Ad
Ad
Ad
Pages: [1] 2 3 ... 5 »   Bottom of Page
Print
Author Topic: Comparisons From DT's 250 and 260 testing in the Library  (Read 7226 times)
Paul2660
Sr. Member
****
Offline Offline

Posts: 1972


WWW
« on: February 11, 2014, 07:01:28 PM »
ReplyReply

After spending some time working up the stitched series from the 250 and 260 shots that were provided by Doug, it's pretty apparent that we most likely have seen the last CCD back.  The 250 files have a tremendous range to them whereas the 260 images seem to suffer once the exposure dropped off.  Since no CF was used on the 32mm Rodenstock, the LCC had a lot of work to do and the noise that is contained in the 260 shot may have been a bit better, however it would have made the 250 image better also the way I look at it. I worked up the series of images that were the shifted series only, not working on the rise and fall as you can see the differences in just the 3 shifted files.  I have attached 4 screen prints showing areas that seem to really stand out.  

The first one is taken from the left shifted portion of each camera.  There is large panel on the far left that at first I thought was wood, however when I looked at the 250 file, you could see what appears to be fabric and then as you move to the right edge, a brass hinge.  The 260 image is so noisy that you really can't make anything out of the details on this large panel until you get to the very right edge.  The support/glass structure on the very left is pretty much gone on the 260 file (partly due to focus and low light) but the same piece of material is much easier to make out on the 250 image.  

The 2nd one shows a close up from the balcony and the railing support.  Here look at the area underneath the balcony, with the 260 image there is really nothing there but you can easily make out details right up to the very back.  What was even more striking was the details around the large plate holding the railing upright in place, and the amount of detail that is shown in the metal of the base of the upright.  This all falls away on the 260 image.  You can also see a lot more of the details in the wood in the back ground.  

The 3rd image is from the curved portion of the balcony.  Looking underneath from the 250 shot here the difference is very impressive.  You can make out several fine details all the way back to the light fixture and see what appears to be some form of a camera next to the light.  Again the detail in the metal work also stands out much better and the patina of the metal also seems to be shown better. In this shot below the balcony you can see the X shaped supports that run throughout the library.  From this angle the 260 picked up a huge amount of aliasing, which I worked on in Capture One with a local adjustment and I found I was able to get a lot of it taken out, but the 250 file from the start has much less of the aliasing.  

The 4th image was taken from the right shift, the near the lowest portion on the floor.  This comparison speaks for itself as the wall in the background on the 260 file has pretty much gone away and the yet with the 250, you easily make out the grain in the wood and see the emblem in the center.  The stand that is hold the glass cover looks much better on the 250 image with almost no noise.  

The images were a challenge to WB, and I know I probably am not close, however I would really have to be in that building myself to get a feel for the room.  I felt there was a overall red tint to the 260 images that was possibly made worse by the aliasing.  The aliasing is extreme enough that it would show up in even a web sized image if not worked on.    The only downside I found to the 250 images was just a bit of banding on the large panel on the left.   I did note that Capture One was a bit shaky working on the 250 files, especially when zooming to 100% and back out.  About 50% of the time on zooming out, the image size dropped to a postage stamp causing me to restart Capture One.  This did not happen when I was working on the 260 images.

Overall the images pretty much tell the future for Medium format digital.  I can't see much difference is color but I there may be some faint differences.   A test in the outdoors would be better for me, but I did appreciate working on this series of images as it's a beautiful subject indeed.  

What is real key in all of this to me: the fact that the Sony chip with it's tremendous DR has now made it to the MF world and the results are as impressive as when in April 2012 the D800 hit the market.  

Thanks again to the Doug and DT for taking the time to provide these examples. 

Paul C
« Last Edit: February 12, 2014, 05:41:45 AM by Paul2660 » Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #1 on: February 12, 2014, 01:18:38 AM »
ReplyReply

I've looked on the raw files for IQ260 and IQ250 32HR, no shift.

