January 29, 2010
My Petavoxel post about the “Great Megapixel Swindle,” from January 19th, has now been viewed more than 60,000 times. (Which I find kind of scary.)
And there were lots of noisy comments elsewhere, complaining that I had stacked the deck unfairly. I made a teensy crop from a cheap camera—and no surprise, it looked bad!
Fair enough. So let’s start over, stacking the deck as much as possible in favor of the photo looking good. Please look at a couple of new images:
(I am grateful to both gentlemen for posting these images under Creative Commons licensing. Paulo was also kind enough to email me his straight-from-camera JPEG; it’s the source of the crops below.)
I choose these because the camera used to take both is the Panasonic Lumix TZ5. This was among the final high-end point & shoot models to receive a full profile from DP Review; so you can have an independent opinion of its quality.
A 9 megapixel model from 2008, the TZ5 has now been replaced by a zoomier 12 Mp version. But the earlier TZ5 remains the second most popular Lumix on Flickr. Its pixels measure about 1.7 microns wide—that’s actually on the large side, compared to today’s typical compacts.
DP Review were rather enthusiastic about this camera, praising the lens in particular. It’s a Leica-branded 10x zoom, with 4 aspheric surfaces and 11 elements, including one of ED glass. This is some rather high-end stuff.
I’d be the first to admit our sample photos look crisp and vibrant at any normal viewing size. And both Paulo and Simon tell me they’re quite pleased with their TZ5’s.
Note that these well-lit shots show off the TZ5 under the best possible circumstances. The sensitivity is at ISO 100 (the lowest setting), which gives the least noise. The shutter speeds are easily high enough to freeze any hand shake at the focal lengths used. But the apertures have been held to the widest possible, which minimizes diffraction.
Even with all these advantages, we begin to see some artifacts that any compact-sensor camera is prone to.
Within the same photo, a white surface in sunlight can be 500 times brighter than a dark surface in shade. That’s about 9 f/stops of difference.
And smaller pixels inherently have lower dynamic range. So even in well-exposed photos like these two, the highlight areas can blow out to a textureless white:
Examining a highlight area in Photoshop, we see that the brightness levels in the selection have completely “hit the ceiling,” so that no detail can be recovered:
So small-sensor cameras will always struggle with high-contrast scenes (unless some tricky HDR processing is used). Yes, shooting in RAW could rescue a bit more from the highlights; but I don’t know of any camera under $350 which offers this option.
What about graininess in the image? (Digital-camera fans use the word “noise,” since we’re discussing electronic signals.)
Even under uniform illumination, photon counts can vary randomly between neighboring pixels. The larger the pixels, the more this averages out. Small pixels inherently have more random brightness variations—that is, higher noise.
So any compact camera must devote some fraction of its processor power to noise reduction, attempting to smooth out those speckles.
Selecting higher ISO sensitivity settings means the pixel noise is amplified even more; and so, noise-reduction must get even more aggressive. Ultimately, this causes a loss of detail, as DP Review’s TZ5 test clearly showed (scroll down to the hair sample, especially).
But our ISO 100 samples should be best-case scenario for noise. Yet even so, the NR processing is leaving some “false detail” in areas that should look smooth:
The effect is stronger in the darker parts of an image. And as you raise ISO, it will just become more obvious.
Might these simply be JPEG artifacts?
These samples are not highly compressed files—the JPEGs are about 4 MB each. Compression with JPEG can add visible “sparklies” around high-contrast edges; but flat areas of color should compress very cleanly. I say this texture comes from noise (or, noise reduction).
A camera of 9 megapixels would seem to promise rather high resolution. But this camera’s sensor is about 1/3rd the area of an aspirin tablet. So is there truly any detail for all those pixels to resolve?
Lets look at some 100% crops near the center of the photos, where lens performance ought to be at its best.
The high-contrast edges here show a lot of “snap.” But the camera’s own sharpening algorithm may take some of the credit for that.
Looking at the lower-contrast details in the shadows (like the hanging tags), the impression is noticeably softer. I feel that the optical resolution is starting to fall apart before we reach individual pixels.
The TZ5 lens gains a whole f/stop when zoomed to the wide end. At this setting, detail is a bit softer still—notice the grille in the archway. At a 4.7mm focal length, depth of field is enormous; so I don’t believe we’re seeing focus error here.
Note that the Airy disk at f/3.3 is about 4.4 microns (for green light). As I discussed before, diffraction means that no lens, no matter how flawless, can focus a pinpoint of light any smaller than this.
