Samyang 8 mm f3.5 fisheye CS lens

Shaving and test report

Following is an early short review about the Samyang "aspherical" lens which I got imported from Poland on the 20th of July. This review is intended for reading by spherical panograhy enthusiasts.

Other brand names (but very same product)

On the web, I have found the same product (certainly made by Samyang) alternatively proposed under many other brand labels such as Bower, Polar, Rokinon,Walimex and Falcon. Opteka has choosen to arbitrarily affecting it with a 6.5 mm focal length and the very same product with a remodelled, sexy and "Pro" look is also labeled " Vivitar 7 mm f3.5 (Series1) fisheye"!

I also have discovered that the same lens is sold on eBay by Bestpriceoptics (Saddle Brook, NJ, USA) without any brand label at all. Bestpriceoptics sells also simultaneously the same objective under the Rokinon name!

We can expect Digital Innovations and others to be also brand names for the fisheye soon....


The instruction manual describes in length how to specifically use the lens on Nikon (Fujifilm), Pentax (Samsung), Sony (Minolta), Konica Minolta (Minolta) Dyna Series and Canon cameras. This document also mentions that the lens design is patented: on the contrary to Pentax (zoom 10-17 mm), Nikon (10.5 mm) and Canon (15 mm), I was unable to find any related Patent file on the USPTO data base.

Build quality

The overall build quality is IMHO excellent and the street price is 1/2 to 1/5 compared to similar other fisheyes (230 Euros; 22% VAT included in my case).

Contradicting a previous review that I recommend to read, the focus and aperture rings aren't loose at all and moreover aperture setting intervals is actually 1/2 stop (f/5.6 to f/22) on the final production model. The omission of the f/4.0 setting value is IMHO fully acceptable: right after f/3.5 comes f/5.0. 
The focus setting ring and the corresponding scale are better than on the Nikkor or Tokina, but there is no scale for depth of field: the instruction manual is erroneous on this last point.

Shaving of the sun shield hood for use on a Full Frame DSLR

Shaving operation was very easy to perform and event less. The sun shield hood is made of thick black molded plastic that encompasses ~12 mm of the front part of the metallic cylindrical barrel of the lens. The lens front barrel is consequently now trimmed to a more reasonable 70.5 mm down from the original (fat) 76 mm in diameter. A complex groove is subsequently created:


Test results

No-Parallax Point location

The NPP location on the lens axis for 90°, 60° ( 3-Shots panorama) and 45 (4-shots panorama) are indicated on the image above. Please click on the thumbnail image to view.

Distortion (Image radial mapping on a FF sensor (or on the diagonal of an APS-C (r= 1.5) sensor

Early tests and measurement on the (shaved) Samyang 8 mm fisheye have been giving some stunning results. That lens imaging features are mostly awesome when it's mounted on the EOS 5D. The angle of view is restricted to ~182° @ f/3.5 (it is barely 180° at smaller aperture openings), and the projection method looks rather unusual: that might be the reason for the nice overall imaging performance.

The projection method looks to belong to the stereographic projection "kind". Some authors define stereographic projection differently from other others. Hence the mathematical formula may differ from a source to another.

Moreover the curvature of the graph is clearly opposite to the common fisheye lenses of the past 20-30 years (most of them are of the equi-solid angle kind):

You can hereabove get an idea of how the Samyang fisheye mapping compares with some other fisheyes of similar focal length range. The difference is dramatically obvious!.

Click on the image to open or download the larger PDF file.

Focal length

The focal length of this lens is subject to controverse on some forums: Vivitar claims it to be a 7 mm fisheye and all others sellers of the same product seem to think it to be a 8 mm of focal. Unfortunately there is no one unique way to determine the focal length of a fisheye especially if the projection isn't exactly of a canonical type (The curve can be somewhere between the curves of two canonical types) and the practical value for focal length also depends on the maximum angle of view allowed by the sensor (and by the sun hood shape that was limiting the angle of view before shaving it).

So, what are the estimated focal length and projection method of the new Samyang fisheye?

We have determined above that the projection method has to be of a y = k tan (theta/k) rather than the usual y = k sin (theta/k) relation. Note that the latter relation characterized most (if not all) existing current fisheyes offered for sale on the market.This new lens is therefore first of its kind: the stereographic projection was mostly the production from computerized outputs. The design and production of true stereographic lenses for DSLR had been thought to be very complex when chromatic and other optical aberration would be well corrected. Samyang seems to have done a breakthough since the lens happens to be really affordable in face of comparable fisheye lenses.

A pair of values (f and k) should determine completely the lens characteristicsof the lens and we should find a way to determine the pair of parameters value that resembles the more with the observations.