The IQ250 is slightly lower exposed in raw values (like 1/3 stop), but in terms of exposure time it's ISO100 3 seconds vs ISO50 6 seconds on the IQ260.

In RawTherapee you can turn off all noise reduction and all hot pixel filtering, and the amount of hot pixels in the IQ260 file is a lot, it's peppered like a 30 second exposure on my Aptus 75. Still the file says only 6 second exposure, so I'm wondering if there's something wrong with the IQ260? It seems unlikely that it should be as noisy as it is.

When pushing the shadows the noise difference between the IQ250 and IQ260 is huge, the IQ250 is as clean as we expect any Sony Exmor sensor to be, but as far as I can see the IQ260 is much more noise than we would expect from a recent CCD back. On the other hand I've not seen multi-second exposures in previous examples, maybe the good DR from an IQ260 only is obtainable with sub-second exposure times?

The attached image is from the IQ260 image, with all noise reduction turned off. Note the hot/dead pixels, the black, white, red and blue dots scattered all over the shadowy parts, and a few in the brighter too.

I'd like to see an explanation why the IQ260 is this noisy, if it's normal or not.

(concerning IQ250 and shift, I'm still very much skeptical about color stability, but I need more testing with my own algorithms etc, my workstation crashed yesterday so I can't do much advanced testing until it is restored)
« Last Edit: February 12, 2014, 02:58:58 AM by torger » Logged
BernardLanguillier
Sr. Member
****
Online Online

Posts: 8359



WWW
« Reply #2 on: February 12, 2014, 01:33:03 AM »
ReplyReply

It would be interesting to throw in the mix a spherical stitch from a D800 + Zeiss 55mm f1.4 with the same angular coverage and a flat projection corrected for verticals.

I am really not impressed by the image quality in the corners of any of those images, at least on the 47mm XL images I checked in detail.

Cheers,
Bernard
« Last Edit: February 12, 2014, 04:23:16 AM by BernardLanguillier » Logged

A few images online here!
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #3 on: February 12, 2014, 02:50:56 AM »
ReplyReply

It would be interesting to throw in the mix a spherical stitch from a D800 + Zeiss 55mm f1.4 with the same angular coverage and a flat projection corrected for verticals.

I am really not impressed by the image quality in the corners of any of those images.

You should be impressed by the 32HR, it's kind of the best there is. But of course proper stitching (ie turning the camera) will always outperform in terms of corner performance. If it doesn't, just put on a longer focal length and stitch even more. I have no doubt that the D800 spherical stitching plus remapping to rectalinear would provide the best result in this stitching scenario, it would be more files to keep track of though, but for the difference in cost you can probably come up with some good workflow Smiley

I think tech cam is best when you don't need to stitch in the general case, and just stitch occasionally. Reduced corner performance for these extreme wide angle shots is rarely a problem in production. These are just test shots, stitching together a large shot to test all sorts of shifting in one image and push beyond the limit so you can see where it is. Shifting 30mm in a 90mm image circle is nothing you would do normally. Of course, if you stitch all the time in production I'd suggest to strongly consider a spherical head.

Also consider that these are f/9 shots. f/11 would improve corners. The older symmetrical 35XL and 47XL performs pretty bad at f/9, they're designed to be shot at f/11, and for large shifts I'd recommend f/16. Personally for the shooting style I have I prefer to have symmetrical designs optimized for smaller apertures, and make a slight sacrifice in corner performance, going all in for resolution just leads to costly heavy overkill designs (with added distortion too). I just want the large format shooting experience, but with a digital back. If that experience is replaced with just a larger mirrorless camera and otherwise the same as smaller format I'd probably will exit MF, to me it's very much about the emotional experience of shooting.