But a 4.4 micron-wide blur covers almost four of the TZ5’s pixels. And at any smaller aperture, the Airy disk expands even further. Thus, I remain skeptical that packing in pixels more densely than in the TZ5 would extract more true, optical detail over what we can see here.
Now, let me be clear.
Paulo tells me, “I love my TZ5. It’s my everyday use camera (being much lighter and more compact than the 500D).”
There’s a need for cameras like this. Among pocketable models, it’s probably among the nicest available.
But even under these best-case circumstances, we start to see some limits imposed by its 1.7 micron pixels. And since the TZ5’s release in 2008, pixels in point & shoots have only shrunken more.
It’s not an accident that DP Review made the final winner of their “enthusiast” roundup Panasonic’s own LX3—a small camera, but with a larger sensor. Its pixel density of 24 Mp per sq. cm is half of today’s worst-case models. The result is that the LX3’s “high ISO performance puts most competitors to shame.”
I’m not against small cameras. I’m not even against high megapixels, if you have a genuine need for them (assuming the sensor is large enough).
But today, pixels have shrunk too far. It’s time to stop.
January 28, 2010
The model designations of Panasonic’s recent Micro Four Thirds cameras can be a bit confusing. In order of release so far, there’s the G1, the GH1, and the GF1.
The GH1 is the largest, and most video-oriented of the three. The newer GF1 is the compact model, lacking the faux-SLR hump on top. The GH1 is a model fewer people seem to know about, at least in still-photography circles.
Anyway, DxO Labs just announced their rating of the GH1 yesterday, and it’s a bit of a bombshell. DxO ranked the GH1 as the best-performing Four Thirds sensor they had ever tested— “Micro” or otherwise. Their discussion of the results makes for quite an interesting read.
DxO, based in France, does very “numbers-oriented” performance testing on camera sensors—but only in RAW mode. This means the quality of the in-camera JPEG conversion is not evaluated.
Depending on what you are trying to photograph, different aspects of a sensor’s performance might matter to you most: Its resolution; its usable dynamic range from dark to bright; or its noisiness in low light. But DxO also attempts to assign a one-number “DxO Mark Sensor” rating, ranking overall image quality.
Anyway, you can look at a direct comparison between the GH1, the original Panasonic G1, and the Olympus E-P1 using this link. (They have not yet tested the GF1. The E-P2 and E-P1 are presumably quite close.)
Anecdotally, the GF1 appeared to lag a bit behind its µ4/3 peers in high-ISO performance. It seems safe to conclude that the GH1 is truly about one f/stop better than the GF1 in low light; that is, the GH1 images look much less grainy at ISO 800 and above.
While pixel size is still the single biggest determinant of image quality, this does demonstrate that sensors in the same size class can still show differences. Different technologies may offer greater pixel sensitivity or lower read-out noise; and it’s clear the GH1 is wringing everything possible out of its 4.3-micron pixels. (in contrast, APS-C cameras might have pixels 5 or 6 microns wide.)
This GH1 news is both exciting and exasperating for me.
It’s exciting because it proves it’s at least theoretically possible for a µ4/3 body to give what I need: A no-apologies ISO 800 setting, for available-light shooting.
I also really like the GH1’s “aspect agnostic” sensor. By making the chip a bit oversize, it’s possible to switch between 4:3, 3:2, and 16:9 proportions without changing the diagonal angle of view. (See the bottom of DP Review’s introduction for a diagram.)
In most 4/3-and-smaller cameras, the 2:3 ratio is just a crop out of a native 3:4 sensor. I prefer the classic film proportions, and would like to use that option without penalty.
On the exasperation side, the GH1 is the largest and most expensive Micro Four Thirds camera now made.
You can’t even buy it unbundled from its expensive zoom—which only opens to f/4.0. Videographers might appreciate its 10x zoom range; but for me, it’s just a boat anchor.
Wasn’t making cameras more compact (while keeping sensors large) the entire point of µ4/3?
After all, you are losing true reflex viewing. And the Four Thirds sensor still pays a penalty from being smaller than APS-C. Several current Nikon DSLRs earn DxO Mark Sensor scores over 70, compared to the GH1’s 63.3. So without a significant size advantage, µ4/3 is hard to justify.
