Multimedia
Blogs
Forum
Affordable HD
Whitepapers
Newsletters
Events
Advertisers
Blogcast
MillimeterHigh-Resolution DV
Mar 1, 2005 12:00 PM, By Steve Mullen
Sony’s HVR-Z1 and HDR-FX1 1080i HDV camcorders ease the transition from DV to HD.
Sony’s HVR-Z1 camcorder shoots 1080i HDV video and features a 14-bit A/D and a 14-bit DXP that help prevent highlights from blowing out.
The Sony HVR-Z1 and HDR-FX1 might be the most elegant camcorders ever designed. Other than the XLR connectors on the Z1, sold by the Broadcast & Professional Systems division, the two camcorders are physically the same. What distinguishes them are the firmware-enabled features provided by the $5,946 HVR-Z1. In a coming issue, we'll examine the many Z1 features and detail the procedures you can use to work with both camcorders. But let's start with a look at a few of the Z1 features and discuss the underlying technology that the Z1 and the FX1 ($3,700) share, with a few insights from my experience using both camcorders.
The primary feature that distinguishes the Z1 from the FX1 is that the Z1 enables you to select between Region 60 and Region 50 modes. Region 60 mode provides NTSC DV/DVCAM, 1080i60, and CineFrame 24 and CineFrame 30. Region 50 mode provides PAL DV/DVCAM, 1080i50, and CineFrame 25. These options, plus the XLR connectors, are features that some videographers simply must have.
The FX1 provides NTSC DV, 1080i60, and CineFrame 24 and CineFrame 30, while the FX1e provides PAL DV, 1080i50, and CineFrame 25. Some Region 60 videographers are buying the FX1e with the belief that CineFrame 25 provides an optimal path to film (see “CineFrame 25 and CineFrame 30” sidebar, p. 18).
The Z1 and FX1 both have an F1.6-2.8, Carl Zeiss Vario-Sonnar T* (anti-reflective coating) lens that accepts 72mm filters. A built-in ND filter has two settings: 1/6 and 1/32. The 12X zoom covers a range of f = 4.5mm-54.0mm. The 4.5mm to 54.0mm lens is equivalent to a 32.5mm to 390mm lens on a 35mm camera. You can zoom via variable servo control, a non-perpetual zoom ring, and a short lever mounted on the ring.
Both camcorders feature an optional peaking setting and a special magnification mode. In my experience, peaking seemed almost invisible, and the magnification mode cannot be used while shooting. Therefore, I relied on AF and One-Push AF (while in Manual mode). Perhaps because of high image resolution, the AF is amazingly accurate.
Pixel math
Sony developed a new 1/3in., Super HAD CCD for its HDV camcorders. Each CCD has 1.12 megapixels that provide an effective pixel count of 1,070,000 pixels — 972 (horizontal) by 1,100 (vertical). Vertical smear level is rated at a very low -107dB. Of the 972 horizontal pixels, 960 are used for capturing an image.
Obviously, 960 pixels are half the 1920 pixels specified in the ATSC standard for 1080i video. To understand this discrepancy, we need to look at “pixel aspect ratio.” The ATSC 1080i standard employs square pixels. (Diagram 1.)
During the time for each line scan, 1920 pixels should be transferred. However, most HD video systems cannot record pixels at that rate. Therefore, DVCPRO HD records 1280 pixels, and both HDCAM and the MainProfile@High1440 (MP@H-14) MPEG-2 encoding used by Sony's HDV camcorders record 1440 pixels. So how are 1280 or 1440 pixels mapped to 1920 pixels? The answer is that each pixel is rectangular — not square. In the case of 1440, each pixel needs to have a 1.33:1 pixel aspect ratio, as shown in Diagram 2.
In order to position correctly Sony's 960 pixels within a scan line that is designed to hold 1920 pixels, the CCD elements must have a 2:1 rectangular pixel aspect ratio — as shown in Diagram 3.
