Jump to content


Photo
- - - - -

Canon EOS-1D Mark III announced


  • Please log in to reply
73 replies to this topic

#21 segal3

segal3

    Powerful Sea Gull

  • Admin
  • 1739 posts
  • Gender:Male
  • Location:Los Angeles, CA

Posted 22 February 2007 - 05:01 PM

If I understand Canon's comments on this camera, all that is happening is more levels of gradation per color within the same dynamic range. This will decrease the need for interpolation or increase it's accuracy either in camera or in converter, but I don't see why it will cure an ill whose major cause is close dynamic parameters.

Check out page 12 of the linked White Paper for the 1D MkIII - it shows an example of a bride in a white gown, and the increase in highlight detail is remarkable when Highlight Tone Priority is used.
Matt Segal - carbonos scuba

#22 Rocha

Rocha

    Salty Dog

  • Senior Moderator
  • 3072 posts
  • Gender:Male
  • Location:California Academy of Sciences, San Francisco, CA

Posted 22 February 2007 - 05:56 PM

There are two problems related to photographing sunbursts in digital, one is blown highlights (due to low dynamic range) and the other is a bad gradation from highlights to shadows, eg. lines of sharp color difference instead of a smooth transition. If you look close enough (or in some real world examples) you will see what I mean. The 14-bit conversion will at least improve the transitions, and the high contrast option does increase the DR by one stop, but it's minimum ISO is 200...

Luiz Rocha - www.luizrocha.com
Nikon D800, Aquatica AD800, Ikelite strobes.


#23 herbko

herbko

    Herbzilla

  • Super Mod
  • 2128 posts
  • Gender:Male
  • Location:Northern California

Posted 22 February 2007 - 08:22 PM

Here's my guess at what the high contrast mode does:

1. meter exposure at ISO 200 setting
2. process the sensor output at ISO 100
3. apply a tone curve to bring the mid tones and shadows back up to ISO 200

This has the net effect of reducing blown highlights at the expense of more noise and banding at the mid tones and shadows. Since the sensor has very low noise to start and now has a 14-bit A/D, the higher noise probably will not be noticed. The two extra bits will give smooth tones to the under exposed and pushed up parts. This essentially the technique for getting a good sunball.

The dynamic range of the sensor is (maximum signal captured before clipping)/(noise level). The new 14-bit A/D does not change that. It just divides the signal into more levels and let you brighten the underexposed regions more and still have smooth tones.
Herb Ko http://herbko.net
Canon 5D; Aquatica housing; 2 Inon Z220 strobes; Canon 100mm macro, 17-40mm ; Sigma 15mm FE, 24mm macro, 50mm macro

#24 stone

stone

    Triggerfish

  • Member
  • Pip
  • 41 posts
  • Gender:Male
  • Location:Hong Kong

Posted 22 February 2007 - 08:29 PM

1.3 crop is a problem! Using FF fisheye like sigma & canon 15mm becomes less fishy. Anyone tried with tokina 10-17 FE on 1DII or 1DIIN?

Here is the link with tokina 10-17 on a 5D:
http://dc.watch.impr...11/17/5039.html

#25 drsteve

drsteve

    Manta Ray

  • Member
  • PipPipPip
  • 477 posts
  • Gender:Male
  • Location:Santa Barbara, CA

Posted 22 February 2007 - 09:20 PM

Here's my guess at what the high contrast mode does:

1. meter exposure at ISO 200 setting
2. process the sensor output at ISO 100
3. apply a tone curve to bring the mid tones and shadows back up to ISO 200

This has the net effect of reducing blown highlights at the expense of more noise and banding at the mid tones and shadows. Since the sensor has very low noise to start and now has a 14-bit A/D, the higher noise probably will not be noticed. The two extra bits will give smooth tones to the under exposed and pushed up parts. This essentially the technique for getting a good sunball.

The dynamic range of the sensor is (maximum signal captured before clipping)/(noise level). The new 14-bit A/D does not change that. It just divides the signal into more levels and let you brighten the underexposed regions more and still have smooth tones.


I think you are right about the high contrast mode, which essentially automates what photographers currently do. I know that when I am shooting a bright scene I apply an exposure compensation of -1 stop to prevent blowing the highlights and then add a contrast enhancing S curve after the fact.

