LensRentals.com

Canon 24-70 f/4 IS Resolution Tests

Posted by

A lot of people have been waiting on the release of the Canon 24-70 f/4 IS, deciding which of the new lenses they wanted, or whether they wanted to upgrade at all. So, when the first batch arrived early this morning we were set up and waiting to do some resolution testing.

The Players

The Canon 24-70 f/4 IS fits in the 'standard zoom' lens category which is fairly crowded. The new Canon 24-70 f/2.8 Mk II and Tamron 24-70 f/2.8 VC lenses are already out. The Canon 24-70 f/2.8 Mk I is no longer produced but is in a lot of photographers' bags already. The Canon 24-105 f/4 IS already offers an f/4 image stabilized zoom with a greater range. It's a lot to choose from and the new lens is going to have to be impressive to sell well at its higher price.

Let's take a look at how this group prices out (these are today's retail prices, but I expect you can get $100 or so off each of them fairly soon).

 

LensPrice
Canon 24-70 f/2.8 Mk II$2,199
Canon 24-70 f/4 IS$1,499
Tamron 24-70 f/2.8 VC$1299
Canon 24-105 f/4 IS$1,091

 

We've previously compared the Mk and Mk II Canons in depth, and compared them to the Tamron f/2.8 VC. A couple of conclusions are already apparent:

  • If you want the best 24-70 f/2.8 zoom at any price, the Canon 24-70mm f/2.8 Mk II outresolves anything else, period.
  • If you want image stabilization with your 24-70 zoom, the Tamron is really very good, and while it doesn't quite resolve up to Mk II standards, it does outresolve the Mk I version (which is itself a pretty good lens, at least when you get a good copy).
  • The Canon 24-105 f/4 IS gives good quality and greater range at a lower price.

I try to identify my expectations going into an evaluation. In this case, given the price, the 24-70 f/4 IS will need to be a better lens than the Canon 24-105 and at least as good as the Tamron at f/4 to justify its price. I'll go further and say it should be better than the Canon 24-70 f/2.8 Mk I at f/4 as well. Anything less would be a failure.

The Usual Disclaimer

This isn’t a lens review. I am not a reviewer. I don’t spend days evaluating a single copy of a lens for all of its traits and characteristics, nor do I take hundreds of really great photos with it and describe how it works in the field. Several of the lenses I tested today are on the way to people who will do just that and their thorough reviews will be available in a week or so.

What I do is test multiple copies of the lens for resolution and other basic stuff. Multiple copies lets us take sample variation into account to some degree, which a thorough review of a single lens can't.

A Quick Comparison

Compared to the Canon and Tamron 24-70 f/2.8 lenses, the new one is a bit smaller at 3.7" long, 1.32 pounds, with a 77mm front filter ring.

 

Left to right Canon 24-70 f/2.8 Mk I, 24-70 f/4 IS, 24-70 f/2.8 Mk II

 

Left to right: Canon 24-105 f/4 IS, Canon 24-70 f/4 IS, Tamron 24-70 f/2.8 VC

 

And with barrels extended

 

It's also significantly shorter than the 24-105 f/4 IS - although they actually look quite alike. I'll add that I really, really, really like the pinch cap on the new lens, which makes it easy to remove the cap even with the hood mounted.

Macro Mode

Not really part of our usual testing here, but the 24-70 f/4 IS has a special switch that allows it to become a near macro (0.7x) at the expense of losing infinity focus. While I usually think of 'macro zoom' as a marketing gimmick (and I'm still not sure about this one), Canon did put their Hybrid IS unit in this lens, just like the one in the Canon 100mm f/2.8 IS L macro lens, so I'll give this one the benefit of the doubt.

 

Macro with 100mm f/2.8 IS L (center 50% of image, resized)

 

Macro mode with 24-70 f/4 IS (center 50% of image, resized)

 

I find the macro mode and small size nice differentiators that may make this lens a good choice for some photographers if the optics are excellent. I should mention, though, that the macro working distance is quite short -- about 2 inches from the front element. Getting it into macro mode is a bit clumsy as you have to hold the switch while rotating the zoom ring. But it's a nice feature.

