7D: 20 shot flight series

[Doug Brown]

There is no doubt that these images suffered from the heavy crop and the resulting lack of pixels. I wasn't trying to illustrate how well the 7D images held up to a heavy crop. I was trying to show that it held focus for an extended period of time against 3 different BGs with a subject occupying a relatively small portion of the frame. To me this is the real test of the 7D's AF for BIF. Had the heron occupied 2/3 of the frame, I'm sure there would have been a lot more detail. But it would have flown across my field of view in about a second. For a test like this, you need the bird to be somewhat far away. When the birds are in close I normally fire off bursts of 2 or 3 frames, but this was a 20 frame sequence. I can show you plenty of 100% crops of flying birds that occupy a substantially bigger portion of the frame and have a lot more detail, but I'm not sure how useful images like that are when evaluating AI Servo AF.

While not tac-sharp by the frame-filling bird standard, the 2 images that I processed from this series look pretty good IMO. I'm going to post a couple of 100% crops of BIF where the bird occupies a large portion of the frame.

Here's a 100% crop of a flying gull; Lightroom to Photoshop with no editing whatsoever. Plenty of detail.

[Doug Brown]

And here's a 100% crop of a Forster's Tern also quite large in the frame.

[Charles Glatzer]

Here's what I mean when I use the term 'sharp' as opposed to 'tac-sharp.' IMO 'sharp' images are ones that look good after processing. In my workflow, soft and OOF images go in the dumpster, but if 'sharp' images have photographic merit, I'll process them. 'Tac-sharp' images are sharp from the unprocessed RAW stage onward. This is a processed version of frame 13, which I labeled as 'sharp.' I think it does pretty well with some NR and sharpening.

Doug,

With all due respect I do not see this as a sharp image.

Chas

[arash_hazeghi]

Doug,

With all due respect I do not see this as a sharp image.

Chas

Chas,

I think that the above samples are pretty darn sharp for a 1.6X high pixel density camera when inspected at pixel level. I guess you are comparing these with the output of 1D(S) MKIII bodies that you have and yes compared to large pixel cameras they have a certain "harshness" or lack of crispness, This has to do with the physics of the sensor with tight pixels and there is not much you can do about it. But the AF in the above examples seems to be spot on.

Here is a 100% crop from a larger pixel camera


100% crop from FF camera


From the earlier post it wasn't clear if the photos in the original 20-shot sequence had similar sharpness (not detail) compared to the ones above. (Sharpness and detail are not exactly the same).

[Robert Empleton]

Hi All,

What i need is a camera that produces clear and crisp images. Not one with another questionable AF and noise lurking where i don't need it. After the AF probs of the MkIII, it is now the noise of the 7D which makes me worried. I understand that these shots were taken in av mode which can explain the extra noise, but i'm affraid my patience is running out. My friend is using his new D3 at the Moto GP this weekend and i'm looking forward to his images.

Regards,

Robert

[Axel Hildebrandt]

Hi All,

What i need is a camera that produces clear and crisp images. Not one with another questionable AF and noise lurking where i don't need it. After the AF probs of the MkIII, it is now the noise of the 7D which makes me worried. I understand that these shots were taken in av mode which can explain the extra noise, but i'm affraid my patience is running out. My friend is using his new D3 at the Moto GP this weekend and i'm looking forward to his images.

Regards,

Robert

I don't think you can directly compare a 18MP 1.6x crop camera with a 12MP full frame camera and expect to see similar noise levels. Comparing a D3 with a 1DsIII and D300 with 7D would be closer in that respect.

[Doug Brown]

Given my distance to subject, I stand by my sharpness ratings.

[Doug Brown]

I find the noise to be entirely manageable. Remember that an 18 megapixel image when viewed at 100% requires the better part of four 30" Apple Cinema Displays to be seen in its entirety. When viewed at a more normal size, the noise becomes much less apparent.

[Joerg Rockenberger]

Remember that an 18 megapixel image when viewed at 100% requires the better part of four 30" Apple Cinema Displays to be seen in its entirety.

Wow. That certainly puts things into perspective. JR

[Manas Khan]

I guess, when we see the unprocessed output of a 7D we should always keep in mind that it's from a 18mp 1.6X crop sensor. They would never be comparable to unprocessed files from a sensor of larger pixels in terms of IQ as well as crispness. That makes even the PP workflow different. What matters is that how the processed (with noise reduction if needed) files from this camera compares to the processed output of other better (as one feels) cameras of the same league.

