1.4X Teleconverter Soft images in AI Servo

[Ollokot]

I was wondering if anyone had come across this before. Using a Canon 400mm Lens with a Canon 1.4X Mark III Teleconverter the images are fine and sharp in One Shot AF operation but are markedly softer in AI Servo AF operation. Take off the Teleconverter and both One Shot AF and AI Servo AF are fine and sharp. Tested it with a different lens, the 100-400 Mark II and with a different camera and results are the same.
Is my Teleconverter faulty.
Thank's for your time.
Pat.

[Manfred M]

Which camera body? If a DSLR, have you tried this with Live-View focus as well as using the viewfinder?

As a general rule, if you are getting a sharp result in one focus mode, the hardware (optical components) are fine but there is likely a calibration issues with a DSLR. Live View should be sharp as the focus is set using the sensor rather than the phase detect autofocus mechanism. If you are having this problem with a mirrorless camera, then some of my assumptions could be incorrect.

Long lenses have a history of getting the focus points "wrong", so what happens if you try this same test using a single focus point? Ideally all this testing should be done using a subject that does not move and gives a clean target to the autofocus mechanism, with the camera and lens set on a sturdy tripod.

[Geoff F]

What about manual focus?

I once tried a converter with my Sigma 150-600 Sport lens and it gave unreliable results with AF although it worked better with manual focus. Same thing occurred with the older 150-500 lens.

[DanK]

What camera? What AF-point selection (single point, etc.)? What's the EV level? For the last, you could give the f/stop, ISO, and shutter speed.

I'm not sure why a difference between focus modes (one-shot vs. servo) would indicate a calibration error because on a canon DSLR, the camera uses the same phase detection system regardless, as long as you aren't using live view.

[Ken MT]

I tried replicating with a 100-400 mm, 1.4 mk3 convertor with one shot and AI focus, on a lens test chart, inside to prevent wind shift, horizontal to the camera (canon 5D Mk4). 1/640 sec, f8.0 spot metering, 2 sec delay, 560mm, ISO 5000. Imported as RAW to photoshop, and viewed at 200x, but couldn't see any significant difference in sharpness. Slight difference in brightness only.

[Ollokot]

Thank you for your replies. Recently I went out into the field to photograph birds after a long layoff due to ill health and was disappointed with images taken using AI Servo hence the test.
My first test was done using two different camera's, Canon 1DX MKII and Canon 5D MKIV with the intention to rule out any problem there might be with the 1DX MKII that I used in the field. Lens used was the Canon 400mm f/5.6 L. Inside, gear on a tripod and settings of 0.5 of a second f/8 and ISO 400, remote cable release used with mirror lockup and 10 second interval between mirror up and shutter release. The images were of a static subject and One Shot mode from both camera's were fine but images in AI Servo where very markedly softer.
I have since done further tests with a static subject taken outside using the Canon 1DX MKII and the 400mm f/5.6 L lens mounted on a tripod, settings 1/1000 sec f/8 ISO 2000 and both One Shot and AI Servo are Fine with and without the converter.
Then one more test inside with and without the converter settings of 0.5 of a second f/8 ISO 400 remote cable release used with mirror lockup and 10 second interval between mirror up and shutter release. Both One Shot and AI Servo Fine.
Am now left pondering and some bit frustrated as to what went wrong out in the field and with the first test.
Here's hoping it doesn't occur again.

[Manfred M]

It is always useful to post some of the captures so that the viewers can get a better idea as to what might be going on. The old adage of a picture being worth 1000 words is definitely true.

The information you have provided gives another clue. A lot of DSLRs will not work on lenses that have a maximum aperture that is smaller than f/5.6. By adding a 1.4x teleconvertor to an f/5.6 you have a native f/8 lens, so this could be contributing to your problem. I have been told that Canon was particularly susceptible to this, but that was some time ago and I have no direct knowledge of this issue.

The other issue is one that I have run into, shooting static birds. Multi-point autofocus can get confused by other material in the scene. Rather than focusing on the bird, the camera can decide that your subject is actually the twig sticking out beside the bird. With the perspective flattening we get with a long lens, this can be especially true as many parts of a busy scene could be "the subject". Switching to a single focus point is an effective workaround there as you tell the camera where to focus.

[DanK]

The extra information is helpful.

I don't know about the 1D, but in the case of the 5D IV, many of the AF points work at f/8. See p. 113 of the manual.

You didn't say what AF point area selection you used. This is of course separate from the mode you select (one shot, servo, AI focus). See p. 112 of the 5D manual, which warns that when AF area is set to zone or automatic, servo may try to switch focus and may fail to focus entirely under some circumstances. This is why I suggested doing your test with single-point or spot AF area. I don't know how multi-point selections work under those conditions.

[Ollokot]

Many thanks for your thoughts.
The centre AF point, this being the most accurate was used doing the tests.The camera and lens combination according to the manual are compatible using the 1.4X MK III Teleconverter at f/8.
Attached are images of the test, this image is the first test where the problem can clearly be seen.
Here is a crop of the images from the first test side by side.

First Test.jpg
One Shot AF to the left AI Servo to the right

Attachment 38015
One Shot AF to the left AI Servo to the right, right side moved slightly donward's so as you can view any difference more clearly.
Don't know why but unable to upload second image, will try in a separate reply.

[Ollokot]

Second Test.jpg
One Shot AF to the Left AI Servo to the right.

[DanK]

With that additional information, I'm stumped.