If all three CCDs were aligned, the output would be 960 pixels per line. By offsetting the green CCD one half-element spacing from the red and blue CCDs, a source of additional luminance information is created. By combining output from all three CCDs, horizontal resolution is increased by 150 percent. And indeed, when 960 is multiplied by 1.5, the result is 1440. Naturally, that leads to the question of how real is the extra resolution obtained by using “pixel offset” technology.
The simple answer is that when resolution tests are performed, the horizontal resolution will be that expected from a CCD that is 1440 elements wide. The complex answer is that with pixel offset technology, the effective horizontal resolution is a function of the colors, the color patterns, and motion of objects in a scene.
The Z1 and FX1 employ “interlace scan dual-line” CCDs. For every upper field, the CCD driving logic selects the top-most row (row 1). The first line scanned out is the sum of each element in CCD row 1 added to the corresponding element in row 2. The second line scanned out is the sum of each element in CCD row 3 added to the corresponding element in row 4. The last line scanned out is the sum of each element in CCD row 1079 added to the corresponding element in row 1080. In this way, the CCDs output 540 lines (one field) of 960 elements that are output and digitized within 1/60 second.
For every lower field, the CCD driving logic selects the second CCD row (row 2) as the first row. The first line scanned out in a lower field is the sum of each element in CCD row 2 added to the corresponding element in row 3. Two immediate benefits arise from summing rows. First, CCDs become 6dB more sensitive to available light. Second, noise is averaged and thereby canceled, increasing the signal-to-noise ratio.
When a CCD adds row pairs, the image is essentially passed through a low-pass filter. (A low-pass filter removes high frequencies representing fine detail.) This filter helps prevent “interlace flicker” and “interline twitter.” Both twitter and flicker are reduced, but not eliminated, when horizontal edges (or lines) are softened by the low-pass filter. Twitter and flicker are two of the most objectionable artifacts of interlace scanning.
A price must be paid for the benefits of row-pair summing. The low-pass filter decreases a field's effective vertical resolution by about 25 percent — from 540 lines to about 400 lines per field. Therefore, effective vertical resolution should be about 800 lines when operating in 1080i60 mode.
Signal processing
Conversion of the three signals from the CCDs is accomplished by a 14-bit A/D. The digital data are then processed by a 14-bit DXP. The wide-word-length converters and DXP prevent highlights from blowing out. These camcorders deliver the best in high-contrast images I have ever seen from 1/3in. CCDs. DXP also supports gain settings of 0, +3, +6, +9, +12, +15, and +18dB. The HVR-Z1 and HDR-FX1 employ the ITU Rec. 709 HD colorspace.
Diagram 4: Gamma Modes
Gamma settings
All video cameras modify the CCD signal with a gamma function. This function can be represented by a gamma curve. (See the dark blue curve in Diagram 4, page 18.) As the CCD input signal increases it is modified in a nonlinear fashion and output. The FX1 offers a single CinemaTone mode, in which the gamma curve (the pink curve in Diagram 4) is relatively linear compared to the “normal” (dark blue curve) “video” gamma curve. This decreases mid-tone scene brightness, which Sony believes creates the look of some film stock.
The HVR-Z1 offers two gamma settings. CineTone 2 (pink) is the same as provided by the FX1, while the CineTone 1 setting provides a gamma curve (light blue) that crushes the mid-tones slightly less. I found that in high-contrast scenes, both CineTone modes drove shadow details to black, thus unnecessarily decreasing latitude — latitude that was already at least three stops short of what I really needed. While you might believe that the point of CinemaTone is to prevent highlights from blowing out, I found that the camera did not overexpose highlights when CinemaTone was not engaged. In fact, AE was usually spot-on. I never encountered a situation in which the FX1's normal gamma did not deliver an optimal exposure — except, latitude was obviously limited for high-contrast situations. Of course, that is why most HD cameras use 2/3in. CCDs.
