50mm Primes that don't suck wide open?

[Steb]

I suspect a large percentage of the AF QC complaints are actually related to the challenge of shooting wide open and nailing/keeping the plane of focus where you want. With a Canon L fast prime the gut reaction is "what am I doing wrong", but with a 3rd party fast prime it's "what did the manufacturer do wrong."

I'm sure that is not the only reason (but it might contribute to the Sigma's reputation). I own the Sigma myself, bought it with huge focus shift issues, sent it in for calibration, got it back in much better condition. But still the AF is not that reliable. Sometimes it just fucks up, sometimes it is slightly off in one direction, sometimes in the other.

I shot a Canon 1.4 for one afternoon and had a completely different experience. It was always spot on and I could trust the AF much more. Without any AFMA adjustments.

The Sigma is very useable after the calibration, but there is still a difference to the Canon. That said, I don't like the build quality of the Canon 1.4 and the Sigma has great IQ.

 

[canikon]

I think you missed the light from outside the f/1.6 cone arrives at the sensor from a steeper angle. Think of each pixel as a short straw with it's active part at the bottom of the straw and all should be clear - in effect the edge of the pixel vignettes each photodiode. I'd expect the cut off to be gradual as the photodiode is shaded.

After reading a recent Panasonic release (http://goo.gl/279uS) I learned that it is typical for current sensors to accept light that impinges between 15 and 20 degrees from the vertical. This corresponds to a light cone of f/1.86 and f/1.37, in general agreement with my measurements.

When you do your test, remember to half-release the lens from the camera so that the camera doesn't know there's an attached lens. Shoot a manual series (raw, full aperture, trade exposure against ISO.) You can then repeat the set with the lens fully mounted and compare images.

The result will then be fairly obvious. If you have software that allows you to compare pixel levels, so much the better.

Thank you very much for further clarifications.
And thank you very much for the sggestion about disconnecting the electronic contacts between lens and camera. From what you say I beleve that you imply that EVEN IN MANUAL MODE with FIXED ISO the camera is cheating with the iso value, and this fact makes me somewhat angry.
Back to technicalities, I am still confused about the above quoted sentences, as I better explain hereby.
You say that "the light from outside the f/1.6 cone arrives at the sensor from a steeper angle".
I have not read the link you provided, I'll do it later, but at moment something comes to my mind.
Given a certain sensor size, the lens is made such as to project a cone of light suitable to cover that particular format. So for EF lenses the cone should cover a bit ove rthe diagonal of a 24x36 frame.
This fact is COMMON to all EF optics, since all of them have to cover tha same sensor format.
So therefore the MAXIMUM angle of incident light cone on the photodiodes is somewhat fixed, and dependent by the values of sensor size and flange to sensos distance.
The only angle that would change with aperture is the angle that say the rightmost ray of light coming from the exit pupil would see when hitting the leftmost pixel on the sensor, and this depends obviously on the exit pupil diameter rather than on the relative aperture.
In this regard as example, the 50mm 1.4 has a fairly "narrow" exit pupil lens, so I don't see any problem at all, while the 85 1.2 has the biggest exit pupil of any lens (they had to fix the contacts on the flass itself!) so therefrore maybe in that case any problem could arise.
This facts leads to the point that I stated above, that the peripheral portion of the sensor is somewhat not sensitive to extremely fast lenses.
I could compute the incident angle for a 50 1.4 and 85 1.2 doing some simple math on the exit pupil diameter, flange to sensor distance, and full frame diameter.
Therefore I don't see what you say, that the maximun angle would depend solely on lens aperture.
Am I thinking the wrong way?

 

[RLPhoto]

F/1.2 - 1.4 - 1.8 are compromises. I wouldn't classify them as suck, I only hold the EF 50mm F/1.0 in that category.

I really wouldn't say the 50/1.0 sucks, but it is a highly specialized use lens. If you're not sure you need the 1.0, you're definitely better off with the 1.2. If you're positive you need the 50/1.0, there's an 80% chance you're wrong, and you'd still be better off the the 1.2.

But... if you need to shoot a candle lit dinner in a coal mine, and don't mind (or like/want) the "dreamy" look the 50/1.0 gives, there really is no substitute.

If you need a 50mm F/1, get the leica... Or at least rent one.

If you can afford Leica 50/1, then you can afford all the fast 50s in question here (as a backup or just for fun) .