My other irritation is this: I thought the superiority of the GH1 was going to stay “my little secret.” As other µ4/3 models came out, I had a fantasy that the GH1’s price would tumble, and I could snap one up as a bargain closeout.
Ah well, I guess that won’t happen now.
January 27, 2010
Gather ’round the campfire, children; Grampa has a story to tell, about a wild and primative time long ago.
In about 1996, a techie friend of mine went out and bought one of the very first digital cameras a normal civilian could find. It cost something insane, like $600; and you held it like a weird pair of binoculars:
It had no removable memory (it could store 48 photos internally), and no zoom lens, either. In fact its lens was “focus free” —meaning anything closer than 4 feet was a blur. It was an electronic Brownie.
Most remarkable is that the images were only 756 x 504 pixels. That means about 380,000 total. (Check the manual if you don’t believe me.)
Yes, children, this was a dark age before megapixels. Today, we’d round that off and call it 0.4 Mp.
I would be the first to admit this was inadequate. You could maybe get away with 756 x 504 on the web, unmagnified; but as for cropping and enlarging those images, there was no hope. Even a 4-inch-tall print could look a little raspy and lacking in detail.
In 1998, people really took notice when the film-photography powerhouse Canon brought out their first “serious” digital camera, the PowerShot Pro 70. It had autofocus, an F/2.0 zoom, and Compact Flash slots. And—it could shoot 1.6 megapixel images.
Those specs seemed impressive enough to outweigh its $1100 introductory price, and its cartoonishly odd appearance:
Brrrr! But hey—at least 1.6 megapixels is four times as good as that sad old Kodak, right?
Well actually… We need to stop for a minute here, and talk about what “resolution” really means.
Doubling resolution means two finely-spaced details can still be distinguished, even when they’re half as far apart. (We’re assuming here that the lens is perfectly sharp, and only the sensor resolution matters.)
To get this doubling, the spacing between the pixels must be scrunched down, so there are twice as many of them per inch.
But remember—a sensor chip has both width and height. To get a doubling of resolution, you must quadruple the number of pixels:
And each time you want to double the resolution… The pixel count quadruples again:
Now we fast-forward to 2006. The explosion of mass-market digital cameras is in full swing. Cameras have shrunk to the size of soap bars, even including a 3x zoom.
And despite this, linear resolution has doubled again, compared to the 1.6 Mp Canon. Now, sensors were up to a crazy 6, even 7 Mp. Six million pixels!
One notable example was the FujiFilm F30, a 6 Mp model:
With some clever sensor design and noise-reduction techniques, its images stayed quite usable even at high ISO sensitivities—startlingly so, compared to its contemporaries.
But what of today? Surely technology should be marching onwards! Don’t we deserve another upwards ratchet in the resolution race?
Well, it’s not that simple. First off, 6 Mp is a plateau which satisfies almost all of the real-world uses we put our photographs to.
We can make excellent 8×10″ prints, view them onscreen (even zooming in considerably), and spare our hard disks the strain of bloated file sizes.
Nikon’s D40 DLSR was one of the runaway success-stories among recent digital cameras; it remains the 3rd most popular Nikon among Flickr users. It was built to be affordable, and is hardly as stout as Nikon’s pro models—but you don’t hear many owners whining about inadequate pixels. Click on a few sample shots and check for yourself.
And you guessed it: Six megapixels.
But, on the remote chance that you really do need more resolution… keep in mind our quadrupling math above. If for some reason, 6 Mp doesn’t satisfy you, the next step up isn’t 10, or 12.
It’s twenty-four megapixels.
The good news is, those cameras do exist! Including a basic lens, I think we can fix you up for about $2400.
So enjoy that extra resolution. I’m sure you have some perfectly good reason to want it.
January 26, 2010
Even as I try to warn people away from high-megapixel point & shoots, some commenters have despaired that my advice isn’t very practical: Virtually every model sold today is 10 Mp or higher.
There’s a reason the last “sensible” point & shoot, the beloved Fujifilm F30/F31, actually went up in price after being discontinued. And even three years later, a used one can change hands for over $200. This despite the fact that aside from the superior 6 Mp sensor, the F30/F31 models were otherwise totally ordinary.
But all of us do need small, take-everywhere pocket cameras. Your cell phone cam can only do so much, with its grainy images and lack of controls.
So what specs would a point & shoot need to have, before I could recommend it? Well, here’s some thoughts—or perhaps just a poignant yearning for the impossible.
