A new Super Wide -Rectilinear- lens:

SAMYANG 14mm F2.8 ED AS IF UMC

Is this new very wide-angle lens any good and... can it be used for panorama photography?

Note: In any case it's really an affordable objective;-)

 

Lens reviews on the Web: high image distortion is reported!

Photozone.de has posted two separate reviews of the Samyang 14mm f/2.8 IF ED UMC-Aspherical lens: one is for APS-C Cameras and the other for Full Format Cameras.

Distortion is hefty: -5% and -5.4% of barrel distortion has been measured for the two respective camera format, with a mustache-style sub-frequency for the second case.

Lenstip.com has posted two successive reviews about the lens but only the recent one is strictly related to the Samyang 14 mm f/2.8 ED AS IF UMC. A former article had been written about the Samyang 14 mm f/2.8 IF ED MC Aspherical (i.e. non "UMC") and Samyang has subsequently withdrawn this model after reading the review (Sept 2009). While the newer lens review is also extremely positive in every aspect but two: besides a critic about vignetting level (that I do not agree upon) you read that the distortion has been here measured to be a high as -5.44% on small (APS-C sensor) cameras whereas on the full frame cameras it increases to the level of –7.21% (!) << while a noticeable moustache distortion appears there as well! >>

Overview of the tests results (others than distortion)

Quote from the Photozone reviewers:

Resolution: Well, we were stunned by the results in this category and I reckon you will join us once you've had a look at the charts below. The Samyang produced nothing short of outstanding resolution figures for a lens in this class. The results are very even across the tested aperture range which means excellent to superb center resolution and very good to excellent borders and corners. Field curvature (flatness of the focus plane) is well controlled.
I reckon that you will not believe us so please have a look at our field image section.

Chromatic aberration: the CA characteristic (color shadows at harsh contrast transitions) came as another surprise. Ultra-wide angle lenses often tend to be so-so here but the Samyang mastered this aspect ... easily. The CAs are well below field-relevance at less than 0.3px at the image borders.

Opinion of this author: After a month of use in various conditions I can confirm the results that are reported on the web. This is not a rumor. It's a fact. The sharpness and local contrast of images from this lens are really incredible, unbelievable, outstanding and formidable and the absence of CA is a godsend. We can use the images from the camera right directly into PTGui without passing by ACR. Great!

Sample image

On the image on the left the huge and complex distortion is visible on the tiled wall of a shower image. On the right, a quick (and dirty) correction of the mustache distortion (a = 0.02423 b = -0.0247 c = -0.102) as well as correction of yaw alignment have been performed before final cropping. All was done with PTGui .

After calibration of the lens with many panoramas, the best combination of coefficients seems to be a~ 0.04 b~ -0.1 c~ -0.02 for the general case (i.e. when the subject is sufficiently far away from the camera)

Here below one can see the effect of the correction applied to the "in-the-shower" panorama. The overlaid MS Excel-made simulation was hacked from Peter Reimer

While being excellent, the correction is here not perfectly accurate because the set of coefficients should apply to photographs which the main subject is at least 2 to 3 meters away. Remember: the movement of the Entrance Pupil tends to change the distorsion with the distance shortness.

Shot 2 sec. at f/11. The distance of the wall "inside the shower" from the front of the lens was about 0.55 meter (20 inches).

Panorama example

Following is an example of Interior Architecture panorama composed of 14 EOS 5D2 images (6@30° + 6@-30° + 1Z + 1N viewpoint) stitched with PTGui Pro to output a 13600 x 6800 pix equirectangular image that was converted and reduced to a 1800 cubic flash SWF file with Pano2VR. The source images were shot in JPEG L with the Canon FF camera (bracket +/- 4 EV of 5 images with a Promote Systems controller) then enfused. Despite the close space (the walls are about 0.55 meter from the camera => Focus at hyperfocal of 0.7 m) the strong mustache type distortion seems to be well corrected and the stitch quality was very good in PTGui.

Least Parallax Point (LPP) location and movement along the lens axis:

YES! The center of perspective moves about 17 mm (!) along the axis on this lens. Measured with the LASER pointer test bench, we could estimate the move with fair accuracy at least when the incident angle of light ray bundle is sufficiently high (i.e. >20°).

LPP Movement on a PDF chart

Click on pict to view or download the PDF chart.

Please note that the accuracy of measurement is inherently mediocre for the lowest part of the Incident Angle range.

The pupil shift of a standard lens is not very well documented in the literature. However some people have reported LASER measurement of the LPP shift of a 28 mm Nikkor lens. Pierre Toscani (in French) has drafted (see Fig. 16 on the page) a very nice illustration of the "somersault" of the Entrance Pupil on the old and discontinued Nikkor 15 mm f5.6. This illustration is truly amazing!

Radial mapping and variation of focal length along the field angle

The focal length of a theoretically perfect (i.e. undistorted image) should be constant. However, the hefty "moustache" distortion that affects the lens makes the focal length to be far from constant along the field:

The variation of focal length amazingly matches the obvious distortion that one can observe on the image!

 

Suggestion for setting the LPP with an APS-C camera (9 images on the horizontal) or a fulframe camera (6 images on the horizontal).