Really, I believe don't so but point here is the Leica 50mm doesn't suck.

http://www.ebay.com/itm/LEICA-NOCTILUX-M-50mm-f-1-11821-E60-/121136252503?pt=UK_Lenses_Filters_Lenses&hash=item1c34488657

http://www.ebay.com/itm/Canon-EF-50mm-f-1-0L-f-1-0-L-f-1LLens-Made-in-Japan-Nr-MINT-/181148016924?pt=Camera_Lenses&hash=item2a2d43611c

Leica M autofocus sucks and Canon 50/1.0L is just as good optically, if not better.

Lol you said the canon 50mm 1.0 was good optically. That's cute.

"Good" and "as good" have different meanings, you know ... , while "as good" and "as bad" are practically the same thing .

Lol because you said the EF 50mm f/1 and good in the same sentence. If you've ever rented one, you'd understand.

 

[mrsfotografie]

I don't think so, that would make pictures taken with fast Canon EF lenses more noisy. It is a fact though that (depending on your camera model) the AF sensor knows it's a fast lens and this the f/2.8 cross points are activated giving you increased AF performance. This has nothing to do with the imaging sensor.

Um... AF performance has absolutely nothing to do with bokeh or detected light. Perhaps you should try to understand the post and then do the suggested experiment?

Remember, the boost is at most 1/2 to 1 stop. As it happens, the image noise does increase - albeit slightly. It takes considerable care to measure changes in the noise floor, particularly below 1600 - 3200 ISO.

Perhaps this 'iso boost' can be explained by in-camera peripheral illumination correction. Did you do the test with this particular feature switched off?

 

[ecka]

F/1.2 - 1.4 - 1.8 are compromises. I wouldn't classify them as suck, I only hold the EF 50mm F/1.0 in that category.

I really wouldn't say the 50/1.0 sucks, but it is a highly specialized use lens. If you're not sure you need the 1.0, you're definitely better off with the 1.2. If you're positive you need the 50/1.0, there's an 80% chance you're wrong, and you'd still be better off the the 1.2.

But... if you need to shoot a candle lit dinner in a coal mine, and don't mind (or like/want) the "dreamy" look the 50/1.0 gives, there really is no substitute.

If you need a 50mm F/1, get the leica... Or at least rent one.

If you can afford Leica 50/1, then you can afford all the fast 50s in question here (as a backup or just for fun) .

Really, I believe don't so but point here is the Leica 50mm doesn't suck.

http://www.ebay.com/itm/LEICA-NOCTILUX-M-50mm-f-1-11821-E60-/121136252503?pt=UK_Lenses_Filters_Lenses&hash=item1c34488657

http://www.ebay.com/itm/Canon-EF-50mm-f-1-0L-f-1-0-L-f-1LLens-Made-in-Japan-Nr-MINT-/181148016924?pt=Camera_Lenses&hash=item2a2d43611c

Leica M autofocus sucks and Canon 50/1.0L is just as good optically, if not better.

Lol you said the canon 50mm 1.0 was good optically. That's cute.

"Good" and "as good" have different meanings, you know ... , while "as good" and "as bad" are practically the same thing .

Lol because you said the EF 50mm f/1 and good in the same sentence. If you've ever rented one, you'd understand.

Nah ... I'm good without it ;D. I just want Canon to get their $#@% together and make a nice EF 50/2 USM Macro.

 

[Steb]

The Sigma 50 1.4 is very good on the Canon 5D3 body. Something about the 5D3 focus system that just makes the Sigma work perfectly. I say this because the same Sigma lens on my old 7D had the usual focus shift inconsistency.

Naaa, you are just lucky... ;D

 

[mrsfotografie]

The Sigma 50 1.4 is very good on the Canon 5D3 body. Something about the 5D3 focus system that just makes the Sigma work perfectly. I say this because the same Sigma lens on my old 7D had the usual focus shift inconsistency.

