(Page 6 of 7 pages for this article « First  <  4 5 6 7 >)

Wednesday, May 21, 2008

Review: Sony HVR-Z7 & HVR-S270 1/3” 3-CMOS HDV camcorders

Sony gets serious about HDV with two fine cameras based on the same core components.

Performance

Optics

The Zeiss zoom is a good lens for the price. Its wide angle of 4.4mm is comparable to that of other 1/3” camcorders (the Z1’s lens starts at 4.5mm, as do the lenses on the fixed-lens Canons; the HVX200’s 13x starts at 4.2mm) and its 12x zoom range is perfectly adequate for most work—you can always change lenses if you want more on the telephoto end.

Fully wide, the lens has a slight but mostly harmless bit of mustache distortion. Between 6-10mm, minor pincushion distortion is present; beyond 15mm there’s no distortion of note.

Maximum aperture ramps smoothly from f1.6 to f2 as you zoom in. There is almost no portholing (edge darkening or vignetting) even at full telephoto; it only affects the outermost edges of the image and is unnoticeable without a waveform monitor—and stopping down to f2.6 evens it out completely.

The lens is adequately sharp even wide open, but doesn’t really come into its own until around f3; at wider stops, it loses some contrast and fine detail. Its sweet spot is around f3.4 or f4; aliasing artifacts on the zone plates of the DSC Labs Multiburst chart carry out to 1200 TVl/ph, showing that the camera is sensor-limited more than lens-limited. Starting at f4.8, diffraction limits come into play, and sharpness decreases as the lens is closed further—a basic limitation of 1/3” HD imaging. Fortunately, the three stages of ND filter let you stay close to the sweet spot in most lighting conditions.

The biggest flaw of the lens is in its chromatic aberration, and even so it’s no worse than average for lenses in this class of camera. Zoomed in fully, it’s nicely corrected, but as one zooms out, more C. A. appears, as radial green/magenta fringing.


The lens at full wide, showing chromatic aberration on a mourning dove (inset at 200% zoom)

I’ve seen C.A. this bad or worse on most other lenses in this price range, as well as on some much more expensive broadcast lenses, though the EX1’s Fujinon and the HVX200’s Leica Dicomar show that C. A. isn’t unavoidable in low-cost zooms.

Of course, if the stock lens fails to please, you can change it out for any other 1/3” lens (Fujinon makes quite a few), or, with adapters, choose from 1/2”, 2/3”, or Sony Alpha DSLR lenses.

Resolution and Detail

These cameras use Sony’s ClearVid CMOS sensors, employing a diagonal array of 1440x810 photosites. When run through Sony’s Enhanced Image Processor (tm), these sensors yield both actual and interpolated image samples, the latter being synthesized from four surrounding photosite values (schematic).

We’ve seen this before using 1/4” sensors on the HVR-V1, where the system gave sharp images with a bit of aliasing, mostly on diagonals; I characterized those images as being comparable to those of the 1440x1080 Canon HDV cameras, though with a slight bit more “granularity” to them.

[Note that marketing materials for the V1 referred to its sensors as “960x1080 (effective)”, but the specs for the camera show them as 1440x810, same as on the Z7 & S270, although in a 1/4” size instead of 1/3”.]

The 1/3” Sony ClearVid images appear to be comparable, but improved: the “granularity” or “grittiness” on near-diagonals is reduced, probably due to both better processing and the higher fill-factor of the larger sensors (just as Panasonic’s 2/3” HPX500 makes a noticeably smoother, less aliased image than the 1/3” HVX200 despite having essentially the same raw resolution). In difficult shots of grass and foliage swaying in a breeze, the HVR-V1’s chips would sometimes betray their nature on fine-lined edges, but similar shots made with the Z7 and S270 looked as clean and naturalistic as those made with an EX1 (using full-res, true 1920x1080 CMOS sensors), and almost as sharp, too—an impressive result, and nicely matched to the 1440-sample-per-line resolution of HDV recording. The only times I’ve seen these cameras limited by their sensors is in the presence of a extremely high-contrast contour, such as a specular highlight, neon tube, or the like, in which case some steppiness can be seen along the edge of the detail.