The IQ250 does not do well with those symmetrical designs unfortunately due to poor and varied angular response. I'm a bit worried that it may have pixel crosstalk too (eg due to low angle photons passing through the red filter gets registered in the green photodiode), but I'm not sure how I could see that in the files.
« Last Edit: February 12, 2014, 02:57:29 AM by torger » Logged
Paul2660
Sr. Member
****
Offline Offline

Posts: 1972


WWW
« Reply #4 on: February 12, 2014, 06:03:36 AM »
ReplyReply

I need to pull down the 60mm tests, as that is a lens I use a lot, and I can't figure out why it did not work as well as it does on the 80MP tests, since it's a back with similar pixel pitch. 

When I look at the the LCC's for the 250 on the 40mm, they look good to and don't show excessive color cast, and but the room being shot is a pretty big challenge, with all the various lighting. 

I agree the 260 shots (at least with the 32mm) appear to be excessively noisy, especially on the shifts and as Doug mentioned there was no Center filter being used and I am pretty sure that would have helped to clean some of the shifted noise up.  Overall the details from the 32mm are impressive to me.   If you look at the center shot only of the 260, the noise is not as bad as it is on the shifts.  It's not up to par with the 250 in the shadows, but not sure the 260 will ever get there. 

Torger, your shot from raw therapy is most interesting.  I am very surprised to see that much noise at only a 6 second shot at iso 50.   I thought the dark frame would remove most of this, as that is why I always felt it was used by Phase One.  The long exposures I have taken with the P45+ and developed in Capture One don't show anywhere near this much noise.  When I opened the 260 shots, Capture One sets the what it feels is the correct noise reduction/sharpening for the raw image based on the back.  Even with the defaults on, there are a lot stuck blue pixels that show up in most of the shadows and even if you take the "single pixel noise" slider all the way to 100%, not all of them are removed.  I am going to setup my 260 this evening and take a few longer exposures and see what I get in the deeper shadow areas. 

I guess the denser pixel size of the 250 creates less aliasing?  The 250 is pretty much clear of it.

I am pulling down the 80MP tests later this morning to see how well that back did in the shadows also. 

Long term, I am wondering who will make a full frame CMOS chip and when?  No doubt to me that Sony has CMOS figured out, but this was clear 2 years ago with the early testing of the D800.  I know that Sony has a patent on a 54MP chip for the 35mm format, but that is most figure a year away from production and even Sony has stated it will not be a inexpensive chip/camera solution.  I am sure Dalsa is working on a CMOS solution, but it will be their first (at least in this market) that I am aware of and I have to wonder if it will have the same DR as the Sony designs.   I wonder if Phase One can get to a full frame CMOS in a year, and if so what the size/MP rating of the chip will be.

Paul C.

 

Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
Richard Osbourne
Newbie
*
Offline Offline

Posts: 27


WWW
« Reply #5 on: February 12, 2014, 06:17:08 AM »
ReplyReply

The problem with stitching is depth of field. You can focus stack but that's a hell of a lot of complication, especially if you're shooting outside and there's any wind at all  - or waves etc. And the longer the lens, the smaller the DOF. Tech cam lenses have startling depth of field - once focused on the hyper focal distance, I rarely change focus for any shot - I know that virtually everything is in focus in a single shot. It has saved me a lot of time.

Re. the IQ250, I'm stunned by the lack of noise and the resolution, but I really need more information about the colours. They are dramatically different between the two backs here but I can't tell if that's a white balance issue. Nature, especially greens, is very demanding on colour fidelity so landscape shots should be interesting. I sold the D800 because I couldn't get accurate nature colours out of it (amongst other issues) - I hope the IQ250 doesn't suffer similarly.
Logged

BernardLanguillier
Sr. Member
****
Online Online

Posts: 8359



WWW
« Reply #6 on: February 12, 2014, 06:41:22 AM »
ReplyReply

The problem with stitching is depth of field. You can focus stack but that's a hell of a lot of complication, especially if you're shooting outside and there's any wind at all  - or waves etc. And the longer the lens, the smaller the DOF. Tech cam lenses have startling depth of field - once focused on the hyper focal distance, I rarely change focus for any shot - I know that virtually everything is in focus in a single shot. It has saved me a lot of time.

