Nifty fifty STM - focus issues, or human error?

I have a question for users experienced with fast primes. Today, I tried to test AF with my 700D (T5i) and my two lenses:

- EF-S 18-55mm 3.5-5.6 IS STM
- EF 50mm 1.8 STM

I noticed an issue with the 50mm. When I open up the aperture to the max, the AF jumps a bit on every other photo.

But here's the catch - when I put the camera to Live View, and focus using contrast-detect only, the focus is nailed every single time. But the moment I revert to phase-detect, the focus jumps every few photos. This was never an issue with my 18-55 lens.

After shooting wide-open, I matched the parameters of 50mm to 18-55 ones - that is, I selected 5.6 aperture (because 5.6 is max on 18-55 at the telephoto end). Phase-detect focusing improved significantly.

My question - is 50mm supposed to jump focus when wide-open? Why it doesn't jump when using contrast-detect?

Note that my camera doesn't have micro-AF adjustment, and doesn't have dual-pixel AF. It uses an old-school 9-point, all cross-type sensor (probably from 60D). I've noticed that phase-detect AF takes more time if set to apertures larger than 3.5.

Testing parameters:

- 700D and 50mm 1.8 STM, subject = a row of books
- global settings = 2 sec delay, wide-open, center AF point (dual cross-type), tripod
- first test with phase-detect, 4 shots, viewfinder = two shots good, two bad
- second test with contrast-detect, 4 shots, Live View, electronic 1st curtain = all good

When using phase-detect, it goes like this (note that camera and subject never moved):

- half press > focused > full press > countdown > shot
- half press > focus jumps > full press > countdown > shot

So is it okay, or should I replace my lens? Is it lens, or camera issue? Few days ago I took a photo of my brother, but it was out of focus (lens was wide-open). I tried to be as still as possible, yet the photo wasn't sharp.

But if we cannot use the widest aperture, what's the point of having it on the lens? Do I really have to resort to contrast-detect Live View to get sharp photos on 1.8? Or is it just how fast primes work? Yes, I know that DoF is very thin, but the camera was on tripod. Most people praise this lens, and I sincerely hope it won't be a big issue during real-life use.

It is not s front back micro adjust issue, however to see what is going on you can diagnose and understand in the same way you remedy a front/back focus issue. I explain...:

-take a tape measure or ruler at a 90 degree angle to use as a subject and put camera on tripod or somewhere stable.

-set aperature to 1.8 focse at say 8 inch marker from whatever distance you chose (something like 5 feet should be okay)

-take a shot, un focus and re-focus take another...repate a couple of times unfocusing and refocusing between shots. a good way to do this is by putting your hand in front of the lens to throw focus off before refocusing.

-Check the tick markers on your measuring tape, you will see exactly where to focus landed and the depth of field

-Repeat at F 2.8, F5.6, F8

What you will notice is

1) if you need to micro adjust (not that important in your case)

2) You will notice that the focus is consistent when the aperature is above 2-2.2Why is that ?

Well, the lens is simply not sharp at 1.8 nor F2.0 really and truly. It is pretty good but not sharp to the degree that the focus system is fully settled. at around F2.8 you should see more focus consistency as the lens itself is more sharp and able to decide on a point to focus on better than F1.8. It is still not perfect as there is some play in the teeth of the focus mechanism but it is much better at those higher apertures.

I have had about 3 or 4 of the nifty fifties they pretty much all exhibited what I mention above. however, I have only tested on a full frame, I am not sure if on an APSC it gets "shape" and has consistency at 1.8 or at 2.8...have a look I am interested to hear your results if you decide to do the geeky test outlined above.

I should also mention that I love this lens, really one of my favorites. It is phenomenal for shooting Panos at F8.

It looks normal to me.  You can expect some variance, especially since focusing is done at f/2.8 (or possibly f/5.6), but the photo was shot at f/1.8.

That's just one of the facts of life.  I also had trouble with this with film cameras.  For perfect viewfinder focusing with fast lenses, you probably need a more expensive camera, or a mirror less camera with EVF.

Most of these type of reports are due to user errors

Test everything from a tripod to minimize this factor.

However this sort of focus imperfections are expected from any fast lens particularly the cheapest

-you dont have to go out to 5.6, should be consistent just above F2

-for test being a little close can help with bias, note that focus should be 1/3 in front and 2/3 behind.