On the charts, the Z7 and S270 cleanly resolve 800+ TVl/ph both horizontally and vertically; arguably they resolve out to 1000 TVl/ph, but mild aliasing from the four-pixel interpolation overlays detail at these spatial frequencies. There’s residual aliasing all that way out to 1200 TVl/ph, but it’s at a low amplitude (at least it shows what the lens is capable of!). As one might expect from the design, diagonals show the most obvious aliasing, as you can see from the chart:


Multiburst chart, 24p @ f/4, 0dB. 100% scale vertically, horizontally upsampled 1440->1920 pixels in Photoshop

Bear in mind this extract was made from an HDV 24p recording at 1440x1080 samples; in E-E mode, the zone plate was as detailed horizontally as vertically, but HDV’s 1440-sample recording low-pass filters the horizontal axis to a maximum of 800 TVl/ph.

Practically speaking, the images offer stunning levels of detail; aside from the occasional slightly steppy edge in extreme-contast situations, there’s nothing in the pictures that indicates they’re made with anything less than full-resolution sensors.

However, that level of detail is compromised at high gain settings; as gain is increased above 9dB, resolution drops: at +21dB, it’s down to 600 TVl/ph. It isn’t just a build-up of image noise; the outer portions of the zone plate simply merge into undifferentiated gray areas as pixel-averaging appears to come into play to keep noise levels down.

The CMOS chips use a 1/60 second rolling shutter, so whip pans and very fast motions may show some tilt or distortion. Practically speaking, it takes a very fast move to show any noticeable tilt; it’s not an issue in most shooting.

Sensitivity, Noise, and Dynamic Range

I measured the Z7 and S270 at ISO 400, regardless of mode. This equals the EX1 in 1080p mode, is a third of a stop faster than the HVX200, a stop faster than the Z1, and two stops faster than the V1.

Some of this speed may be due to improvements in the chips, and some due to more amplification in processing. The cameras’ images are as noisy at 0dB as the EX1’s and the Z1’s at around +3dB to +6dB, and still show about a 3dB difference at +18dB (that the comparative noise performance improves slightly as gain is boosted fits with the apparent resolution loss at high gain: if the ClearVid/EIP combo is using pixel-averaging to improve noise at boosted gain levels, that’s what we’d expect to see). Put another way, the images are about as noisy at 0dB as the HVX200’s; it’s not something that calls attention to itself, but if you look into solid areas, you can see some moving “grain” in the luma channel (chroma noise isn’t an issue until very high gain levels).

If you want cleaner images, the Z7 and S270 offer both -3dB and -6dB gain settings. Going the other way, you can boost gain to +21dB, though I found that going much beyond +12dB was inadvisable due to both noise and resolution loss (I also found that boosting gain after the fact in Final Cut Pro gave me comparable noise but with more resolution; a +9dB scene boosted to match a +18dB scene matched for overall noisiness, but held more detail).

When pointed at a backlit Stouffer 41-step grayscale wedge, the Z7 gave me a solid 8.3 stops of dynamic range at 0dB, ranging from white clip down to the level where step changes were lost in the noise. This performance is comparable to other HDV camcorders. If I dove into the menus and stretched the black gamma, I might see even more; sadly I had returned the cameras before I thought of that test.

Recording Modes and Quality

These cameras record 25 Mbit/second DV, DVCAM, and 1080-line HDV.

In HDV interlaced modes, the cameras shoot 60i, 24p, and 30p as previous Sonys did: 2:2 pulldown for 30p, 2:3 pulldown for 24p. Two 24p modes are available: plain old 24, and 24A, as found on the HVR-V1. In the latter, “the phase of the 60i conversion is reset each time recording starts”, which lets some NLEs handle 24p extraction more easily: for example, if you bring such clips into FCP from CF cards or the HVR-DR60 disk recorder using Sony’s plugin for the the Log & Transfer window, FCP will remove the pulldown and record the 24p clips as ProRes422.

