Over many years, video equipment manufacturers have responded to the trend toward digital video by producing a large number of application-specific digital black boxes. These products were developed to fulfill specific production needs. They were operating at incompatible sample rates, number of bits per sample and quantizing range, and so they are often incompatible with each other. The one thing they do have in common is a link with their analog ancestry: analog composite I/O interconnect ports that make them compatible with all-analog composite production studios.
The continuing trend towards an all-digital studio resulted in the need for digital video equipment industry standards. What resulted was the development of two sets of composite digital standards for studio equipment: the 4fSC NTSC standard and the 4fSC PAL standard.
These standards specify that the analog composite video signal must be sampled at a rate of four times the color subcarrier frequency (4fSC). The number of bits per sample plays an important role in determining the signal quality and the economics of videotape recording, so the standards allow a choice of eight or 10 bits per sample.
In North America, the initial interest in 4fSC composite digital videotape recorders (VTRs) was spurred by the need to replace obsolescent analog composite VTRs with digital VTRs that had analog I/O ports. A number of manufacturers developed such products, identified as D2 (Sony and Ampex) and D3 (Panasonic) digital VTRs. Subsequently, a wide range of compatible 4fSC digital video studio-quality equipment appeared on the American market. In Europe, however, interest in 4fSC VTRs was limited because these VTRs cannot handle SECAM.
This article discusses the sampling and quantizing characteristics that govern these VTRs as specified in the SMPTE 244M standard.
The SMPTE 244M standard sets the sampling frequency at four times the subcarrier frequency, or 14.3181 MHz (14.3 MHz nominal). The sampling clock is derived from the analog signal's color burst. Figure 1 shows the sampling spectrum of the 4fSC NTSC. The shaded area represents the suppressed subcarrier and its sidebands. In this example, the sidebands are limited to ±600 kHz. The sideband bandwidth depends on the NTSC encoder design. The SMPTE 170M standard allows narrow-bandwidth (±600 kHz) and wide-bandwidth (±1.2 MHz) chrominance sidebands.
There is a significant gap between 4.2 MHz, the maximum nominal NTSC baseband frequency, and 7.16 MHz, the Nyquist frequency. Unlike the ITU-R BT 601 component digital standard, SMPTE 244M does not specify the characteristics of the anti-aliasing and reconstruction filters. The manufacturer has the choice of developing complex and costly wideband, brick-wall, ripple-free filters that yield an extended baseband frequency response, or a moderate cost, 4.2 MHz, low-pass filter with a gradual roll-off.
The sampling structure
The SMPTE 244M standard was developed with reference to the original (1953) specifications, which used I/Q encoding instead of the B-Y/R-Y encoding currently used. Figure 2 shows that any chrominance vector can be represented by I/Q or B-Y/R-Y vectors. The original intent of the NTSC standard was to assign different baseband bandwidths to the I signal (1.2 MHz) and the Q signal (600 kHz), thus providing a better resolution of the orange visual information. Figure 3 shows the block diagram of a typical 1953 encoder. The first block of the encoder, the matrix, converts the gamma-corrected E´G, E´B and E´R primary signals into E´Y, E´I and E´Q signals. A low-pass filter limits the bandwidth of the E´I signal to 1.2 MHz and another low-pass filter limits the bandwidth of the E´Q signal to 600 kHz. The E´Y and E´I signals are suitably delayed to match the delayed narrow-bandwidth E´Q signal. The two chrominance components feed dedicated suppressed-carrier amplitude modulators in phase quadrature. An additional subcarrier phase shift rotates the two vectors with respect to the B-Y reference, as seen in Figure 2. A modern I/Q encoder as per SMPTE 170M would use equal-bandwidth I/Q signals, so the low-pass filters would be identical and there would be no need to delay the E´I signal. The I/Q-encoded NTSC signal can be decoded along the I/Q axes, with equal or unequal bandwidths, or along the B-Y/R-Y axis, with equal wide or narrow bandwidths. In NTSC transmitters and receivers, the baseband bandpass is limited to 4.2 MHz. Attempting to decode chrominance signals beyond 600 kHz would result in severe I to Q or B-Y to R-Y crosstalk due to upper chrominance vestigial-sideband effects, so receivers never take advantage of wider-bandwidth chrominance components when they are present. Very few I/Q decoding monitors or receivers were built because the circuit complications would not yield any visible picture improvements. It is surprising that SMPTE 244M was developed using obsolete chrominance signals.