Resolution Results

We tested 22 copies of the Canon 24-70 f/4 IS L at 24mm and 70mm using our Imatest lab. We had numbers from testing 100+ copies of the 24-105 f/4 IS already available. We had recently tested all of our 24-70 Mk I, Mk II and Tamron 24-70 f/2.8 VC lenses, but only at f/2.8, so we repeated those tests on 10 known good copies of each at f/4 so we would be able to compare how all the lenses performed at f/4. There was not time to test everything at f/5.6.

Results at f/2.8

The f/2.8 numbers are posted elsewhere, but I'll repeat them here.  If you're going to pay more money for an f/2.8 lens, but mostly will shoot at f/4, it's worthwhile knowing how much resolution you give up at f/2.8. The numbers are Imatest MTF50 values at the cener, averaged at 13 points over the entire lens, and the average of the 4 corner numbers.

Lensfocal lengthCenterMTF50AverageMTF50Avg. CornerMTF50
Canon 24-70 f/2.8 Mk I70mm710580360
Canon 24-70 f/2.8 Mk II70mm940810480
Tamron 24-70 f2.8 VC70mm740660420
Canon 24-70 f/2.8 Mk I24mm730610380
Canon 24-70 f/2.8 Mk II24mm950820510
Tamron 24-70 f2.8 VC24mm815765430

 

The three 24-70 f/2.8 zoons are all good lenses, but it's obvious from a resolution standpoint the Canon Mk II is the best and the Tamron between the Mk I and Mk II results.

Results at f/4

We have 5 lenses to compare at f/4 and I've added corner resolution, distortion, and chromatic aberration numberes. To keep all that data organized I'l make separate tables for results at 24mm and 70mm.

We'll look at 24mm first.

LensCenterMTF50AverageMTF50AvgCornerMTF50Barrel Dist.CA%
Canon 24-70 f/2.8 Mk I8607354702.0%.05%
Canon 24-70 f/2.8 Mk II10109106152.1%.05%
Tamron 24-70 f/2.8 VC9408155002.8%.04%
Canon 24-70 f/4 IS9508255601.7%.05
Canon 24-105 f/4 IS8907304805%.06%

The f/2.8 lenses, which are all good at f/2.8, sharpen up even further when stopped down to f/4. The Canon 24-105 f/4 IS accounts itself well here, resolving just as well as the original Canon 24-70 f/2.8 lens, although it does show more barrel distortion than the others. The new 24-70 f/4 IS isn't quite what I'd hoped (I was hoping it would match the Mk II f/2.8 lens at f/4), but it's better than the 24-105 or the original 24-70 f/2.8. It's probably a bit better in the far corners than the Tamron 24-70 f/2.8 VC, but otherwise they're about dead even.

One thing that is very good on the new lens is the lower barrel distortion, just under 2%. This probably is most noticeable when compared to the 24-105 f/4 IS, which has pretty bad barrel distortion right at 24mm.

Here is the same data when the lenses are shot lenses at 70mm

LensCenter MNTF50AverageMTF50Avg.CornerMTF50Pinc. Dist.CA%
Canon 24-70 f/2.8 Mk I8056454301.3%.04%
Canon 24-70 f/2.8 Mk II9758205801.4%.05%
Tamron 24-70 f/2.8 VC8907355101.5%.04%
Canon 24-70 f/4 IS9207505251%.05
Canon 24-105 f/4 IS8406804701.2%.05%

 

None of the f/2.8 zooms are quite as sharp at 70mm as they were at 24mm, but the difference is pretty minimal. The 24-105 f/4 IS still does quite well, perhaps a bit better than the original 24-70, although that's splitting hairs. It should be mentioned, though, that the 24-105 starts to get a tiny bit softer after 80mm - at 70mm we're in its sweet spot.

The new 24-70 f/4 IS stays in proportion -- it's about the same as the Tamron, not as good as the Canon 24-70 f/2.8 Mk II, but better than the original Canon 24-70 and the 24-105 f/4 IS.

Sample Variation

I've graphed the center and average resolution for all 22 copies below.

 

Center (horizontal) and average (vertical) MTF50

 

That's a nice, tight pattern at 24mm, but at 70mm things are a bit more spread out. There are two outliers at 70mm. Further testing on the worst one shows it is clearly decentered. The not-quite-as-bad one seems to be a bit decentered as well. I will note that I took these two values out when I calculated the averages above since I want those to reflect good copies of the lens.

