Sony DSC-F505V Digital Camera
("First Look" Review posted 4/28/00)
||2.6 megapixels, 1856x1392 uninterpolated image size|
||Special pre-JPEG interpolation for 2240x1680 image|
||5x Carl Zeiss optical zoom lens, f/2.8-f/8.0|
||True manual focus option|
||Spot, aperture priority metering options|
||Built-in flash, MPEG movie capture option|
Sony has long been a dominant player in the digital camera field with their Mavica(tm) line of floppy-disk based cameras. At the high end of the market though, Sony has developed a compelling line of products, incorporating high-quality Zeiss optics and advanced features found on few competing camera models.
Late in 1999, Sony introduced a 2 megapixel design with an incredibly sharp 5x zoom lens by Carl Zeiss. The DSC-F505 was hugely popular, and Sony's problem seemed to be largely one of trying to satisfy the level of demand for the product. Recently, the 505 has been in very short supply, with most dealers out of stock. This has led to speculation that Sony was about to upgrade the unit. Now, the speculation has ended, with Sony indeed announcing an upgrade, the DSC-F505V.
The new model is somewhat unusual though, in its resolution specifications. The camera sports a 3.34 megapixel Sony CCD sensor chip (the same chip essentially *everyone* making 3 megapixel digicams is using at the moment), but only about 2.6 million of its pixels are actually being used. This reduced "effective pixel" count caused much speculation on the 'web as to what might account for it, but the answer in fact appears to be fairly simple: Rather than re-engineering the entire optomechanical system of the F505, Sony simply dropped the 3.34 MP sensor into the original body. As is commonly the case, the internal design of the camera was set up to mask the original 2.11 megapixel array slightly, to allow for dark-current calibration of the CCD. Some simple back-of-the-envelope calculations reveal that the same size mask applied to the 3.34 megapixel sensor would yield about 2.7 million "effective" pixels. Allowing for errors in the calculations due to different sensor geometry and layout of the active elements, a final pixel count of 2.6 million seems very reasonable.
Beyond the masking of the CCD array though, Sony's decision to use interpolation to produce an optional image size of 3.7 megapixels touched off the usual internet firestorm over interpolation, marketing claims, etc. Part of the purpose of this "First Look" review is to inject a little reality into the heated discussion of the merits (demerits?) of interpolation, particularly as it applies to the DSC-F505V.
First though, a couple of quick notes about the camera itself: Physically, it is identical to the original F505, which should please many of those who've been waiting to buy one of those units. Beyond the case and optics though, Sony has added several significant enhancements to the electronics relative to the original model. Here's a quick rundown:
- 12 bit digitization: This is a pretty significant improvement,
IOHO - We're still shooting our test subjects, but early results seem to show
that the increased digitization accuracy has indeed resulted in significant
improvements in subtle color and tonal rendering, particularly in strong highlights.
- Improved aperture & shutter control: Another pretty significant
improvement. The F505V has full 1/3 stop control over shutter and aperture,
giving much more precise control over these exposure parameters.
- External flash option: Although it uses a proprietary Sony flash
connection (Supports the same HVL-F1000 external flash unit that we're in
the process of reviewing with the Sony S70.)
- TIFF file format option: The new F505V can save images in uncompressed
TIFF format at all but the largest (interpolated) image size. - A much-requested
- Improved manual focus operation: This is a feature brought over
from Sony's new "Y2K" Mavica line. Whenever you touch the focus
ring on the new F505V in manual focus mode, the display immediately shifts
to a magnified view. This makes it much easier to focus manually using
the LCD for feedback. (Although the little in-focus indicator dot is still
- Improved processing speed: The new camera electronics use a
special Sony single-chip CPU design that eliminates the need for separate
"scratchpad" and "buffer" memory. The result should be
faster cycle times and (possibly) lower power consumption. The higher level
of integration could also lead to lower levels of image noise.