The recent Canon S90 proves it’s physically possible to put an f/2.0 lens and a 1/1.7″ sensor into a pocket-sized camera. It’s not quite as slim as Canon’s old Elphs. But it’s still smaller than the venerable 1979 Olympus XA, the breakthrough model which first showed the world how fine a shirt-pocket camera could be.
The S90 includes All Mod Cons: A jumbo 461,100-dot LCD; face detection; RAW option; and a movie mode (though not in HD, weirdly).
The so-called 1/1.7″ sensor size is actually 7.6 x 5.7 mm. The S90’s pixel density of 23 Mp/sq. cm means each pixel is about 2.1 microns wide.
Canon gets credit for scaling back the megapixel count in some recent cameras, compared to earlier models (the S90 is 10 Mp). But to my thinking, diffraction and noise still make 2.1 micron pixels pretty borderline. But is that a problem for the image quality, really?
Well, DPReview hasn’t tested the S90 yet. However its big sister, the Canon G11, recently got the full DP Review workup. And since both cameras apparently share the same sensor and Digic 4 processor chip, the results should be similar.
Unfortunately when you look at the tests at different ISOs (scroll down to the green feathers), you see that by ISO 800, the noise-suppressing algorithm is also blurring away lots of fine detail. And Canon is actually overstating its “800” speed slightly—in reality it’s closer to 640.
For our daily snapshots we rarely need a bazillion megapixels, not for any real-world use. You can make crisp 8×10″ prints, or get a nice magnified view on your computer, with only 6 Mp. (And even then, you’d still be ahead of James Cameron!)
Let’s say you took the chip dimensions of the S90, but held it down to 2828 x 2121 pixels (6 Mp total). Each pixel would be 2.7 microns wide—65% more area than those in the S90. That’s a significant difference. High ISOs wouldn’t need such aggressive anti-noise smoothing then.
But-but-but… Fewer pixels! Wouldn’t you lose detail doing that? No—at least not at any smaller lens opening than f/4.5. By that point, diffraction blur is much larger than 2.1-micron pixels.
Could we also hope that dropping to 6 Mp would also knock some bucks off the S90’s $400 price? That’s awfully steep, considering how inexpensive an entry-level DSLR is today.
Okay, here’s my next crackpot request: No zoom lens.
I realize from a marketing point of view, this sounds insane. Isn’t a 4x zoom better than a 3x zoom, and a 12x zoom best of all?
The problem with a zoom is, as jack-of-all-trades, it is master of none. A zoom is inevitably larger than a single-focal-length lens, and not always as sharp.
But the biggest problem is that zooms cripple the maximum aperture. For many typical ones, f/3.5 is the brightest f/stop. (Yes, they might be a little better at the widest zoom setting. But when a lens is labeled something like f=8–24, 1:2.8–5.9, the latter numbers tell how the the widest f/stop dims as you zoom in.)
You can make a single-focal-length lens (often known as a “prime”) much faster. Like two stops brighter. I think a smart marketer might get some mileage out of the promise, “gather four times as much light!”
A small image format actually makes it easier to design a fast lens, compared to e.g., one for DSLRs. If you look into closed-circuit television cameras, you’ll discover oodles of f/1.4 or even f/1.2 lenses that cost less than 90 bucks.
If you compare how much sharper a point & shoot image looks at ISO 200 versus 800, it raises an interesting thought. Your extra two stops of lens brightness might let you crop the image quite a bit harder, while still getting adequate detail. In effect, you could “zoom” after the fact, at home on the computer.
A wide maximum f/stop would also let you throw backgrounds a bit out of focus, if desired—a tool creative photographers appreciate. (Admittedly, any DSLR will be better at this.)
And lenses starting from f/1.7 would permit a greater range of possible apertures, before you hit the diffraction limit (about f/5.6 or smaller, with 2.7 µm pixels) and begin to lose sharpness.
But I admit, the no-zoom option is probably something grandma wouldn’t go for.
So that can be the special “Petavoxel” edition. And believe me, I’d pay for it.
January 25, 2010
So, this new movie premiered a little while ago. You might have heard about it. The title was Avatar.
The mixture of live action and computer graphics sent reviewers into a hyperventilating frenzy. Even those who thought the story was a bit zero-dimensional found its imagery “dizzying, enveloping, vertiginous” and “epochal” among a bunch of other adjectives.
It was also legendarily expensive to create. Depending how you figure it, developing the technology and then completing the film probably went north of 300 million dollars.