Naaa, you are just lucky... ;D

Same luck here, it focuses perfectly on my 5D2, even with the MkII's lame AF system

 

[dtaylor]

Since the sensor doesn't actually detect light from a fast cone, it can't contribute to the bokeh

I'm not sure where you read this, but it's demonstrably false. I'm looking at two 7D + Sigma 50mm shots on my screen right now, f/1.4 and f/2, same subject/distance/etc., and the difference in bokeh is quite clear. There are various sites online which demonstrate fast lens bokeh using both FF and crop, and you can confirm the difference. This guy directly addresses the point:
http://photo.stackexchange.com/questions/4459/do-dslrs-play-games-with-iso-when-used-with-fast-lenses

The idea that DSLRs can't see light cones beyond a certain point became popular with an open letter at Luminous Landscape:
http://www.luminous-landscape.com/essays/an_open_letter_to_the_major_camera_manufacturers.shtml

I remember reading the letter and doing the test where you defocus and shoot at f/1.4, f/2, and then slightly rotate the lens and compare the shots. Yes, the DSLR is playing with ISO at f/1.4. But it's also quite clear that f/1.4 has more exposure then f/2, i.e. the detached lens shot is between the f/2 and f/1.4 attached shots.

I'm not convinced that this is primarily the result of digital sensors and the angle of the light seen on SLRs. Why?

* The same claim was made in relation to vignetting long before the LL article, but 35mm film shot in a SLR vignettes just as much. Most people just never make the comparison to find out.

* Logically the claim leads to the conclusion you made about bokeh, yet bokeh is observably quite different. The camera can't fake that.

What no one has done, to my knowledge, is very, very carefully shoot and develop 35mm film and check the density to determine the t-stop of a lens at f/4, f/2.8, and f/2, and then compare with f/1.4 to see if there's a shift. That's because this would require equipment most do not have along with meticulous work. It's easy to do these kind of tests digitally, and not so easy to do them and do them right with film. (Looking for vignetting is easy, but precisely determining the exposure received by the film to compute t-stop and compare with digital is not.)

It would not surprise me in the least to discover a t-stop shift at fast apertures due to some factor inherent to lens design and the registration distance in a SLR. We treat f-stop as if the resulting change in exposure is absolute, but it's not. T-stop varies between lenses and with subject distance using the same lens. It wouldn't surprise me to learn that it shifts at fast apertures as well.

Nor does it surprise me that camera companies would compensate for this in firmware.

 

[sdsr]

Only you know whether you need or want a 50mm prime. Since you already have one, see how you like that focal length on your 5DIII when you get it. Also, spend a week or two using only your 24-105 L and see where most of your shots are taken. You may prefer 35mm or wider, or 70mm or longer, or may have, say, three most-often-used lengths - if so, and assuming a prime will offer you something the zoom doesn't, get whatever the closest primes are.

I don't have a lot of experience with Canon 50mms. My 50mm 1.8 was sharp enough wide open to create sharp enough photos in very low light, but I didn't like its bokeh and sold it. Bought a 50mm 1.4, but it often front- or back-focused, or just plain missed altogether (as in nothing at all was properly in focus anywhere in the image); I returned it for a replacement, which is excellent - no focusing problems at all and it's certainly useable wide open (superior bokeh to the 1.8 too). After being impressed by the Sigma 35mm I tried their 50mm 1.4 in the hope that I would land a good copy, but mine seemed to combine all the flaws I had read about - not one photo I took while testing it out was in any way acceptable, so I returned it for a refund in the hope that Sigma tries again before too long (or Canon for that matter - I wouldn't object too much if they added one to their line of IS primes). In the world of M43 the 50mm equivalent, the Panasonic/Leica 25mm 1.4, is impressive even wide open, but of course that doesn't help here....

 

[GMCPhotographics]

In preparation for my upgrade to a 5d3 (from a 60D) later this year I'm trying to compile a list of potential prime lenses to get between the 35 and 100mm range. I've already got a 24-105L and will sell my non-L 70-300 to get the L version. Also have a 50mm 1.8 so I'll need to get a 85 1.8 to get my portrait capability back.

Then the question remains: Do I keep the little plastic toy (which never use wider than 2.8 anyways) or just get a pancake (which is actually pretty usable wide open, vignetting aside)? Do I even need a 50? I know the 50mm 1.4 isn't great wide open either and has fragility issues but at least it'll work with the automatic CA corrections in the 5d3 (I shoot RAW+jpeg). The sigma 50 1.4 is...really big for a 50mm prime, has AF quality control issues and won't work with auto CA corrections. Rumor mill has it that the sigma 50 is due for a rebody to the "art" line they've got going now but probably won't get an optics refresh... Oh yeah, the 50mm 1.2 is a no thanks I'll keep the money. Everything 50mm seems to be a compromise and you'd think after so many decades of people using such a prolific "normal" lens the designers would have perfected the formulas by now.

Do I even need a 50?