As Figure 4 illustrates, the sampling instants coincide with the peak positive and negative amplitudes of the I and Q subcarrier components. The upper part of the drawing shows that these sampling instants provide an adequate representation of the B-Y/R-Y information. Given a sampling frequency fS = 14.3181 MHz (nominally 14.32 MHz) and a horizontal scanning frequency fH = 15,734.25 Hz, the number of samples per total line is equal to fS/fH = 910. The digital active line accommodates 768 samples. The remaining 142 samples comprise the digital horizontal blanking interval.
The quantizing range and its implications
Figure 5 shows the relationship between analog NTSC signal levels and eight-bit and 10-bit sample values of a 100/7.5/100/7.5 color bars signal. The 10-bit approach provides 1024 digital levels (210), expressed in decimal numbers varying from 000 to 3FF. Digital levels 000, 001, 002, 003 and 3FC, 3FD, 3FE, 3FF are protected and not permitted in the digital stream. This leaves 1016 digital levels, expressed in decimal numbers varying from 4 to 1019, or in hexadecimal numbers varying from 004 to 3FB, to represent the video signal. The sync tip is assigned the value 16 decimal or 010 hexadecimal. The highest signal level, corresponding to yellow and cyan, is assigned the value of 972 decimal or 3CC hexadecimal. The standard allows a small amount of bottom headroom (some call it footroom), levels 4 to 16 decimal or 004 to 010 hexadecimal, and top headroom, levels 972 to 1019 decimal or 3CC to 3FB hexadecimal. The total headroom is on the order of one dB, and allows for misadjusted or drifting analog input-signal levels. This reduces the signal-to-RMS quantizing-error ratio (S/QRMS) by the same amount. The theoretical S/QRMS of a 4fSC device with analog I/O interfaces is given by the following formula:
S/QRMS(dB) = 6.02n + 10.8 + 10 log10 (fS/2fmax) - 20 log10[Vq/(VW-VB)]
n (number of bits per sample) = 10
fS (sampling frequency) = 14.32 MHz
fmax (maximum baseband frequency) = 4.2 MHz
Vq (quantizing range) = 1.3042 V
VW (white signal amplitude) - VB (blanking level) = 0.7143 V
Given the above values, the calculated value of S/QRMS for a 10-bit system is 68.10 dB.
In an eight-bit system, 254 of the 256 levels (01 through FE) are used to express a quantized value. Levels 00 and FF are protected and not permitted in the data stream. The calculated theoretical value of S/QRMS for an eight-bit system is 56.06 dB.
D2/D3 digital composite VTRs appeared on the market at a time when Betacam SP composite analog VTRs were in the process of capturing the market. Betacam SP proved to be the more popular format, especially for newsgathering, because of the availability of a field unit featuring a piggyback camera, as well as complete compact editing systems. To avoid multiple NTSC decoding/encoding picture-quality degradations, editing suites used an S-Video connection with separate luminance and chrominance paths. The performance figures of the D2/D3 VTRs were superior, especially if parallel or serial digital (143 Mbits/s) interfaces were used. To this effect, several manufacturers offered digital composite production switchers with serial digital interfaces. In most cases, the D2/D3 VTRs were used as drop-ins in an NTSC analog composite environment. The appearance of competitively priced component digital video equipment has tilted the market in favor of component digital video.
Michael Robin, former engineer with the Canadian Broadcasting Corp.'s engineering headquarters, is an independent broadcast consultant located in Montreal, Canada. He is co-author of Digital Television Fundamentals, published by McGraw-Hill.
Send questions and comments to: firstname.lastname@example.org