- Improved interpolation algorithm: One consequence of the increased processor horsepower appears to be that the Sony engineers could undertake much more sophisticated image preprocessing. Specifically, they were able to include a mode that interpolates the image data directly as it comes from the CCD array, rather than after it's gone through a JPEG process first. Sony's claim for this technique is that it improves detail rendition, and reduces image noise up to 16x. (Actually, the 16x claim is a little unclear, as it most likely also involves the impact of the improved digitization accuracy.)
Other than these changes, the new F505 works virtually identically to the original.
We'll be working on our usual leave-no-stone-unturned review of the new model,
but in the meantime, you can get a pretty good idea of its capabilities by referring
to our previous DSC-F505 review. We'll also
be posting sample images from the new F505V in the Comparometer fairly shortly,
so you should find samples there as early as 5/1/00.
Pixels, Interpolation, and Image Quality
It ain't just about pixels!
As noted above, the reason we've broken our promise to ourselves of "no more 'First Looks'" (they burn up time we could better spend getting the full reviews out) is because there's been such a storm of controversy over Sony's use of interpolation. Thus, the main thing we're going to talk about here is how the resolution of the new camera stacks up against both 2 and 3 megapixel models. Before diving into that discussion though, we want to make the strong point that there's more to choosing a digicam than counting the pixels! It might be a peculiarly American thing, but it seems that people always want to find one "goodness" rating for a product, and then buy whichever model has the highest score. In the digicam arena, people tend to look at resolution, or more simply megapixel counts as that universal goodness rating. It's our position that you need to consider all that you're going to do with a camera, compare features and capabilities carefully, and then buy the product that best matches all your needs. DON'T just decide you're going to buy a 3 megapixel camera and then start narrowing down the field: You may miss out on a feature that's important to you, just because you weren't willing to look below some arbitrary threshold on the resolution front. What's more, as we're about to see, resolution is more than a matter of counting pixels anyway: A camera with a lower effective pixel count may actually produce sharper images and more detail than another one with a larger sensor!
So, having just cautioned everyone to not get too fixated on pixels and resolution, that's exactly what we're going to talk about for the rest of this (short) review.
For the purposes of this article, we though it would be useful to compare digicam images from five different cameras, and both the in-camera pre-JPEG interpolation of the F505V and post-interpolation using Photoshop's(tm) bicubic spline resampling. The samples we compare are as follows:
- Sony S70 (3.14 million effective pixels)
- Casio QV-3000EX (3.24 million effective pixels)
- Sony DSC-F505V internal interpolation (2.6 million effective
- Sony DSC-F505V external interpolation in Photoshop (2.6 million
- Olympus C-2500L (2.34 million effective pixels)
- Sony DSC-F505 (original model, 1.92 million effective pixels)
What we did
We used as our primary comparison image shots taken of our ISO-12233 standard resolution test chart. In our shooting of this target, the images are very carefully framed to include just the active target area, making the scale of all images identical. In the case of cameras with different aspect ratios (width to height ratio), the standard calls for adjusting the framing to match the height of the frame to the height of the active area of the target.
To make it as easy as possible to compare the output of these various cameras, we resampled all of the images up to the same final image size, 1680 pixels high. (This matched the vertical dimension of the largest image, that from the F505V's internal resampling option.)
What?! We resampled the images? (We can hear some of you saying.) Well, yes - There really didn't seem to be any more valid way of comparing the resolution on an apples to apples basis. Bicubic interpolation is pretty darn good, and as you'll see from the samples, produces results very close to that from Sony's super-sophisticated raw-CCD-data approach. Also, the two 3MP cameras we tested were thus resampled by the same (small) amount, since they both produced 2048x1536 pixel file sizes. As far as comparing images between cameras, you're welcome to ignore the sample of the F505V's internal interpolation, in which case its files received much the same treatment as the Oly C2500, the other "2.5 megapixel" camera in this comparison.
Actually, after looking at the results, we're pretty confident that differences between images resulting from interpolation are much smaller than the inherent differences in resolution we're seeing.