So for shooting the live-action sequences, Cameron probably used some pretty high-end gear—right?
It might surprise you learn that those astounding, compelling images were shot using 2.2 megapixel cameras.
I’m not making this up. Avatar’s 3D camera rigs were developed by Cameron’s director of photography Vince Pace. In a fascinating article, he talks about the technology, and mentions that most of Avatar’s live-action 3D shots were done with pairs of Sony F950 HD video cameras. Download the brochure, and you can confirm that these models use (three) 2.2 Mp sensor chips (one for each RGB channel)*.
This might be less surprising when you consider that “full” HDTV is also not that many megapixels. Go ahead, multiply 1080 by 1920 (I’ll wait). Just a hair over two million, no?
Yet we all seem to feel that 2 Mp looks pretty darn nice, even when blown up to a 52″ screen.
Typically, computer graphics for theatrical releases are rendered at a what the industry calls “2K” resolution. This means 2048 pixels across the frame width. The vertical pixel count varies, depending on the aspect ratio. But for 16:9 proportions, it’s under 2.4 megapixels total.
Let that sink in for a moment. An industry that spends billions of dollars creating sock-o imagery, intended to be seen at billboard dimensions, thinks two megapixels is fine.
Oh yes, but won’t they kick that up even higher, the minute new technology permits?
Well, perhaps. There is some movement in Hollywood to introduce “4K” as a new standard. This means 4096 pixels across the width of the frame. District 9 was shot in 4K (despite the dearth of theaters so far which can project 4K). And CG films are sometimes rendered in 4K for IMAX release.
That pushes you up to around 9 megapixels per frame.
So you would imagine that James Cameron, who adores whizzy new imaging technology like no one else on earth, would be all over 4K.
Ah, actually not. Towards the end of this interview in Variety, Cameron says that, “4K is a concept born in fear [….] I would vastly prefer to see 2K/48 frames per second.” (Today’s film standard of 24 frames per second has trouble showing fast motion smoothly.)
Nine megapixels? Cameron doesn’t need it.
*Yes, this means the camera resolves color edges better than a 2.2 megapixel Bayer-filter sensor. But what our eyes mostly care about is the luminosity resolution.
January 24, 2010
Lets say you want to own the ultimate digital camera. You want the image quality to be absolutely state-of-the-art; something that can see even better than the human eye.
You hire yourself some really smart engineers, and you go to work. So, what kind of camera specs do you end up with?
Well, the little gadget I’m referring to is the “Wide Field Camera 3.” It was installed during the Hubble Space Telescope repair mission last May (after a several-year delay, while the Space Shuttle fleet was grounded).
And the specs are kind of interesting. Does it boast a bazillion megapixels? It does not.
The larger of the two detectors, the wideband UVIS, includes 16.8 megapixels (4102 x 4096). Considering that back down here on earth, you can buy a 56 Mp digital back for a mere $30,000, perhaps that’s surprising.
The secret is that each pixel is really, really big: 15 microns wide. That gives each one quite a large area to absorb light; and it allows each pixel to store lots of electrons (for a wide dynamic range from darkest to brightest).
Each UVIS pixel is about 7 or 8 times the area of the ones in your typical consumer DSLR. But compared to mini point & shoots, it’s even more ridiculous. Their teensy pixels are about 1/100th the area.
Anyway, the WFC3 does take pretty nice pictures:
Even when price is no object, remember it is still fiendishly hard to make a large CCD completely free from flaws. Instead of one big chip, WFC3 needed to use two, pushed side-by-side.
And even the chips carefully selected to fly in the UVIS instrument have some pretty obvious scars (er, “prominent local structures”) that must be subtracted out of the final image.
Anyway, if you decide to you can’t live without this ultimate in digital cameras, I do have a couple of cautions:
And even though delivering the camera was a little slow, it wasn’t exactly cheap. You better budget a little more for shipping: 1.1 billion dollars.
At that price, you might as well go back to shooting film.
January 24, 2010
So, here’s what happened in these past few days here at petavoxel:
This is all great and flattering, of course. But I am a little concerned that 90% of those folks only saw my “diatribe” post. That article did kind of a shaky job of explaining the technical issues, compared to my other ones about noise, and diffraction. There were also two followups that clarified some points, and decoded camera pixel sizes. But the “Swindle” headline got the hits.
Ah well, that’s the internet for you.
But in more positive news—I promise I have absolutely nothing to say about the Apple tablet.