LLP practical location and Entrance Pupil center shown as overlay on the lens construction drawing:

 

The construction of this recent lens Samyang 14 mm f2.8 is much simpler than e.g. the venerable and discontinued Nikkor 15 mm f5.6 (see fig 16). This streamlining is of course mainly due to the inclusion of an aspherical lens.

Anyhow some similar principles are applied on both design, including a bulging front glass element surface, and consequently the two caustics of the respective Entrance Pupil center obviously resembles each other.

 

Effect of large Pupil shift on a single image (viewed when objects are very close to the camera).

When an object is very close to a fisheye lens spectacular bowing of straight edges appears on a single image, the bowing depends on the distance and is directly the result of the shift of the pupil when the incident angle varies. The fisheye distortion is not corrected at all as it is what make the intrinsic feature of a fisheye lens.

On a similar way the observed Samyang lens movement along the axis implies a correlated movement of the entrance pupil. Amazingly the shift of the pupil has only a very moderate impact on parallax in contrast with the fisheye lens. It could be not noticed at all if one does not look carefully enough.

Test set-up

and this is the resulting image shot on a Canon EOS 5D Mk2 below:

The distance A is larger than B but barely larger!

Comparison of the image radial mapping with some fisheye lenses

Please view this complex PDF overview where the boldest solid red line show the New Samyang curve.

The Samyang 14 mm lens is a standard ("rectilinear") lens. Well, that what it's supposed to be;-)

Comparison of the image IQ with a venerable very wide angle lens: Samyang 14 mm f2.8 ED AS IF UMC (2010) Vs Nikkor 15 mm f3.5 AI-s (1976)

The Nikkor lens was a highly praised lens in the 70's and it is still a rare and expensive lens to buy. Despite its troublesome proneness to catch flare in many situation, it was said to be very sharp. Pierre Toscani has let me used one of them for testing (Thx to Pierre).

Let's see what it's up to when it's compared with one of the sharpest lens of the present time (Fall 2010).

See the images below. Both were shot at a distance of 1.20 meter from the sensor plane. As expected, the Samyang lens (left) shows a larger field of view than the Nikkor (right). Both approximate coverage are illustrated on the image on the left. The same conversion from RAW was simultaneously applied to both the Nikkor and Samyang images with ACR (Notice a slight magenta cast on the former and yellow cast on the latter).

Flare:

Well the images show it all: the Nikkor is very sensible to light coming sideways at about 60 degrees oblique on both sides of the camera. This plague has been what refrained the sale and the use of the lens in its era. The Samyang is quite immune.

Distortion:

The Samyang exhibits about three to four times as much distortion as the Nikkor and the shape is more complex but similar (moustache). That was reported already above in this article.

Vignetting:

The vignetting of the Nikkor is very present at f/3.5 and still important at f/22. Much less vignetting on the image by the Samyang especially when the lens diaphragm is set to a low aperture.

Chromatic aberration:

The image from the Nikkor is affected by a very distinct and high level of (Magenta- light Green) lateral color. The Samyang has almost no visible chromatic aberration (hard to believe, but true).

Sharpness and contrast:

Three zones (A, B and C) were selected.

  SAMYANG 14mm @ f/2.8 Nikkor 15 mm @ f/3.5 SAMYANG 14mm @ f/8.0 Nikkor 15 mm @ f/8.0 SAMYANG 14mm @ f/22 Nikkor 15 mm @ f/22

B

A

C

There are three decades that separate the two lenses. The more recent Korean lens design incorporates two aspherical components. Impressive progress of the technology for a very affordable lens..

 

Conclusion

After 3 months of use, I have found that this lens is excellent. The contrast and sharpness are sensational especially at the center of the image. I completely understand and share the enthusiasm of the testers from LensTip.com!

The high and complicated distortion is real. It can be corrected but then the angle of view shall be reduced after subsequent necessary cropping in general photography. We would not recommend this lens for architectural photography and especially not for interior architecture where the subject with straight edges might be quite close to the camera.

This lens is excellent for stitching panorama photographs though. But the panorama output resolution advantage is rather meager when the number of images to make a complete 360 x 180 panorama is taken into account. A 15 -16 -17 mm fisheye lens is IMO better suited for "exclusive" panorama photographing. No less than 12 images on a FF camera are needed to close a spherical panorama. Adding one Nadir and one Zenith to make 14 images in total is probably better for outputting a 13600 x 6800 pix. Update: I have fortuitously discovered that with adequate care, 6H +1Z + 1(or 2) N is enough to complete the 360 x180° panorama. The overlapping area is really scarce in some place but that's feasible!

When one wishes to use a single versatile lens for both standard photography with very wide lens (e.g. for landscape and reportage photos) in addition to high resolution panorama photography, then the new Samyang is one of the best possible buy (the Canon EF 14 mm f/2.8L USM II for example is priced about x10 the Samyang). A direct comparison with the much praised Nikkor 14-24 mm f2.8 @14 mm is planned to verify if the Samyang is indeed superior (cf LensiTip.com assertion)r.

A new page was written about the shift of the LPP along the axis of lenses (in general) following the above experiments and results.

Michel Thoby

September 20, 2010

Rev 11 march, 2011