When we try to place the measured radial mapping of the Samyang fisheye lens on a set of theoretical projection graphs, the best fit that can be found when trying to match the Samyang measured data against theoretical graphs of all lenses appears to be y= 3*f*tan (theta/3) in association with f= 8.7 mm:


Click on the image to open or download the PDF file.

You must know that this way of determining the possible projection law cannot give a bi-univocal relation. Other couple of values for f and K could yield also a decent fit. While I could not make sure that y= 3*f*tan (theta/3) can be called "Stereographic", this is sufficiently close to y= 2*f*tan (theta/2) for a practical interpretation.

What's the practical advantage of this projection?

To better let the world know about the fine peculiarities of the "stereographic projection" you may just change this somewhat ackward and scientific name and give it an other sexier label, such as "Tailored Distortion" for example. Then you may claim that this is new and revolutionary. You could probably easiliy get granted of a patent for its design in the USA but you could not register the appealing title though as this was already done (Tailored Distortion is TM by this Company). End of rant.

Image Quality

The IQ is as good as with my best other fisheyes and tops them all at f/3.5 especially near the edges of the image circle, thanks to the specific "very low" radial compression in this area coupled with "sharp" vignetting. Take note that the image circle diameter is exactly the same as from a shaved Nikkor 10.5 mm.  The Samyang has however 2 mm and 20° of focal length and FOV respectively less than the Nikon lens!
Sharpness and contrast are very good and inherent TCA can be easily corrected. Flare and colored ghosting seem to not be worst than usual (dedicated test yet to be performed). Light fall-off is insignificant and vignetting is "clean". 

IMO the Samyang fish is the ideal lens for fast 3-shots panos on a FF DSLR camera. Stitching yields 7250x 3625 pix equirectangular (EOS 5D sensor). PTGui and AutoPano are at ease with the new fisheye projection but the user may expect rather unusual optimization parameters values! 

I dare to predict a huge success for this product in the panographic community despite the diagonal restricted to 167° on a (r=1.6) DSLR sensor;-) 

Sample images

Zip file (37 MB): Three photographs of the interior of a room in raw format. Shot at ~120° intervals on the horizontal plane. SAMYANG 8 mm Fisheye, Focus 1 m ; aperture f/8

Example of panoramas

Interior of a room

Panorama Flash (3.7 MB): made with the three sample images.

Stitched and blended with PTGui 8 pro (7300 x 3500 pixels). There is a slight glare and some light flare ghosts at the Zenith as well as a flare colored pattern on the Nadir of one image. This was caused by some of my finger prints that I had probably and inadvertently left on the front element although I had carefully cleaned its surface (Note that the aperture was set at f/8). These artifacts disappear almost completely on the photographs shot on the same scene at full aperture i.e. f/3.5. The focus distance was deliberately and manually set at the 1meter mark on the distance ring/scale.

Sunflowers fields in the French south-western country (Sunny day)

This was done with four shots around from the top of a 4 meters high pole f/11, 1/20 sec.,100 ISO

And the same scene six weeks later, afer harvesting.

Same setup but f/5.6, 1/1000 sec.


Illustrated comparison with three other fisheye lenses

The subject is to view the appearance of the different images of photographs shot with the same camera (Canon EOS 5D) at the same place and from the same spot but with different lenses: The camera was placed exactly at the same location as above for the panorama, but in landscape mode rather than portrait.

The (shaved) Nikkor 10.5 mm, the (shaved) Samyang 8 mm, the Sigma 8 mm f4.0 or the (shaved) Tokina 10-17 mm @10 mm were successively mounted on the camera.

The photos were downloaded in JPEG format from the camera and resized down by PS CS.

Two distinct object (flash) movies show one images series each: the direction of shooting differs by 90° from one series to the other. One series aims at the door to the garden and the second aims at the target for playing darts, both having a corner of the room in common on the right or left of the image respectively.

The radial mapping, the angle of view, the light fall-off and the chromatic aberration are put in relation.

You may click on the images to run the movies:

How do the Nadir views compare?

The unveiled projection geometry of the Samyang has ignited some hot debate on some forum about its advantage compared to the other more classic fisheye lenses.

Some see very clear and sharp Nadir view while another seasoned panographer pretends that there is no difference to be expected when all processing is correctly done. IMO there was a need for an experimental comparison.

Comparison testing

3 images for the same "3-shots panorama" has been shot by an EOS 5D successively with a Nikkor 10.5 mm @~f/8, a Samyang 8 mm @f/8 and a Tokina zoom @~10.8 mm and f/8. The camera was hung from the ceiling at the tip of a monopod where a Nodal Ninja R1 clamped around the lenses.

