Full format DSLR, or not?

[ErikKaffehr]

Hi,

I recently bought my first FF DSLR an Sony Alpha 900 with a Sony SAL 24-70/2.8 ZA lens. I had some doubts before buying and I want to share some initial experience.

To begin with the image quality is impressive. There has been a lot of discussion about the high ISO quality (or lack thereof) of the Alpha 900 images. High ISO shooting is not what I normally do, and I cannot compare to other DSLR's but I share the impression that the Alpha 900 is less than optimal for high ISO work. Regarding the 24-70/2.8 my personal impression is that it is much better at shooting real life pictures than test charts. My guess, based on testshots, may be that it has less than ideal flat field corner performance at close distances but I'm not in the lens testing business so I won't evolve in commenting that.

The real question for me was if I would see a benefit in sharpness from the Alpha 900 over the Alpha 700. I felt that the Alpha 700 was good enough for reasonably good A2 prints. So the two questions I had was how much better A2s I would get with the Alpha 900.

Three different comparisons were made. The first was based on downloaded images "Imaging Resource", for the Alpha 700 I only found JPEGs, but I found "raw" images for the Alpha 900. I made A2 prints from these. Initially I couldn't tell them apart, later I learned to see the difference and could tell them apart from 1 meter distance.

When I got the camera, but without the 24-70/2.8 lens I shot a comparison with my Alpha 700 with a similar result. First I had difficulty to tell my prints apart, after learning to see differences I could tell them apart with some reliability.

In the third comparison I photographed a large painting on display at a parking lot. It contains a lot of fine detail. First I was doing some pixel peeping. Both images were scaled to 70 cm width at 360 PPI that is 9920 pixel wide using bicubic softer. The images were compared in PS CS3 at actual pixels. The difference in the image quality was striking.

http://www.pbase.com/ekr/image/107619976/original  (A900 to left and A700 to the right)

Now I cropped the image into half and printed on A4 which corresponds to A2 for the full image and printed from Lightroom 2.1 (because 2.2 doesn't work for me    ) on my Epson SP 3800 at 2880 DPI. Settings in LR were Landscape Preset capture sharpening,
480 PPI output and medium sharpening for glossy.

The result was that I could not see any difference, period!

Both prints were scanned at 300 PPI and compared in Photoshop at actual pixels. Here I could see a significant difference.

http://www.pbase.com/ekr/image/107823207/original

Yesterday I asked a couple of colleagues to check out my prints. Like me they could easily tell apart my first test but not the second one.  On the second sample (based on the painting) my elder colleague who has much experience of professional printing, as he worked at one the best professional labs in Sweden, actually preferred the A700 print which according to him might give a sharper impression. He also pointed out that the A700 print had slightly higher contrast and was more bluish, a difference lost on me. My friend also explained that the higher contrast can make a print look sharper.

When we compared the scanned images on screen we all agreed that there was a significant difference.

The conclusion from this very limited testing my be that:

1) APS-C and 12.5 MPixel sensor with good technique can give nearly optimal A2 prints.
2) Technique and subject may play a significant role, some subjects may be more critical than others
3) There is no doubt that the images from FF and 24.5 MPixels contain more and better detail, the question is if it can bee seen in a print
4) FF and 24.5 MPixels allow for more cropping and larger prints

The difference between APS-C and FF was in this case less than what I would have expected. Possible explanations may be that the Alpha 700 may have a very weak AA (Anti Aliasing) filter. This may be indicated by the extremely high "extinction resolution" reported by "DPreview" for most Sony Alphas. Another explanation may simply be that Lightroom does a very good job on scaling and ouput sharpening.

Some high res sample pictures from the Alpha 900 are available here:

 http://www.pbase.com/ekr/a900_samples

Here are some: 24-70/2.8 ZA test pictures

http://www.pbase.com/ekr/2470za_test1

And some Konica-Minolta 28-75/2.8 test:

http://www.pbase.com/ekr/km_28_75_test

All samples are in full size JPEGS, to see full image click on "original" size button below the picture.

