by Jarred Land and Barry Green
Today’s video cameras are not what they used to be. Features like 24P recording, adjustable gamma settings, widescreen 16:9, and now high-def imaging are now on the market. For aspiring filmmakers, commercial producers, wedding/event videographers, and other shooters, there are three main cameras that are garnering all the attention: the new HDV Sony FX1, the almost-as-new Canon XL2, and the legendary Panasonic DVX100A (which is less than a year old, at the time of this writing).
Panasonic has sold a boatload of DVX100’s and DVX100A’s, they’re being used by networks, commercial producers, music video producers, independent filmmakers (and aspiring filmmakers) and videographers. A film shot on the DVX100 won the Best Cinematography award at the Sundance Film Festival. Clearly established as the camera of choice for many, it now finds its position challenged by two newcomers: Canon’s long-awaited XL2, and Sony’s first foray into consumer high-definition video, the HDR-FX1.
The cameras all share some factors in common – they’re all able to record in DV format, they all use three 1/3” CCD’s, they record on DV tape, and they all cost less than $5,000. How does a camera buyer choose among these three?
We decided to put these cameras to the test. We both got on a plane and headed to NY city, with the three cameras in hand mounted side-by-side on a tripod, with the aim to shoot identical footage of resolution charts, color charts, low light tests, frame rate tests, panning tests, motion tests, still-shot tests, latitude tests, anything and everything we could think of. And while we were at it, we got to learn the cameras quite well, learning how each performs and the differences in thought process and form factor that go into the cameras. What we found was, these cameras are in many ways more different than they are alike. For the potential buyer, we believe that’s a good thing – the more different they are, the easier it will be for you to determine what features are most important for your particular needs, and what features aren’t as important, and then you can select the camera that most closely fits your priorities.
The most glaring differences are these: the Sony shoots high-definition (HD) interlaced video, whereas the Canon and Panasonic shoot standard-definition (SD) 24P and 30P progressive-scan video (as well as SD interlaced video). The Canon also features interchangeable lenses, but costs about 33% more than the other two cameras, and is more of a shoulder-mount camera. The DVX and XL2 feature professional audio subsystems, the FX1 has a very limited consumer-style audio subsystem. The DVX and FX1 feature great wide-angle lenses, where the Canon has superb telephoto capabilities.
As far as which camera is “best”, that really depends on your intended usage. We’ll examine the cameras themselves, then we’ll examine the footage the cameras produce, and finally we’ll make some judgment calls as to what camera is best suited for particular applications.
FORM FACTOR:
Both the FX1 and the DVX share a similar form factor, and they look pretty much alike. Both are primarily hand-held cameras with flip-out LCD screens. Most DVX/FX1 users will find themselves primarily using the LCD, and less frequently using the viewfinder. The XL2 is built very differently – it’s much bigger than the others, and is meant to be primarily used with the viewfinder. The XL2 doesn’t have a conventional flip-out LCD; instead it provides that you can move the viewfinder’s magnifying glass out of the way, providing access to the tiny 2” LCD screen. More on the viewfinder later, but back to the body shape: the XL2 is bigger, heavier, and looks more like a typical broadcast camera: the viewfinder fits like a broadcast/ENG camera’s viewfinder, and the zoom handgrip is more like an ENG camera’s handgrip. If physical impressiveness of the camera package is important to you, the XL2 is the more broadcast-looking camera of the three. Some clients will want to see a camera that looks like something more substantial, and while it’s not in the same league as a Sony DSR300 or DSR570, the XL2 certainly looks more like it than the FX1 and DVX do. For wedding/event videography this could be an advantage.
If small size/stealth is more your preference, the FX1 and DVX are much smaller than the XL2. The FX1 looks beefier and thicker than the DVX, although it’s “airy”, meaning that while it looks a little bigger than the DVX, they weigh the same. If your style is to shoot “guerilla” without needing to get a permit, you’ll probably get further with an FX1 or DVX than you will with an XL2.
LENS:
No discussion of the XL2 can even start without mentioning its main claim to fame: it has the ability to use interchangeable lenses. The DVX and the FX1 have a fixed lens that cannot be removed or exchanged. Canon provides a small selection of lenses that offer unique features, such as a 3x wide-angle lens and a 16x manual-control lens. On the surface this would seem to be a huge advantage for the Canon
In practice, however, this advantage is not nearly so clear-cut. While it’s true that you have the option to interchange lenses, no one lens does all the jobs one would want. Each lens in the Canon lineup requires that you sacrifice some (potentially) important feature. For example, Canon’s standard 20x zoom lens has superb optical image stabilization and autofocus, but an annoying non-repeatable, imprecise-control servo zoom and servo focus control, with no provision for true/precise manual focus and no feedback whatsoever about zoom or focus position. You can buy the 16x manual lens and get excellent manual focus and excellent manual and power zoom, but you will have to give up optical image stabilization and autofocus – they just don’t exist when using the 16x manual lens! And neither lens offers a real iris ring. So you can go to the 14x manual lens and get a real iris ring, real manual zoom and real manual focus, but then you have to give up autofocus, power zoom, neutral density filters and optical image stabilization! And none of these lenses offer a true wide-angle field of view, so to get that, you can buy the Canon 3x wide-angle lens, which offers an excellent very wide angle: but then you give up all telephoto reach (the 3x has only a 3x optical zoom), image stabilization and precision manual focus/manual zoom again. You cannot have it all, at the same time. And if you choose to buy these lenses to give you the options, you will spend thousands of dollars more for the complete XL2 kit – an XL2 with the 20x lens, 3x wide-angle, and 16x manual carries a “street price” of approximately $7,000 – or enough to buy both of the other cameras in this comparison. So while interchangeable lenses is a feature the other cameras don’t have, you’ll have to weigh just how important/useful this feature is for your intended use. If the lens is a deciding factor, please realize more than 75% of XL1 users never bought or even have used more than the standard lens... but the fact that you can if you decide to lures many buyers to the Canon camp.
