Canon 100-400mm always seems soft.

[Simon Wantling]

Hi, i've seen others here use the canon 100-400mm lens. I'm not sure why but it's never as sharp as I expected. I'm using it with the canon 7d, but even with fast shutter speeds , the images aren't that sharp. Is there any advice you could give or have anyone had a similar experience.

Thanks

Simon

[Daniel Cadieux]

Hi Simon,

I have that lens as my main workhorse and it is plenty sharp. Although there are reports of soft copies out there, most problems are user error or tweakable adjustments. Check the following:

- Have you checked for the need of micro adjustments?
- Do you have a filter on...if so, I know this lens isn't very happy with one on (especially cheap generic ones)
- When handholding with IS on, are you using proper handholding techniques?
- Are you using a tripod, and if so is IS turned off? (it should)
- If used on a tripod with IS off, are the images sharp or sharper?
- Are the images sharp when stopped down (say, f/7.1 or f/8)? Some copies are reputed to be soft wide open, but sharp stopped down...

- Do you have any examples, including exif information?

[Roger Clark]

Hi Simon,

Compared to Canon's MTF charts the 100-400 lens has much lower MTF than those for fixed focal length lenses. I have a 100-400 and find my 300 f/4 with a 1.4x TC produces sharper images than my 100-400 at 400. A 7D has pretty small pixels so requires the best optics to deliver the detail the sensor is capable of recording.

I also believe there are variations in quality of the 100-400. Mine is softer than a friend's at 400 and tested identically (on tripod with mirror lock up, daylight fast exposure).

Be sure to remove any filter, even expensive ones. The large aperture combined with the long focal length will magnify any imperfection, which means even the best filter will likely cause some distortion.

Roger

[Simon Wantling]

Thanks for the tips Daniel. I suspect it needs some micro adjustment. I also have fairly large crops sometimes as well which can't help. Looked at your gallery. Wonderful photos.. Could you tell me what was the shooting distance on your photos. Just interested to know what the ideal distance between the camera and the subject is.

Thanks

[Daniel Cadieux]

Thanks Simon! Yes, distance to subject can be another factor. When cropping, if I end up with anyhting less than 75% of full-frame I just won't post it or put it on my website (I could crop more in many cases, but that is just a personal thing that I have). If you do crop heavily, yes, your images may appear soft if they aren't criticalyy sharp to begin with. This is another reason some examples could help us help you.

Have you photogrpahed with another long lens with better results, have you had some very sharp images with your 100-400, or are they ALL soft?

[Chris Ober]

Moved to photo gear forum

[Charlie Woodrich]

I had that problem and sent it to Canon. It was back focusing; they adjusted it, and it's great now.

[Larry Handal]

Mine is razor sharp. I check micro adjustment every once in a while though. If I have plenty of light I never hesitate to use it and found it at 400mm indistinguishable from my 400f5.6L which I sold shortly thereafter.

[David Billingsley]

Mine is also razor sharp - although I did send to to Canon when new for calibration.

[Ed Cordes]

My copy has excellent sharpness. This lens will never be as sharp as the 70-200 f2.8 or 400 F5.6. However, it is versatile and this makes up for the downsides. When I first got the lens in 2004 I had to send in under warranty to fix a noisy AF motor. Since then it has been in twice for cleaning after especially dusty Africa trips. Each time I asked that it be calibrated to exact specs. I now use a 7D and 1DMK4. With these bodies I have utilized the micro adjustments. I find the 100-400 to be an excellent zoom offering super tele capability in a "relatively" light weight "reasonably" inexpensive, manageable lens.

[Flavio Rose]

An amusing factoid: Lensrentals.com was selling four used copies of the 100-400L a few days ago. They rated the sharpness of each one (in the center wide open I suppose, can't guess the focal length). The ratings ranged from 14 lp/mm to 22 lp/mm and were all different.

[Larry Handal]

Both 400f2.8LIs they have for sale resolve 20/20 the same as the better 100-400s. That shows the quality that you can have if you have a good copy of this lens.

[Roger Clark]

Both 400f2.8LIs they have for sale resolve 20/20 the same as the better 100-400s. That shows the quality that you can have if you have a good copy of this lens.

I'm not sure what metric they use to get that resolution, but modern DSLR sensors resolve on the order of 4 times that level, so none of these lenses would be considered sharp. In contrast (pun), check the Canon MTF charts.
http://www.usa.canon.com/cusa/consum...ef_lens_lineup
There is a world of difference between the 100-400 and 400 f/2.8.

Roger

[Flavio Rose]

I'm not sure what metric they use to get that resolution, but modern DSLR sensors resolve on the order of 4 times that level, so none of these lenses would be considered sharp.