I hope that this info may be helpful for some people considering full frame in general or specially the Alpha 900.

You may also check this excellent article by David Kilpatrick: http://www.photoclubalpha.com/2008/10/16/d...d-an-alpha-900/

Best regards
Erik Kaffehr

[Bill Caulfeild-Browne]

Very interesting, and I'm not surprised at your findings. The value of the a900 lies in (1) the ability to make very large prints like A1 without uprezzing and (2) the very important ability to crop as little as a third or a quarter of the frame and still get an acceptable print. This is very important to me as I photograph wildlife that is invariably too far away for the lens I happen to have on the camera when I see it!

I realize that seeing JPGs on a screen isn't ideal, but here is my very first shot with the a900. It was taken with the 24-70 Zeiss at ISO 800 and is a crop of roughly the middle third of the pic. It  easily enlarges to Super B.

Bill

[attachment=10811:_Custom_Name_015.jpg]

[ErikKaffehr]

Hi,

Thanks for comments! Nice cat!

Erik

Very interesting, and I'm not surprised at your findings. The value of the a900 lies in (1) the ability to make very large prints like A1 without uprezzing and (2) the very important ability to crop as little as a third or a quarter of the frame and still get an acceptable print. This is very important to me as I photograph wildlife that is invariably too far away for the lens I happen to have on the camera when I see it!

I realize that seeing JPGs on a screen isn't ideal, but here is my very first shot with the a900. It was taken with the 24-70 Zeiss at ISO 800 and is a crop of roughly the middle third of the pic. It  easily enlarges to Super B.

Bill

[attachment=10811:_Custom_Name_015.jpg]

[Ray]

Erik,
You've obviously taken a lot of time and gone to a lot of trouble comparing images at the output stage, even making scans of the prints. This is more trouble than most people are willing to take. It's just so easy to compare images on one's monitor at 100%, which is pixel for pixel, ie. one monitor pixel represents one image pixel. At such close inspection and magnification it's easy to see all sorts of resolution and noise advantages that one image may have over another.

However, in practice and on the print, as your tests show, such differences will not be noticed. In order to see on the print what you see on the monitor at 100% you would need to make a print about 4ft x 6ft, depending on the size and resolution of your monitor.

My current monitor is an old 19" Sony CRT set at a resolution of 1280x1024. Doing a bit of basic maths, the screen is about 14.4in wide. 1280/14.4=89 ppi.

An A900 image at 89 pixels per inch is about 46"x69", if my calculation is right, or 1150mm x 1725mm. My Epson 7600 can handle a maximum width of 610mm, or to put it another way, although my printer is perhaps wider than most, it can make a print hardly bigger than one quarter the size of an A900 image viewed at 100% on my monitor.

[ErikKaffehr]

Ray,

I think it is actually quite complex and the issue is involving both eyesight and viewing distance. Normally I would look at a large picture at a longer distance. I also wear correction glasses. Because I'm near-sighted I can do pixel peeping just removing my glasses. Perception is also involved, I'm pretty sure that the eye/brain can be trained to observe even small differences once they have been found.

Another observation I have made is that FF with 20+ MP is simply much more demanding. It's normally not a big issue for me, because I'm not a fast shooter. But depth of field is going to be incredibly short with a 24-70/2.8 at full aperture, this would be much less of a problem with the 16-80/3.5-4.5 zoom I have on the APS-C.

Best regards
Erik

Erik,
You've obviously taken a lot of time and gone to a lot of trouble comparing images at the output stage, even making scans of the prints. This is more trouble than most people are willing to take. It's just so easy to compare images on one's monitor at 100%, which is pixel for pixel, ie. one monitor pixel represents one image pixel. At such close inspection and magnification it's easy to see all sorts of resolution and noise advantages that one image may have over another.

However, in practice and on the print, as your tests show, such differences will not be noticed. In order to see on the print what you see on the monitor at 100% you would need to make a print about 4ft x 6ft, depending on the size and resolution of your monitor.

