Canon 24-70 f2.8L - Bad copy?

[Marcel]

Took some photos of my neice last night, and while processing them, someone on IRC pointed out one of them wasn't that sharp.

(Whole scene was lit with flash, hence the f8) - Zoomed into 1:1 in Lightroom.

Anyway, I did some testing this morning. Splayed my tripod legs out and got low to the floor. Wired remote, plus mirror lockup, one shot focus, and central focus point.
Pointed at the grass outside, and placed a dandelion over the central focus point, and shot at f2.8, f4, f5.6, f8, f11, f13 etc

Here's f2.8 (1:1 view in LR)

Quite poor considering. But there's nothing that's actually 'sharp' which means its not front focusing or back focusing.

Here's f4.0 (1:1 view in LR)

and finally f8.0 (1:1 view in LR)

Naturally upon seeing this I was a bit shocked, so I took a cornflakes packet into the garden for another test. AF, One Shot, Focused on the 'guidelines' word with centre spot only active, then turned the lens to MF.

f2.8

f5.6

f8.0

Look at the difference! At this point, I'm thinking 'this is quite poor, must be a really soft copy. My sigma is often sharper than this'

So I tried something. I turned on live view, turned AF off, and zoomed in and focused manually....at 2.8

This is what I got

f2.8 (Focusing manually)

Still not brilliant, but much better than the previous AF attempt I think.


Here is a comparison between the AF 2.8 and the MF 2.8

The difference between manually focusing is quite noticable tbh.

Thoughts?

 

[dazzajl]

If the first shot of your niece is unsharpened, then it's not really too bad. I'm often shocked at just how poor 35mm kit is sometimes when you start to pixel peep.

As for the other stuff, there are some worries there aren't there. It really does look like you have two issues compounding the results there. The MF is way ahead of your AF isn't it and that's not at all right. Then as you point out with the 1st dandy, there isn't really a sharp zone at all, so it's not front or back focusing.

If you're happy with the tests so far, then the next step has to be to see if the camera is giving you the AF problems or not. Back to the garden to do it all with another lens I guess. Fingers crossed it's just the lens and one thing to ship off to the menders.

 

[puddleduck]

Even the MF version doesn't look as sharp as I'd expect (I'm using a Tamron 28-75 and Nikon 24-70 f/2.8 as reference here).

There is definately an AF error, but this could equally be a body issue too.

 

[digitalfailure]

I have a similar thread going on poor focus with a 24-70L but after an hours chatting with Canon and their repair representive's the suggestion is that my 5D body is at fault and the 24-70 is really highlighting things, more so than other lenses.

I'll be taking the 5D which appears to back focus by 15" when a subject was at 7.5ft away when using the 24-70 to H Lehmans in Stoke on Friday morning.

Apparently it's not the first time they've seen a drastic focus fault when the 24-70 and 5D are paired. He did say that 5cm back focus at 1meter is something he'd seen before a few times

 

[HoppyUK]

Not convinced by the dandelion test. The focus point on my 40D is much bigger than the square in the viewfinder, and always picks up the nearest object.

But the cerial packet is pretty convincing. Live view is incredibly accurate and I doubt that any AF system can match it (they work within DoF tolerance after all) but it should do better than that.

To see what the AF is up to, how about reshooting the cerial packet but with a 1ft ruler propped up vertically beside it, so that at the bottom it leans a couple of inches in front of the face of the cornflakes box, and a similar distance behind at the top. Just lean it against a second packet.

When you AF on the face of the cornflakes, the corresponding point on the ruler should also be it's sharpest point. If it's not, and another part of the ruler is in clearer focus, then you know if it's front or back focus, and by how much.

But while this is all very well, the original pic that started this was shot at f/8, which should be enough to cover any small focusing error. So is this an AF question, either camera or lens related, or a 'bad copy' lens question as originally posed?

 

[Marcel]

Hoppy, it's both (AF and poor copy - related).

Reason being if it was front or back focusing, the dandelion would be sharp at *some* point, it clearly isnt sharp anywhere.
But the cornflakes indicate an AF problem (maybe its poor *and* has an AF problem?).

The live view focusing wasn't AF, it was manual, I turned on Live view, zoomed in, and manually focused

 

[Marcel]

Hmmm interesting.

I've just set up a focus test chart on my desk, lens on tripod roughly pointing down at 45 degrees...you know, the usual.

This is AF at f2.8 (I've highlighted with a red square to show where the focus point was).

This is the shot that came out :

This is the same shot, but using Live View to Manual Focus

 

[dazzajl]

OK then Marcel, two questions for you. What does the same test look like if you focus through the viewfinder with this lens and what happens if you use another lens?

 

[puddleduck]

You definately have a large optical problem here - even the one that is reasonable well focused is far too soft. The CA to me suggests that a major lens element - probably aspheric - needs to adjusted.