Only you can answer your last statement.
To be fair a 24-70 f2.8 L does so much and so well, a lot of guys don't need the primes in a post high iso world. it can do all the depth of field isolation at the 70mm end or by getting closer and shooting wider. It's MFD is amazing and it's bokeh is first rate.
I use a 35 f1.4 and 85 f1.2 a lot. But prefer the f2.8 zoom trinity if I can, light permitting. On sunny days, the primes are too bright and my 5DIII bounces off it's 1/8000th max shutter speed....lol!

There isn't a great 50mm, sorry but that' the truth. The best is certainly the 50 f1.2L but it's not that great. It's ok and it does the job best of all. But it's lacking compared to the 35L and 85L. We are all hoping for a 50L which is worthy of the moniker...but that's probably just a fools hope. I think the 50L was a big wake up call for Canon and every lens since has been stellar.

 

[noisejammer]

Since the sensor doesn't actually detect light from a fast cone, it can't contribute to the bokeh

I'm not sure where you read this, but it's demonstrably false. I'm looking at two 7D + Sigma 50mm shots on my screen right now, f/1.4 and f/2, same subject/distance/etc., and the difference in bokeh is quite clear.

Apologies for trimming your post.

First, I'm well aware of the LL open letter and have referred people to it previously - here, on POTN and elsewhere.

I observed four effects -
1. Few people bothered to go to read the reference and digest its content.

2. It refers to a DXO study and there's a rabid hatred of DXO measurements, particularly since they tend to be quite scathing over Canon offerings.

3. Few people who read the letter did any testing but this did not prevent them from commenting that it was a lie - perhaps a result of point 2.

4. There was a lot of nastiness - particularly from owners of 50L and 85L lenses - perhaps because they had a lot of dollars sunk in the lenses.

Because of this, I've changed my approach. I invite people to perform the experiments and simply judge for themselves. It cuts through the internet noise quite effectively.

Bokeh
Ok, so let's assume you've now performed the experiment using a 7D.

If you did it properly, you will have noted that the image is dimmer when the camera could not communicate with the lens. I venture that the same number of photons hit each pixel (within photon noise and your ability to align the tests) so presumably something else gets changed when the camera communicates with the sensor.

There is simply no way that a photon that is not detected can plausibly contribute to the bokeh. (If you doubt this, please suggest a mechanism.) Therefore, if there's a difference between in the image quality between (say f/1.4 and f/2), it can't be light that's doing it.

I strongly suspect that there is some image processing going on inside the camera to deliberately soften the image from lenses that have large apertures. It may be possible to demonstrate this idea by installing an aperture mask on a wide open lens and seeing whether the image quality changes (say between f/1.4 and f/1.8.)

Finally - another reference that may be of interest http://goo.gl/5We9r . Note how the 7D loses 0.65 EV when the sensor is illuminated by a f/1.4 lens. This strongly suggests that the pixels do not respond to light that's coming from a cone which is faster than f/2. Similarly, a 5D2 and 1D4 will lose 0.44 EV at f/1.4.

regards etc.

 

[dtaylor]

Ok, so let's assume you've now performed the experiment using a 7D.

You don't have to assume. I explicitly stated this.

If you did it properly, you will have noted that the image is dimmer when the camera could not communicate with the lens.

I noted this. (Honest question: did you read my entire post?) But I also noted that it was brighter then f/2. At least with my Sigma 50mm f/1.4 the detached f/1.4 shot looks about 1/3 of a stop darker then the attached f/1.4 shot, which means it is brighter then the f/2 shot.

There is simply no way that a photon that is not detected can plausibly contribute to the bokeh.

Agreed. And since bokeh changes, the photons are being detected. Therefore your theory is incorrect.

Please review the link I provided: http://www.canonrumors.com/forum/index.php?topic=15824.45;topicseen
Please also see: http://www.wlcastleman.com/equip/reviews/50mm/bokeh/bokeh.htm
And: http://www.wlcastleman.com/equip/reviews/aps-c_port/bokeh/85mm/index.htm

Note that in every case the bokeh...specifically the defocus of detail outside the plane of focus...changes as one would expect all the way to f/1.0.

Therefore, if there's a difference between in the image quality between (say f/1.4 and f/2), it can't be light that's doing it.

We're not observing a difference in image quality, but a difference in defocusing which can only happen if the sensor is seeing the entire light cone.