Your comment about the dynamic range is also correct. I think that a lot of people think that dynamic range is the max/min signal (where min is 1 bit), which would be correct for a noise free sensor. In real sensors, the dynamic range is limited by the sensor noise, which is a combination of well depth and readout noise. Canon claims that the noise is lower due to their improved microarray which leads to better light gathering efficiency. However, this remains to be seen. I suspect that the noise performance will not be much different than the 5D which has pixels that are 8.2um. The new sensor has pixels that are 7.2um with the improved microarray. Therefore the wells are shallower, but get more light. It is probably a wash.

One place they might have made improvements is in the readout noise. I know that my 5D shows patterned noise in the shadows, which must be a flaw in the readout circuitry.

Edited by drsteve, 22 February 2007 - 09:21 PM.

...STeve
www.flickr.com/photos/drsteve/
Canon EOS 5D3, Aquatica Housing, Inon Z220+Z240 strobes, Sea&Sea TTL, ULCS Arms


#26 Glasseye Snapper

Glasseye Snapper

    Tiger Shark

  • Member
  • PipPipPipPip
  • 538 posts
  • Gender:Male
  • Location:Edmonton, Canada
  • Interests:Fish ID & behaviour and photos thereof

Posted 22 February 2007 - 09:36 PM

The 1D Mk X series cameras are overkill for underwater use -- no one needs 10fps underwater unless they're shooting action in available light!!


But with the improved low-light focusing sensitivity and supposedly low-noise high ISO settings this may be just the camera to get the most out of ambient light and filter photography. In particular, it may allow the use of somewhat longer focal length lenses to capture the action of fish behavior. I experimented with a 135mm F2 lens in January. Results weren't too good due to motion blur and too shallow depth of field but I wasn't clever enough to boost ISO (still had it set to 100). Because I did this on my last dive I couldn't see what ISO800 or 1600 would do to prevent motion blur and improve depth of field. I think it could work with a Canon 20D at shallow depth and, of course, a 5D or the new 1D Mark III should do a lot better.

Here is an example of a close-up of a longsnout seahorse shot with a magic filter at about 10-12m depth. I have to check the camera settings but probably shutter 1/120sec, aperture 2.8. This shot is still not really sharp and you do a lot better with flash but I think with a low-noise high-ISO camera and a bit more ambient light interesting opportunities exist.

LongsnoutSeahorse5813_1200.jpg

Longsnout seahorse, Bari reef, Bonaire
JPEG straight from camera, no cropping or adjustments (just rotated 90 degrees and reduced resolution)

Bart

Edited by Glasseye Snapper, 22 February 2007 - 09:44 PM.

Olympus OM-D EM5/Nauticam, 12-50mm & 60mm macro
Sea&Sea 110a, iTorch, GoPro3 BE

#27 bmyates

bmyates

    Great White

  • Team Wetpixel
  • 973 posts
  • Gender:Male
  • Location:Seattle, WA
  • Interests:UW Photography, motorcycles.

Posted 23 February 2007 - 08:49 AM

...I'm excited for the new 16-35 lens...

...The 1D Mk X series cameras are overkill for underwater use -- no one needs 10fps underwater unless they're shooting action in available light!!


Ditto both points.

I've got both a 16-35 and 17-40, and although they're both "good" lenses, I'd love a "great" wide angle zoom for FF!

I have a 1D MkIIN, and it's 8fps are great for topside action (e.g., moving cars, motorcycles, sports, etc.), but have little application underwater. As someone said, the 1D Mk X cameras are really designed for professional sports photographers. It will appeal to those who make their living shooting "on the margin" where even the slightest increase in FPS is of value. For the rest of us, the marginal utility (if you'll pardon the economics jargon) of an extra few fps just doesn't exist.

Bruce Yates
www.UnderwaterReflections.com
Canon 5DMkII in Aquatica, 1DsMkII in Seacam, G15 in RecSea...Inon Z240's...too many lenses
"If at first you don't succeed, try, try again. Then quit. No use being a damned fool about it." WC Fields


#28 davidrodkeller

davidrodkeller

    Lionfish

  • Member
  • PipPip
  • 74 posts

Posted 23 February 2007 - 03:45 PM

Check out page 12 of the linked White Paper for the 1D MkIII - it shows an example of a bride in a white gown, and the increase in highlight detail is remarkable when Highlight Tone Priority is used.

I read that, but I also read the sentence about increased noise and you cannot have both. Stretching left data and compressing right data will deliver better highlights but generates a noiser image. You can do that in post now.

Because they don't talk about actual capture advances, I suspect HTP is really just a metadata tag generated by the camera software rather than a true expansion of the camera's capture ability. It's probably a very nice adjustment for the given example but white dress highlights doesn't quite replicate the drastic contrast of an underwater sunball/sunburst.

Please keep in mind that my comment was in response to suggestions that 14-bit would be of some cure to sunball/sunburst ills. And there doesn't seem to be anything in the available literature to indicate that blooming and then the resulting visible fringing have been mitigated.
Thanks,
DRK

#29 davidrodkeller

davidrodkeller

    Lionfish

  • Member
  • PipPip
  • 74 posts

Posted 23 February 2007 - 04:49 PM

There are two problems related to photographing sunbursts in digital, one is blown highlights (due to low dynamic range) and the other is a bad gradation from highlights to shadows, eg. lines of sharp color difference instead of a smooth transition.

I understand why we get what we get, but here is my thinking: why would 14-bit processing eliminate or mitigate blooming and the resulting visible fringing? Or why better defined and gradated color fringing (which probably won't be the reality anyway) is an improvement. As we know better defined errors or failures are one of the pitfalls of digital capture.

I just notice that Canon has made no claim to the effect that the sensor has been enabled to bucket more charge. And that is what we are really looking for, at least those of us who do find sunballs/burst on current digital systems displeasing. I recognize that Herb, Paul and a few others disagreed with my earlier assertion that sunballs/bursts from current technology are bad (I may have said crap :) ).
Thanks,
DRK

#30 herbko

herbko

    Herbzilla

  • Super Mod
  • 2128 posts
  • Gender:Male
  • Location:Northern California

Posted 23 February 2007 - 11:56 PM

I understand why we get what we get, but here is my thinking: why would 14-bit processing eliminate or mitigate blooming and the resulting visible fringing? Or why better defined and gradated color fringing (which probably won't be the reality anyway) is an improvement. As we know better defined errors or failures are one of the pitfalls of digital capture.

I just notice that Canon has made no claim to the effect that the sensor has been enabled to bucket more charge. And that is what we are really looking for, at least those of us who do find sunballs/burst on current digital systems displeasing. I recognize that Herb, Paul and a few others disagreed with my earlier assertion that sunballs/bursts from current technology are bad (I may have said crap :) ).



It's often necessary to underexpose to keep from blowing out a sunball. The rest of the pic can be brought back up to a proper exposure at post processing. A low noise sensor gives you the good signal-to-noise necessary to push the underexposed water and subject a couple of stops and still not be noisy. The limit in the 5D in this process is not the noise of the sensor but quantization of the dark parts when brighten 2+ stops. Going from a 12-bit to a 14-bit A/D will make this quantization/banding problem much better.

In summary, it takes a high dynamic range sensor (large charge bucket and low noise) and fine quantization levels(more bits).
Herb Ko http://herbko.net
Canon 5D; Aquatica housing; 2 Inon Z220 strobes; Canon 100mm macro, 17-40mm ; Sigma 15mm FE, 24mm macro, 50mm macro

#31 dmoss

dmoss

    Sting Ray

  • Member
  • PipPipPip
  • 222 posts
  • Location:Too far from the ocean

Posted 24 February 2007 - 07:01 AM

I'm excited for the new 16-35 lens.


The lens does look interesting. Disappointing that it's designed with an 82mm filter thread. That means a new set of filters for topside and if anyone wants to experiment with diopters.....I'm not sure if there are any out there. Maybe the lens performance will be so good we won't need diopters.
David
Canon 5D, Ikelite


My Flickr Photos

#32 davidrodkeller

davidrodkeller

    Lionfish

  • Member
  • PipPip
  • 74 posts

Posted 24 February 2007 - 10:29 AM

It's often necessary to underexpose to keep from blowing out a sunball. The rest of the pic can be brought back up to a proper exposure at post processing.

