstream ‘barrel distortion’ correction only mildly corrects anything, and cuts off the frame edges to boot. But you should be buying this lens BECAUSE it does distortion to get you to 180 degree coverage!

There’s also a “warp” feature in Photoshop to help straighten to just about any magnitude you wish, but then you’ll need to crop the borders of what’s left. Lightroom distortion corrections, however, can turn this into a rectilinear super-wide lens that looks quite good.

Resolution Testing

My goal is to enable you to assess resolution between different lenses in a standard, scientific way. I also give you the information (see my MTF Cliff’s Notes article) to be able to repeat these tests for yourself.

The following resolution tests were done using a Nikon D7100 (24 MP) with unsharpened 14-bit compressed RAW format.

Resolution measurements are in MTF50 lp/mm. This measurement represents how many image line pairs can fit inside a millimeter before the white-to-black chart transitions degrade to 50%; e.g. “turn to mush”. For me, anything beyond about 30 lp/mm is fine, and beyond 50 is outstanding. Higher-resolution sensors yield higher measurements, much like you’d expect.

This is a really tough lens to measure. It drove the MTF Mapper software crazy. Due to the fisheye nature of this lens, I cannot measure right to the corners of the frame. I show some photographs of what the chart looks like below, so you understand what difficulties the measurement process has. Also, the (large) A0 resolution target I used had to be only 18 inches from the camera sensor to fill the frame. As a result, the resolution numbers are for a pretty close focus distance.

I use a (free!) program called MTF Mapper from here to measure lens resolution. The download site also has files for printing out the resolution targets (mine are A0 size on heavy glossy paper (‘satin’ finish seems to work just as well), dry-mounted onto a board). This program is covered in more detail in another article, but suffice it to say that this is really great stuff; it’s comparable to ‘Imatest’ in the quality of the MTF measurements, and it uses the “slanted edge” technology similar to ‘Imatest’, also. The author of MTF Mapper, Frans van den Bergh, really knows his stuff. Visit his site and give him the praise he deserves.

The chart design used for resolution tests orients all of the little black squares to be ‘slanted’ but they’re generally aligned in meridional and sagittal (think spokes on a wheel) directions to correlate better with the usual MTF plots you’re familiar with. There’s often a dramatic difference in sharpness between these two directions, and the chart photographs show it clearly.

The meridional/sagittal differences are what “astigmatism” is all about. This lens is better in the sagittal direction. The meridional direction is terrible until about f/5.6, and it nearly disappears as if by magic around f/8. Strangely, the bad meridional performance at wider apertures was only on the right-hand side of the image, as if the optics were tilted. If it were simply tilted optics, though, then the sagittal values would also have been bad.

What the resolution target looks like. Mine is mounted ‘upside down’.

Finally, I’m getting around to some actual resolution results.

Tests were done with manual focus, IR remote, and a really big tripod.

Resolution Results

The following measurements need some explaining. The MTF Mapper program has some problems analyzing the whole resolution chart, due to the image distortion. As a result, some of the plots below have a ‘butter fly’ effect. Also, the distortion means that the chart doesn’t extend to the left/right frame edges. This means that there isn’t any “corner” resolution data.

First, here are some close-ups showing the meridional-direction problem I spotted. The problems are mostly seen on the right-hand side of the frame.

Strange meridional performance at f/3.5 on one side of squares.

Meridional performance on frame edge substantially improved by f/8.

In the following plots, another anomaly I noted is that the “sagittal” plot indicates a bad center, which is not correct. The center is actually the highest resolution, as can be seen in the “annotated’ resolution chart photos.