The Z1 also includes a Black Stretch mode (green dots) that alters the normal video gamma to provide more shadow detail. I found it very useful in high-contrast situations because it seemed to increase latitude by almost one stop. Moreover, the Z1 should have a near-linear gamma mode for shooting video that will go to film.
Camera audio
Two-channel audio can be obtained from the built-in stereo mic or via an external mic. The Z1 has dual XLR connections and has dual audio level controls. The FX1 provides a 1/8in. mini-jack, and its audio level control adjusts the level for both channels simultaneously. I found that automatic gain control (really only a limiter) delivered perfect audio — with peaks at -14dB. Switching to manual, if I raised peaks to 0dB, the audio was obviously distorted. One should consider -10dB to be the limit for manual level control.
DV audio is carried as PCM data, while HDV audio is carried as MPEG-1 Layer 2 data (MP2). This precursor to MP3 is, like MP3, a “perceptual” encoding system that discards audio information that DSP computations indicate will be masked by other audio information. In short, it is a lossy encoding system that might not meet all field-recording requirements.
How does it look?
The Z1 and FX1 are also accomplished DV/DVCAM camcorders. Both camcorders offer “over-sampled” NTSC (or PAL) video. You can select between 4:3 and anamorphic 16:9 aspect ratios.
So how does the 1080i60 HDV image look? The first time I saw a sample of 1080i, I viewed it on a 1920×1200 23in. Apple Cinema HD Display. The sample looked very much like the video you would get from a Sony PD170 or VX2100. However, what was amazing was that the 1080i video was playing in a near full-screen window with the same image quality you would get from DV — only if it were playing it in a relatively tiny 720×480 window.
Next, I viewed HDV 720p30 (JVC) and 1080i (Sony), as well as 1080i from HDnet and Discovery HD and 720p60 sports from ESPN and ABC. Although there is a very distinct colorimetry difference between the Sony and JVC HDV camcorders — Sony provides the analytically accurate color the company is known for — there was no major difference in fine detail.
When, however, both HDV samples were compared to broadcast HDTV, there were major differences in apparent sharpness and overall picture impact. HDV simply lacks the quality obtained from the far more expensive equipment. Therefore, I cannot confirm the claims made by some Sony HDV buyers that FX1/Z1 video looks better than “any SD camera” and “rivals the images produced by a CineAlta.”
If you want to understand why there should be these differences, you'll find this article on Modulation Transfer Function (MTF) to be very informative:http://bg.broadcastengineering.com/
ar/broadcasting_hdtv_lenses_mtf. Simply put, the “apparent sharpness” of an image cannot be directly predicted by either the CCD or the recording resolution specifications. It requires that contrast remain high — in both lens and camera — as image detail becomes increasingly fine. To achieve a high MTF requires very high-quality, and thus very expensive, components.
Bottom line, the HDR-FX1 and HVR-Z1 deliver exactly what DV shooters need as much of the world transitions from SD to HD: high-resolution DV as provided by “HDV.” Whatever use you now make of a Sony PD170 or VX2100 for NTSC/PAL video productions or filmmaking, you'll be able to use a Z1 or FX1 for HD productions. Not only can you obtain high-resolution video from these camcorders, you can do so at a cost not appreciably different from that of their lower-resolution brothers.
See the next issue for more about working with the HVR-Z1 and HDR-FX1.
CineFrame 25 and CineFrame 30
In CineFrame 25 and CineFrame 30, de-interlacing is used to create video that has a temporal resolution of either 25fps or 30fps, respectively. De-interlacing can be accomplished in many ways, and image quality depends on the type of de-interlacing that is employed.
De-interlacing begins with the dropping of every odd field. Because the camera uses interlace scanning, CCD output vertical resolution drops significantly, but light sensitivity remains the same as for 50i/60i. Because only the even field is employed, there are no interlace artifacts. The odd field of the 1080-line output frame is “interpolated” from lines in the even field.
When CineFrame 30 was selected, I found that a 1/60-second exposure looked better than 1/30-second exposure because it was far less blurred. The 30fps video, obviously softer than 1080i60, was pleasing because it did not have the strobing look of 720p30 from the JVC HDV camcorders.
