1DS Mark III - Lens Resolving Concerns

[Mark D Segal]

Speaking as a 1DSMkI owner who certainly can see optical problems on many 'L' series lenses using a MkI, I think that this is one of the most pertinent questions that needs answering. [a href=\"index.php?act=findpost&pid=137082\"][{POST_SNAPBACK}][/a]

I also use the 11.1 MP 1Ds, and I also know that on a 1Ds there is visible sample variability within L lens models. This means the lens is already the limiting factor of image sharpness, in terms of the resolving power the optics can deliver. Hence if the lens is the binding constraint, whether the sensor has 11 MP or 22 MP won't change the optical sharpness of the image at any given resolution (defined as pixels per inch). What will change dramatically however between a 1Ds and a 1Ds MkIII is the resolution itself - by about 38%. This makes for a tremendous improvement in high-resolution cropping capability.

[Ray]

One question appear to me.

If "the resolving power of current L series lenses are not enough to cope with the increased resolution of the new 1DS Mark III. There has been talk for a long time that the resolving power of current L lenses already struggles to keep up with the 1DS Mark II sensor" is ture then, how do we expect medium format lens to keep up with 22MP-39MP digital back?   
As we know 135 lens should have higher lens resolution than medium format lens. 
[a href=\"index.php?act=findpost&pid=134543\"][{POST_SNAPBACK}][/a]

Jeff,
The new Digitar lenses with reduced image circle suitable for the cropped format MFDBs seem to be better than most 35mm lenses. In fact, some of them appear to be on a par with the Zuiko lenses for the 4/3rds format, if you can believe the MTF charts.

If you are using a lens which is at least as good as any 35mm lens, with a sensor which is larger than 35mm format but which has a similar pixel count, then the larger sensor will produce the sharper image. Ie. the larger sensor is less demanding on the lens and requires a slightly lower lp/mm figure, which in turn translates to a higher MTF of captured detail.

For a 22mp 35mm DSLR to match the quality of a 22mp MFDB with a sensor double the size, the 35mm lens needs to be better than the MF lens. Some 35mm lenses are better. However, the best of the new digital lenses for MFDBs are very expensive and do appear to be better than most 35mm lenses.

If you also take into consideration the smaller apertures required for same DoF, then these Digitar lenses probably have an additional advantage. For example, if you want a shallow Dof then you might be using f4 with a 75mm lens, with a 48x36" sensor, as opposed to f2.8 with a 50mm lens on FF 35mm.

With most 35mm lenses, sharpness diminishes at wide apertures. There are a few exceptions but it's generally true. My Canon 50/1.8 and 50/1.4 are not particularly sharp at f2.8 and probably significantly less sharp than a 75mm Digitar lens at f4 (if there is a 75mm Digitar prime, if not then the nearest option).

[Ken R]

I think the problem is blown out of proportion.

The resolution difference between the 1Ds MkII and MkIII is:

4992x3328 vs. 5616x3744

That's a paper increase in angular resolution of ... hold on to your knickers! ... 12.5%.

My prediction is that the buyers/testers will see a slight increase in perceived resolution (though not 12.5%), and perhaps a very slight degradation in perceived per-pixel quality for some lenses, but there will be no disasters.

There are other and better reasons to purchase the 1Ds MkIII than resolution, and worrying about the camera "out-resolving" the lenses seems comparatively unimportant to the very tempting improvements between the models.

Disclaimer: I don't own a 1Ds (I or II), nor have I used any for a significant period of time.
[a href=\"index.php?act=findpost&pid=134338\"][{POST_SNAPBACK}][/a]

Exactly, some people get confused, they say "12mp is double 6mp" but thats not the case exactly, since 6MP is 3000x2000 pixels and 12MP isnt 6000x4000, that would be 24mp obviously but all pixels arent equal and the new 1Ds mk3 should have very high quality pixels at that so the increase in pixels is nice but if the pixel quality improves on top of that then you start to see significant improvements in IQ.

IMHO for every 800 pixels or so added in both directions of a sensor you start seeing increased detail easily. 600, yes maybe a little no question, 400 humm not much...

[vgogolak]

One question appear to me.

If "the resolving power of current L series lenses are not enough to cope with the increased resolution of the new 1DS Mark III. There has been talk for a long time that the resolving power of current L lenses already struggles to keep up with the 1DS Mark II sensor" is ture then, how do we expect medium format lens to keep up with 22MP-39MP digital back?  
As we know 135 lens should have higher lens resolution than medium format lens. 