A trio of "assorted" images has been inputed in PTGui Pro: one image from the Nikkor, an other from the Samyang and the last one from the Tokina, each separated in yaw by 120 deg. from the others. The 3-layer PSD output was then processed in PS-CS to crop three equal 120 deg. parts and then in the PTools "Adjust" plugin to center the panorama at the Nadir. No blender was involved.

The resulting panorama is shown

as a flat JPG cropped image of the Nadir area (i.e. center of the equirectangular image):

or alternately as a Flash VR panorama (1.3 MB):



Sensitivity to flare

The Samyang shaved fisheye is prone to show flare ghosts when the light is entering the front lens in certain oblique direction. There seems to be however no (or at most a moderate) loss of contrast or haze in this situation. More testing shall be devoted to assess this dreaded possible problem even if it would occur rarely.

Examples: (Shots at f/11aperture stop setting)

A panorama with multiple small hexagonal colored dots:

New (5-Aug 2009): I was able to reproduce consistently this phenomenon . The small ghosts (of circular or hexagonal shape) appear mostly when the aperture stop is set to high value (small diaphragm opening like f/22) but I still need some time to analyze the results.

With this new findings, I feel compelled to warn the reader about an evident sensitivity of the Samyang lens to flare. The importance of this drawback must be investigated.

A specific page on the Flare topic including comparison with other fisheye lens is being created.

Click to view the VR panorama (Flash 2.3MB)

Almost the same panorama when the sun has move a little bit and was then hidden behind the bell tower of the church:

Click to view the VR panorama (Flash 2.3MB)

Focus, distance of focus and sharpness

The same subject was shot 13 times at the same distance of 1.5 meters (+/- 2 mm) from the sensor plane vertical axis. The Tif images (converted from RAW) were layer stacked in PS CS4. Then adequate crop was done to reveal the subject at 13 places against the same background before flattening the image.

Distance set on the focus distance scale & ring = 1.5 m; f/3.5, 1/250, Mirror locked up. While the aperture setting is at fully open, everything seems to be as sharp as it can be but this test is not conclusive.

To find a better instructive information, a specific test involving peering to about 300 photographs of a single target was performed to draw up thefollowing flash object movie:

2D Mapping of the in-focus zone

You may read a how to draw up such a map!

Color keys:

Too soft
Slightly soft
Ugly bokeh
Badly out-of-focus
If need be only

Shooting with the distance of focus ring set at just a tiny bit below the "Infinity" mark is probably the best all around setting that can be. Closing the aperture ring to f/5.6 or f/8 and every thing should be as sharp as it can be!

The protective cap after shaving

The cap is well designed as it has (like many other standard cap that grip inside the lens filter thread) two buttons to radial push and thus firmly secure the cap by two lugs that grip inside the sun hood grooves. That function is totally neutralized when the sun hood is sawn off.

The -monstrous- lens cap could be used again for protection on the shaved lens by inserting the (cut) hood inside the cap (with some tape to fill the tiny gap) but I would firmly advise not to do so: the "protuberant shape" of the cap is a dreaded temptation for the user to inadvertently lift the camera by hand grabbing it around the cap when the camera is lying on its back (on a table for instance). The camera and its attached lens shall then eventually fall on the floor... but after a second without warning signal. This is a lesson learned by seasoned panographers: this very same problem happens with all of the FC-E8 and Coolpix camera combos for instance. 

The lens cap then must now be adapted to functionally and securely fit with the new configuration and this might be the only real drawback (when one accepts the lack of Auto-Focusing and of EXIF due to the absence of electrical contacts). 

The fisheye for night sky watching and astronomy

Some astronomers use a fisheye lens mounted on a DSLR to make either photography or time lapse videos.

I have been contacted by one astronomer that inquired about the possibility to focus at infini with a Canon sensor on a camera deprived from its IR filter. The consequence of removing this IR filter is a change of the back focus distance of about 0.8 mm: it's as if the lens was moved away by 0.8 mm from the camera mount.

As some lenses have been designed with the infinity mark on the focusing distance on the scale having a small extra angle of turn beyond infinity before being stopped to cope with this situation, we decided to make a test and see if the Samyang lens could still get a crisp image when set to infinity and when the lens is temporarily mounted with a 0.8 mm gap between the camera and the back of the lens mount: the result was unambiguously negative. The whole image is badly blurred @f/3.5. An optical corrected adaptor must then be used to make the Samyang fisheye suitable to a camera from which the IR filter has been removed.

Close-up range and D.O.F. adaptation

The foreward [4-lens optical group] element can be unscrewed very easily. The thread is about 7 mm long before total dismounting.

Under his own responsability, the user therefore may select to temporarily extend slightly the lens to reduce the lower limit of depth of field : I have tested this way to verify that an object actually touching the glass surface of the front element shall be sharply phototgraphed...

Michel Thoby

Issued 23 July 2009

Last revision 14 December 2009