The Z1's CineFrame 25 can be transferred to film easily. However, the reduced effective vertical resolution in CineFrame 25 makes it less than an optimum choice for video-to-film applications. Rather, CineFrame 25 can be used to create widescreen progressive PAL DVDs. Likewise, widescreen progressive NTSC DVDs can be made using CineFrame 30. In these applications, the loss of vertical resolution will not be harmful.
CineFrame 24
According to Sony, in CineFrame 24 mode, the camcorder “synthesizes a 24fps temporal rate — not 24p — video from 1080i60.” Because interlace scanning is used in CineFrame 24, light sensitivity remains the same as for 60i. (Region 50 camcorders do not offer CineFrame 24.) Sony has claimed this is accomplished by using the de-interlacing process used in the CineFrame 25 and CineFrame 30 modes — with a clever twist.
Diagram A shows the capture time (in milliseconds) for each field of 59.94Hz interlaced video. The yellow cells indicate those fields that are selected to become 24fps video frames. As you can see, unlike in CineFrame 25 and 30, both even and odd fields are selected. Nevertheless, each field is time-coherent and so has no interlace artifacts. It does, however, have significantly less effective vertical resolution.
Diagram B shows the capture time (green text) for each frame (in milliseconds) of 23.976Hz video. By appropriately selecting fields (blue text) from the 59.94Hz interlaced frames using a 1:1:1:2 cadence, every 10 fields (five interlaced frames) are converted to four non-interlaced fields. These four fields do not fall on exact 1/24-second intervals. The error (red text) indicates, in milliseconds, the error between the time of exact 1/24 intervals and the selected field. These fields are captured 20 percent earlier than they would normally be. The unequal 24fps sampling time is one of the factors in the look of CineFrame 24.
Once the 24fps fields have been selected, they are expanded to 1080-line frames using interpolation. One of the results of interpolation is that each frame has a different effective vertical resolution — not true of 24p. When CineFrame 24 is selected, you should select a shutter speed of 1/60. (Strangely, neither camcorder offers the option of a 1/30-second shutter speed in CineFrame 24 mode.) This high shutter speed prevents the natural motion blur that results from the 1/48 shutter speed typically used with 24fps video. Therefore, CineFrame 24 video has a strobing look.
The combination of lower effective vertical resolution, uneven 24fps cadence, non-equal frame-to-frame vertical resolution, and strobing yields “24fps” video that looks very different from 24p video, which Sony describes as a film look.
As shown in Diagram C, 2:3:2:3 pulldown is applied to the 24 samples to generate 1080i60 video that can be encoded to MPEG-2 and recorded to tape. (Six times each second, four samples are converted to five frames, yielding 30 frames.) With 2:3:2:3 pulldown, two split frames — number 3 and number 4 in bright red text — are included within every five video frames.
You can edit 2:3:2:3 video as though it were 1080i60 video. Simply do not cut to a split frame. Alternatively, you can apply reverse 2:3:2:3 pulldown in post to recover 24 frames for each second of video. While the four samples can be recovered, the recovered frames will continue to have temporal errors as well as low vertical resolution.
There are alternatives to CineFrame 24 — and CineFrame 25. For a film look, software applications work with 1080i60 or 1080i50 video. Other software applications can create, from 1080i60 or 1080i50 video, 24fps and 25fps video, ready for editing and transfer to film. Such applications de-interlace video while preserving maximum effective vertical resolution for each frame. Shooters will have to experiment and use their practical and aesthetic judgment to get the results that will best suit their projects. From that perspective, the HDV format — and the inherent features of any manufacturers' camcorder — must always be regarded as a starting point. In upcoming articles, Video Systems will further explore this topic.
feedback
To comment on this article, email the Video Systems editorial staff at vsfeedback@primediabusiness.com.
Continue the discussion on “Crosstalk” the Millimeter Forum.