For the rest of this article we will ignore the optional lenses and focus instead on the lens that Canon bundles with the camera, the fluorite 20x. The Canon lens has great glass, an excellent power zoom, and by far the longest telephoto reach of any of these cameras. For sports/events/nature photography, this can be a big advantage for the Canon. The longer lens also lets you get much shallower Depth of Field effects than the other cameras do. However, this lens comes with the worst manual control options – the servo zoom and the servo focus system offer no feedback whatsoever about lens position, and this lens is the most difficult to repeat zoom or focus moves (unless you pre-program the moves into the lens’ electronic memory system, a system shared – and expanded upon – by the FX1). When compared against the last generation of cameras (like the Sony PD150), the Canon lens controls would have been considered completely acceptable, but the bar has since been raised. It’s no understatement to say that we hated manual zooming/focusing with the XL2 lens, as compared to the DVX or even the FX1. The power zoom on the XL2 is luxurious, with 16 different speeds.
While not necessarily restricted to the lens itself, we’ll put discussions of the iris/f-stop ring in the lens section as well. The XL2 has the cheesiest iris control of all the cameras tested; it has a clicky stepping button, whereas the other cameras offer manual dials/rings that make changes smoother, easier, and with less danger of bouncing the camera during an iris change. For a professional camera designed by a company famous for its lenses, it’s really curious why the lens of the XL2 features the most amateur controls of all these cameras.
The FX1’s lens is a built-in system with a very wide-angle focal length and a nice 12x optical zoom. On the surface it looks to be directly inspired by the DVX’s lens: same 4.5mm wide-angle, same 72mm filter diameter, a manual zoom ring marked in focal length with hard stops and a zoom pin, and whereas the DVX lens is designed/certified by Leica, the Sony’s lens is designed/certified by Zeiss (both legendary German optics companies). However, there is a dramatic difference between the FX1’s “manual” zoom and the DVX’s: the FX1 “looks” like it has a real manual zoom, but in fact it’s a servo-driven motorized zoom, dressed up with a manual-looking ring. It acts and performs like a servo zoom. That’s not necessarily all bad, but it’s certainly not as nice as the actual mechanical linkage provided on the DVX. It’s definitely a step up from the XL2’s infinitely-spinning zoom ring however. When in manual mode the FX1’s zoom ring feels luxurious and smooth, and it does have hard stops at minimum and maximum. However, the lens is not directly linked to the action of the lever, and it’s possible to get the lever ahead of (or behind) the action of the lens. Also, when in power zoom mode, the ring doesn’t move! This means that the ring doesn’t reflect the actual zoom position of the lens. You could manually zoom to full wide angle, where the focal length of the lens would read as 4.5mm, and then switch to servo zoom and execute a motorized zoom, but the lens ring would still read as 4.5mm. There is no feedback of lens position when in servo zoom mode (other than an amateurish little bar graph with a “W” on one end and a “T” on the other!) For focus, the FX1 offers an improvement over the manual servo ring on the XL2: the camera reads out the focus distance of the lens in meters in the viewfinder, giving the user a good idea of where the lens is currently being focused. This is a great improvement over the XL2, which gives the user no clue as to where the focus position of the lens is. However, the distance readout can become inaccurate if you attach a wide-angle or telephoto lens adapter. The lens is reasonably sharp and looks good overall, although exhibiting some chromatic aberration, primarily showing a mild green fringe on high-contrast vertical elements on the left side of the screen and a purple fringe on similar elements on the right.
Unlike the XL2, the FX1 lens lets you have most nearly all the desired features at once: autofocus, image stabilization, power zoom, decent manual focus and a reasonably good manual zoom (better than the XL2’s at least). The lens provides multiple power-zoom speeds, and its fastest speed takes about 2 seconds to traverse the full range from full wide to full telephoto.
The FX1 has a superb-feeling manual iris dial, a large metal wheel that lets you dial in iris changes that are calibrated to ¼ of an f-stop. The changes are somewhat noticeable, but much better than whole-stop increments.
The DVX’s lens is not interchangeable, and it has the shortest focal length of the group. It also has the fewest power zoom speeds. That’s where the negatives stop, as it all gets good from there. The DVX provides true manual zoom with an actual mechanical linkage between the lever and the lens, it provides a very wide-angle 4.5mm field of view, a decent range from the 10x optical zoom, and easily the best focus ring of the bunch. The DVX provides the most comprehensive feedback to the user about lens and focus position: for zoom, the lens ring is marked in focal length, and a precise 00-99 scale shows up in the LCD/viewfinder to tell the user exactly how far into the zoom’s travel it is currently positioned. The focus ring is precise and repeatable and, while not offering hard stops at minimum and maximum, it acts with all the efficiency of a true manual focus ring. You can even convert it to hard stops with an optional adapter from Century Precision Optics. Even without it, you can execute smooth and precise completely repeatable focus moves. It provides focus position readout on a similar 00-99 scale, which not only lets you determine precisely where your focus marks should be, but the system is also adaptable to wide-angle and telephoto lens adapters (using lens adapters causes the focal plane to change, which would throw off a feet-and-inches focus scale, but doesn’t affect a numerical percentage scale… it’d be nice to have both though).
The DVX’s iris ring is the smoothest-acting of the three; the iris is calibrated in increments as small as 1/6 of an f-stop, so changes in iris can be executed smoothly and seamlessly.
As far as lens goes, the XL2 has by far the longest reach, the DVX is the hands-down winner for control and feedback, and the FX1 fits somewhere in the middle.
VIEWFINDER/LCD:
We found the XL2’s viewfinder to be the poorest of the group. It’s a huge improvement over the XL1, but it is completely and thoroughly outclassed by the DVX and the FX1. Both the DVX and the FX1 offer superb large easy-to-read high-resolution LCD’s. The XL2 offers a tiny 2” LCD panel, and the viewfinder consists of a magnifier over that panel. It may look externally like a broadcast camera’s, but in performance and usability the FX1 and DVX were much, much nicer to use.
All the cameras offer zebra exposure markers (zebras let you know what elements of your picture exceed a specified level of brightness, and are very handy for setting proper exposure). The FX1 and the XL2 offer one adjustable set of zebras and the DVX offers two adjustable sets, so you can check for skin tone exposure and also for overexposure. The FX1 and the DVX offer “peaking”, which is a viewfinder focus assist mechanism that sharpens outlines around objects that are in focus. Curiously, the FX1 will allow you to have zebras OR peaking, but not both at the same time. The DVX offers both at the same time. The XL2 doesn’t offer peaking at all (unless you get the additional $1550 FU-1000 B&W CRT viewfinder, but if you do, the peaking is much more useful on the FU-1000 than on the FX1 or DVX, and the viewfinder is much better on the whole, except that you lose any ability to monitor color).
The FX1 offers a very nice “Expanded Focus” option. This zooms in on the picture in the LCD, extracting the center patch and displaying it at full resolution in the LCD, making precise focusing much easier. This is not some sort of “digital zoom”, it’s more of a full-resolution extraction that allows you to see the frame as it really is. Hopefully the other manufacturers will take note and offer this type of function on their future cameras, as it’s very nice to have. One minor complaint, Expanded Focus is not available while actually recording – it would be nice to have the option.
For viewfinder/LCD, the FX1 is the clear winner, with the DVX a very close second, and the XL2 a distant, distant third (although $1550 additional could reverse the XL2’s ranking, with the excellent optional FU-1000 viewfinder).
IMAGE CONTROLS:
All three cameras offer quite a good degree of control over the image. All offer some manner of controlling the chroma level, chroma phase, sharpness/detail, and some provision for selecting gamma. The FX1 offers the least amount of image control, but it’s still an impressive amount for a so-called “consumer camera” (Sony markets the FX1 through its consumer division, reserving the $5946 HVR-Z1 for its professional division). The XL2 offers more control than the FX1, but curiously its menus offer the least feedback. Whereas the DVX and FX1 offer precise numerical readouts of the image settings, the XL2 forces the user to “guess”, using an amateurish bar graph. Trying to repeat settings using this system was tedious, imprecise, and annoying. With the DVX and FX1, you could write down your settings exactly, such as “Detail Level +4”, but with the XL2 you would have to try to find the middle point and then count the number of times you click the up/down buttons to get your setting. Completely amateur, and does a disservice to this camera – the images it produces deserve better. It’s also somewhat of an enigma. Why does the camera offer some controls on excellent manual dials, while relegating important items like image control to a consumer-ish bar graph system? The biggest dial on the camera is for auto-exposure compensation, a large physical wheel that lets you select 0, ½, 1, 1 ½ stops over/underexposure etc. Why not give the user professional feedback on their menu settings? Hopefully Canon will address this if they ever release an XL2s.
As far as image controls go, the DVX offers by far the most control. It offers basically all the settings the other cameras do, but goes much further, offering 7 different gamma curves, four different color matrices, more options for vertical detail and overall offering the widest variety of image settings. I have gone through and actually replicated/simulated 12 of the looks from “Magic Bullet Movie LooksTM” in-camera in the DVX – that’s how extensive the menu settings are.
For variety of image controls, the DVX is the hands-down winner, with the XL2 coming second (and would have been a close second, had the feedback to the user been better) and the FX1 coming in third (although still offering a nice array of image settings, and good feedback on them).
AUDIO:
The DVX has the best audio system of any prosumer camera. The signal-to-noise ratio, pre-amps, connectors, line/mic switch, phantom power, it’s all there. Audio is one of the best features of the DVX.
The XL2 is quite close to the DVX in connectivity, lacking only a line/mic switch on the XLR inputs. However the XL2 enjoys a distinction all its own: it can record four tracks of audio simultaneously, something none of the other cameras can do. Granted it’s in DV’s lower-quality 12-bit recording mode, but still the option exists.
The audio section on the FX1 is one of its weakest points. The FX1 has the poorest-designed audio system of any “prosumer” camera: no XLR jacks, all input is done through a 1/8” unbalanced mini-jack; no phantom power provision; you cannot control the audio level of each track individually (the audio level dial controls both tracks simultaneously) and you can’t disengage the audio limiter! We didn't’t even bother testing the audio system on the FX1, we’ll leave that to dedicated audio experts, but the compromises involved do not make us optimistic. DVX and XL2 tie for first, and the FX1 didn’t even show up for the race.
TIME CODE:
The XL2 and the DVX both have well-rounded SMPTE timecode support, including preset and regen, free run and rec run, drop-frame and non-drop frame. The FX1 has no timecode options; it runs only in rec run mode, like any consumer camera. Sync’ing cameras to timecode would not be possible with FX1’s, whereas DVX’s and XL2’s could be sync’d with free run timecode, making long-form events much easier to synchronize. DVX and XL2 tie for first, FX1 phoned in sick.
PRICE:
The FX1 and DVX are approximately the same price, with the DVX having a slightly lower street price. The XL2 is much more expensive; the standard configuration costs approximately $1000 more than the others at street prices. DVX nips the FX1 for first, the XL2 is a distant third.
THE IMAGES:
Besides general physical configuration of the camera, the imaging is the other biggest difference between them. Here we can group the XL2 and DVX together, since the imaging between them is much, much more similar than either is to the FX1. The differences are night and day, and that goes both ways: the FX1 has high-definition interlaced video, and the DVX/XL2 offer filmlike standard-definition video. Which one is “better” depends on your intended usage.
First, let’s look at what everyone’s most curious about: the HDV picture of the FX1. The FX1 produces a high-definition interlaced image. Watch it on a high-def display, and you will say "wow". Very impressive. It really gives that "looking through a window" feeling. When viewed on a standard-def display, it’s nowhere near as impressive: it looks about like a lower-cost 3-CCD video camera. But if you intend to display your work on a high-def display, and don’t mind it looking like “video”, the FX1 is in a class all by itself. It’s not in the same league as the HDCAM and VariCam products, but at 1/20th to 1/30th the price, it’s an amazing accomplishment.
For film looks, the FX1 footage looks nothing whatsoever like film. It is a high-definition video-like picture all the way, with lower latitude, less shadow detail, and a razor-sharp image (on still shots). But it’s nothing like film, and the CineFrame 24 mode is completely useless as a film emulation.
On the other hand, the DVX made its reputation on its 24P shooting capability and its filmlike CineGamma and CineMatrix settings. The DVX produces standard-def images that look exceptionally filmlike, the most filmlike images you can get from a prosumer camcorder. The 24P frame rate is a vital component to imitating the look of motion picture film (note, we’re not worried about the “movie look” here, we’re talking about imitating the look of celluloid film, which runs at 24 frames per second). Combined with the DVX’s CineGamma and CineMatrix and deep latitude, the DVX’s picture is the most filmlike of the bunch – but it’s standard definition.
The XL2 follows on the heels of the DVX in being the only other prosumer camera to offer 24P. And it does it well. The XL2’s image is nice and sharp, with filmlike motion, but the gamma and matrix curves don’t look as filmish as the DVX.
An observer watching the footage from the three cameras might say:
FX1 (on high-def monitor): “Wow, that’s sharp video!”
DVX: “That looks like a movie!”
XL2: “Did you shoot this on film? It looks like it might be film…”
Many interested camera buyers were hoping that the Sony FX1 would be the “holy grail” for independent filmmakers, offering high-definition resolution and 24-frame shooting speeds. Unfortunately, this just isn’t the case. The Sony has an extreme video look to it, and there’s no way a casual or trained observer would think it looked remotely like film. You could maybe hack-job it with post effects to make it somewhat filmic, but you'll be sacrificing resolution every step of the way, and resolution is what this camera's all about. When trying to simulate the film look, fields get dropped and blended, which lower resolution, and the motion will never be the same as true 24fps capture. The DVX and the XL2, on the other hand, actually GAIN resolution when shooting in 24P mode, which narrows the gap between them and the FX1 quite a bit (when talking about resolution on film-simulation footage). And the DVX and XL2 will still look much more filmish than the FX1.
Consider your targeted display as well: for corporate users exhibiting their work on a private HD plasma screen (in their lobby, at trade shows, etc) high-def may be important. But for distribution to the consumer, standard-def holds a nearly-insurmountable edge for the near future. There is no high-def distribution medium (other than broadcast); the distribution medium of choice is DVD and will likely remain DVD for the next several years. You cannot distribute high-def content on DVD, you would have to down-rez it to standard-def resolution. And both the XL2 and the DVX make better-looking DVD footage than the down-rezzed FX1 does. Down-rezzed FX1 footage put on DVD has the poorest low-light response, the least latitude, the muddiest colors, and the lowest actual resolution. The XL2 and the DVX just look better on DVD. If you’re an indie filmmaker, instructional video producer, or are otherwise looking to produce content on DVD, the DVX and the XL2 make better looking images for DVD.
For film transfer – the jury’s out on that. The question is, will HDV’s higher resolution make up for the field-blending/resolution robbing process of converting its inherently-interlaced image into progressive 24fps footage? And if so, will that footage look better, when transferred to film, than native 24P footage from the DVX or XL2, after they’ve been blown up to film? We don’t know the answer yet, but we intend to find out – we’re having our footage blown up to film to find the real answer. Speculation will fall by the wayside compared to actual results, and we intend to get actual results.
Which to choose – HD resolution or filmlike 24P? If you want to make DVD’s that look like film, the DVX is the clear winner here, with the XL2 a close second. The FX1 is not in that race at all, it’s off in the stands buying popcorn and watching. If you want high resolution, the FX1 is cleaner high-res footage than up-rezzed DVX or XL2 footage. If you want native interlaced HD footage, the FX1 is far superior. For content destined for an interlaced HD display, the FX1 leaves the other two cameras staring at their shoes and mumbling to themselves.
For aspiring filmmakers and commercial producers, we think the DVX is the much better choice. While it’s true that many Hollywood features and television series are now being shot on high-def, it’s important to note that practically nobody shot movies on interlaced high-def, they waited until they could get high-def with 24P. There are plenty of interlaced high-def cameras available that don’t shoot 24P, and you rarely hear of anyone attempting to make a film with them. It was the introduction of 24P on the Sony CineAlta that started the HD/24P shift in feature film production. It is the combination that makes for high-def filmlike footage. And currently there is no prosumer-priced camera that offers that combination. Until Panasonic or JVC or Canon or some other manufacturer introduce a prosumer HD model with 24P, you’ll have to choose which is more important to you: high-def resolution, or 24P motion. For film simulation, it is our contention that the SD 24P cameras deliver a much more filmlike look than you will ever get from post-processing 60i high-def HDV footage. Especially when the target delivery medium is DVD – the SD cameras make wonderful filmlike DVD’s, much better than the HDV camera does.
LOW LIGHT/SENSITIVITY/LATITUDE
Download a quick lowlight wmv here
The cameras offer different approaches to CCD pixel counts. Canon and Sony both went with megapixel high-density CCD’s, and Panasonic employs much larger CCD pixels, just enough to cover the frame. A byproduct of tinier pixels has typically been that they’re less able to gather light, are more susceptible to smearing, and exhibit less latitude. However, the Canon and the Sony are more recent products, so the question becomes: have they somehow overcome these limitations?
In short, no. The Sony, with the smallest pixels, offers the lowest latitude of all the cameras, approximately five and a half stops. The XL2 fares a little better, picking up perhaps another half-stop of latitude. The Panasonic is much more sensitive than both of them,with even wider latitude; in our testing the Panasonic was capable of resolving approximately seven stops of latitude, which puts it a stop wider than the XL2 and 1.5 stops wider than the FX1.
As for sensitivity, it was no contest. The Panasonic is about one stop faster than the XL2 and usually two full stops faster than the FX1, sometimes more. That means that to get equivalent brightness to the Panasonic, you’d need twice as much light for the XL2, and four times as much for the FX1!
In side-by-side shooting at night, the images from the Panasonic were sharp and bright, and you could see well into the shadows. On the XL2, the shadows encroached quite a bit further into the picture, and on the FX1, more of the picture would be completely black, whereas the DVX was capable of resolving detail much further into the shadows and showing much more of the picture. The FX1 handles overexposed highlights a little smoother/cleaner than the DVX does, when transitioning from max exposure to overexposure. However, the DVX has more range.
In daylight settings, the XL2 and FX1 would frequently be set at f/1.8 and the Panasonic would be at f/3.4 to f/5.6! To deliver the same exposure the Panasonic requires half to ¼ as much light. This is a major advantage to the Panasonic.
As for low-light image noise, the Panasonic exhibits quite a bit more noise than the FX1 and XL2. The FX1 and XL2 signals look quite clean, even after adding electronic gain. The XL2 and FX1 both exhibited some noise in the dark sections of the picture, but the DVX showed noticeably more. At 18dB of gain the DVX’s image was quite a bit noisier and blotchier than the FX1 at 18db, although the FX1 at 18dB showed much flatter color, whereas the DVX at 18db had substantially richer color in the subject. The XL2 responded similarly to the FX1 – much lower noise but also less color than the DVX.
The FX1 and XL2 both look to be employing significant detail coring to smooth out noise patterns. The DVX can be set up to do something similar, but in our testing we had all settings at 0/middle/default, so we didn’t explore that option. With detail coring cranked up in the DVX’s menus, the difference in noise may have been less. Given the choice of some noise vs. 2 additional stops of sensitivity and wider latitude and richer color, we’d go for the latitude and sensitivity, although you the reader will have to make your own choice. The XL2 splits the middle: not as sensitive as the DVX, but more sensitive than the FX1; not as much latitude as the DVX, but more than the FX1; and less noise than the DVX, comparable to the FX1.
Tiny CCD pixels are also more susceptible to flare and smear. Indeed on some FX1 footage, shot at night, car headlights would smear significantly, making a huge vertical spike that covered the entire height of the frame – similar behavior was exhibited on low-light tests on the candle flame. The DVX’s larger pixels mean it didn’t exhibit such behavior – the DVX resolved bright highlights cleanly when both the FX1 and XL2 rendered vertical streaking, but the XL2 did less of it than the FX1 did.
Lens Flare on FX1
For low light performance without using gain, the DVX wins in a knockout, the XL2 makes the standing 8-count, and the FX1 gets dragged off the mat by the custodian and takes up knitting. The FX1 is not in the same ballpark as the other two cameras when it comes to low-light performance and latitude. When using gain the contest narrows down significantly, as the DVX’s superior attributes (sensitivity and color reproduction) are offset by more visible noise, where the other cameras gain brightness with not much increase in noise.
DVX vs FX1 sensitivity
THE FILM LOOK
This has already been discussed, but let’s go through it again for clarification. The DVX, XL2, and FX1 all feature abilities that claim to emulate the look of film. Film captures the image all at once, as opposed to the “interlaced scanning” system of video. Film also runs at 24 discrete frames per second, versus the 60 interlaced fields per second of video. Contrary to much internet discussion and “film vs. video” wars, the “film look” has nothing to do with lighting, cinematography technique, art design, depth of field, or story – those all belong to the category of the “movie look”. The “film look” is simply this: does the image look like it was shot on film? There are miles and miles of 8mm and Super 8mm home movies that don’t look anything like a “movie”, don’t have any lighting or cinematography skills, have infinite depth of field, and don’t have a story at all, but it all looks like “film” because it is indeed film! The film look is an integral part to the movie look, but the movie look is much more: the movie look means lighting and art direction and cinematography and staging your shots and dolly and crane moves and so many other things. The actual “film look” is but one component.
So to emulate the look of something shot on film, the camera has to deliver the look of 24 frames per second, scanned all at once (not interlaced), and has to have a color/exposure response (or “gamma”) approximating film.
All three cameras claim to offer these options: the Sony offers CineFrame 24 “to emulate the look of film” and “CinemaTone Gamma”. The XL2 offers a Cine gamma setting and real 24P frame rates. The DVX offers real 24P scanning, plus three different Cine gamma settings. How do they perform?
The FX1 fails the test. Cineframe 24 is completely useless as a film look technique. The FX1 shoots 60 interlaced fields per second, and only 60 fields per second – it uses a field-dropping/field-blending technique to try to simulate 24 frames per second out of that. And it doesn’t do it well. Cineframe 24 shows unnatural motion artifacts and even double-images on the screen. Side-by-side with the DVX and the XL2, the DVX and XL2 exhibit filmlike motion, and the FX1 footage makes you ask “what’s wrong with that camera”? Furthermore, because of the field-dropping/field-blending Cineframe 24 technique, it can’t even be used for still shots, because it causes pulsing and flicker and jitter on stationary objects, especially lines that show high contrast.
The XL2 and DVX, on the other hand, offer true variable-frame-rate, progressive-scanning CCD’s that run at 24 frames per second. They render motion that looks exactly like film motion. I have run tests where the DVX was strapped to a film camera, shooting the identical scene side-by-side, then the film footage was telecine’d to video and brought into the editing program and split-screened against the DVX footage to compare. The results were identical motion: the DVX motion is identical to film motion. And the XL2 exhibits the same type of motion. For film motion, the XL2 and DVX tie, and the FX1 punts.
For gamma, the XL2 and FX1 both offer a “cine gamma” type of feature, and the DVX offers three different cine gamma settings. The FX1’s CinemaTone gamma darkens the overall picture, crushing the shadow detail and eliminating some of the precious little latitude the camera has. The DVX’s CineGamma, on the other hand, flattens out the contrast and stretches the range, adding a half-stop of extra exposure latitude at the top end of the scale. The DVX with CineGamma and CineMatrix and 24P looks very filmlike. The XL2 is a bit in the middle, gamma-wise: it looks fairly filmish, but still a little video-ish. But both are far superior, filmlook-wise, as compared to the FX1.
The FX1 is an interlaced high-def camera. Trying to make it perform as a 24P film-look camera is kind of like strapping a fairing and a sidecar onto a motorcycle and saying “there: it’s just as good as a car.” It’s not the same thing at all.
DROPOUTS
Why is there a section on dropouts here? All three cameras use the same tape, so dropouts should be the same on each, right? Well, no. The XL2 and the DVX (and the FX1 in DV mode) all handle dropouts the same, but the FX1 in HDV mode handles them very differently, and it’s a significant-enough issue that you should be aware of it.
All three cameras use miniDV tape to record on. And miniDV tape occasionally exhibits dropouts – moments in time when the recording on the tape “misses” somehow -- a little fleck of dust on the tape perhaps, or a tiny bit of the magnetic media flakes off the tape… it could happen for various reasons. But it’s a known quantity, and it happens on DV cameras at varying intervals depending on the quality of the tape being used and the state of the camera’s heads (dirty heads are more prone to it, perhaps).
On DV, dropouts typically take the form of little corrupted blocks appearing in a frame, maybe five tiny 4-pixel by 8-pixel blocks spread throughout the frame. Some people may not even notice them happening. When they do happen, they can usually be repaired in post if it’s necessary to have pristine dropout-free footage. These corrupted blocks can make it look like there are little bits of the picture that are “sparkling”, so these dropouts are sometimes referred to as “sparklies”.
Occasionally you’ll encounter a “venetian blind” style dropout, where the picture breaks up into about 9 thick rows, and half the rows “freeze” action where the other rows continue moving, and this might go for maybe half a second. I’ve never had a “venetian blind” dropout on a DVX, but used to get them occasionally on earlier cameras like the Sony VX1000 and TRV-8. Speculation is that the “venetian blind” is caused by dirty tape heads or by mixing tape lubricants; keep your heads clean and use top-quality tape and you may (hopefully) never encounter a “venetian blind” dropout.
The ratio of “sparklies” to “venetian blinds” is probably at least 50 to 1. Venetian blind dropouts are bad, but very rare. Sparkly dropouts are much more common, but much, much more insignificant.
On HDV, it’s entirely a different story. EVERY dropout causes an entire 15-frame segment to be lost. You’ll be watching your video, everything’s going along fine, and then suddenly the screen “freezes” for a full half-second! That’s a dropout on HDV.
DV handles dropouts differently. Every frame in DV is treated individually and distinctly, and if a dropout hits a frame, it’s contained within that frame and doesn’t necessarily affect any other frames (although a big dropout could, perhaps, if the affected area on the tape happens to stretch across frames). But HDV uses MPEG-2 compression, which groups frames together into groups of 15. Every frame but the first one relies on the information from prior frames. If any of those frames gets corrupted by a dropout, they ALL get corrupted, and the camera just drops all 15 frames in the group. If you’ve ever been playing back footage from a non-linear editor and seen how bad your footage looks when it drops a frame on playback, imagine how bad it would look to drop 15 frames in a row. Every dropout on HDV is that bad. Whereas on DV it might have been a tiny little sparkly dropout, on HDV it causes a half-second freeze-up on the screen every time.
Now, dropouts aren’t a too-common occurrence… in shooting about 90 minutes of source footage on the FX1, we got hit with one dropout on recording, and about six more times when putting the footage back out to tape. The frequency of dropouts is not too bad, but the severity of them is devastating – if you were shooting a wedding or a concert, and one of those dropouts hit, you may be completely hosed.
With the FX1, dropouts must be prevented at all cost. It is utterly vital that you use the very, very highest-quality tape you can get. Sony introduced a new $18 tape with the introduction of the camera, perhaps in response to the potential and severity of HDV dropouts. I’d recommend using it, at the very least. In fact, I would be extremely reluctant to use an HDV camera on a paying job without some sort of redundant recording in place – either a second HDV deck chained by firewire, duplicating the first recording, or preferably something like a Firestore hard-disk recorder (if only an HDV-compatible recorder existed; hopefully soon one will be brought to market). HDV can be a wonderful high-resolution thing if you like the interlaced video look, but if you get hit by a dropout (such as during a wedding, during the vows or something) you will be hating life – and probably fired.
16 x 9
One of the buzzwords surrounding the XL2 was its introduction of “native 16x9” recording. 16x9 refers to the shape of a widescreen television set, 16 units wide by 9 units tall, or just under twice as wide as it is tall (mathematically, about 1.78 times as wide as it is tall). Conventional television sets are 12 x 9, or (mathematically reduced) 4x3. All 1/3” CCD video cameras have traditionally used 4x3 CCD’s. The XL2 promises “native” 16x9, and so does the FX1. The DVX shoots 4x3, but can also generate an electronically stretched 16x9 image. The XL2 and FX1 shoot 16x9 but can also generate 4:3 images by using smaller portions of their CCD’s.
Which is better? It depends on what you intend to display your footage on. If you want to show your work on a widescreen 16:9 television, a native 16:9 chip would do a better job than converted 4:3 footage (all other things being equal, which they rarely are). If you’re displaying your footage on a 4:3 TV, a 4:3 chip will give you more image area to work with than converted footage from a 16:9 chip.
The XL2 shoots 16:9, but it uses a trimmed-down portion of a 4:3 CCD to do it. This means that in sheer terms of CCD real estate, the XL2 uses the same amount of CCD space as the DVX does, although the XL2 has more pixels packed into that amount of space, which lets it deliver full native resolution (the DVX gets 16:9 by electronically “stretching” its image to become widescreen, at lower resolution). The FX1 has CCD’s that are actually shaped 16:9, so it uses the full surface of its chip to produce 16:9 footage.
In 4:3 mode, both the XL2 and the FX1 lose 25% of their CCD to make a 4:3 image, whereas the DVX uses the full surface of its CCD. In 4:3 mode, the XL2 and FX1 perform like ¼” CCD cameras, whereas the DVX has the full use of its 1/3” CCD.
16:9 is a hot buzzword, but is it practical? In today’s marketplace, a wild guess and some web research show that approximately 90% of the installed base of televisions in the USA are 4:3. This is changing, and 16:9 sets are becoming more common in stores (about half the sets at the average Best Buy or Circuit City are 16:9 shaped). So 16:9’s market share is growing in the US, but it’s still quite small – the overwhelming supermajority of people are still watching TV on 4:3 sets. As for HDTV, all HDTV broadcasts in America are 16:9, but HDTV’s represent only about 3% to a maximum of 10% market penetration (according to the difficult-to-find statistics I found on the web).
16:9 would be great, if it didn’t compromise your 4:3 footage, but it does. 4:3 would be fine if it would let you also get great 16:9, but it doesn’t. The two are mutually exclusive, basically square pegs and round holes. So how do the cameras perform in 16:9 mode, and how do they perform in 4:3 mode?
In 16:9 mode, in HD mode, obviously it’s no contest – the FX1 delivers the high-def HDV picture, whereas the other cameras are standard-def. But for down-rezzing to DVD display, the XL2 delivers the sharpest 16:9 mode. The DVX and FX1 are close behind; the FX1 still shows more resolution after its HD->DVD down-rez than the DVX does on its squeeze-mode up-rez, but the FX1 is still an interlaced picture versus the DVX’s filmlike picture. However, a note on the DVX’s squeeze mode: if you’re shooting 24P in 16:9 mode, the DVX still delivers the maximum resolution that a television can safely display without interline flicker! It delivers as high-resolution as a “native 16:9” Sony PDX-10. So even though the DVX’s 16:9 mode isn’t “native”, if you’re shooting progressive-scan, it’s still very competitive and as high-resolution as a TV can safely display without risking interline flicker. No doubt that the XL2 in 16:9 mode is a little sharper though. In practice, they’re all very close in terms of overall resolution when put on a DVD.
In 4:3 mode, the DVX is king. The XL2 and FX1 do a good job of subsampling their CCD’s to deliver 4:3 images, and they do look good, but the bigger pixels of the DVX drink in the light, delivering better latitude, sensitivity and contrast, while having more than enough resolution to overpower a television display (if that’s what you want to do).
For television commercial work, instructional videos, etc., all broadcasts (except the few HDTV broadcasts) in the USA are done from a 4:3 source. The DVX delivers 4:3 footage better than the other cameras. If you want to broadcast local commercials shot by an FX1 or XL2, you’ll have to either shoot in 4:3 mode, or downrez/letterbox the image to fit within a 4:3 signal (which would give you a letterboxed widescreen-looking image on a 4:3 TV). The DVX still looks better, since its image will not have gone through a digital resize process.
When 16:9 TV’s are more common, and when US broadcasters switch to broadcasting 16:9 images, the balance of power will shift. And when (or if) that happens is anyone’s guess. For now, for today’s market, 4:3 CCD’s are a practical choice, even though they don’t have the “sexiness” of widescreen CCD’s. In the near future, say a few years from now, that may not still be the case. But we don’t see the 4:3/16:9 issue as being of too much importance yet. If 16:9 CCD’s delivered just-as-good 4:3 performance as a 4:3 CCD, obviously the choice would be a no-brainer, but they don’t, and today’s television landscape is still 4:3 (in the US; Europeans have a different set of rules to play by, European broadcasters are broadcasting 16:9 signals and 16:9 is much more important in Europe).
DEPTH OF FIELD
How do these cameras compare, regarding depth of field (and more specifically, how suitable are they for getting the “subject-in-focus, background-out-of-focus” look?) Does the FX1 do any “magic” that allows it to get shallower Depth of Field (DOF)?
No. All three are 1/3” CCD cameras (although the utilized portion and shape of that CCD varies depending on which camera and what mode you’re in). For any given focal length and aperture, they will exhibit identical depth of field. As a practical matter, in a room or other confining space, the FX1 will have just as deep depth of field as the others. In fact, the XL2 is the camera most capable of getting the shallow depth of field look, because its lens is the longest (at 100+ millimeters, the depth of field will be much shorter than the FX1’s maximum of 54mm or the DVX’s maximum of 45mm). However it’s not always easy to stage your shots such that you can get far enough away from your subject to use the Canon’s long telephoto reach. But when you can, the XL2 is the king of shallow DOF looks. In 4:3 mode, the DVX delivers the shallowest DOF for any given subject size, unless you can get far enough away to use the extreme telephoto of the Canon. Under similar circumstances all the cameras will give identical DOF, but under extreme circumstances where the Canon can bring its long lens into play, the Canon wins the shallow-DOF comparison.
HIGH-DEF: NOW? OR LATER?
The big question comes down to the FX1’s high-def resolution. As a camera, the DVX is far superior for most uses to the FX1, and the XL2 enjoys some significant advantages over the FX1. But for interlaced high-def, the FX1 is quite superior to the others. So regardless of all other issues, you have to answer three questions about high-def before you can decide between these cameras:
1) How important is high-def to you today?
2) Does this camera do what you want?
3) Must you buy today? If not, do you think this is the only high-def camera you’ll have to choose from?
1) High-definition is cool technology, that’s for sure. But it’s met with extreme indifference from the consumer. High-def has been around in some form or other for approximately 20 years, but only now has it caught any “fire”, and even then it’s only in about 3% of the households in the US. The most optimistic figures I’ve ever seen for HDTV estimate it’s in as many as 10% of US households. Contrary to popular belief, there is no government mandate that the US switch to high-def (the FCC has mandated a switch to Digital Broadcasting, but not high-definition; stations can broadcast standard-def signals and still be in complete compliance). Some people think HD is inevitable, some think it’s a non-event. Answer the question about how important it is to you, and how successful you think it will be, and how soon – and that will help you determine how much priority to put on the FX1’s HDV capabilities. The most optimistic estimates I’ve seen place HDTV as appearing in 50% of American households by the end of 2008. Standard-def will be with us for many years to come, and DVD will remain a viable distribution option for years and years. Think about it this way: the DVD player is the fastest-adopted format in consumer electronics history… people snapped up and/or converted to DVD faster than they did to CD, VHS, or even television itself. And it still took SIX YEARS for DVD players to reach the point of appearing in just half of American households. Even if HDTV were to catch on and rapid adoption were to occur, how soon will HD-DVD players become a substantial presence in the marketplace? 5 years (even faster than DVD itself?) 8 years? Standard-def DVD will be with us for quite a while. And for standard-def DVD production, the FX1 is not as capable as the other cameras in this comparison.
2) If this camera doesn’t do what you want, why buy it? If the DVX or XL2 do what you want, why buy the FX1 instead? They’re very different. If the FX1 does what you want, why buy the DVX or XL2? They’re very different. In many ways the DVX and XL2 are more alike to each other than they are to the FX1, primarily because of the high-def interlaced issue. The FX1 is an interlaced camera that shoots high-def, lower-latitude video. It doesn’t shoot filmlike images, and it’s not nearly as professional in features or feedback as the DVX. So if you want an FX1, make sure you’re buying it for what it is, and not what you think it “should be.” For an indie filmmaker customer, who typically doesn’t have a lot of cash, think about it this way: why spend a lot of money you don’t have, to buy something that doesn’t do what you want, just to get a feature (HD) you don’t need? Should you really sacrifice high-quality audio, professional lens control, 24P shooting, gamma adjustment, low light performance, latitude, picture controls, and other features, just to get high-definition recording? Especially when, in nearly all likelihood, your project will be distributed at standard-definition on DVD?
3) The appeal of the FX1 is compelling – it’s high-def and it’s today. But that high-def comes at a steep price in other ways – poor audio connectivity/controls, narrow latitude, weak low-light response, and a video-only look, as well as significant loss of resolution on moving shots, and the dropout issue. If there were never going to be another high-def prosumer-priced camera, one might be tempted enough by high-def resolution to settle for the other shortcomings. However, there are other manufacturers who haven’t released their product yet – Canon is a member of the HDV forum, and they know how to do 24P. JVC has already announced a 2/3” shoulder-mount 24P HDV camera, although it will be a lot more expensive. And Panasonic, godfathers of prosumer 24P, have yet to announce their intentions. While the FX1 makes some very nice interlaced HD footage, I have to wonder if filmmakers/commercial shooters who jump on the FX1 are going to regret it once competing products get announced. If you have to buy now, and the FX1 does what you want, then buy it – it’s a very nice camera for what it is. But if you don’t have to buy now, or the FX1 doesn’t do what you want, do yourself a favor and wait to see how things shake out. Further complicating matters is that Europe has yet to decide on an HDTV standard – some headlines announced at IBC that the EBU would be adopting 720/50P, which would make this camera incompatible in Europe! Apparently that announcement was premature or that decision was rescinded, and a formal selection of interlaced or progressive has not been made yet for Europe. If 1080i is selected, the FX1 would be a safe choice; but if 720p is selected, the FX1 may not be nearly as useful in Europe.
CAMERA USAGE RECOMMENDATIONS
The FX1 is a very nice HD interlaced camera, but that's all it is. If you were shooting a reality series, something where you need that interlaced "immediacy" look, the FX1 would be the first choice (actually the forthcoming “pro” model, the Z1, would be the first choice, but it's $2300 more than the FX1). If interlaced 1080i HD is your priority, the FX1 would be the first choice. For filmmakers... if you want HD, wait for Canon or JVC or Panasonic to respond with a real progressive-scan 24P HD camera. Big-screen filmmakers don't use interlaced HD to make movies, they waited until Sony introduced the 24P CineAlta. Trying to use the FX1/interlaced is a step backwards, in my opinion. However, the Z1 could be nice, with CineFrame 25 converted down... but then you're talking about a $5946 camera. CineFrame 25 offers a de-interlaced image with about 575 lines of resolution, so you’d have the same vertical resolution as on a PAL DVX or XL2, but twice the horizontal resolution.
For precision, control, and filmlike footage, the DVX is still the camera to beat. And it carries the lowest street price of the three cameras, has the longest history and the most established base of users (great for q&a and support). The 16:9/4:3 issue is becoming more noticeable, but right now still approximately 90+ percent of the TV's in the USA are 4:3, and neither the FX1 or XL2 can deliver the 4:3 image that the DVX can, and a letterboxed 4:3 image will give the look of 16:9 on a 4:3 set, without downsampling (which both the FX1 and XL2 would require).
For lens options, telephoto reach, and “appearance”, the XL2 is boss. Shooting a concert or live event where you can’t get close to the subject, or weddings or other situations where people are hiring the camera and expecting it to look like they paid money for it, the XL2 has the advantage. For nature type work, the XL2 enjoys a substantial advantage.
For progressive/24P footage, both the XL2 and the DVX gain resolution in that mode, whereas the FX1 loses resolution when trying to simulate 24P, so the gap in resolution closes between them. With the FX1, for a filmmaker, you're having to settle for a camera that doesn't do what you want, just so you can pay extra to get a feature you don't need (HD resolution). Five years from now when HD distribution is more realistic, that situation may need to be reviewed, but right now DVD is the reigning distribution method, and the DVX and XL2 make better-looking DVD's than the FX1 does. Transferring to film is a rare thing, but we intend to submit footage to a film transfer house to see just how well it does.
As a camera, regardless of the footage, the DVX is the most professional of the bunch. The others have consumerish compromises that the DVX just doesn't. As far as footage goes, the FX1's is easily the highest-resolution for HD, (especially on still shots, moving shots are another story) but the DVX's is the most filmlike, with the XL2 a close second. And when downrezzed to DVD resolution, the FX1 isn't any higher-res than the XL2 or DVX, so for DVD release, the FX1's high-def isn't even an advantage. (I know this conflicts with what some people have been saying; all I can say is, I've got identical footage from all three cameras here, I'm making DVD's with it using Vegas to do the conversion, going from the original 1440x1080 .m2t transport stream, using "best" video quality, and that's the results I'm observing.)
The cameras are all very different, and have some majorly different features, which is a good thing -- it actually makes it a lot easier to choose between them. Pick your priorities, then pick the camera that suits them best. They're all three absolutely incredible cameras, but they have very different "personalities" and each is suited to certain tasks much better than the others. For reality shows or high-def interlaced productions, the FX1 would be the choice. For events/sports/concerts, I'd probably choose the XL2. For weddings it’d be a tossup between the filmlike ability of the DVX vs. the bigger/more impressive camera size of the XL2. For the indie filmmaker or TV commercial producer, the DVX is still king.
Charts XL2
Charts FX1
Charts DVX