Twenty-two lenses is a pretty small sample to make further comments. I'll have more to say when we have seen 60 or 70 copies.

Addendum, January 14: 

After a head's up from our friends at SLRgear.com and a couple of other users, who saw lower resolution in their copies at 50mm, we went back and retested a couple of dozen 24-70 f/4 IS at 3 focal lengths (24, 50, 70mm) instead of our usual two. (Most, but not all, zooms have lowest resolution at one extreme or the other, so we focus our testing there.)

We did find that 50mm resolution was slightly lower than 70mm for every copy. The center / weighted average at 50mm for the 24-70 f/4 IS was 875 / 700, compared to 920 / 750 at 70mm. Not a huge drop, but it was consistent. This is a bit surprising, but not a total shock. Some wide angle zooms exhibit similar behavior and the dip in resolution isn't extreme.

One thing that may be important to those of you who shoot around 50mm a lot, though, is that we also tested the Canon 24-70 f/2.8 Mk II and the Tamron 24-70 f/2.8 VR at 50mm. Both of these lenses were as sharp at 50mm as they were at 70mm, at both f/2.8 and f/4.

Conclusions

Obviously this hasn't told us a thing about autofocus accuracy, bokeh, or a dozen other things that have to be considered when choosing a lens. Just like you, I'll be waiting for more complete reviews to tell us about that.

On the basis of this information, though, I'm . . .  well, I don't know what I am. This is a good lens, but I at the price point I'd probably prefer the f/2.8 of the Tamron VC to the new Canon's f/4. The macro feature is nice and will certainly pull some people towards the Canon.

This is only a sample of 22 copies, but the sample variation at 70mm is a bit bothersome. I don't feel comfortable making any statements about it, though, until we've seen another 40 or 50 copies. This might just be a couple of bad lenses in a small sample.

My bottom line is I sit here thinking the prices need to settle down a bit. If I was considering upgrading to one of these lenses I'd probably hold off a few months and see how the prices change.

 

Roger Cicala

Lensrentals.com

January, 2013

54 Responses to “Canon 24-70 f/4 IS Resolution Tests”

Paul Doran said:

Isn't the issue with the tamron that it is quite slow to AF compared to the Canon?

Andrey said:

Nice review!

John said:

Ive shot with Tamron lenses in the past with success. They are lenses which are workable. However, as noted in many reviews out there their focus system has issues. First its slower than the Canons. Second it is not always accurate or predictable. Another issue with the Tamron is bokeh which is different from the Canon.

These issues probably will not matter to most photographers but in the high demanding wedding field I will settle for Canon only lenses. Not saying Tamron is a bad lens but when you have your reputation on the line you need all the advantage you can get. If the bokeh doesnt look right or the focus doesnt seem right an angry couple can just go file a lawsuit or trash you out in an unfavorable review.

steve said:

Very interesting. This kind of data really helps the decision. I am the ideal person for the 24-70 F4 IS. Here is why.
- I find the stabilization is worth more than just 1 stop. You really can hand hold many more photos successfully. I need IS basically.
- I also find that letting a lens stop down to f2.8 is worse than f4 and a bumped up stop to the ISO. Thats only if you have more ISO to spare. As it turns out, I usually do. I shoot daylight handheld and only when I shoot in shadows do I hit my max aperture.

I am very impressed that the sharpness is better with the F4 IS than with the Tamron at f2.8 AND f4. That is a huge deal. Usually you need to stop down 1 stop to get to a better quality, but this f4 lens is better wide than the tamron. When things get much darker, I am on a tripod and can get to f5.6 or f8.

I think by f8, you would not know the difference in these lenses, they are all very good. I have been shooting with the f4 24-105 for a very long time and I am also surprised to see that it resolved better than the 24-70 mk1 which I never heard anyone comment on before, so thats interesting.

JonJon said:

the traumflieger guys claim to do the worldwide most ACCURATE tests if you read the website.

well.... to me it´s seems as they do a comedy show?!

a testchart and imagelab software .. well i have that at home too.

but im not so bold to say my test are the most accurate in the world.

LensRentals Employee

Roger Cicala said:

Hi Pat,

No, I'm afraid we don't. The table feature we have wouldn't hold it all.