My current monitor is an old 19" Sony CRT set at a resolution of 1280x1024. Doing a bit of basic maths, the screen is about 14.4in wide. 1280/14.4=89 ppi.

An A900 image at 89 pixels per inch is about 46"x69", if my calculation is right, or 1150mm x 1725mm. My Epson 7600 can handle a maximum width of 610mm, or to put it another way, although my printer is perhaps wider than most, it can make a print hardly bigger than one quarter the size of an A900 image viewed at 100% on my monitor.

[Ray]

Ray,

I think it is actually quite complex and the issue is involving both eyesight and viewing distance.

Yes, of course. But it's your eyesight in both cases, whether viewing an image on the monitor or viewing the print. What you see on the monitor at 100% is approximately what you would see on a 4ftx6ft print from the same distance you view your monitor (except for the peripheral view which is cropped off the monitor).

Everyone's visual accuity varies to some degree, but it's generally true that the further away you view the print the less detail you see and the less the need for a high resolution camera for, say, A2 size prints. From a viewing distance of, say, 6-12ft, depending on eyesight capability and print size, a 2mp camera might well be sufficient, especially if it had a Foveon sensor.

[ErikKaffehr]

Ray,

Thanks for putting my comment in perspective ;-)

Erik

Yes, of course. But it's your eyesight in both cases, whether viewing an image on the monitor or viewing the print. What you see on the monitor at 100% is approximately what you would see on a 4ftx6ft print from the same distance you view your monitor (except for the peripheral view which is cropped off the monitor).

Everyone's visual accuity varies to some degree, but it's generally true that the further away you view the print the less detail you see and the less the need for a high resolution camera for, say, A2 size prints. From a viewing distance of, say, 6-12ft, depending on eyesight capability and print size, a 2mp camera might well be sufficient, especially if it had a Foveon sensor.

[Ray]

Ray,

Thanks for putting my comment in perspective ;-)

Erik

Erik,
I haven't even started   .

The chief characteristic of photography is that one can capture a 3-dimensional scene and reproduce it in 2 dimensions with an extremely faithful representation of detail. The only distinction (to the eye) between the photographic image and the real scene that was captured, is the 2-dimensional aspect of the captured image and the finite amount of detail present on the print or display.

We know from our experiences that any real object reveals finer and finer detail the closer we view it. Use your reading glasses from about 10" away and we like to see in the print the crisp and sharp detail we would see viewing the real object or subject. Use a magnifying glass, you see even more more detail when viewing a real object. Use a microscope, you see yet more detail. Use an electron microscope and you can even see the individual molecules.

We like high resolution cameras because they can represent more accurately what we see in reality. The closer we get, the more detail we see, hopefully.

I once had the experience a few years ago of visiting a neighbor who had a large print in the hallway which was very impressive and caught my attention. From an initial viewing distance of 10ft or so, the thought occurred to me that this was a print from an MF camera. (I don't know why this thought occurred to me. It seems very strange that anyone would think about such matters, but there you go. There are strange people in the world.)

Okay! I lie. I do know why the print reminded me of something originating from MF format. It had great 'local conrast', something you can create in PS with the Unsharp Mask; pixel radius 60, amount 30, and variations on those amounts.

From a viewing distance which was definitely greater than twice the diagonal of the print, the print looked great. There was a sofa immediately below the print, next to the wall, well-positioned to prevent pesky visitors with an interest in photography from getting too close. But I wasn't deterred and clambered onto the sofa to get a closer look.

Well, I was disapointed. This print was clearly from 35mm film. On close inspection, it was all fuzzy. What a let-down!

[ErikKaffehr]

Thanks Ray!

I'm looking forward to episode #2.

Erik

Erik,
I haven't even started   .

The chief characteristic of photography is that one can capture a 3-dimensional scene and reproduce it in 2 dimensions with an extremely faithful representation of detail. The only distinction (to the eye) between the photographic image and the real scene that was captured, is the 2-dimensional aspect of the captured image and the finite amount of detail present on the print or display.