These cameras add true progressive HDV recording modes, recording 30p as 30 frames, and 24p as 24 frames, with no pulldown into interlaced formats. While these progressive formats are not backwards-compatible with older gear (currently, only the Z7, S270, and the M35 VTR can read them), they offer a more efficient use of the fixed bit rate, and in 24p especially the resulting data stream doesn’t need to be “reverse telecined” to extract the original 24 frames. For example, using Final Cut Pro 6.0.3 and feeding it a mixture of interlaced and progressive clips from the same tape, FCP captured all the interlaced material as 1080i60 regardless of camera frame rate, while the progressive 30p and 24p material showed up in the bin as 1080p30 and 1080p24 respectively.

In terms of image quality, 1080-line HDV is still pushing the limits at 25 Megabits/second, but the true progressive 30p and 24p seemed a bit more robust in the “shakycam conniption” test: I whip-pan, dutch-tilt, and snap-zoom the camera wildly and spasmodically, trying to break the codec. While I could cause compression conniptions—severe, mosaic-like blocking defects over one or more frames—in all modes within a couple seconds, the 24p progressive recording seemed to break up perhaps 30% less frequently than the 60i recordings, while the 30p footage was halfway between the 24p and the 60i footage.

In real-world footage, as with other HDV cameras, single frames show compression artifacts if you look for them (see the 1:1 multiburst image, and look at the areas around the text and the edges of the patterns; I chose a JPEG compression level that didn’t visually degrade the image beyond what HDV had done to it), but in normal playback the images look fine, with perhaps a bit more visual noise as scene detail and motion increase. It takes pretty radical frame-to-frame changes to cause enough breakup to be seen in normal playback, and I didn’t record any real-world shots that showed any problems, despite having used the Z7 for a run’n’gun documentary.

The cameras can shoot interval footage, and in DV modes they offer single-frame recording as well. They also have the same “Smooth Slow Recording” feature seen on the V1: they’ll shoot short bursts (3-12 seconds) of low-resolution, high-speed footage (about 4x real time, or 120fps) to a memory buffer, then record it to tape. It’s a useful trick for quick-and-dirty motion analysis (golf swings and the like), but the images are of poor-enough quality that they won’t intercut with normal images: imagine a Fisher-Price Pixelvision shooting heavily-compressed 120fps in color, and you have a rough idea of how the images look.

Audio

The supplied shotgun mike records clear, vibrant sound, including a full range of frequencies (no, I don’t have a frequency plot, I just know I got a lot of good-sounding audio). As is usually the case, the on-camera shockmounts are barely adequate to isolate the mike from handling noises, though they’re better at isolation than a built-in mike would be.

Channels can be set individually to mike or line levels, and to auto or manual gain. When AGC is used, it can be coupled across two channels (useful for correlated, stereo channel recording) or decoupled so that each channel is controlled separately (useful for uncorrelated inputs, like separate mikes on two different subjects). I used the Z7 on a doc, with the shotgun feeding one channel in AGC and the other in manual gain; both sounded very good, and I didn’t notice any nasty audio artifacts in either track (but I didn’t do a formal evaluation of the sound quality, so it’s quite possible I missed something).

The S270 offers four channels of audio, of which the first two will be audible in older HDV players—shades of Hi8 days, when we (sometimes) had two channels of AFM and two of PCM! In my limited testing, channels 3 & 4 sounded as good as channels 1 & 2.

When I captured the 4-channel clip in FCP 6.0.3 using FireWire, I got channels 1 & 2, but not 3 & 4, and FCP’s HDV capture preset doesn’t let me change anything about how it works (I could work around this by capturing using an AJA or Blackmagic Designs card, of course. Ingesting from CF card using Log & Transfer would probably preserve all four channels, too, but I didn’t test this).

Sony tells me that all four channels are present in the i.Link data, so you may have better luck with other NLEs—I’d expect Vegas to support it ASAP, because it’s Sony’s product, and Edius, just because those Edius folks always seem to be on top of the newest formats before anyone else.