Pat said:

Roger,
Do you have a page on the site that shows the resolution data for all the lenses you've tested? I'm interested in that!

LensRentals Employee

Roger Cicala said:

Hi Don,

I wish I did, and I've actually given it a lot of thought. I would assume a 3 per second tremor would be most useful, since that's the human neromuscular sweet spot for tremor, but I'm not sure of the variation or magnitude in each of the 5 planes (since we'd want to compare 3 and 5 axis IS systems, too). Imaging Resource has a tool they use, but I think it's still rather crude.

Right now the best I've come up with is the "Roger drinks two cups of coffee and tests the IS system" but I don't think it's very reproducible :-)

Don said:

Roger,

Forgive me if this has already been covered but do you have a standardized (mechanical)hand held motion equivalent test that can evaluate the IS performance parameters of the lens under test. Such a test can be used to compare non IS lens performance under motion conditions. It should be simple to design test platform for single, dual and three axis testing.

For those who shoot on a tripod IS is useless but for those hand holding such tests would be of interest for comparisons. I assume that the best performing lens, without IS, would appear near useless in comparison to a IS lens operating within its useful range.

Don

Uli said:

Forget Traumflieger
Some comments mentioned "Traumflieger" and the up to 15 minutes inflated clip of EF 24-70 4 IS and asked for the translation from the German. I am German and watched this clip - a total waste of time. I will nothing translate, it is too sad... 15 Minutes long murmured those Traumflieger-Guys some irrelevant talk about Macro-snapshots of fungi and other Murcksquatsch. They change with dumb hands some lenses, for example the 100 Macro and talked in an diffuse and diffractive way about this and that.
If the informational lack of quality of that dipweed-made clip is representative for their judgements, then you can trust more some playing with dices. I am very angry about that stupid clip and feel offended of that compatriot. Unfortunately I have not the power to expatriate them to restore German honor, so I only can excuse me for this Blödsinn. Yours Uli

LensRentals Employee

Roger Cicala said:

AC you can buy an Imatest set up if you want - although it has a significant learning curve.
The problem is your Imatest results and mine will be slightly different since lighting, test charts, shooting distance, camera used, etc. all are variables.

In general I recommend just doing some good testing at home: http://www.lensrentals.com/blog/2010/11/how-to-test-a-lens

ac said:

Is there a way to measure these MTF scores on a given lens at home. What software/hardware would I need? I'd like to ensure I have a sharp copy

Nqina Dlamini said:

Great read.
The feedback to comments is also lightening.
Thank you Roger

David said:

Roger, I think you need to write UV filters article. Everything you've done so far has been to greater depth than almost anywhere. Why not get some good data out there?

george said:

When Canon puts forth a 40 mega-pixel camera and 2 meter wide images are printed, it will matter how sharp these lenses really are. Until that day arrives all this talk is over splitting hairs of quality.

Samuel said:

Roger, thank you once again for your efforts in doing these elaborate tests. I'm wondering if you have also tested the 24-70 f4 IS in macro mode for resolution. Does the corner quality suffer a lot?

L.P.O. said:

Noah Chen said:
> Everyone forget about the “distortion.” Canon’s 24-105
> is totally unusable for portrait or architect shooting.

I'm afraid I fail to find any information in this sentence.

1) I fail to think of any way that 24-105's distortion characteristics would make any difference in portrait work, let alone make it useless? (If you answer this, remember not to confuse normal 24-mm-elongated-heads-in-the-corners-distortion with anything specific to the 24-105.)
2) DPP as well as pretty much any RAW image converter can do distortion correction. So can the in-camera RAW developer in Canon newest DSLRs, including the whole current FF product line: 1D X, 5D Mark III and 6D. So I fail to see how the 24-105's distortion would have any effect with architecture - unless you insist on shooting JPEG-only with an old body.

Regards,
- L.P.O.

LensRentals Employee

Roger Cicala said:

David, I apologize in advance, I completely hate that article. It's the only thing on the Lenstip site (which I love) that I find totally worthless and used as misinformation.

Begin rant>

I have no so much an issue with Lenstip's results, but rather I think it's data in search of a purpose: nobody buys UV filters for their UV blocking capacity. That was only necessary in film. So I could care less about UV radiation blocking that they so carefully rated.