We know from our experiences that any real object reveals finer and finer detail the closer we view it. Use your reading glasses from about 10" away and we like to see in the print the crisp and sharp detail we would see viewing the real object or subject. Use a magnifying glass, you see even more more detail when viewing a real object. Use a microscope, you see yet more detail. Use an electron microscope and you can even see the individual molecules.

We like high resolution cameras because they can represent more accurately what we see in reality. The closer we get, the more detail we see, hopefully.

I once had the experience a few years ago of visiting a neighbor who had a large print in the hallway which was very impressive and caught my attention. From an initial viewing distance of 10ft or so, the thought occurred to me that this was a print from an MF camera. (I don't know why this thought occurred to me. It seems very strange that anyone would think about such matters, but there you go. There are strange people in the world.)

Okay! I lie. I do know why the print reminded me of something originating from MF format. It had great 'local conrast', something you can create in PS with the Unsharp Mask; pixel radius 60, amount 30, and variations on those amounts.

From a viewing distance which was definitely greater than twice the diagonal of the print, the print looked great. There was a sofa immediately below the print, next to the wall, well-positioned to prevent pesky visitors with an interest in photography from getting too close. But I wasn't deterred and clambered onto the sofa to get a closer look.

Well, I was disapointed. This print was clearly from 35mm film. On close inspection, it was all fuzzy. What a let-down!

[KenS]

...

1) APS-C and 12.5 MPixel sensor with good technique can give nearly optimal A2 prints.
2) Technique and subject may play a significant role, some subjects may be more critical than others
3) There is no doubt that the images from FF and 24.5 MPixels contain more and better detail, the question is if it can bee seen in a print
4) FF and 24.5 MPixels allow for more cropping and larger prints

The difference between APS-C and FF was in this case less than what I would have expected. Possible explanations may be that the Alpha 700 may have a very weak AA (Anti Aliasing) filter. This may be indicated by the extremely high "extinction resolution" reported by "DPreview" for most Sony Alphas. Another explanation may simply be that Lightroom does a very good job on scaling and ouput sharpening.
 ...

Best regards
Erik Kaffehr

Erik,

I am wondering if there may have been some other variables that could have degraded the A900 image more than the A700 image (even though the A900 appears sharper on screen at high mag is it 'optimal').

1- You used two different lenses.  What f-stops were used for your comparison?
    I haven't examined your jpg images in detail (function of f-stop) but  theory would suggest f/5.6 might  be optimal for both lenses.  If lenses were stopped down
    beyond f/9 or so diffraction effects may be at play, degrading contrast and resolution.
2- Was a sturdy tripod used for both the A700 and A900 images?
3- Mirror locked up?
4- What about focus errors?
    Choice of f-stops and distance from target will of course affect depth-of-field as well as diffraction effects and provide a tradeoff with focus errors.
    Focus bracketing and/or manual focus might be worth doing as well as comparisons at several f-stops.
5 -Wind, causing vibrations of either test target or camera?

Just some things to consider,
Ken

[jjj]

I realize that seeing JPGs on a screen isn't ideal, but here is my very first shot with the a900. It was taken with the 24-70 Zeiss at ISO 800 and is a crop of roughly the middle third of the pic. It  easily enlarges to Super B.
[attachment=10811:_Custom_Name_015.jpg]

May I suggest that photos of cats, for those of us who have a feline beastie to hand are very good subjects for showing how well a camera performs, as fur is very hard for cameras to capture nicely. This shot certainly shows how good the Sony is.

[aaykay]

You may also check this excellent article by David Kilpatrick: http://www.photoclubalpha.com/2008/10/16/d...d-an-alpha-900/

I think what David is stating is that he needs *more* DOF in the particular shots he has displayed.  For *more* DOF, you need to shoot with lenses where the focal length number is really wide.  It is not merely the case of achieving the same angle of view ("equiv" Focal length).  

In other words, shooting with a REAL 16mm FL in an APS-C (yielding an equiv 24mm FL in FF terms), will yield a deeper DOF than a real 24mm on FF (yielding the same angle-of-view as the 16mm on APS-C), when stopped down the same.   This is the reason why P&S cameras typically have enormous DOF, since the REAL focal lengths at which the shot is taken (if we ignore the "equivalent Focal Length" portion), is around 5mm or so on the wide-end.....which obviously yields DOF that stretches from the tip of one's nose to infinity.