I was also not sure of what metric lensrentals.com uses, and tried asking Roger Cicala, who suggested it is like dxomark's but dividing the answer by two.

I would note that the MTF chart for the 100-400 on the Canon website you pointed to goes only to about 22 lp/mm (if lp/mm are the units on the horizontal axis).

A further small point: if pixel pitch is 5 micrometers, at 100 lp/mm (horizontally or vertically) the black lines are one pixel wide and the white lines are one pixel wide. In a typical testing scenario, however, there would be a random offset between the images of the black and white lines on the sensor array and the rows/columns of pixels. Thus, you can expect the black and white lines to each stand astride two rows/columns of pixels, and I don't think from any practical perspective much would be resolved at 100 lp/mm (remember there's also Bayer interpolation, noise, and issues arising from the exact color of the white line). I would say in practice that at most 50 lp/mm can be resolved meaningfully with a 5 micrometer pixel pitch. Still a far cry from 20 lp/mm.

Finally, as I understand it (and please correct me if I'm wrong), to turn an MTF chart into a lp/mm value, you need to choose a percentage or fraction transfer function value as your cutoff (as well as picking a particular curve on the MTF chart). Dxomark uses 20%. Not sure what lensrentals.com uses.

[Larry Handal]

Sorry I opened my mouth! LOL

The pix just looked the same when I shot both lenses side by side with same settings. Good enough for me.

[Roger Clark]

I would note that the MTF chart for the 100-400 on the Canon website you pointed to goes only to about 22 lp/mm (if lp/mm are the units on the horizontal axis).

Flavio,
The horizontal axis on Canon's MTF charts is distance from the lens axis (center of the sensor frame). Thus 22 mm corresponds to the image quality in the corner of a 35mm frame.

The thick lines are 10 line pairs/mm. The thin lines are 30 line pairs per mm. Black lines are lens wide open, blue are f/8. Solid lines are radial detail, dashed lines are perpendicular detail to the radial direction.

I mainly pay attention to the thin black lines as that represents the detail wide open. The distance between a set of solid and dashed black lines is bad (the the set of thin black lines). For example, the wider the separation, the more stars would be stretched into oblong blobs (and similarly destruction of fine detail in an image). The lower the curves, the worse the contrast of that fine detail with the perception of lower image quality.

Canon's MTF charts are not the norm, but are ingenious in the presentation and tell a lot more than a standard MTF chart.

A further small point: if pixel pitch is 5 micrometers, at 100 lp/mm (horizontally or vertically) the black lines are one pixel wide and the white lines are one pixel wide. In a typical testing scenario, however, there would be a random offset between the images of the black and white lines on the sensor array and the rows/columns of pixels. Thus, you can expect the black and white lines to each stand astride two rows/columns of pixels, and I don't think from any practical perspective much would be resolved at 100 lp/mm (remember there's also Bayer interpolation, noise, and issues arising from the exact color of the white line). I would say in practice that at most 50 lp/mm can be resolved meaningfully with a 5 micrometer pixel pitch. Still a far cry from 20 lp/mm.

Yes, I agree that alignment of the lines can be a factor, but statistically lines rarely line up with the sensor, both spatially, and angularly, so one actually gets close to the theoretical max. By this I mean the lines are usually rotated and offset a little. Lens distortion usually helps in this regard.

Finally, as I understand it (and please correct me if I'm wrong), to turn an MTF chart into a lp/mm value, you need to choose a percentage or fraction transfer function value as your cutoff (as well as picking a particular curve on the MTF chart). Dxomark uses 20%. Not sure what lensrentals.com uses.

Well, it is really a continuum and the finest details are also important. It is convenient for people to cite a single value. But if one deleted all the detail that had lower than 20% MTF, the image would be noticeable degraded in fine detail. What is actually important for image quality is the shape of the MTF curves. Some low MTF levels can be compensated by unsharp mask or real sharpening, but it is harder to compensate for some image destructive MTF shapes, like separated meridional and sagital response (the radial and perpendicular response).

Roger

[Jim Cash]

I have 100-400 that I was always frustrated with. I started looking real closely at the images and figured out that it was mostly the left side of horizontal images that was soft and the lower part of vertical images. I did some test shots to confirm this and found out that it was true.

I surmised that it was an alignment issue. I read on a forum somewhere on the internet that a number of these lenses have had issues with the alignment of the elements because of the unusual push-pull design which causes them to become misaligned after a while.

I sent the lens to Canon, explaining what seemed to be happening, and $350 and 2 weeks later (it was out of warranty, of course) I received what works like a brand new lens. Now my 100-400 is sharper at 300mm that my 300mm f4L. It is also sharper than my 70-200.

You might consider sending it to Canon. $350 may seem like a lot, but it's cheaper than a new lens.