Jeff
[a href=\"index.php?act=findpost&pid=134543\"][{POST_SNAPBACK}][/a]

Not so sure; I have Leica glass, R and M and Lots of Contax 645 and hasselblad.

In general I find the MF kenses do as well or better than the 135 set up
Here is an expl from a recent euro shoot. I am quite pleased with the ctr and edge performance of this 'discontinued' glass (one that MR seemed to think was not up to the P45!  :-))

it is a random 80mm 2.0 lens (I used 35, 45, 140 and 210 and they all performed well). Lots of even better examples. Here the focus is maybe a bit different ctr to edge as well.

anyway, this could filla billbd (with GF of course)

Victor

[Marc Schultz]

I think the comparison on the difference between the 1DS MKII and MKIII in terms of resolution being only 12.5% at 4992x3328 for the MKII vs. 5616x3744 for the MKIII is an interesting point for consideration.

It seems like there will be a bit of noticeable added detail if one were to upgrade from 16MP to 21MP, but the bigger reason for one to jump from a MKII to a MKIII in my opinion would be if it adds another 30% cropping capability, which it won't since the jump in resolution is not that great. And as it is now, people rarely have to push the print size capability limits of a MKII for 90% of all print applications either so more net pixels are not really in much demand at this point.

So again, the 2 main reasons to step up from a MKII to a MKIII in terms of added resolution are: 1 - Added cropping capabilities. 2 - Added visible detail.

In both cases though I believe neither will be a significant enough increase to warrant an upgrade. If though you are sitting there with an 11MP 1DS then the upgrade to a MKIII seems much more warranted.

[John Sheehy]

Add me to the list of those more excited by the 14-bits than the 21MP. Sure, I'm happy to get the extra MP, but the extended tonal range will have more impact over the current 1Ds2 images, IMHO.[a href=\"index.php?act=findpost&pid=134422\"][{POST_SNAPBACK}][/a]

There is no extended tonal range in either of the two 14-bit cameras Canon has already released.  What makes you think the 1Dsmk3 will be any different?

[Mark D Segal]

What is the definition of "extended tonal range"? If you mean blacker blacks and whiter whites I would also be surprised to see any difference. But bit depth is not so much about that - it is more about the number of tonal levels between black and white. The difference between 12 bit and 14 bit is exponential - specifically 2^14 rather than 2^12, or 16,384 levels instead of 4,096. IN PRINCIPLE, this difference should provide a higher quality raw file (less filling to do by the raw converter graduating from 14 bit to 15+1 compared with graduating from 12 bit to 15+1). Before passing judgment on it, one would need controlled experiments to see on paper the extent of any visible image quality improvement.

[D White]

If one is expecting Hasselblad "V" series with Zeiss lenses to please them more than a Canon DsII, they will likely be disappointed. Having a DsII with 12, (mostly L-series), lenses and a Blad with 7 lenses from 40 to 500 APO, (including  the superachromat), it is disappointingly evident how significant the aberrations are in the Zeiss optics. Vastly more chromatic problems are seen on film with the Zeiss than Canon lenses on a more demanding digital sensor. It would not be worth me acquiring a digital back for my Blad with the limitations of the optics. I wish I was wrong.

As far as whether it is worth upgrading to a mark III, with people expressing concerns over the ability of the current Canon optics to do justice to the new sensor, the sum of the parts is greater than any one advancement. The complete package of more photo sites, improved AD conversion, dust removal, etc, is sure to take image quality to new levels regardless of the exact lens. The improvements in the AD converter alone are likely why some feel the 5D is better than the DsII.
I would just go for the dust removal alone if it actually works.

More photo sites have an exponential benefit in less up-ressing needed for large output.

So I ordered a DsIII the very day it was announced.

[budjames]

I'm looking forward to the 1Ds MkIII for a lot of reasons already mentioned here, but after using my new Canon 40D for the past week, the larger LCD screen is on the top of my list.

I had a hard time viewing images and menus on the smaller screen of my 1Ds MkII and 20D bodies.

Bud James
North Wales, PA

[David Anderson]

I've ordered as well, can't wait to see what the II can do with lenses like the 85 1.2 and 135 f2.