The portions he has unfortunately not put any emphasis on, in his article, are situations when you need *less* DOF, where a large sensored product holds an edge over a smaller sensored product.

Bottomline, the upcoming 16-35 f/2.8, at the 16mm end, should yield a deep enough DOF (without stopping down too much) for almost all applications where a deeper DOF is desired.  A 12-24 Sigma would go even further, since the *real* focal length is now gone down to 12mm on the wide-end.  There is no substitute for a really low FL number (not "equivalent"), if a deeper DOF is the desired outcome.

[ErikKaffehr]

Ken

Thank you for your input, I added my responses below.

I'm not complaining about sharpness. Both pictures are very sharp at the pixel level. When scaled to 9920 pixel width there is a large advantage to the 900 which is expected. What was less than expected, at least for me, that very little of the large difference that was visible in uprezzed images was visible in the print.

I made a similar test using images from "Imaging Resource" but in that case the A700 image was a JPEG and the A900 image a "raw". The results in that case were similar, but the difference between the two was more obvious.

My own suggestion is simply that A700 in combination with Lightrooms image processing pipeline is good enough for A2. The images were printed using the Lightroom pipeline, capture sharpening, uprezzing to 480 PPI and output sharpening for glossy.

Best regards
Erik

Erik,

I am wondering if there may have been some other variables that could have degraded the A900 image more than the A700 image (even though the A900 appears sharper on screen at high mag is it 'optimal').

1- You used two different lenses.  What f-stops were used for your comparison?
    I haven't examined your jpg images in detail (function of f-stop) but  theory would suggest f/5.6 might  be optimal for both lenses.  If lenses were stopped down
    beyond f/9 or so diffraction effects may be at play, degrading contrast and resolution.

Erik: f/8 on both
2- Was a sturdy tripod used for both the A700 and A900 images?
Erik: Velbon Sherpa Pro 630 and RRS BH-40 head.
3- Mirror locked up?
Erik: Yes.
4- What about focus errors?
    Choice of f-stops and distance from target will of course affect depth-of-field as well as diffraction effects and provide a tradeoff with focus errors.
    Focus bracketing and/or manual focus might be worth doing as well as comparisons at several f-stops.
Erik: Focusing errors were entirely possible. Autofocus was used. The painting was essentially flat, so DOF was not really an issue.
5 -Wind, causing vibrations of either test target or camera?
Erik: Possible but not probable.

Just some things to consider,
Ken

[BJL]

First, thanks Erik for the comparisons. The French magazine Chasseur d'Image takes a similar approach (printing at equal size from different cameras and then scanning for their published comparisons) and comes to a similar conclusion: up to about A3, 12MP more or less matches any higher pixel count; A2 prints can benefit visibly from more than 12MP.

The value of the a900 lies in ... (2) the very important ability to crop as little as a third or a quarter of the frame and still get an acceptable print. This is very important to me as I photograph wildlife that is invariably too far away for the lens I happen to have on the camera when I see it!

If the need to crop is due to not having a long enough lens to fill the frame, the only thing that will help (other than a longer lens) is more resolution in lp/mm probably through smaller pixels, not more pixels of the same or greater size on a larger sensor.

Photographing the same subject from the same distance with the same focal length, a subject that fills the 12MP A700 frame fills a bit less than 11MP worth of the A900 frame, and in general, any "same focal length, same subject distance" image cropped to the same FOV will give a somewhat lower pixel count with the A900 than with a A700, Nikon D300 or Canon 40D, and distinctly less than with a Canon 50D or any recent 4/3 DSLR. (For now, the new 12MP E30 is the DSLR leader in lp/mm of resolution, and thus in "telephoto reach per mm of focal length".)

It is when one needs more than 12MP after cropping, or better low light performance, that the new 24x36mm sensors paired with larger lenses have a true advantage.