I don't have a lot of hard fast rules when shooting digital, but one of them is that I don't underexpose by 2 stops for the purpose of bringing it back up in post. Expanding thin (left) data just doesn't work. My preferences is to always compress thin data and stretch fat (right) data. Plus my suspicion is that 14-bit processing will accentuate the problems inherent in stretching thin data. It's going to let you really see how little data was applied for the lower levels.

I agree with you that coupling greater than 12-bit processing with a sufficiently hi dynamic range sensor is the holy grail for these really high contrast shots, but in the case of sunbalss I don't see any benefit from 14-bit processing alone.
Thanks,
DRK

#33 Drew

Drew

    The Controller

  • Video Expert
  • 10595 posts
  • Gender:Male
  • Location:GPS is not reliable in South East Asian seas

Posted 24 February 2007 - 10:22 PM

That looks very promising! Although enabling it does constrain the ISO to 200+, which is not ideal for sunballs.

I'm excited for the new 16-35 lens. One of my friends has already promised to buy it, and I'll give it a test once he gets it (whenever it ships).

The 1D Mk X series cameras are overkill for underwater use -- no one needs 10fps underwater unless they're shooting action in available light!! But the new technology embedded in the camera is really exciting, and once it trickles out to the rest of their line I'm hopeful that we'll see marked improvements in image quality across the line-up.


Well I can think of at least 3 places where 10fps would be trés useful esp with a mustard filter :) But I'm not happy about the Li-Ion battery. It may save a few grams but environmentally and lifespan, it sucks.
At $3500 or less at street, it'll be interesting to say the least. I can imagine the 1DMk2N just dropping in value.

Drew
Moderator
"Journalism is what someone else does not want printed, everything else is public relations."

"I was born not knowing, and have only had a little time to change that here and there.


#34 herbko

herbko

    Herbzilla

  • Super Mod
  • 2128 posts
  • Gender:Male
  • Location:Northern California

Posted 24 February 2007 - 11:19 PM

I don't have a lot of hard fast rules when shooting digital, but one of them is that I don't underexpose by 2 stops for the purpose of bringing it back up in post. Expanding thin (left) data just doesn't work. My preferences is to always compress thin data and stretch fat (right) data. Plus my suspicion is that 14-bit processing will accentuate the problems inherent in stretching thin data. It's going to let you really see how little data was applied for the lower levels.

I agree with you that coupling greater than 12-bit processing with a sufficiently hi dynamic range sensor is the holy grail for these really high contrast shots, but in the case of sunbalss I don't see any benefit from 14-bit processing alone.


14-bit A/D does no good if the sensor it's sampling does not have enough dynamic range. I thought that was clear from my last message. I think you developed your rule from using sensors with insufficient dynamic range.
Herb Ko http://herbko.net
Canon 5D; Aquatica housing; 2 Inon Z220 strobes; Canon 100mm macro, 17-40mm ; Sigma 15mm FE, 24mm macro, 50mm macro

#35 StephenFrink

StephenFrink

    Great Hammerhead

  • Industry
  • PipPipPipPip
  • 711 posts
  • Gender:Male
  • Location:Key Largo, Florida Keys

Posted 25 February 2007 - 11:29 PM

The lens does look interesting. Disappointing that it's designed with an 82mm filter thread. That means a new set of filters for topside and if anyone wants to experiment with diopters.....I'm not sure if there are any out there. Maybe the lens performance will be so good we won't need diopters.


I'm actually glad the new 16-35 is a new filter size, and hope that means dramatic new design, rather than a subtle tweak of existing 16-35. I'm excited to try, for sure. Other thoughts:

1. Looking at the back I see it is substantially different (from robgalbraith.com). Seacam is making a MKIII housing, details to be determined shortly:

1880_mkiii_preview_01.jpg

2. The need for 10-frames per second comes up more often than I would have thought, and is nice to have. With the enhanced chip performance of new MKIII, this should be a very capable camera, even with the 1.3 crop.

T49P032D.jpg
Stephen Frink - www.stephenfrinkphoto.com
Publisher - Alert Diver Magazine
Distributor/North America - Seacamusa.com
Travel - Waterhousetours.com

#36 davidrodkeller

davidrodkeller

    Lionfish

  • Member
  • PipPip
  • 74 posts

Posted 26 February 2007 - 08:43 AM

14-bit A/D does no good if the sensor it's sampling does not have enough dynamic range. I thought that was clear from my last message.