Another point is that, people who have not worked with 18mp image files might find it difficult to correlate the sharpness of a raw file (converted but not processed) to the final sharpness than can be taken out from it by pp. Perhaps that's the reason many people here who use 1.3X or FF body (with bigger pixel sensors) feel that the files marked as "sharp" and "tack sharp" by Doug as unusable.

To me, if a camera with tighter pixel density is capable of producing comparable images I would go for it. As it would give me more liberty on tighter crop. I have been using 20D + 100-400mm for bird photography and many a times feel that if I could have more pixels in my cropped images they would make large better prints. For example attaching a (purple) heron in flight image here.... this is a crop out of the 8.6 mp 20D image....



regards,
Manas

[Jim Fenton]

Doug...

I think I might need to re-think my comments regarding your sharpness ratings which I commented on yesterday as there is one thing which I hadn't considered:

Is there any level of in camera sharpening applied or an in RAW conversion application or are these presented at what would be the equivalent of my comvertiong a Nikon RAW file and turning off sharpening completely?

One comment I will make after looking at the Forster's Tern image is that it is lacking both in sharpness and detail to my eyes.

Does this mean that Chas and I are in agreement? :)

[Don Saunders]

Doug,

Your flight sequence of photos I found very positive. Based upon wanting to see consistent AF tracking across multiple busy backgrounds, the new AF performed much better than a XXD series body. I saw a post of BIF on another forum that were also excellent. The photographer and their technique does make a difference!

[Doug Brown]

There was no in-camera sharpening applied, nor was there any sharpening applied in post-processing (Lightroom or Photoshop). These are pure RAW images with JPEG conversion only. I think the Forster's Tern image illustrates perfectly why it's so hard to judge sharpness at 100% magnification with an 18 megapixel sensor. Have a look at the processed image. What do you think of sharpness now?

[Desmond Chan]

I think the Forster's Tern image illustrates perfectly why it's so hard to judge sharpness at 100% magnification with an 18 megapixel sensor.

I'm not sure I understand this statement. It seems to suggest that it is hard to judge sharpness because it is an 18 megapixel sensor. If so, it begs the question that if it's not an 18 megapixel sensor, then it is or may not be hard to judge sharpness at 100% magnification? If yes, why?

Have a look at the processed image. What do you think of sharpness now?

A photo that is not sharp originally can be made to appear so in post-processing. Showing it in smaller size also helps.

I still think it is better to compare sharpness using raw files from the same camera at 100% magnification.

[Robert Amoruso]

Doug,

As I have some interest in purchasing this camera, I am interested in seeing these results. I agree with Chas that displaying the AF points along with the image would have been helpful. Perhaps I missed it but how many AF points where being use and are one or all on the bird during the image shooting.

BTW, thanks for doing this, I am not trying to be argumentative; just inquisitive.

[Doug Brown]

I agree with you Desmond that if you're talking about evaluating the relative sharpness of 2 frames from the same camera body at 100% magnification. However, as I said a little earlier, a 7D RAW file requires just over two 30" Apple Cinema Displays in width and just over two of them in height to fit the entire image at 100% magnification. The displays have a native resolution of 2560x1600, and a 7D file has a resolution of 5184x3456. A Cinema Display measures 27" wide and 18" tall including the bezel. A 100% view of a 7D RAW file would be about 4 1/2 feet wide and over 3 feet tall. I view my monitor from a distance of about 16". Is that an appropriate viewing distance to evaluate the sharpness of an image the size of which I've just described? I would argue that it isn't.

When I view the unprocessed RAW file of the Tern in Lightroom 2 using the 'Fit' to screen size, the image measures 18" wide by 12" tall on my monitor. It looks razor sharp without any sharpening applied. The image also looks razor sharp in the 1:2 view which produces an image about 25" wide and almost 17" tall.

My question to you is this: is an unprocessed RAW file that appears tac-sharp at 25"x17" (but not at 54"x36") from a distance of 16" really sharp? I'm sure that for some the answer is no but for me the answer is a definite yes. I normally post at 1024 pixels wide and print at 13"x19". The Tern looks great at both of those sizes, and would probably look great at 24"x36". Any larger than that has no practical application in my photographic world.

[Mark Theriot]

Great series and serves exactly the purpose you intended to present.

I think your ratings are spot on considering settings and hope you continue to post more data with specific qualitative objectives. It helps me learn.

Mark

[Doug Brown]

Hi Robert. I was in AI Servo mode and was using focus point expansion with a diamond of 4 points (3, 6, 9, and 12 o'clock positions) around the central AF point. I had tracking sensitivity set to slow.

[Morkel Erasmus]

Doug your explanation makes great sense regarding sizing of images...looking forward to more of this discussion :)

[Doug Brown]

As a followup, I analyzed all 20 frames for whether the AF group was on the bird. There were 3 frames where the bird was not being hit by at least one of the AF points. Two of those 3 frames were OOF (frames 10 and 12). The other was frame 9, which I labeled as tac-sharp. Perhaps the camera was still holding focus because I had tracking sensitivity set to slow. In the rest of the frames I had at least one AF point on the bird. Here's what frame 12 looked like (too bad because I like this wing position the best :o).

[Doug Brown]

It also deserves mentioning that I paid a shutter speed price as the bird moved from the sky to the green BG; it dropped from 1/2000 to 1/800. That slowdown also impacted image sharpness.

[Jay Gould]

Hi Doug, you have done some wonderful extensive testing of the 7D - very impressive and very helpful. Would you post your C.Fn settings (especially III) either here or in the Gear section? I think a discussion regarding the various critical settings, using your settings as a base start, would be very interesting. Thanks,

[Doug Brown]

So in 2 out of 3 OOF frames, operator error is to blame :D:o (focus points not on bird and shutter speed a little too slow). The third OOF frame is sandwiched between the other two OOF frames, and I'm not sure that with tracking sensitivity set to slow the camera is allowed to adjust AF that quickly.

[Desmond Chan]

Thanks Doug for the clarification! That's what I was looking for.

A 100% view of a 7D RAW file would be about 4 1/2 feet wide and over 3 feet tall. I view my monitor from a distance of about 16". Is that an appropriate viewing distance to evaluate the sharpness of an image the size of which I've just described? I would argue that it isn't.

I would argue you can as long as you know what a sharp image looks like from that viewing distance based on your experiences with images that size viewed from that distance. Usually I judge if the image is sharp based on the part that the camera is focused on.

When I view the unprocessed RAW file of the Tern in Lightroom 2 using the 'Fit' to screen size, the image measures 18" wide by 12" tall on my monitor. It looks razor sharp without any sharpening applied. The image also looks razor sharp in the 1:2 view which produces an image about 25" wide and almost 17" tall...[snip]
My question to you is this: is an unprocessed RAW file that appears tac-sharp at 25"x17" (but not at 54"x36") from a distance of 16" really sharp? I'm sure that for some the answer is no but for me the answer is a definite yes.

No doubt about that. Sharpness is relative and subjective. I'd argue that two individuals viewing the same image from the same distance under the same conditions could still come up with two different conclusions regarding the sharpness.

[arash_hazeghi]

I agree with you Desmond that if you're talking about evaluating the relative sharpness of 2 frames from the same camera body at 100% magnification. However, as I said a little earlier, a 7D RAW file requires just over two 30" Apple Cinema Displays in width and just over two of them in height to fit the entire image at 100% magnification. The displays have a native resolution of 2560x1600, and a 7D file has a resolution of 5184x3456. A Cinema Display measures 27" wide and 18" tall including the bezel. A 100% view of a 7D RAW file would be about 4 1/2 feet wide and over 3 feet tall. I view my monitor from a distance of about 16". Is that an appropriate viewing distance to evaluate the sharpness of an image the size of which I've just described? I would argue that it isn't.

When I view the unprocessed RAW file of the Tern in Lightroom 2 using the 'Fit' to screen size, the image measures 18" wide by 12" tall on my monitor. It looks razor sharp without any sharpening applied. The image also looks razor sharp in the 1:2 view which produces an image about 25" wide and almost 17" tall.

My question to you is this: is an unprocessed RAW file that appears tac-sharp at 25"x17" (but not at 54"x36") from a distance of 16" really sharp? I'm sure that for some the answer is no but for me the answer is a definite yes. I normally post at 1024 pixels wide and print at 13"x19". The Tern looks great at both of those sizes, and would probably look great at 24"x36". Any larger than that has no practical application in my photographic world.

Hey Doug,

Thanks for reposting the sample with AF points superimposed, this is what I always do to make sure it was not my error :D

I agree with you that for practical purposes except maybe for printing you always downsample the images but I don't believe that is the right method to evaluate the AF performance on its own as opposed to general camera performance. Consider this example, you have two cameras A and B with the same AF sensor A is 6 mpixels and B is 18. If you shoot an identical sequence with both and then downsample B to match the size of A the slightly OOF shots will appear what you call "tac sharp" when compared to the 6 mpixel camera. In this case do you think it is a fair statement to say camera B had better AF than camera A?

This is very similar to evaluating lens sharpness which is always done at 100% regardless of output size, do you ever downsample your photos when you evaluate lens sharpness? If so we have to accept that all lenses suddenly become sharper with a higher mpixel camera.

