Is STM more accurate than USM

My answer, Maybe.

Because I love a 50mm, when the new 50mm f1.8 STM came out I had to buy one.

First, as alway I did the lens calibration on all three of my bodies. All on my lenses get varying calibrations on different bodies. The STM required no adjustment on any on the bodies. It was right on.

I used it a wedding last weekend.

This is a situation I shoot in all the time. People moving, in very dim light. I do a wedding just about every weekend June thru September. At the end of the ceremony it always comes to this. No light, weird color from LED DJ lights. It's an AF nightmare, AF longshot. You shoot 5 frames to get 1 that's just right. MY assistants have never even been able to get even that, maybe 1 in 10.

I have used a 50mm f1.4 USM, Sigma 50mm f1.4 hypersonic (not the Art) and the Canon 50mm f1.2L. I always shoot at f1.8, even with a 1.4 or 1.2 lens both to get a bit better odds and to have DOF that isn't razor thin. Last Saturday I used the new 50 STM. We were shooting with the cameras we always use. The only thing different was the lens.

It is way slower to focus, not snappy at all, but much more accurate. When it hits, it is exact. I got most in focus. I could see it happening right in the viewfinder. It was slow, but then got it without hunting. To see what would happen in someone else's hands I gave it to my assistant and she hit at least as well as I have ever done with other lenses.

Has anyone observed this or is it a fluke?

Tom

The Canon EF 50mm f/1.4 is not a real USM lens. It uses an inferior micrometer instead of a ring mechanism that is used in almost all Canon USM lenses. The 70-300 IS (the non-L version) is another micrometer lens. The 50L focuses much better, but it has never been known as Canon's fastest focusing lens.

EDIT: The following post is based on erroneous beliefs, mine. Please disregard. Left intact for reference.

The way I understand things, STM has the potential to be more accurate than USM. STM is an iterative process between the lens and the camera, allowing for very fine final adjustments. USM/PD autofocus is a one-time determination-and-go focus. I field-tested my 24-105 STM a few days ago. If I intentionally go from very close to very far, the focus is noticeably slower than USM. However, in real life, nobody shoots like that. I found it quite speedy enough in AI-Servo for darting squirrels and running children. The OOF shots were, as usual, my fault. I find the 24-105 to be dead-accurate on both my 60D and 6D, although I find the focal length a bit stupid for apsc.

GeoKras1989 wrote in post #17595260
The way I understand things, STM has the potential to be more accurate than USM. STM is an iterative process between the lens and the camera, allowing for very fine final adjustments. USM/PD autofocus is a one-time determination-and-go focus. I field-tested my 24-105 STM a few days ago. If I intentionally go from very close to very far, the focus is noticeably slower than USM. However, in real life, nobody shoots like that. I found it quite speedy enough in AI-Servo for darting squirrels and running children. The OOF shots were, as usual, my fault. I find the 24-105 to be dead-accurate on both my 60D and 6D, although I find the focal length a bit stupid for apsc.

I thought that was a matter of newer bodies and newer lenses, including some USM lenses (like the 35/2 IS), not just STM? That was how Roger at LR described it in a series of articles.

LonelyBoy wrote in post #17595565
I thought that was a matter of newer bodies and newer lenses, including some USM lenses (like the 35/2 IS), not just STM? That was how Roger at LR described it in a series of articles.

I am confused. What is a matter of newer bodies/lenses? Focus accuracy? USM is fast, and for the most part, very accurate. That said, STM has the potential to be more accurate, more often, because it is an iterative process. USM lens/camera combinations make ONE calculation, then adjust focus. That provides the speed. STM lens/camera combinations make A (as in 1) calculation, adjust, check focus, adjust, check focus, repeat until focus is achieved. This iterative process provides the potential for more accurate focus, but it is slower.

Potential and accuracy are only valid up to a point. Once something is 'in-focus', it can't get more 'in-focus'.

Qualify all of this with a field test of my 24-105 STM. It is certainly slower, especially from MFD to infinity, but nobody shoots like that. In practice, it is nearly (possible actually) the equivalent of the 70-200 f/4 IS USM in AI-Servo. If tracking one subject, like a football/basketball/ho​ckey player, the STM will match USM. If going from near-field to far-field to catch something shot-worthy, the STM will seem incredibly slow.

LonelyBoy wrote in post #17595565
I thought that was a matter of newer bodies and newer lenses, including some USM lenses (like the 35/2 IS), not just STM? That was how Roger at LR described it in a series of articles.

I haven't found that. One the same bodies my 35f2IS doesn't hit with the same pinpoint accuracy as the 50 STM does. But, the STM lens is noticeable slower getting there.

Tom

GeoKras1989 wrote in post #17595582
I am confused. What is a matter of newer bodies/lenses? Focus accuracy? USM is fast, and for the most part, very accurate. That said, STM has the potential to be more accurate, more often, because it is an iterative process. USM lens/camera combinations make ONE calculation, then adjust focus. That provides the speed. STM lens/camera combinations make A (as in 1) calculation, adjust, check focus, adjust, check focus, repeat until focus is achieved. This iterative process provides the potential for more accurate focus, but it is slower.

Potential and accuracy are only valid up to a point. Once something is 'in-focus', it can't get more 'in-focus'.