Perhaps it wasn't you who said 14-bit will appreciably improve sunbursts. My apologies for the inaccurate cite.

I think you developed your rule from using sensors with insufficient dynamic range.

That specific rule or determination is the result of how the software applies the collected data. Your example of underexposing would be fine if the cameras applied data to lower levels using the same protocols they use for upper levels. And at the risk of stating the obvious: if the camera applies half of captured data to the highest exposure, half the remaining to the next highest, half the remaining to the next highest, and so on down the line it is clear that if we want to capture most image data at underexposed values we will not be working with anything close to clean or sufficient data. You just can't stretch that thin dirty data in post without ending up with very high levels of noise and other unpleasing results specific to digital capture. That's one of the problems with sunball shots, the sunball itself is using up so much of the total possible data that there is very little left for the more pleasing exposed areas of the image. Underexposing only exacerbates the problem.

But yes, it would be fair to conclude that many of my personal rules for digital capture are the result of the insufficient dr of cameras on the market as well as the data spread currently used. Your Canons and my Nikons are equally guilty of both. No current camera proportions data differently at this point in time, including the mkIII.
Thanks,
DRK

#37 Rocha

Rocha

    Salty Dog

  • Senior Moderator
  • 3072 posts
  • Gender:Male
  • Location:California Academy of Sciences, San Francisco, CA

Posted 26 February 2007 - 10:17 AM

I don't have a lot of hard fast rules when shooting digital, but one of them is that I don't underexpose by 2 stops for the purpose of bringing it back up in post. Expanding thin (left) data just doesn't work.


It doesn't work because there is very little information in the lower levels with a 12bit image, but this may change for the better with 14bit processing...

My preferences is to always compress thin data and stretch fat (right) data. Plus my suspicion is that 14-bit processing will accentuate the problems inherent in stretching thin data. It's going to let you really see how little data was applied for the lower levels.


I am not sure I agree with that, check the attached figure. The lower levels will have more information (and less banding after post-processing) in a 14bit image. High levels will hold better too.

14bit.jpg

So, this is how I see it (correct me if I am wrong): looking at the table above, say you expose the sun area correctly (exposure zone 1) and want to bring detail back to an area that was 4 stops underexposed. With the 12bit images you will have 256 levels available, with the 14bit image you will have 1024. The 14bit image will look better after post-processing, right?

Luiz Rocha - www.luizrocha.com
Nikon D800, Aquatica AD800, Ikelite strobes.


#38 herbko

herbko

    Herbzilla

  • Super Mod
  • 2128 posts
  • Gender:Male
  • Location:Northern California

Posted 26 February 2007 - 10:44 AM

So, this is how I see it (correct me if I am wrong): looking at the table above, say you expose the sun area correctly (exposure zone 1) and want to bring detail back to an area that was 4 stops underexposed. With the 12bit images you will have 256 levels available, with the 14bit image you will have 1024. The 14bit image will look better after post-processing, right?


Nice chart. That's what I've been trying to say.
Herb Ko http://herbko.net
Canon 5D; Aquatica housing; 2 Inon Z220 strobes; Canon 100mm macro, 17-40mm ; Sigma 15mm FE, 24mm macro, 50mm macro

#39 Rocha

Rocha

    Salty Dog

  • Senior Moderator
  • 3072 posts
  • Gender:Male
  • Location:California Academy of Sciences, San Francisco, CA

Posted 26 February 2007 - 03:51 PM

Not that I care anyways (max ISO that I regularly use is 400), but for those who do, here is a sample of the new Mk3 at ISO 6400:

http://www.imaging-r...3hSLI6400NR.HTM

Should look much better at 1600 and 3200...

Luiz Rocha - www.luizrocha.com
Nikon D800, Aquatica AD800, Ikelite strobes.


#40 segal3

segal3

    Powerful Sea Gull

  • Admin
  • 1739 posts
  • Gender:Male
  • Location:Los Angeles, CA

Posted 26 February 2007 - 03:55 PM

Should look much better at 1600 and 3200...

That already looks ridiculously good for 6400...yes, a fair amount of chroma noise, but the preserved detail in the scale on the right is excellent.
Matt Segal - carbonos scuba