Neither of these is a reliable long-term backup mechanism. In fact, they're a relatively risky medium-term solution as well.
As for the cost of storage, well, you can
Ok....I'm listening (this is new). What is a safe way to backup?
As for the cost of storage, well, you can
There is no absolutely 100% safe way to back up data. Vincent Bockaert has some articles about storage issues on this web site--see http://www.dpreview.com/learn/?/key=issues and http://www.dpreview.com/learn/?/Glossary/Storage/Comparison_01.htm . He recommends archiving two independent backups (one magnetic and one optical, or two magnetic, or two optical), and, for additional piece of mind, one more backup (optical or magnetic). Whatever level of protection you feel comfortable with, having 1.5MB "Normal" files as opposed to 3MB "Fine" files just means that you can buy, as in my example, a 250GB hard drive rather than a 500GB, and 50 rather than 100 DVD-Rs. Again, the difference in cost is trivial--and the 3MB "Fine" jpeg from the F31fd is still much smaller than, say, the 11MP RAW file from a Panasonic DMC-FZ8.
It's an interesting world where one considers "double the cost" to be "trivial".
Yeah ... and RAW shooters say the same thing about quadruple
This is an interesting question ... I currently have an off board hard disk that is synchronized in realtime by Allway Sync, a very nice shareware application.
That's because there isn't any difference at all. Most of us have probably done the same test, pixel peeped at 100%, 200%, and 400% even, and found no difference whatsoever!OK- I just took two shots.
One on F, the other on N.
I pulled both into photoshop and seriously tweaked the lighting
until I could see posterization in the sky.
Yes- I could see some posterization.
But- I still couldn't tell any difference between the two images.
So, I need to rescind my earlier statement.
I think the challenge that we have before us is to actually take
two side by side images. One on N the other on F, and post 100%
crops that indicate any difference at all. So far, I've tweaked
and examined an N vs F and haven't revealed any noticeable
difference at 100%.
The challenge is before us!!!
--
I had to run the 6m fine / 6m normal / 3:2 quality test on my new
f31fd, to decide for myself.
I could not see any difference just looking at the photos.
So then I did a few test shots for use with Norman Koren's IMAtest. I
respect his work; his tools helped me detect a wierdly defective
camera once.
So, here's a quantitative answer for the difference:
3:2 mode: 1661 LW/PH, around 4 mp "ideal" pixel count
6m normal: 1911 LW/PH, around 4.8 to 4.9 mp "ideal" pixel count
6m fine: 1924 LW/PH, around 5.0 mp "ideal" pixel count
Discussion
does not have the full width needed to accomplish 6.1mnp in 3:2
- I heard somewhere that 3:2 mode interpolates pixels, as the camera
ratio. That may explain the very measurable difference for 3:2 mode.
difference I cannot imagine anyone actually detecting the difference
- However measured, normal vs fine is different, yes, but such a tiny
other than looking at a single pixel.
I think I'll avoid 3:2 mode, because yes I am picky. 20% loss of
effective resolution may not be really visible, but it's enough to
make me nervous.
But I see no reason to use 6m fine mode. None at all. 100% more
storage for 2 percent more resolution? Gimmeabreak
I agree 100%. I took that a bit further in my testing, since I am not always going to have the luxury of shooting in multiple modes
... I wanted to know even the hidden differences, that I might not see or notice in any given shot.It's a 15-20 percent loss of actual resolvable, usable resolution (exactly how much depends on which measure you use.)
I'd rather crop and upsample myself as needed, rather than have the camera do it. That's all.
If you intend to crop to 3:2 format anyway (which I prefer, to match my dSLR), if you took images in the 4:3 format and cropped them, you'd loose 13% (by going from 2848×2136 to 2848×1898). I think that 13% is basically what you're seeing in the tests.
Here's the key thing. The SuperCCD design has all of its photosites arrayed in a grid turned 45° from square -- sort of like the pattern of a chain-link fence. In order to produce the regular square-pixel JPEG output in any resolution, these photosites are interpolated into pixels.
If you take a 4:3 format image and crop it to print 3:2, you print it at 13% lower ppi than the 3:2 format. If you're printing at 4×6, it's not gonna matter anyway because you probably don't have 500+ dpi from your printer anyway -- but if you were to go to 18×12, you'd be at 168ppi for the 3:2 in-camera version and only about 157ppi for the 4:3 one. Consider that in two dimensions, and there's the almost 15% difference (but in reverse).
If you're thinking about the "lost" pixels, then no. The test does not look at the entire photo. It looks at a small area near the center and detects the usable resolution (very neat trick: it examines a sharp edge, at an angle!)
This has nothing to do with SuperCCD. All digicams (except Foveon) have photosites arrayed in a Bayer pattern grid, with R, G, B near but not on top of each other. And they all "interpolate" photosites into pixels. The 45 degree trick gives finer resolution in the same size sensor, but doesn't affect the basic "interpolation" to pixels.
But those "ppi" were upsampled in the camera. They are not "real." They do not contain 13% more information. So all we've done is perform two resamplings for the 3:2 format to whatever print resolution we're using, and only one resampling for the 4:3.
(I do, but in people-testing, most don't notice anything beyond 250dpi, and nobody yet has been able to distinguish beyond 300dpi
)Since everything always is resampled for display or print, I think of it as resampling once by staying at 4:3 (i.e. it is as close to raw mode as I can get), or resampling twice if I use 3:2 in the camera. I like preserving sharpness as long as I can.
This doesn't actually contradict what I said. The 3:2 format images do indeed have less information per pixel, but the amount less is approximately (I think probably within margin of error of the test) equal to the amount by which the image is bigger than 4:3-> cropped-to-3:2. In other words, it's a little bit wasteful of space on Fuji's part to make the 3:2 mode scale up to 3024×2016, but doesn't really gain (or lose) anything.
There's definitely interpolation, of the same sort you get if you take a picture and rotate it 45° in Photoshop. Except presumably better, because this probably all happens at the same stage as the Bayer interpolation.
Which makes sense, although it is actually getting interpolated twice no matter what.
Yeah. The reason I don't is that I often do want 4×6 prints, and if I leave the camera at 4:3 it's easy for me to forget when composing the shot. Then I end up having to decide later whether heads or feet are more important.
Agreed!