From what you posted, the main difference between the two tests was shutter speed. Since the blurry image was the longer shutter speed, slow AF can't be the culprit.

It's very hard to see what's going on with such a tiny posted image (can you use a link to post to a larger one), but I do see one thing that strikes me as weird. Look at the noise in the top band of the right half in test #1. The noise is more blurred near the center. To a lesser degree, that's true further down in the image as well. Since the noise is generated by the internals of the camera--the sensor and the circuitry--that shouldn't be so, if AF is the problem. Or am I off base?

[Manfred M]

With that additional information, I'm stumped.

From what you posted, the main difference between the two tests was shutter speed. Since the blurry image was the longer shutter speed, slow AF can't be the culprit.

It's very hard to see what's going on with such a tiny posted image (can you use a link to post to a larger one), but I do see one thing that strikes me as weird. Look at the noise in the top band of the right half in test #1. The noise is more blurred near the center. To a lesser degree, that's true further down in the image as well. Since the noise is generated by the internals of the camera--the sensor and the circuitry--that shouldn't be so, if AF is the problem. Or am I off base?

Dan - I don't think that we are looking at noise, but rather roughness of the surface on the building, like a stucco, with the painted sign.

In AI-Servo, I don't think the camera needs to acquire a focus lock before taking the picture (or at least that can be disabled). Given the cable release, tripod, mirror lockup, etc. I wonder if my initial thought about focus not being acquired properly might not be the issue here? I use Nikon's equivalent (with back-button focus)in virtually 100% of my work and find that I can miss focus from time to time.

[DanK]

Manfred,

Ah, I hadn't thought about that because this is a setting I don't change. Canon DSLRs allow you to adjust the extent to which the camera prioritizes focus rather than prompt shutter release (pp. 136 and 140 in the 5D IV manual). That could do it. I would suggest setting the AI Servo all the way to focus priority if it's not there already.

Dan

[Ollokot]

Manfred,
Yes it is the surface of what is a spreadable butter carton and not noise that you are seeing,
Dan,
The AI servo is set all the way to focus priority.
I keep wondering if some camera shake could have been introduced or focus wasn't achieved although it was confirmed in the viewfinder, but can't see how.
Just have to put it down to one of those oddities and hope it doesn't occur again.

Pat

[DanK]

I did wonder about motion blur. With a focal length that long, it doesn’t take much.


Sent from my iPad using Tapatalk

[William W]

. . . Recently I went out into the field to photograph birds after a long layoff due to ill health and was disappointed with images taken using AI Servo hence the test.

My first test was done using two different camera's, Canon 1DX MKII and Canon 5D MKIV with the intention to rule out any problem there might be with the 1DX MKII that I used in the field. Lens used was the Canon 400mm f/5.6 L.

Inside, gear on a tripod and settings of 0.5 of a second f/8 and ISO 400, remote cable release used with mirror lockup and 10 second interval between mirror up and shutter release. The images were of a static subject and One Shot mode from both camera's were fine but images in AI Servo where very markedly softer.

I have since done further tests with a static subject taken outside using the Canon 1DX MKII and the 400mm f/5.6 L lens mounted on a tripod, settings 1/1000 sec f/8 ISO 2000 and both One Shot and AI Servo are Fine with and without the converter.

Then one more test inside [assumed camera on a tripod] with and without the converter settings of 0.5 of a second f/8 ISO 400 remote cable release used with mirror lockup [hence assumed Static Subject] and 10 second interval between mirror up and shutter release. Both One Shot and AI Servo Fine.

Am now left pondering and some bit frustrated as to what went wrong out in the field and with the first test.

Hi,

I am commenting only on the above testing procedure.

I believe all three tests are with bias and flawed.

I have not found it explicitly stated, though it is numerously implied: Canon EOS, "AI Servo" function should not be used with a Static Subject - and - Camera on Tripod.

A colleague (in the USA) and I tested this extensively around 2008, using five different EOS Cameras, all mounted on a tripod and shooting a static test target, placed at 45 degrees to the lens's axis.

I used 135L/2 and 70 to 200/2.8L (i.e. both sharp lenses) and Michael used 70 to 200/2.8L and (I think) 85/1.8.

The test comprised > 1000 images, each.

The results showed front or rear focusing (i.e. inaccurate Auto Focus) just a bit greater than 20% of the total number of images taken in "AI Servo" Mode.

"One Shot" Mode images were accurate at a rate just slightly greater than 98%.

A result of those tests lead us to an hypothesis that the EOS AI Servo sometimes 'hunts' for movement, when there is none; i.e. neither a movement caused by the hand holding of the camera nor caused by a mobile subject.

Whilst the many of the technologies of EOS Camera have developed since 2008, I would bet a Mars Bar that the basic functionality of "AI Servo" has not changed.

WW

[DanK]

Bill,

Thanks. Very helpful. It makes sense that this mode would be programmed to search for motion, since its purpose is to track focus points.

On a vaguely related note: some people use servo and back button focus to try to combine the benefits of servo and one-shot mode. The argument is that you can simply release the back button when you want servo to stop operating. I've tried it, and I find it difficult because I'm never quite sure where the focus point is. For the same reason, when the slight noise isn't an issue, I typically leave the confirmation beep turned on when I am using one-shot mode.

Dan

[Ollokot]

Bill,
Thank you for your very informative reply. Front or rear focusing was playing on my mind and I was even contemplating AF Microadjustment but with One Shot mode showing accurate focusing decided against it.

Pat.