[arash_hazeghi]

Doug,
One more question, now that you have had the 7D for a while, how many full blown AF shots do you get in normal tracking sequences? The slightly OOF shots and tac-sharp vs less sharp discussions aside, do you honestly not get frames that focus sensor is right on the bird and yet AF is full blown like this?




Thank you.

[Doug Brown]

Hey Desmond. I'm not saying that I can't tell whether a 7D image is sharp when viewed at 100%. What I'm trying to say is that if it looks slightly soft at 100% (4 1/2 feet by 3 feet) from a distance of 16" but it looks sharp at 36" x 24" (or whatever the maximum output size I prefer), then what practical relevance does the softness at 100% have? IMO it has essentially no relevance other than to the pixel peeper lurking inside of me. And if I were printing it at 4 1/2 feet by 3 feet, I suspect it would look quite sharp at an appropriate viewing distance (no less than several feet away).

[Doug Brown]

Hi Arash. Of course I get frames like yours. We all do! It's just hard to interpret the significance of a single frame with a red dot on the bird. Did you have sufficient shutter speed? Was your focused locked at the time of capture? What was your AI Servo tracking sensitivity set to? Did you have IS on or off? Where was that red dot relative to the bird a second or two before you captured the image? When I look at my 7D image files, there are a handful of frames that I feel should have been in focus but are not; I've never owned a camera where that wasn't the case. Overall though, I'm very pleased with both my percentage of keepers and the quality of the images.

[Robert Empleton]

G'day Doug,

It has now become a whole lot clearer that the 7D could be a great camera. And i think that if we only look at the end result, it becomes an amazing camera. It's just the fact that the original size is so huge, that it becomes difficult to see it's true potential.
I'm looking forward to see more positive results.

Regards,

Robert

[arash_hazeghi]

Hi Arash. Of course I get frames like yours. We all do! It's just hard to interpret the significance of a single frame with a red dot on the bird. Did you have sufficient shutter speed? Was your focused locked at the time of capture? What was your AI Servo tracking sensitivity set to? Did you have IS on or off? Where was that red dot relative to the bird a second or two before you captured the image? When I look at my 7D image files, there are a handful of frames that I feel should have been in focus but are not; I've never owned a camera where that wasn't the case. Overall though, I'm very pleased with both my percentage of keepers and the quality of the images.

Thanks Doug,
This one was in single AF mode (no expansion) with tracking sensitivity set to slow 1/1600sec ISO 400. The focus was locked at the perched hawk, when hawk started flying the camera initially tracked for 1-2 frames but the next 10 frames which contained the peak of flight action were all like this. This was with 400 f/5.6 so no IS.

From yours and others' samples and even a few of my own, we all know that it is possible to produce tac sharp and high quality flight shots with the 7D. However, the question is what is the probability of producing an acceptably sharp photos with a 7D compared to its predecessors (40/50D) and other cameras such as 1D(S)MKIII given the same conditions. How much the ratio of sharp over OOF shots has really improved with the new AF system without considering the extra benefit you get in the larger number of pixels. In my experience, with raptors and conditions that I had described earlier this ratio was noticeably lower than 1D series and Nikon cameras, but my camera might have been defective or out of calibration.

After comparing your results from 7D with that of your 1DMKIII and 50D can you confidently state that on average, it produces a higher ratio of "acceptably sharp" photos when viewed at 100% ?

[Doug Brown]

After comparing your results from 7D with that of your 1DMKIII and 50D can you confidently state that on average, it produces a higher ratio of "acceptably sharp" photos when viewed at 100% ?

I don't have enough experience with the body to make claims like that. My initial impression is that it acquires focus as quickly as and tracks better than either the 50D or the 1D Mark III. I only have a handful of series where I took a burst of more than 2 or 3 frames. I'll next post a burst of 6 frames of a Belted Kingfisher taken over about 1 second. With the exception of the first frame, where the bird was slightly below and to the right of the focus point grouping, I kept the bird pretty close to the center of the frame.

[arash_hazeghi]

Another point is that if you use 4-point expansion mode, DPP software only shows the center of the expansion group that was chosen, even if the camera used one of the 4 expansion points. In expansion mode, when no AF point is displayed in red in DPP, it means that the camera did not pick anything.




In zone AF or all-point AF DPP does show the point(s) that acquired focus.

So when looking at photos taken in expansion mode, if any of the 4 points from the expansion diamond around the selected red point is on the subject, the camera should have theoretically grabbed the subject.

[Doug Brown]

Here's a full frame version of the image to give you an idea of how large the bird was in the frame. The camera techs were: 500mm + 1.4x, f/6.3, 1/5000, ISO 800, hand held.