 

[Craig Lamson]

If one is expecting Hasselblad "V" series with Zeiss lenses to please them more than a Canon DsII, they will likely be disappointed. Having a DsII with 12, (mostly L-series), lenses and a Blad with 7 lenses from 40 to 500 APO, (including  the superachromat), it is disappointingly evident how significant the aberrations are in the Zeiss optics. Vastly more chromatic problems are seen on film with the Zeiss than Canon lenses on a more demanding digital sensor. It would not be worth me acquiring a digital back for my Blad with the limitations of the optics. I wish I was wrong.

As far as whether it is worth upgrading to a mark III, with people expressing concerns over the ability of the current Canon optics to do justice to the new sensor, the sum of the parts is greater than any one advancement. The complete package of more photo sites, improved AD conversion, dust removal, etc, is sure to take image quality to new levels regardless of the exact lens. The improvements in the AD converter alone are likely why some feel the 5D is better than the DsII.
I would just go for the dust removal alone if it actually works.

More photo sites have an exponential benefit in less up-ressing needed for large output.

So I ordered a DsIII the very day it was announced.
[a href=\"index.php?act=findpost&pid=138126\"][{POST_SNAPBACK}][/a]

You should mount your Hassy on a Canon and compare.  They Zeiss glass falls down when compared to some of the Canon glass.   I tried my older CF 50mm, 100mm and 150 and compared them to the Canon 24-70, the 90-tse, 70-200 f4 and an old Nikon  (via adapter) 55mm micro.  The best Ziess of the bunch compared to the 35mm glass was the 150.  I was suprised, given that the Hassy glass was only using the sweet spot in the center of he image circle.

[John Sheehy]

What is the definition of "extended tonal range"?

I can't remember how to thread these messages, so I don't know who you are replying to without a quote, but if you are asking me, I meant just about anything that the other person could have possibly meant.  The noise in the RAW files is too high for 14 bits to be of any real use.  The best ISO 100s (Canon 1D* series, 40D, Fuji S* cameras) are only worthy of about 12 bits.  12 bits is even overkill for ISO 100 on noisier cameras, and higher ISOs on all of them.  The D2X at ISO 1600 is only worthy of about 7 bits.

If you mean blacker blacks and whiter whites I would also be surprised to see any difference. But bit depth is not so much about that - it is more about the number of tonal levels between black and white. The difference between 12 bit and 14 bit is exponential - specifically 2^14 rather than 2^12, or 16,384 levels instead of 4,096. IN PRINCIPLE, this difference should provide a higher quality raw file (less filling to do by the raw converter graduating from 14 bit to 15+1 compared with graduating from 12 bit to 15+1). Before passing judgment on it, one would need controlled experiments to see on paper the extent of any visible image quality improvement.
[a href=\"index.php?act=findpost&pid=138096\"][{POST_SNAPBACK}][/a]

Your control has to compare 14-bit data truncated to 12-bit, to 14-bit, because the real issue is not "is it worth using a 14-bit ADC", but rather, "is it worth storing the two extra bits", and the answer to that is no, unless you are doing astrophotography or something similar and are stacking multiple images, in which case adding all those images together will clear up a bit or two from mostly noise to mostly signal.

Now, the extra two bits may force some converters to use more bit depth in the conversion process, but this does not prove that the two bits are meaningful, and they could have just as well been 00 (or 10 to keep the mean from changing).

[Marc Schultz]

Where Hasselblad lenses work well on a Canon is for portraits. Canon lenses are way too sharp and show too many imperfections in the skin normally. Hasselblad lenses will be more favorable to skin on a Canon body in general. The 120mm Hasselblad macro lens is a good choice for portrait on a Canon body. The CF lenses will be of course better as well than the older C and other older series Hasselblad lenses. The issue is now you are down to manual focus. You will also see a big drop in viewfinder brightness, especially if you are stopped down to anything smaller than F/5.6

Personally I found it very difficult to work with, but many have talked about favorable results with this configuration.

[Mark D Segal]

I can't remember how to thread these messages, so I don't know who you are replying to without a quote, but if you are asking me, I meant just about anything that the other person could have possibly meant.  The noise in the RAW files is too high for 14 bits to be of any real use.  The best ISO 100s (Canon 1D* series, 40D, Fuji S* cameras) are only worthy of about 12 bits.  12 bits is even overkill for ISO 100 on noisier cameras, and higher ISOs on all of them.  The D2X at ISO 1600 is only worthy of about 7 bits.
Your control has to compare 14-bit data truncated to 12-bit, to 14-bit, because the real issue is not "is it worth using a 14-bit ADC", but rather, "is it worth storing the two extra bits", and the answer to that is no, unless you are doing astrophotography or something similar and are stacking multiple images, in which case adding all those images together will clear up a bit or two from mostly noise to mostly signal.

