Revelation about F828 files......

Interesting find Lin. I've noticed much the same thing, but not been able to identify the root. Thanks.

Now if only it didn't take 12- 15 seconds to save a RAW file. This camera desperately needs a smarter processor and a bigger buffer.

Michael

[BJL]

Thanks Lin.

   Could this be as simple as too much JPEG compression? I note that Sony's highest quality "fine" JPEG files are only about 2.4MB, whereas for comparison the highest quality "SHQ" JPEG (2.8:1 compression) of the 5MP E-1 is bigger at 3.1MP, despite the lower pixel count. The numbers suggest that Sony is offering at best 6:1 compression.

   The whole picture suggests that their MP count has got a bit ahead of what they can currently handle well in storage and processing speed at their chosen price level, unless one chooses the slow, careful option of TIFF or RAW output.

   (It strikes me that in camera processing is often the limiting factor for pixel counts in DSLR's too, perhaps even more than noise issues: for example, it seems that Canon can go to 8MP in the 1D mark II now primarily because they have been able to increase processing speed.)

[Lin Evans]

I've been closely examining these crops and have noticed some additional issues.

If you look carefully at the white areas of reflected light (small square in the yellow bead, etc.) you will note that the edges of the reflections are cleanly defined in the RAW converted tiff and rather indistinct in the jpg.

Also, and I think more importantly, there is yellow color in the jpg at the same area where the tiff has white. At first glance it appears that the jpg has a smoother appearance of sort of a gradient effect, but this is a deception. What is actually happening is that adjacent four pixel "clumps" of yellow are "bleeding" into adjacent areas and transcending the pixel boundaries.

This not only introduces color distortion, but actually destroys detail in the luminance channel as well. Notice that on the Tiff crop black lines delineating the tiny glass bead boundaries are continuous and somewhat uniform in contrast. On the jpg side, the identical area of these lines are "broken" in regular, repeating patterns. This appears to be the result of these four pixel "clumps" again bleeding onto adjacent detail and creating distortions.

Definitely there appears to be issues here which Sony needs to address. If I save a PhotoShop jpg from the RAW converted tiff, both the edge detail and surface detail are preserved no matter the degree of compression. The four pixel "clumps" do not return so this issue is one which results from the camera's firmware algorithm which converts the RAW capture to either tiff or jpg because both have this four pixel "clump" issue. It's not captured that way, the camera introduces the issue when saving either as a tiff or jpg.

Best regards,

[Lin Evans]

Some of the problems which have plagued the F828 concern edge artifacts which, upon really close examination, show a block or "clump" of four pixels which define boundaries. When prints are greatly enlarged, this tends to result in edge aliasing of some degree which detract from the very excellent images this camera is capable of producing.

Sony didn't provide much in the way of a RAW converter, but with Adobe PhotoShop 8 there are a host of supported RAW formats which allow significant improvements in the ability to convert RAW files from numerous digital cameras including the F828.

In playing with this I discovered that it's apparently the Sony jpg algorithm which produces the annoying "stairstep" aliasing. A file captured in RAW and converted by PhotoShop to a jpg doesn't have this issue and represents a substantial improvement over the Sony jpg as saved by the camera. Fortunately, Sony saves in both a jpg and RAW format when shooting RAW. This makes it very easy to compare the results without intervening variables.

Following are tiny crops enlarged 500 percent which clearly show the remarkable differences in edge smoothness and detail between RAW captures and jpg captures with the F828. Below the crop are links to a jpg converted from RAW and it's jpg as "saved" by the camera for those interested. I tried to match the color, tint, levels, etc., as closely as possible, but they are not perfect since Adobe RAW has numerous settings which can be used. But it's still quite easy to see the differences. These differences probably would not matter too much in small prints up to 8x10 or so, but if you make really large prints, shooting in RAW and converting to either TIFF or JPG will clearly result in superior detail and edge sharpness.

Hopefully, Sony will see fit to make a firmware upgrade to this camera to correct the crude jpg conversion they are using.




http://www.lin-evans.com/samples/raw.jpg
http://www.lin-evans.com/samples/jpg.jpg
Best regards,

[Lin Evans]

Hi Michael,

It really would be nice had Sony elected to provide a decent buffer to avoid the lengthy between shots wait when shooting RAW. It seems many camera manufacturers stop just short of a finished product - guess that's what keeps the "new and improved" models selling, but it's sure annoying.

I can't take credit for discovering this issue, someone on another forum mentioned it a few days ago,  but remembering Canon's problem with their early RAW converter which created a similar pattern in reds when converting from RAW directly to jpg with the 10D made me interested enough to purchase PhotoShop 8 for the Adobe RAW which supported the F828 and investigate it further.

As I recall, Canon's jpg algorithm had some flaw which was later corrected, and unless Sony has a rather obscure reason for needing to produce this clump of four pixels in the conversion process, it would seem that a firmware "fix" could be provided making the use of RAW more optional than necessary for the best enlargement potential.

Best regards,

[Scott_H]

Now if only it didn't take 12- 15 seconds to save a RAW file.

That's about twice as fast as my E-20.  

Does the camera totally lock up while it is writing?