[Doug Brown]

Frame 1: only a very small portion of the bird (the tip of the tail feathers) is in the AF area

[Doug Brown]

Frame 2: doing a little better with placement of the bird in the AF zone

[Doug Brown]

Frame 3: a little low but still in the zone

[Doug Brown]

Frame 5: still holding focus

[Doug Brown]

Frame 7: subject is slightly to the left of center but still in the AF zone

[Doug Brown]

Frame 7: bird now to the right of center but still in the AF zone

[Desmond Chan]

Hey Desmond. I'm not saying that I can't tell whether a 7D image is sharp when viewed at 100%. What I'm trying to say is that if it looks slightly soft at 100% (4 1/2 feet by 3 feet) but it looks sharp at 36" x 24" (or whatever the maximum output size I prefer), then what practical relevance does the softness at 100% have? IMO it has essentially no relevance other than to the pixel peeper lurking inside of me. And if I were printing it at 4 1/2 feet by 3 feet, I suspect it would look quite sharp at an appropriate viewing distance (no less than several feet away).

Agreed, Doug !

[Doug Brown]

The bird was quite small in the frame, so there's not a lot of detail. But I think the camera held focus quite well. This series was done with the same AF settings as the Green Heron series.

[Doug Brown]

Were you using Spot AF?

[arash_hazeghi]

Thanks Doug for posting this, let's go back to frame #3 I guess it has the following merits:

1) This would be the keeper of this series due to wing position :D
2) It is not the first or 2nd frame so there was enough time for the servo mechanism to stabilize.
3) The King is covering an entire AF area.

I am also making the following assumptions, please correct me if I am wrong:

These were shot in JPEG mode and therefore some plumage detail is lost to NR.
Comparing the size of AF frames with the 100% crops I posted these have been downsampled by a factor of 2. So the Kingfisher was originally ~1200 pixels wide.

I guess this is where are differences come from

I agree that none of these is a fully blown AF like the one I posted but again these are against a plain blue BG and I expect a modern DSLR to easliy lock focus in these conditions, it is true that the kingfisher is flying at high speed, but the relative tangent of the tracking angle does not change by much so the camera doesn't need to vary focus distance rapidly.

Although the focus is locked in these, even after downsampling by a factor of 2 the photo lacks the kind of detail that you would expect. I would expect to get a good 1024 pixel web keeper from a photo in which the bird itself is ~1200 pixels wide, in this case it seems that AF although not fully blown did not have the precision to deliver a final 1024 pixel web keeper.

[arash_hazeghi]

Were you using Spot AF?

No, I used expansion AF for sky shots (as shown in the Kestrel sample) and single AF point (not spot) for foliage shots.

[arash_hazeghi]

For better or worse I usually have good luck tracking small birds against busy BGs, here is an example of a Curlew which is ~ 2X smaller in the frame than the Kingfisher example, I can still get a small web keeper out of this by including some habitat and cleaning up the noise, given that this was with a 50D, I think it is reasonable to expect 7D to deliver equal or better results.




50D + 400 f/5.6 ISO 400

[Doug Brown]

Here's what I got out of this frame, which was shot in the RAW format. The frame I posted with the focus points displayed was a screen capture from DPP; they have a softer look than the same images viewed in Lightroom. I would call this image suitable for web posting, but nothing much bigger than that. Pretty good detail in the processed image.

[arash_hazeghi]

yup this one sure looks good.

I look forward to hearing your final verdict when you get a better feel to compare keeper ratio to 50D and 1D(S)MKIII. The key point here I guess is how much the new AF system has really improved.

[john crookes]

When you send the photo from lightroom to photoshop are you using the default settings in lightroom or are you shutting down like this to include zero contrast linear curves and no noise reduction

if you are at lightroom defaults you are adding some sharpening

[Rich Williams]

Someone mentioned this earlier and I will say it again. Image sharpness is highly subjective. One only has to look through a few images in this forum and read the comments to realize this. I do think that this post does show what it intended to show and that is the auto focus on the 7D is not bad at all. This comes from a "totally objective" Nikon user.

[Jay Gould]

Doug, when you indicate that the images were imported into LR with no processing and then exported to PS for sizing did you turn off "Detail" in the develop module BEFORE you did the import?

LR automatically applies some sharpening to every image through the Detail module.

I have turned this off by default so that the first round of sharpening is by my choice and is generally when I export to PS and apply Pixel Genius Capture Sharpening after applying Topaz Denoise. The Topaz Denoise video tutorials suggest that applying Denoise should be the first step in the workflow process.

Thoughts?