I strongly suspect that there is some image processing going on inside the camera to deliberately soften the image from lenses that have large apertures.

Neither my observations nor the observations linked above are related to softness. Defocus ability is not softness and cannot be simulated by the camera firmware.

Finally - another reference that may be of interest http://goo.gl/5We9r .

That's a dead link.

Note how the 7D loses 0.65 EV when the sensor is illuminated by a f/1.4 lens. This strongly suggests that the pixels do not respond to light that's coming from a cone which is faster than f/2. Similarly, a 5D2 and 1D4 will lose 0.44 EV at f/1.4.

Bokeh / defocus change as expected with the change in aperture. This means the pixels are responding to light from a cone which is faster then f/2.

Either the cause of the shift in t-stop is something unrelated to digital, or the sensors see the full light cone but with some intensity loss at certain angles. Or some combination thereof. But the idea that the sensor cannot see any portion of the light cone beyond f/2 is demonstrably false.

 

[skitron]

The Sigma 50 1.4 is very good on the Canon 5D3 body. Something about the 5D3 focus system that just makes the Sigma work perfectly. I say this because the same Sigma lens on my old 7D had the usual focus shift inconsistency.

Naaa, you are just lucky... ;D

Same luck here, it focuses perfectly on my 5D2, even with the MkII's lame AF system

Ditto, mine works perfectly. Had to send it to Sigma, but it came back spot-on at any distance.

For the price it's very good. I do like the asthetic of the Canon 1.2 better, but then that difference is +$1100...

 

[AJ]

Here's the thing about the 40/2.8, in my humble opinion.

This lens could easily have been speced at f/2. However it isn't. Canon made the diaphragm to open up only to f/2.8 so that it's sharp "wide open". This satisfies the testchart-shooters, and ensures the lens gets good reviews.

Personally I like a lens that opens a little wider, even if that results in a little softness. I like my 50/1.8 and I accept the fact that it's not razor sharp at f/1.8. However it is cracking sharp at f/2.8 so that's what I stop down to when I need more sharpness. It's all about knowing the characteristics of the lens and working with it.

Here's the other bit about fast lenses: it's all about what's not in focus, i.e. bokeh. Sharpness in the center still matters, somewhat, but sharpness in the corners wide-open is completely moot for most purposes (except perhaps astrophotography).

 

[Pixelsign]

there is only one really sharp 50:
Carl Zeiss Distagon T* 1,4/55

 

[Sporgon]

Here's the thing about the 40/2.8, in my humble opinion.

This lens could easily have been speced at f/2. However it isn't. Canon made the diaphragm to open up only to f/2.8 so that it's sharp "wide open".

Not so. I think you'll find the Canon 40mm design is identical to the 40mm Pentax M f2.8 of 1975, produced as a budget 'kit' lens for the diminutive ME and MX, although the term 'kit lens' hadn't been coined then.

The Voightlander 40 is different.

 

[mrsfotografie]

Here's the thing about the 40/2.8, in my humble opinion.

This lens could easily have been speced at f/2. However it isn't. Canon made the diaphragm to open up only to f/2.8 so that it's sharp "wide open".

Not so. I think you'll find the Canon 40mm design is identical to the 40mm Pentax M f2.8 of 1975, produced as a budget 'kit' lens for the diminutive ME and MX, although the term 'kit lens' hadn't been coined then.

The Voightlander 40 is different.

The Voightlander 40 is an interesting lens, I've got one. I only use it on special occasions though, when I like to go MF and portable. It's a great little lens to help boost creativity because I find 40 mm (on full frame) a challenging focal length. Reasoning behind that is that 40 mm is as 'normal' as you can get so it takes extra care to make interesting photo's.

 

[paul13walnut5]

If you are able, check out this weeks Amateur Photographer magazine (UK) big group test review of 50mm lenses from all the main manufacturers and third parties such as Sigma and Zeiss. The results are quite surprising.
They do a months free trial for the ipad version if thats any use...

 

[RGomezPhotos]

I tried both the Canon 1.2 and Zeiss 1.4. I bought the Zeiss because of it's contrast characteristics. I also like the bokeh compared to the Canon. I think it's also sharper. I remember the Canon being a good deal soft at big apertures. But the Zeiss is MF only. Which is fine for me.

The only 50mm that's going to trump both of these is a Leica with a Canon adapter. And that lens will cost more than both of these combined. But the Zeiss 55mm that's coming out soon will be even pricier than that!