 

[mmcp42]

I would have said from the dandelion picture that it's actually focussing on the grass that is closer to the camera (front-focussing?)

the later pictures seem to confirm this

 

[InaGlo]

Marcel, where did you get the 'focus test' from ... looks a better one than we are using here.

 

[Marcel]

It's from www.focustestchart.com

I'm gonna try a different lens in a minute (will actually try my sigma 24-70...havent sold it yet )

 

[Marcel]

Just tried my Sigma 24-70 f2.8 and here are the results.
f2.8, Auto Focus

f4 Auto Focus

f8, Auto Focus

And finally, f2.8 using Manual Focus

I think I know what this looks like, but I want someone else to say it

 

[puddleduck]

I think I know what this looks like, but I want someone else to say it

Its your camera body.

 

[dazzajl]

but I want someone else to say it

You get to lower the postage cost per unit by sending off a camera and lens.

 

[Marcel]

Thats what I was thinking. Just going to ring a friend, see if I can pinch his 450D for testing....

 

[puddleduck]

Thats what I was thinking. Just going to ring a friend, see if I can pinch his 450D for testing....

For reference - and this is from a Nikon and Tamron 28-75 f/2.8, this is how my Tamron looks (light is going here, so its a bit under)

100% crop

​

Not sure how this lens compare to your Canon ones (in the Nikon world the Tamron is a super performner), but I think you should be seeing this sort of sharpness (or close to) from at least your 24-70 L, on the long end, wide open.

 

[Qwerty69]

Thats what I was thinking. Just going to ring a friend, see if I can pinch his 450D for testing....

Fully agree that the only way to be 100% is to try both lenses on another camera body.

 

[Marcel]

puddleduck. To be honest, that's the sort of sharpness I was expecting from my Canon. It's an L lens which is a professional lens.....I should be able to use the camera at f2.8 without worry.

I've got the 450, will be using that for a test in about half an hour.

 

[Marcel]

I've also just noticed a post in digitalfailure's thread.

One simple check from a tripod in single shot mode is to focus on something and then focus again, if the lens adjusts it could indicate a calibration problem as it clearly didn't get it right the first time

This is something that my camera does when in AI Servo I can focus on something, and if I hold the focus, keeping the camera and the subject still, I often see the lens 'twitching'...like it can't make it's mind up.

 

[puddleduck]

puddleduck. To be honest, that's the sort of sharpness I was expecting from my Canon. It's an L lens which is a professional lens.....I should be able to use the camera at f2.8 without worry.

I'm not familiar with your lens at all, but something is clearly amiss. I'd certainly expect to be able to use any decent f/2.8 lens wide open, and I'd only expect to be stopped down for more DoF - never for sharpness.

Be interested to see how you find things with a different body... I'm sure your body is at least part of the problem.

 

[Marcel]

Shots taken, just downloading and importing them now.

 

[Marcel]

Well here are the results.

450d, Canon 24-70, @f2.8, Auto Focus

450d, Canon 24-70, @f8, Auto Focus

450d, Canon 24-70, @f2.8, Manual Focus

And the Sigma

f2.8, Auto Focus

f8, Auto Focus

f2.8, Manual Focus

Do those Canon examples at f2.8 look acceptable now? I can see clearly the difference between it and the Sigma now, but are they acceptable enough? Where would you say the plane of focus is?

 

[HoppyUK]

Does the 450D come with a lens? It would be good to have a control image, to see what it looks like with a 100% functioning camera and lens.

Apart from the focusing error, the CA on the Canon 24-70 is scary!

I also notice that you haven't managed to absolutely nail the focus even on manual with live view on maximum magnification. Shows how critical that test is Frankly, I think it's unrealistic to expect a lens designed for distant subjects to get that right every time.

Edit: Oops! Crossed post.

 

[puddleduck]

Your Sigma looks about as bad as the one I'm returning to Sigma in my current "bad lens" thread, in all honesty.

The Canon looks sub-par too.

Neither are acceptable to my eyes, as they are low in contrast and lack critical sharpness. I know these are 100% crops, but I think you could find better samples of both lenses (at least I'd hope so)

IMHO an f/2.8 lens should be very usable at its open aperture, otherwise you are paying a price / weight premium for no reason, if its not usable at its open aperture.

 

[HoppyUK]

Do those Canon examples at f2.8 look acceptable now? I can see clearly the difference between it and the Sigma now, but are they acceptable enough? Where would you say the plane of focus is?

Really need a shot with a different lens to compare. Neither lens looks brilliant in that test.

Canon lens looks front focused 2mm to me, but that may well be within tolerance. Even with live view it's not easy to get much better than that.

 

[puddleduck]

Doesn't Liveview do contrast based AF? In theory at least, it should be about as sharp as you are going to get...

 

[Monkey]

Doesn't Liveview do contrast based AF? In theory at least, it should be about as sharp as you are going to get...

not on the 40D no, they added that on the 50D

 

[mrgubby]

Apart from the focusing error, the CA on the Canon 24-70 is scary!.