Re. the IQ250, I'm stunned by the lack of noise and the resolution, but I really need more information about the colours. They are dramatically different between the two backs here but I can't tell if that's a white balance issue. Nature, especially greens, is very demanding on colour fidelity so landscape shots should be interesting. I sold the D800 because I couldn't get accurate nature colours out of it (amongst other issues) - I hope the IQ250 doesn't suffer similarly.

True, it doesn't work for all subjects equally easily.

As far as colors go, WB can be challenging to get perfect on the D800.

Cheers,
Bernard
Logged

A few images online here!
Paul2660
Sr. Member
****
Offline Offline

Posts: 1972


WWW
« Reply #7 on: February 12, 2014, 06:43:49 AM »
ReplyReply

The problem with stitching is depth of field. You can focus stack but that's a hell of a lot of complication, especially if you're shooting outside and there's any wind at all  - or waves etc. And the longer the lens, the smaller the DOF. Tech cam lenses have startling depth of field - once focused on the hyper focal distance, I rarely change focus for any shot - I know that virtually everything is in focus in a single shot. It has saved me a lot of time.

I agree, with the wides, once you have the hyperfocal figured out, the shifting is great.  I love to shift, 2nd main  reason I use the tech camera in landscape situations, 1st being the quality of the wide lenses compared to the comparable DF+ mount wides, (28mm, 35mm 45mm etc)

Paul C 
Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #8 on: February 12, 2014, 07:04:25 AM »
ReplyReply

Regarding stitching and DoF:

You could tune the hyperfocal to your taste on your manual focus DSLR lens and put a tape there so you can find the setting. Then it becomes no different from tech cam lenses in terms depth of field. Of course, if you define CoC related to pixel pitch and you stitch a 500 megapixel image with a long focal length you're going to have DoF issues, but also more resolution. If you stitch to get the same resolution as with the tech cam system you have no more or less DoF issues.

You do get a DOF advantage with the tech cam stitching in scenes where tilt is applicable though. However I've done multi-row DSLR stitching and when tilt is applicable it often works out to stitch the lower row on a closer distance and the longer farther away. You really need to be focused when doing this type of stitching though, there's no room for mistakes.

With spherical stitching you also get a curvature of the depth of field, while a tech cam lens might be flatter, which in some instances might be better.

The attached wide image is a stitch I made few years ago (before I got my Linhof Techno), it was made with a Canon 7D (18 megapixel APS-C) using a 50mm lens, I used a closer focus distance for the lower row, the finished image is about 90 megapixels. The square image is also a Canon 7D multi-row stitch made on a nodal ninja click-stop head but the closest distance is past hyperfocal so I could focus on infinity. Since I got the Techno I have so far not stitched a single image.
« Last Edit: February 12, 2014, 07:34:43 AM by torger » Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #9 on: February 12, 2014, 07:05:35 AM »
ReplyReply

Regard IQ250 aliasing. I see quite a lot of aliasing in the 32HR image I've looked at, look at the metal bars in front of the books, lots of false colors. But there's no particular problem of the back, this is what happens when shooting with f/9 with a very sharp lens at this pixel pitch.
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #10 on: February 12, 2014, 07:12:41 AM »
ReplyReply

Regarding colors I find color accuracy to perhaps be the least important in landscape photography, but color separation is important, and high dynamic range is good to have. Color accuracy is near impossible anyway as the light conditions (ie the illuminant) is so far away from where the camera was profiled that you can't get accurate. During winter up here close to the polar circle the light conditions are so special that the concept of white balance partly breaks. Color becomes interpretation in post-processing. I generally strive for a natural look, but profiles generally won't help much.

Color separation helps though and I think the IQ250 should have CFAs designed for that to a stronger extent that the D800.
Logged
Paul2660
Sr. Member
****
Offline Offline

Posts: 1972


WWW
« Reply #11 on: February 12, 2014, 07:30:59 AM »
ReplyReply

Regarding stitching and DoF:

You do get a DOF advantage with the tech cam stitching in scenes where tilt is applicable though. However I've done multi-row DSLR stitching and when tilt is applicable it often works out to stitch the lower row on a closer distance and the longer farther away. You really need to be focused when doing this type of stitching though, there's no room for mistakes.