Light transmission is nice to know, I guess, but I'm much more worried about is the glass perfectly flat, does it have tiny ripples that will affect distortion, what is the reflective % of the coatings involved. Bad light transmission is bad, but good light transmission isn't necessarily good. This is a very blunt tool - if a filter is bad fine, it's bad. But if it's light transmission is OK, that doesn't mean it's a good filter.

Showing flare by hand in the sun or with a streetlight is nice, but that's a very blunt tool, too. Conditions were changing (inevitably as the test was done). Again an awful result would be pertinent, but all the decent results don't mean a whole lot. Did you actually look at the comparison shots on the various filters? They're all slightly different. Therefore they are largely meaningless.

Plus the early part of the article says it all: All the photographic tests were conducted with a D200 Nikon camera and a Sigma 17-70 mm f/2.8-4.5 DC Macro lens. Really? You would have to have a very, very poor filter to show me much effect with that combination. How about we try a Nikon D800 and a Zeiss 21mm f/2.8? I'm sure comfotrable a filter isn't going to ruin the image quality of a D200 with a Sigma 17-70 variable zoom lens.

I'm sure their results are accurate as far as they went, those guys are meticulous. I just don't think they have any use and like so many articles, people don't read it critically, they just say " Lenstip says this is the best filter".

Now, if they decide to measure scratch resistance or shattering force, that would be interesting since people buy those filters for protection. I'd also love to know about binding of the threads in lenses at different temperatures. If you have an aluminum filter holding ring on the lens, is aluminum or brass filter ring less likely to bind? What about a steel lens? What about plastic?

Or if someone would get a simple laser and measure reflectance at different angles (Every filter will be pretty fine on parallel light entering directly into the front element, it's the angled light that matters. And reflectance will change at different wavelengths and angle of entry - what might do well under Tungesten lighting might not under LEDs, or in sunlight.) Perhaps measure corner coma or other aberrations with the various filters?

Plus we need to do this for both wide-angle and telephoto lenses since the angle of view is so different. What might be fine at 135mm might not be fine at 24mm.

/rant off>

Roger

george said:

note on filter use, while working a wedding one time, a guest turned very quickly while holding a lit cigarette and smashed it out in my 24mm leica asph. Needless to say, the loss of a B+W MRC filter was worth it. (this was a total accident by an elderly woman).

on sharpness of the 24-105, mine is really sharp, and in practical event use the corners are almost always on a different plane of focus, so critical corner sharpness is just not that important. when i am working on architecture and landscape, i try to use prime manually focused lenses. (zooms always have some compromise).

the new 40mm is really a good lens, almost no distortion and tack sharp to the corners. (if you want perfection by only prime lenses).

remember many jobs only require web or 8x10 print sizes and all of these lenses will yield images that surpass that requirement by a long shot.

Noah Chen said:

Everyone forget about the "distortion." Canon's 24-105 is totally unusable for portrait or architect shooting. I guess Tamrom wont be better neither. This might be the selling point for 24-70 F4!

LensRentals Employee

Roger Cicala said:

Larry - I don't want to start a 3 page forum-type argument but here goes. If it's a truly high-quality UV filter, then it should have very little effect except perhaps when shooting lights. But if it's a truly high quality UV filter it may cost as much as it would cost to simply replace the front element.

If it's a $30 Tifhoya or whatever, then yes, it will probably affect image quality.

Full Disclosure: I'm always hesitant to buy an item I know is the highest margin item in the store. And filters are just that. So take what the camera salesman says with a grain of salt. A lot of our employees are ex-camera store and will talk about how their reviews came down to how many filters they sold, or that they got no commission on cameras and lenses, but $5 for every filter sold, etc. It's left me really anti-filter.

Larry said:

On a lens like the new 24-70 f/4L (or the 24-70 f2.8L II, what is your opinion of using or not using a high quality UV filter? Does it affect image quality at all?

LensRentals Employee

Roger Cicala said:

L.P.O. - first thank you for the head's up. Fixed the it's.

My numbers seem to reflect the 24-105 gets a little softer in the corners after 80mm, but those are imatest numbers, shot at 20-30 feet in our lab. It wouldn't shock me if the lens behaves differently at longer shooting distances.

Leave a Reply