Picture Profiles

The cameras offer six sets of custom presets, called Picture Profiles. These let you change the way the camera renders its images:

  • Vary the master black level, along with R, G, and B black levels individually.
  • Choose a gamma setting from Standard (like 709 gamma but with slightly compressed highlights), Cinematone 1, Cinematone 2, ITU709, ITU709 with a toe slope of 5 instead of 4.5 for increased shadow separation, PD to better match DCR-PD-series camcorders (slightly crushed, contrastier shadows), and x.v. for x.v.Color (wide gamut) recording (like 709 with slight black stretch). Cinematones have depressed midtones, as seen on other Cinematone-capable Sonys.
  • Black gamma: change the shadow gamma range between low/medium/high, and vary black stretching or compression over a 15-step range.
  • Knee: Choose auto or manual knee (on the S270, this is done with the bars/cam/dcc switch); set the auto knee’s sensitivity and maximum allowed white level; set the manual knee’s setpoint and slope.
  • Color mode: Choose between standard, 709, and two Cinematone matrices, and mix between the selected matrix and standard over an 8-step range. I found the standard matrix looked very much like 709 on other cameras, while 709 seemed oversaturated.
  • Color level (saturation) and phase (hue).
  • Color depth: for each of the six primary and secondary colors, vary the brightness of that color while keeping its saturation constant: the waveform monitor changes as you tweak a color, while the vectorscope doesn’t!

    Color depth: red at -7, 0 and +7 settings

  • Color Correct: pick one or two colors in a scene, and vary their hue and saturation independently, or vary their R gain and B gain independently.
  • WB Shift: vary the camera’s idea of white, either by warm/cool and green/magenta controls (thus giving you both “color temperature” and “tint” controls), or by R gain and B gain.
  • Detail: vary the detail level over 15 steps; turn detail on or off; change H/V balance; positive-going/negative-going enhancement balance; black and white limits; crispening; and highlight detail setting. In sub-$15,000 cameras, only the PMW-EX1 gives you more control of the detail signal.
  • Skintone detail: choose a color to undo or limit the effects of detail enhancement on.
  • Name the profile; copy it to another profile setting, or reset it to defaults.

Other Functions and Features

  • You can set an assignable button to PHOTO, and grab still images to Memory Stick even as you’re recording to tape or CF card.
  • Up to 99 camera setups can be saved to a Memory Stick, and moved between cameras.
  • Shot Transition lets you store two combinations of focus, zoom, and exposure settings and move between them in a programmed manner. It’s like a shotbox, or a camera-parameter motion-control system, great for “hitting marks” in a repeatable manner.
  • Focus Marking replaces most viewfinder displays with a linear focusing scale. You can set one or two focus marks on the scale, and the camera will show you the focus marks and the current focus position, much like you would with a follow-focus control. Very nice, but its modality—displacing most of the other onscreen displays—is distressing.
  • The cameras offer clear scan shutter speeds as well as fixed, preset shutter speeds.
  • You can choose the image area that the auto-exposure system uses.
  • You can vary the responsiveness/speed of auto-exposure and auto-white-balance tracking.
  • You can enable x.v.Color mode, which overrides many of the image adjustments, to record in an expanded-gamut mode. The visual effect on a normal display is that saturation appears to drop slightly; on a wide-gamut display, the saturation range of colors is expanded.
  • HD pictures can be downconverted to SD on all outputs (including i.Link) in letterbox, squeeze, or edge-cut modes. Downconversion quality is very good.
  • DV/DVCAM can be recorded in 4:3 or 16:9 modes.
  • Timecode is fully tweakable: Preset vs. Regen, DF vs. NDF, etc. (note that timecode is always 30fps timecode; in progressive 24p recording, the clip timecode is discontinuous, skipping one frame number in five. This doesn’t seem to cause any problems in FCP 6.0.3, but your mileage may vary).

Next: Conclusions, Pros/Cons/Cautions, Other Resources

Cameras

(Page 6 of 7 pages for this article « First  <  4 5 6 7 >)



My First Shoot with the Sony F35

Art Adams | 11/16- 06:41 PM

Dymo DiscPainter Review

Matt Jeppsen | 10/31- 09:10 PM

Scarlet Killers - Attack of the HD DSLRs!

Steve Hullfish | 10/30- 01:47 PM

PMW-EX1 Firmware 1.11

Adam Wilt | 10/21- 04:15 PM



Name:

Email:

Location:

URL:

You must be registered to comment.

Smileys

Remember my personal information

Notify me of follow-up comments?

Submit the word you see below:




Advertisements
















Copyright 2008 ProVideo Coalition LLC