I totally agree, the advantage of adding just 1/2 of a degree of tilt can often give you up to 5 more feet of sharp, not smeared image right to the corner of the shot when using a tech camera.  In my environment the light is changing constantly or the wind is blowing and working by your 2nd method of switching focus just won't work many times. plus the issue parallax comes into play unless you are level with a nodal stitch setup, but even then with a 35mm MF and a nodal stitch, it won't always lineup perfectly or close to perfect without a good bit of warping which most of the software tools do very well now.   

Paul C. 
Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
Richard Osbourne
Newbie
*
Offline Offline

Posts: 27


WWW
« Reply #12 on: February 12, 2014, 07:41:44 AM »
ReplyReply

I've found that some DSLR lenses - including some M645 lenses - play very nicely and have a very useable HFD. Canon 17-40 (at the 40 end) and 50 1.2L (ok but not great), and M645 55AF (especially good) come to mind. But others - latest Phase 80LS, or the 75-150 - seem to have far less DOF, even using the HFD. I'm guessing that some lenses are designed for sharpness/DOF and some for bokeh, but I'm no expert on lens design.

So, I shoot single row panoramas with the 55AF and IQ260 and the HFD DOF is extremely good - best of any camera/lens combo I've tried. I used to have to focus bracket with the old 55MF. Multi row I haven't tried though I suspect it would work well with this lens. Previous attempts with the 1DsIII and 50 1.2L weren't so successful - stitching multi-focused rows was not successful.

Regarding stitching and DoF:

You could tune the hyperfocal to your taste on your manual focus DSLR lens and put a tape there so you can find the setting. Then it becomes no different from tech cam lenses in terms depth of field. Of course, if you define CoC related to pixel pitch and you stitch a 500 megapixel image with a long focal length you're going to have DoF issues, but also more resolution. If you stitch to get the same resolution as with the tech cam system you have no more or less DoF issues.

You do get a DOF advantage with the tech cam stitching in scenes where tilt is applicable though. However I've done multi-row DSLR stitching and when tilt is applicable it often works out to stitch the lower row on a closer distance and the longer farther away. You really need to be focused when doing this type of stitching though, there's no room for mistakes.

With spherical stitching you also get a curvature of the depth of field, while a tech cam lens might be flatter, which in some instances might be better.
Logged

torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #13 on: February 12, 2014, 08:28:53 AM »
ReplyReply

I've never heard that a specific lens design can have more DoF than another, seems to defy physics. But as DoF is subjective I guess that say if a lens renders with better local contrast one could accept more out-of-focus blur and thus get an impression of more DoF. The bokeh and how it transitions from sharp to out of focus can probably also affect the impression of having more or less DoF. Interesting indeed, could be something for me to investigate.
Logged
JoeKitchen
Sr. Member
****
Offline Offline

Posts: 743



« Reply #14 on: February 12, 2014, 09:26:53 AM »
ReplyReply

I would be interested in seeing a long exposure comparison between the 250 vs 260.  Something is the range of a minute or more. 
Logged

Joe Kitchen
www.josephmkitchen.com

"Photography is one percent inspiration and ninety-nine percent moving furniture."  Arnold Newman
"Try not to be just better than your rivals and contemporaries, try to be better than yourself."  William Faulkner
Paul2660
Sr. Member
****
Offline Offline

Posts: 1972


WWW
« Reply #15 on: February 12, 2014, 12:52:05 PM »
ReplyReply

Not sure what is going on with the 260 as after looking at the full jpg provided the noise is much better at the top stitch, maybe more light as it got to the ceiling, but on the completed image is combined as a strange blob. 

Paul C.

Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
Stefan.Steib
Sr. Member
****
Offline Offline

Posts: 421



WWW
« Reply #16 on: February 12, 2014, 12:58:29 PM »
ReplyReply

Thanks a lot Doug for that good piece of work.

Very interesting ! One thing that is definitely not expected (well I suspected it, but now there is proof) - the CMOS beats the CCD´s in about any aspect besides that the IQ280 is bigger.
The IQ260 is a disappointment, grainy, bad shadow details, color noise and actually less quality than the 50 Mpix at 100 ASA though it was shot at 50 ASA. The IQ280 is on 35 ASA and just reaching the level, probably better with more light and less critical brown tones (compliment Doug for this honesty. It is the ultimate test - brown with lowlight shows everything !)

One hint: the standard settings in C.O. 7.2 are oversharpened (for all 3) and the 260 needed addition of moire.

This needs to settle down a bit, but in my brain it already forms a BIG question: if the CMOS is SO much better, who will stil buy a CCD ?

? ? ?

Greetings from Germany
Stefan
Logged

Because Photography is more than Technology and "as we have done this all the time"
www.hartblei.de     www.hcam.de    www.spectralize.com
torger
Sr. Member
****
Offline Offline

Posts: 1590


« Reply #17 on: February 12, 2014, 01:59:38 PM »
ReplyReply

Not sure what is going on with the 260 as after looking at the full jpg provided the noise is much better at the top stitch, maybe more light as it got to the ceiling, but on the completed image is combined as a strange blob. 

The IQ260 images in the shoot seem to vary a lot in noise level. So what you see is a low noise image stitched together with a high noise image. Stitching algorithms don't do straight seams so it may look like a random noise blob but if you look carefully you can follow the seam also outside the dark area.

Why the images vary in noise I don't know. Maybe there was different exposure times and/or ISO shots made and those where mistakingly stitched together. Only Doug can answer that. Something is wrong though. Too much noise for an IQ260, and way too large variation.
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7915


WWW
« Reply #18 on: February 12, 2014, 02:38:03 PM »
ReplyReply

Hi,

Chris Barret has posted a comparison of IQ 260 vs Sony Alpha 7r, and the IQ 260 was quite noisy compared to the A7r, so it comes as little surprise. That comparison was at 100 ISO, however.

It was known for long that CMOS with on sensor converters has much lower shadow noise than CCD. But CCD friends have always stated that CCDs have some magic properties, like the emperor's new cloths. So we just found out that the emperor has no clothes and a lot of underbody fat.

Best regards
Erik

Thanks a lot Doug for that good piece of work.

Very interesting ! One thing that is definitely not expected (well I suspected it, but now there is proof) - the CMOS beats the CCD´s in about any aspect besides that the IQ280 is bigger.
The IQ260 is a disappointment, grainy, bad shadow details, color noise and actually less quality than the 50 Mpix at 100 ASA though it was shot at 50 ASA. The IQ280 is on 35 ASA and just reaching the level, probably better with more light and less critical brown tones (compliment Doug for this honesty. It is the ultimate test - brown with lowlight shows everything !)

One hint: the standard settings in C.O. 7.2 are oversharpened (for all 3) and the 260 needed addition of moire.

This needs to settle down a bit, but in my brain it already forms a BIG question: if the CMOS is SO much better, who will stil buy a CCD ?

? ? ?

Greetings from Germany
Stefan
Logged

Fine_Art
Sr. Member
****
Offline Offline

Posts: 1153


« Reply #19 on: February 12, 2014, 02:59:08 PM »
ReplyReply

That is a bit of revisionist history. CCD used to be better than CMOS. There is a reason all the high end astrophoto cameras were CCD. The kodak kaf8x00 is still sold with -50oC cooling based on that was about as good as it got for low light low noise photography. If you look up the specs they are quite open about 16 e- noise floor. Now CMOS is at 2 e- according to BJL (I have not seen the spec sheets). That does not mean that it was that way when everyone was using 6-10MP cameras!
Logged
Pages: [1] 2 3 ... 5 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad