samyang 14 vs. canon 10-22 vs. ts-e 24 ii
On Wednesday I took delivery of a new Bower 14mm f/2.8 lens.
You wouldn't know it from the B&H listing (which warns it's an APS-C-only lens) or even the box itself (which labels it a fisheye), but this is a rebranded Samyang 14mm f/2.8 full-frame rectilinear lens.
Samyang is a relatively new Korean lens manufacturer with a growing line of good, fast, affordable, full-frame primes. The catch is they are fully manual—they have no electronics or motors. That means not only is focus manual, but so is stopping down the aperture.
But, since this is an ultrawide lens, that's not as big a deal as you might think. My Canon EF-S 10-22mm f/3.5-4.5 has autofocus and auto-aperture and yet I still use it mainly in full manual mode on a tripod—mainly because it is difficult to level and center an ultrawide composition handheld. Those who buy a $1700 Zeiss 21mm f/2.8 or $1700 Nikkor 14-24mm f/2.8 + $200 G-to-EOS adapter also have only manual focus and aperture.
Yeah, but it's worth it because the Zeiss and Nikkor are so sharp, you say? Well, let's see what ~$400 gets you.
Top: Canon EF-S 10-22mm f/3.5-4.5 @ 10mm on a Canon Rebel XSi
(16mm-equivalent FOV)
Middle: four-frame shift-stitch of a Canon TS-E 24mm f/3.5L II on a Canon 5D Mark II
(15-and-change-mm-equivalent FOV)
Bottom: Samyang 14mm f/2.8 on a Canon 5D Mark II
(14mm FOV)
Test notes
- All frames taken at f/11, 1/180s, ISO 100.
- Tripod not moved between frames, not even when shifting.
- All frames manually focused in Live View @ 10x magnification on the base of the central tower. See Observations section about field curvature.
- All frames RAW-processed in DPP with Standard Picture Style, Daylight WB, zero noise reduction, zero sharpening, no lens aberration corrections.
- 10-22 frame brightened in DPP 2/3rds of a stop to match others. (It was a clear day and the shots were taken minutes apart so I guess this is either light loss in the lens or a reported ISO/sensitivity difference between the XSi and 5DII.)
- TS-E and Samyang frames resized with Photoshop Bicubic to match vertical pixel count of EF-S frame.
- Four-frame stitch with shift axes 30 degrees off horizontal yields a 1.59:1 aspect ratio; hence that frame is a little wider than the others.
Observations
- All three frames are remarkably similar at 12 megapixels (almost 10"x15" @ 300dpi).
- The Samyang is slightly yellower than the other two.
- At first I thought the EF-S was cooler than the TS-E, but the more I look at it the more I
think the L lens is just more saturated.
- The TS-E shows slightly more vignetting in the corners than the other two (keep in mind this is at full shift).
- The Samyang has a hefty amount of moustache distortion. This can be almost completely removed by selecting the Samyang 14mm f/2.8 profile in PTLens.
- As advertised, the Samyang has almost zero CA. Possibly even less than the vaunted TS-E with both movements centered.
- This particular lens appears to be slightly decentered, with the lower right corner of this frame slightly blurry. Other frames (at different focus ring orientations) have the left side blurry or nothing blurry at all (since it only affects the outer 10-15% of the frame, if the blurry region is oriented to the top or bottom it will be out of frame).
Stuff not shown in pictures
- Sizable vignetting wide open. I'm talking 50 f/1.4 and 24-105 f/4 levels of vignetting.
- Focal length visibly changes with focus distance. It is an internal-focus lens, after all.
- It has a markedly different field curvature than the EF-S. This can be seen by moving focus from the near grass to the middle tower foundation. The EF-S requires almost no change - i.e. both are in focus wide open. With the Samyang, only one or the other is in focus all the way up to f/22.
- The TS-E displays a similar focus behavior on the same subject. I'll go out on a limb and guess that both the Samyang and TS-E have a near-flat field, while the EF-S has strong field curvature, with the center of the frame focused much farther away than the corners.
- If the Live View preview is any indicator, evaluative metering is way, way off. Going by the histogram yields a frame 3-2/3 stops overexposed. [Edit 9/8/2010: Live View metering way off in Manual exposure mode, but metering accurate in Av with and without Live View.)]
- It is already sharp wide open—almost all the way to the corners.
- Bokeh is decent by ultrawide standards. It's still got nisen ("bright ring") bokeh, but it seems smoother than the Tokina 11-16mm f/2.8's bokeh.
- Wide-open aperture markings seem optimistic. Switching from f/3.3 (half a stop smaller than f/2.8) to f/2.8 lets in less than a quarter stop more light in the center (corner exposure is completely unchanged due to vignetting). Going from f/4 to f/3.3 (again, nominally half a stop) lets in a third stop more light in the center (and about 3/100ths of a stop in the corners). Overall, going from f/4 to f/2.8 only lets in a little over half a stop more light in the center. So in terms of light gathering, this lens is really about an f/3.3. And that's only in the center.
Closing thoughts
So these are just preliminary findings from a couple hours shooting, but here's how I see the Samyang 14mm f/2.8 so far:
Pros
- good build quality (closer to the TS-E than the EF-S)
- easily out-resolves the 5DII (I use moiré to tell what's in focus)
- sharp wide open
- sharp almost all the way to the corners (although see below about decentering)
- next to no CA
- relatively flat field of focus
- smaller and lighter than other ultrawide, full-frame lenses
- cheaper than all competitors; less than a fourth the cost of similarly-sharp ultrawides
Cons
- manual focus
- manual aperture
- no EXIF (the camera doesn't know there's a lens attached)
- heavy vignetting at larger apertures
- heavy moustache distortion
- focal length visibly changes with focus distance
- yellowish color balance
- metering way off
[Edit 9/8/2010: Live View metering way off in Manual exposure mode, but metering accurate in Av with and without Live View.)]
- no front filters
- cheap, slightly-awkward lens cap (although I've seen worse)
- perhaps not the greatest anti-reflection coatings
- nisen bokeh, but better than the Tokina 11-16 f/2.8
- more like f/3.3 in practice
- the one I got was decentered
Useful for
- Landscape (may need PTLens)
- Architecture (needs PTLens)
- Astrophotography (Milky Way, Aurora Borealis)
In short, if you get a good copy and factor in an extra $25 for PTLens this Samyang is an unbelievable value. Any excuses about how it's inherently expensive to build a sharp ultra-wide—let alone one that's sharp wide open—have just been Heismanned. Now I just have to find a properly-centered one.