Without further ado, here are two (rather largish) images showing composites of clips from our resolution test target shots. (We think the results are fairly interesting) (all megapixel (MP) numbers are "effective" pixel counts):
Resolution test targets: Horizontal resolution axis
(All images scaled to ~3.7 megapixel level)
Resolution test targets: Vertical resolution axis
(All images scaled to ~3.7 megapixel level)
These are all fairly artificial images of a test target: For more typical "real-world"
subject matter, click here to see
samples taken from our "House" poster images. Framing on these isn't
as tightly controlled, so there's more variation in image size, but the results
are generally quite consistent with the res-target clips above.
So, what did we see here? Well, a few things. First, Sony's fancy in-camera interpolation technique does indeed seem to be better than bicubic spline interpolation in Photoshop after the fact. But only slightly... In fact, it's entirely possible that one of the more sophisticated interpolation schemes (for instance Mitchell or Lanczos, as implemented by Mike Chaney in Qimage Pro) would produce better results. The DSC-F505V definitely produces improved results when compared to the original 1.92 megapixel model. It also seems to win in the resolution department over the Olympus C-2500L, if only slightly. Furthermore, no surprise, Sony's own S70 3.14 megapixel digicam (also with a Zeiss lens on it) beats the F505V. What perhaps is a surprise though, is that the DSC-F505V is noticeably sharper than the 3.24 megapixel Casio QV-3000EX. We think this last is important for a couple of reasons.
There's been some considerable consternation on the 'net that Sony appeared to be preparing to market the DSC-F505V with a "3.3 megapixel" label on it. Along with everyone else, we were concerned about this, since the camera is really only using 2.6 million of those pixels. We did hear (just today) from Sony US though, that while the camera will have "3.3 megapixel" silkscreened on its side (as did our sample unit), all of the retail packaging will clearly state 2.6 megapixels. "All the retail packaging" in this case includes the product box, the flashy peel-off labels on the side of the camera, and the "hang tag" that will be attached to the retail demo units. This seems like a very reasonable approach, and certainly no consumers should be misled into thinking that the camera works other than as it does. The 3.3 megapixel silkscreening could still be a bit confusing, but at best it'll just give the users a little more bragging rights to friends and bystanders attracted by the 505V's unique body styling! ;-) Seriously, it sounds like Sony will be taking great pains to make clear the actual effective CCD resolution of the new unit, and are to be applauded for that.
The real problem, as we see here, is that megapixel counts are almost meaningless as a determiner of digicam performance, other than in the most general sense. (As witness the comparison above of the DSC-F505V with the Casio QV-3000.) (Before we get jumped on by all the QV-3000 enthusiasts though, we'll immediately agree that there are a lot of good & valid reasons to buy the QV-3000, most of which have nothing to do with resolution... Exactly my point!) What's really needed is an objective industry-wide standard that will let manufacturers, users, and journalists talk about cameras' resolution figures in meaningful terms that are understandable to non-experts. This was the original intent of the ISO-12233 resolution test target & protocols, but its most correct and objective usage produces "Spatial Frequency Response" curves that would at best be very difficult for the average consumer to understand, and potentially even more confusing/misleading than the current reliance on pixel counts.
We've been giving this issue a lot of thought, and discussing it with other digicam-oriented webmasters, to see if there might be some light that we could collectively shed here. So far, nobody's found the magic formula, but we're encouraged by the interest and spirit of cooperation that's thus far been evident. Stay tuned on this one...
Meanwhile, what about the DSC-F505V?
So, the obvious question is "What do we think about the DSC-F505V?" Fortunately, that question at least has an easy answer: It looks like a really super digicam! The original DSC-F505 was already a great product, and we have many emails from ecstatic F505 owners to prove the point. Our email traffic gives us a fair slice of how users respond to digicams, and the original 505 had more than it's fair share of enthusiastic email in our in-box. In developing the F505V model, Sony has added a number of excellent enhancements, including higher resolution, faster processing, better tonal reproduction (more accurate digitization), reduced noise, and easier operation (the manual focus improvement). The net result is clearly another significant step forward for the digicam world: Stay tuned for our full review!