[Scythels]

 

Scythels, you're right, especially with the fact that the worst Sigmas are better than the best Canons

Not really -

 

Sigma #1 is better than all the canons, but by a small margin.  It is fairly representative of the average 35A. 

 

Sigma #2 has enormous astigmatism spikes (almost 80% in some regions of the frame) and would look very soft, even if it maintains good sagittal resolution.

 

Sigma #3 is poorly assembled and heavily tilted.

 

Sigma #4 has decentering as bad as the first 23/1.4 Klaus got for testing - left side is under 1/4 as good as the right.

 

On the other hand, buyers might get something like sigma #5, which is nearly diffraction limited at the tested aperture (f/1.4) right out to the corners. 

 

Here's the resolution gain with several zeros... I still don't know the best way to represent this.

 

http://i.imgur.com/VQ3CXzm.png

 

All of these tests are without the influence of a normal camera.  The test device works with collimated light and the result is measured with a diffraction limited microscope lens at f/1 (that is - the contribution of the measurement device is miniscule)

 

----

 

 

But less annoying than the AF variation of the camera/lens system itself

 

PDAF relies on the lens and body being aligned with each other.  Essentially, if the rear nodal point of the lens isn't just where it is expected to be (and the difference between perfect and very off is hardly a millimeter) the geometry "fails" and light won't come to a focus on the beam splitter in the PDAF system at the correct depth - resulting in front or backfocus tendancies.  It is a limit to the quality of manufacturing of the manufacture.

 

If your body's mount is 1mm too far forward, all of your lenses will front focus unless they are very far off/too shallow on the mount.  Unless your body is right where it should be, it's a crapshoot for if your lenses will line up with your body correctly (and because anecdotal stories apparently are very valuable - my 6D, 35 IS, and 70-300L are all perfectly aligned with each other out of the box).  Whether your sigma or your nikon or your canon or your pentax or your sony A is properly aligned matters very little.  Micro adjust it and move on.

 

If focus errors are different at different distances, it's a failure of the lens/lens motor.  The camera has a LUT which contains information on the drive speed of the AF motors of lenses (I could get into more detail, but it's outside the scope of this thread) - first party manufactures obviously don't include third party mfg data.  The camera expects the lens to focus at a certain rate, if the lens doesn't it's a failure of the lens to meet the camera's expectation and focus will miss.  If focus is randomly off in the way that the tested 50A at TDP is/was - the focus system wasn't designed with finely enough toothed gears and where the lens 'settles' lacks "proper" precision.

 

Canikon will charge a higher premium than Sigma for two reasons:

 

1. High profit margins are profitable
   
2. Buyer psychology - if canikon make the $2,000 70-200 telephotos, consumers equate that with quality- they need to maintain a reputation as the quality brand. (See also: white lenses at sporting events being built in marketing material for canon)
   

They also have tighter tolerances, at least canon's lenses are better engineered from a physical standpoint, and as well engineered optically.  The 35L isn't all that much worse than the 35A optically, and it's 14 years older.  The 24-105L trades evenly with the 24-105A (quality corners where the 24-105A lacks them), and is 7 or 8 years older.  While being smaller and lighter.  Planar lenses have much less engineering freedom (to change one element requires changing its partner behind the diaphragm) and also much less effort.  Canon can do a 50mm distagon as well as sigma.

 

Don't get me wrong, sigma is doing great things at great prices, but they are not eating canikon's cake en totale.