Qualify all of this with a field test of my 24-105 STM. It is certainly slower, especially from MFD to infinity, but nobody shoots like that. In practice, it is nearly (possible actually) the equivalent of the 70-200 f/4 IS USM in AI-Servo. If tracking one subject, like a football/basketball/ho​ckey player, the STM will match USM. If going from near-field to far-field to catch something shot-worthy, the STM will seem incredibly slow.

http://www.dpreview.co​m …phase-detection-autofocus

Both USM and STM lenses use a 'Closed Loop' iterative process. Let's not resurrect this myth again...

The STMs have an advantage in CDAF in the 70D because of its Dual Pixel AF technology

MakisM1 wrote in post #17595714
http://www.dpreview.co​m …phase-detection-autofocus

Both USM and STM lenses use a 'Closed Loop' iterative process. Let's not resurrect this myth again...

The STMs have an advantage in CDAF in the 70D because of its Dual Pixel AF technology

Informative article. Thank you. Now I can add that to the list of things I know that are not true. Like tomato juice kills skunk funk. I will edit my previous posts in a bit.

I am glad I didn't know this about the skunk... We had one 'mark' our car when we were looking for a house in NJ (it was the house we eventually bought and put in the terms of the contract to remove the skunk from its nest under the deck). The car smelled for the rest of the week!...

MakisM1 wrote in post #17595872
I am glad I didn't know this about the skunk... We had one 'mark' our car when we were looking for a house in NJ (it was the house we eventually bought and put in the terms of the contract to remove the skunk from its nest under the deck). The car smelled for the rest of the week!...

Erroneous post edited.

Skunk funk killer recipe follows. I use it on the dogs about once a year. They each manage to corner a skunk about that often.

1 quart hydrogen peroxide.
1/4 cup baking soda.
1 tsp Dawn (brand optional, I think) liquid dish soap.

Mix to suspend insoluble soda in foamy peroxide. Liberally apply to skunked-dog. Let stand 15 minutes. Rinse thoroughly. Re-wash dog with your favorite pet shampoo (peroxide is hard on the skin/coat). Your dog can sleep in bed with you that night, skunk free. Same recipe also works on cat urine on the carpet. Voice of experience.

To keep this relevant, I am sure the mix will clean skunk funk off of USM and STM lenses. Try it at your own risk.

MakisM1 wrote in post #17595714
http://www.dpreview.co​m …phase-detection-autofocus

Both USM and STM lenses use a 'Closed Loop' iterative process. Let's not resurrect this myth again...

The STMs have an advantage in CDAF in the 70D because of its Dual Pixel AF technology

Okay, so why does my STM lens seems like it's more accurate. I think it is more accurate. Is it sample variation? Random happenstance.

Tom

GeoKras1989 wrote in post #17595900
Erroneous post edited.

Skunk funk killer recipe follows. I use it on the dogs about once a year. They each manage to corner a skunk about that often.

1 quart hydrogen peroxide.
1/4 cup baking soda.
1 tsp Dawn (brand optional, I think) liquid dish soap.

Mix to suspend insoluble soda in foamy peroxide. Liberally apply to skunked-dog. Let stand 15 minutes. Rinse thoroughly. Re-wash dog with your favorite pet shampoo (peroxide is hard on the skin/coat). Your dog can sleep in bed with you that night, skunk free. Same recipe also works on cat urine on the carpet. Voice of experience.

To keep this relevant, I am sure the mix will clean skunk funk off of USM and STM lenses. Try it at your own risk.

Since I started this thread, I feel the need to contribute something. My dog feels it his business to kill all the skunk kittens in our woods, and I have significant experience. Forget the make your own. Nature's Miracle Skunk Remover knocks it right out. I keep a quart handy.

He's probably out hunting now.

Tom

kachadurian wrote in post #17596018
Okay, so why does my STM lens seems like it's more accurate. I think it is more accurate. Is it sample variation? Random happenstance.

Tom

My guess is that the 'step' on the STM motor is much smaller than the 'step' in the MkI/MkII, therefore for any given distance it is easier to find a spot within AF tolerance and stop hunting.

Example: Say that the AF tolerance is 4 inches and MkII has a step of 12 inches while the STM has a step of 1 inch (this is an example, the numbers are made up).

Say that a subject is approximately between 6 and 7 ft away.

Chances are that 1/3 of the time the subject distance is such that the step falls between 6' 8" and 7' and will focus on the far end (7 ft) within AF tolerance, the other 1/3 probability is that it will fall between 6' 0" and 6' 4" and it will focus in the near end (6 ft). But if the subject is between 6' 4" and 6' 8" inches, the MkII will hunt interminably not being able to land between 6' and 7'. The STM of course will have no trouble finding the subject anywhere between 6' 0" and 6' 12" inches landing within 1 inch (of the required 4).

As others have said in this thread, it depends on the lens and the USM implementation. I don't doubt that the f/1.8 STM might be more accurate than the f/1.4 USM lens which used a rather poor micromotor USM implementation.

However, the fact that Canon still uses USM AF in its new super-telephoto lenses should mean that the USM variants those lenses use are still better than STM at pure autofocusing (video excluded).

if I had to guess, STM will be more accurate. USM have very short throws and small amount of steps, so accuracy will be in question. Compare the 50L to 85L, the 50 has very short movements, while the 85 has much longer (and slower) precise movements.