Have to agree with that :thumbsdown:.

It's a shame you didn't post the examples with the make hidden, it would have been amusing to see all the "Thats typical Sigma quality" posts

 

[digitalfailure]

I know I shouldn't feel happy, but I'm glad my 24-70/body combo isn't the only one that's sub par

I'll be taking both the body and the 24-70 with me on friday to have them checked out.

Got to admire the effort you've put into this thread Marcy

I wonder if the varying lighting and shutter time is having any bearing on the results and the lenses ability to lock focus.......some of the lighting looks murky tbh.

For a test to be 100% accurate you need a control in the form of a known good and repeatable conditions.

 

[Marcel]

Here's the 'control' image.

Canon 450d with Kit Lens, @ 55mm, f5.6

 

[HoppyUK]

Kit lens wins! At least it looks like it. Ouch.

Need to see two side by side images, 24-70 and kit lenses, at f/5.6, same focal length, autofocused. Exposure and white balance adjusted so that the paper is clean white, and not orange. And can we see a bit more of the image area?

 

[Marcel]

Sure thing. I'll do it now
40d or 450?

I deliberately left the white balance as shot (auto) as I wanted the shots to be completely unmolested.

 

[HoppyUK]

450D. Never mind the white balance, just give it another two stops so to show any CA more clearly.

 

[pearce_jj]

out of interest what AF mode are you using? My 5D consistently front-focuses in servo mode if used with stationary objects. I sent it off to be calibrated and it is better but it does just seem to assume the subject is moving towards you if set to that mode.

 

[Marcel]

450D, Canon 18-55, @ 55mm, f5.6
450D, Canon 24-70, @57mm, f5.6

Or both side by side.....(cropped more though)
http://www.marcelbooth.co.uk/m/grabs/Lightroom-20090429-001828.jpg

pearce_jj, I'm using One Shot

 

[Jonny]

I'm looking into replacing my 24-105 with the 24-70 at the moment and i know people say it's just the odd one 24-70 that has a horror story with softness/focusing but now i'm actually researching it there's ***** loads. And this topic furthers my point!

I'd love the f/2.8, but when i've got a sharp copy of the 24-105 i'm starting to question if it's worth risking it? (Don't mean to thread hijack )

 

[HoppyUK]

Still under exposed but that's more like it.

Looks like a camera AF problem with the 40D then. At least with the 24-70... I think

You need to check that the lens isn't soft though. It's very hard to judge from these tests for all sorts of reasons, but it doesn't look great at f/2.8. If there an element that's been decentred, it will show up with uneven levels of sharpness if you blow up and compare the four corners of an image shot at f/2.8. They should all be equally sharp, and although they will probably look pretty mushy anyway, if one or more corner is markedly worse then that nails it.

It's easy to check this by shooting something small and reasonably distant, like a road sign or car number plate (something like that). Careful focus and exposure with a high shutter speed, then lock it all in manual and shoot four images with the target in each of the four corners.

I blame your neice

 

[Marcel]

LOL she does have alot to answer for, grrrrr

I forgot to overexpose the shots (And was already downloading them when I remembered).

I'm going to give Lehmanns a call in the morning, see what they say.
I do think the 40D is clearly backfocusing with the 24-70, as show in the first series of shots. Maybe it's a little bit of the body, and a little bit of a lens problem too?

df contacted Lehmanns about his 24-70 and 5d, and its a problem known to them (wrt the 5d anyway), in that the 24-70 and the 85 1.8 are lenses that show up this particular problem with the 5d.

Im wondering if this is a similar thing with the 40D. I've always 'wondered' about quite a few of my shots, merely blaming it on user error, or 'missing' the focus at wide apertures, and moreso, blaming soft / cheaper lenses.

Maybe it's not been me all along?

 

[HoppyUK]

LOL she does have alot to answer for, grrrrr

I forgot to overexpose the shots (And was already downloading them when I remembered).

I'm going to give Lehmanns a call in the morning, see what they say.
I do think the 40D is clearly backfocusing with the 24-70, as show in the first series of shots. Maybe it's a little bit of the body, and a little bit of a lens problem too?
df contacted Lehmanns about his 24-70 and 5d, and its a problem known to them (wrt the 5d anyway), in that the 24-70 and the 85 1.8 are lenses that show up this particular problem with the 5d.

Im wondering if this is a similar thing with the 40D. I've always 'wondered' about quite a few of my shots, merely blaming it on user error, or 'missing' the focus at wide apertures, and moreso, blaming soft / cheaper lenses.

Maybe it's not been me all along?

"Maybe it's a little bit of the body, and a little bit of a lens problem too?"

There's a problem, but it's not 100% certain what. To check that out thoroughly, you need a good body and a known good 24-70 L to compare. Best thing is to ship it off I guess.

Keep us posted on what Lehmann's say. Interested to know what the problem is exactly with this incompatability thing