Now, the extra two bits may force some converters to use more bit depth in the conversion process, but this does not prove that the two bits are meaningful, and they could have just as well been 00 (or 10 to keep the mean from changing).
[a href=\"index.php?act=findpost&pid=138171\"][{POST_SNAPBACK}][/a]

John, yes I was asking you what you meant by "extended tonal range". LAB 0 is black and LAB 255 in 8 bit mode is white. To my way of thinking that is the full extent of the tonal range. Were you trying to say that neither the previous nor current generation of high end digital cameras are capabale of capturing this extent of the tonal range, but only something less? I'm asking because I simply don't understand what you were trying to say. To me, if we're talking bit depth, it is mainly a story about what happens to the quality of the tonal range between the end points - this isn't so much "extension" as "quality of gradation" between the end points. I've never heard that the bit depth affects the extent of the range, but I have seen demonstrations that it does affect the quality of the tonal gradation within the range.

Now if you are saying that the additional theoretical quality of smoother tonal gradation gets lost in practice because of noise, that is another story and one that is certainly worth contemplating - in the first instance by seeing the results of some controlled tests. We are constantly being told that successive generations of sensors going from the 1Ds to the 1Ds MKii to the 5D are showing improved performance with respect to noise. Is this development now being arrested in the interest of adding more pixels to the same sensor dimensions, or is the noise reduction technology proceeding apace? I suppose we won't realy know the answer to that question for the 1DsMKiii until we see rigorous test results.

[Ray]

According to Bob Atkin's tests of the 40D, the 40D exhibits very slightly more noise than the 20D at all ISO's with noise reduction off, but very slightly less noise than the 20D with noise reductions on.

To my eyes, there's not much in it either way so any improvement is really only of pixel-peeping proportions, in my view. However, the noise reduction employed by the 40D does not appear to reduce resolution in any way.

Perhaps Canon have been able to implement such noise reduction as a result of the higher bit depth.

[bjanes]

To me, if we're talking bit depth, it is mainly a story about what happens to the quality of the tonal range between the end points - this isn't so much "extension" as "quality of gradation" between the end points. I've never heard that the bit depth affects the extent of the range, but I have seen demonstrations that it does affect the quality of the tonal gradation within the range.

[{POST_SNAPBACK}][/a]

Mark,

After your masterly essay on curves, I hesitate to challenge any of your statements, but bit depth definitely affects dynamic range when linear integer encoding is used, as it is in raw files. This is clearly explained on [a href=\"http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/]Roger Clark's web site.[/url]--see figure 4. Roger is using the engineering definition of dynamic range: full well electrons/read noise in electrons. He states that each additional bit on the ADC adds one f/stop of potential dynamic range. However, in practice, this potential may not be realized because of noise.

Norman Koren also has a table showing the bit depth necessary for a given dynamic range, but approaches the problem somewhat differently from a photographer's viewpoint, requiring a certain number of levels in the darkest f/stop to avoid banding. His criterion is that this darkest zone must contain 8 levels to avoid banding. By these criteria, a 12 bit linear ADC could give 9 f/stops and a 14 bit device would give 14 f/stops, presuming that the sensor is capable of such performance.

If logarithmic encoding were used, coding efficiency would improve and fewer bits would be required to avoid shadow banding and allow expanded DR. Even with 16 bit integer linear encoding, dynamic range falls short of HDR, and HDR encoding efficiency is greatly improved by Log and floating point formats as Gregory Ward discusses on his stie. Gamma encoding (as opposed to linear) also needs fewer bits to record acceptable shadow levels, and this is shown in Norman's chart.

In some of Bruce Fraser's writings, he took your viewpoint and used an analogy comparing dynamic range to a stair case. The height of the stairs relates to dynamic range and the size of the steps to the bit depth. Adding more steps decreases the size of the steps but does not change the height of the stairs. However, If one adds more steps and does not change their size, the height of the stair case (dynamic range) would be increased.