-FYI focos and recompose at 1.8, 1.2, 1.4 is going to give you poor consistency it is really hard to nail shots when moving and trying to keep the focal plane fixed especially with a depth of field of centimeters (depending of course on how close you are to your subject). Getting that eye pin sharp EVERY time is just not going to happen under 1.8.

-looks like the issue however is sorted, glad if what I suggested helped.

TMHKR wrote:

archiekar wrote:

-you dont have to go out to 5.6, should be consistent just above F2

-for test being a little close can help with bias, note that focus should be 1/3 in front and 2/3 behind.

-FYI focos and recompose at 1.8, 1.2, 1.4 is going to give you poor consistency it is really hard to nail shots when moving and trying to keep the focal plane fixed especially with a depth of field of centimeters (depending of course on how close you are to your subject). Getting that eye pin sharp EVERY time is just not going to happen under 1.8.

-looks like the issue however is sorted, glad if what I suggested helped.

Well, I'm not sure it's consistent before at least F3.5. Focus jumps are still noticeable, but not as noticeable as with F1.8 and F2.0. The more I stop down the lens, the jumps are less noticeable (there are virtually no jumps by F4.0).

If I keep half-pressing the shutter button several times (while holding focus point on target), the lens jumps a tiny bit at each half-press. It attempts to refocus, but instead it just jumps a millimeter here, a millimeter there (half-press = focused, half-press = unfocused, and so on). I believe that's just how it works.

Yes it is in part how it works, the focus ring has play as do the teeth that control it so there is a lack of precision. The jump is quite fast and I believe it uses an inerpolation algo of some sort so it is not just a move in one direction but rather just past and back.

I have a copy of the 50 F1.8 ii (not STM) that becomes quite good at F2.2 and consistent shooting a 5D mkii if that makes a difference.

We are really nit picking here, what we get out of this lens is truly great, and that statement does not need the " for the money" thing before it...even though it is a bargain at its price around $100. After understanding its limitations you can still use it to effectively capture many types of scenes SO enjoy it, I do, hope you do now as well!

archiekar wrote:

TMHKR wrote:

If I keep half-pressing the shutter button several times (while holding focus point on target), the lens jumps a tiny bit at each half-press. It attempts to refocus, but instead it just jumps a millimeter here, a millimeter there (half-press = focused, half-press = unfocused, and so on). I believe that's just how it works.

Yes it is in part how it works, the focus ring has play as do the teeth that control it so there is a lack of precision. The jump is quite fast and I believe it uses an inerpolation algo of some sort so it is not just a move in one direction but rather just past and back.

Well, no.  Manual focusing does indeed have some play on at least some STM lenses, but TMHKR's comment has nothing to do with the focus ring.

I've just been testing my new 50 STM and am struggling  with the same issues. 
I tend to think it is not down to mechanical play since switching to live view nails focus every time, and using old-fashioned SLR AF nails focus on average a bit less than every other time, depending on light level (based on preliminary tests). Either that or the way the algorithm works excites the mechanical play and leads to the way focus doesn't come to rest at the same point each time you half press. 
My tests were done at maximum aperture but I didn't notice an improvement when I tried a few with the lens stopped down after reading some comments here (other than a deepening of the zone of focus which somewhat masked the problem in some instances).

47872Mike wrote:

I've just been testing my new 50 STM and am struggling with the same issues.
I tend to think it is not down to mechanical play since switching to live view nails focus every time, and using old-fashioned SLR AF nails focus on average a bit less than every other time, depending on light level (based on preliminary tests). Either that or the way the algorithm works excites the mechanical play and leads to the way focus doesn't come to rest at the same point each time you half press.
My tests were done at maximum aperture but I didn't notice an improvement when I tried a few with the lens stopped down after reading some comments here (other than a deepening of the zone of focus which somewhat masked the problem in some instances).

Yes, live view's AF system works better in challenging situations where the light isn't perfect and the focus is razor thin.

I think the expensive "L" lenses handle poor lighting better and control fringing better. That is one reason why they are worth the money. Cheaper lenses throws off the optical viewfinder auto focus system since they don't handle the light as well (blue/purple fringing and contrast).

My 50mm f1.4 hits all the time on my EOS R, which is always live view type AF. It missed almost all the time with when mounted on my T5i, which was through the optical viewfinder.