Bill

[John Sheehy]

According to Bob Atkin's tests of the 40D, the 40D exhibits very slightly more noise than the 20D at all ISO's with noise reduction off, but very slightly less noise than the 20D with noise reductions on.[a href=\"index.php?act=findpost&pid=138275\"][{POST_SNAPBACK}][/a]

Perhaps the camera was under-exposing?  And, nore noise where?  Read noise (shadows) and shot noise (mainly visible in midtones) are independent of each other.

[Mark D Segal]

Mark,

After your masterly essay on curves, I hesitate to challenge any of your statements, but bit depth definitely affects dynamic range when linear integer encoding is used, as it is in raw files. This is clearly explained on Roger Clark's web site.--see figure 4. Roger is using the engineering definition of dynamic range: full well electrons/read noise in electrons. He states that each additional bit on the ADC adds one f/stop of potential dynamic range. However, in practice, this potential may not be realized because of noise.

Norman Koren also has a table showing the bit depth necessary for a given dynamic range, but approaches the problem somewhat differently from a photographer's viewpoint, requiring a certain number of levels in the darkest f/stop to avoid banding. His criterion is that this darkest zone must contain 8 levels to avoid banding. By these criteria, a 12 bit linear ADC could give 9 f/stops and a 14 bit device would give 14 f/stops, presuming that the sensor is capable of such performance.

If logarithmic encoding were used, coding efficiency would improve and fewer bits would be required to avoid shadow banding and allow expanded DR. Even with 16 bit integer linear encoding, dynamic range falls short of HDR, and HDR encoding efficiency is greatly improved by Log and floating point formats as Gregory Ward discusses on his stie. Gamma encoding (as opposed to linear) also needs fewer bits to record acceptable shadow levels, and this is shown in Norman's chart.

In some of Bruce Fraser's writings, he took your viewpoint and used an analogy comparing dynamic range to a stair case. The height of the stairs relates to dynamic range and the size of the steps to the bit depth. Adding more steps decreases the size of the steps but does not change the height of the stairs. However, If one adds more steps and does not change their size, the height of the stair case (dynamic range) would be increased.

Bill
[a href=\"index.php?act=findpost&pid=138294\"][{POST_SNAPBACK}][/a]

Hi Bill,

Glad you enjoyed the Curves essays, and by all means - there is no end to what there is to learn, so I am very pleased you have brought these factors and references to my/our attention. I did indeed have the Bruce Fraser analogy in mind.  Now, from what you relate, it would appear that depending on sensor performance, the added bit depth should contribute to more DR, which is what we want, isn't it? And I guess that again means we will just have to wait for the 1DsMKIII to see how that particular package of hardware and firmware behaves in these respects.

[vgogolak]

If one is expecting Hasselblad "V" series with Zeiss lenses to please them more than a Canon DsII, they will likely be disappointed. Having a DsII with 12, (mostly L-series), lenses and a Blad with 7 lenses from 40 to 500 APO, (including  the superachromat), it is disappointingly evident how significant the aberrations are in the Zeiss optics. Vastly more chromatic problems are seen on film with the Zeiss than Canon lenses on a more demanding digital sensor. It would not be worth me acquiring a digital back for my Blad with the limitations of the optics. I wish I was wrong.


More photo sites have an exponential benefit in less up-ressing needed for large output.

So I ordered a DsIII the very day it was announced.
[a href=\"index.php?act=findpost&pid=138126\"][{POST_SNAPBACK}][/a]

And a lot of pwople buy audio based on low harminic distortion.

Yet the top end, glorious sounding stuff can have just average spectcs

same with glass. I cant say why Leica and Zeiss produce better more exciting images, for some people, but they do. If you cant see the difference, then stick with Canon
if you DO see the difference, boy it is hard to go back!

Victor

PS Knocking Hassey glass also seems a losing proposition. This is no 'emperor has no clothes situation' not after 150 years it aint

[tgutgu]

Does the 1Ds MkIII make it easier to turn on Mirror Lock Up than having to dig through menus on the MkII?

Bud James
[a href=\"index.php?act=findpost&pid=136876\"][{POST_SNAPBACK}][/a]

From what I understand from the recent previews, there isn't any mirror lock up button. Funnily the 40D seems to be better in this regard, as you can access 3 custom configurations from the main dial.

I've configured a custom configuration with mirror lock up plus 2s self timer, so that I can set mirror lock up as quick as with a dedicated button (even more effective as more than one adjustment can be done with one single step).

See my other post today.

Kind regards

Thomas

Munich, Germany