After years of hype, high-definition television (HDTV) is finally becoming a viable commercial reality. The technology is actually a subset of the broader digital-television (DTV) standard now employed by many stations. The rise of such innovations in broadcast studios has lead to the design of more and more digital components and subsystems for this growing market. Testing them has become a pressing issue.
Although this article focuses on the measurement needs of the design engineer working with HDTV, it also should be of interest to those working with other types of telecommunication signals. The wideband HDTV signal has many of the electrical characteristics of other wideband serial-digital communication signals, such as the SONET STS electrical and ITU SDH signals. To drive optical fiber, those signals are converted to light.
Just prior to transmission, HDTV signals are compressed into a 19.39-Mbit/s serial-bit data stream. This process enables RF transmission to the home in existing 6-MHz television channels. The compressed MPEG data stream may contain several video and audio programs. It's best analyzed with a specialized MPEG data analyzer.
Measuring compressed data signals that have been transmitted as RF signals requires specialized instrumentation. A spectrum analyzer may be required, as well as a vector signal analyzer and constellation analyzer.
On the other hand, equipment designed for use within the video-production studio handles the HDTV video as a serialized, 1.5-Gbit/s digital signal without any data compression. The electrical interface can be measured with a wideband, general-purpose oscilloscope. The HDTV data signal is distributed with nearly 100% data integrity on a single 75-Ω coaxial cable up to distances of 300 ft. It uses the serial-digital SMPTE 292M format standardized by the Society of Motion Picture and Television Engineers (SMPTE). This 1.5-Gbit/s data-signal format is used to interconnect HDTV equipment, including cameras, routers, videotape and disk recorders, switchers, and special-effects equipment. It also works with picture monitors and video test equipment.
Designing equipment to the SMPTE 292M input/output format allows loss-free distribution of the digital, uncompressed HDTV signal within the television studio. A single 75-Ω coaxial cable can interconnect the video signal's 30-MHz luminance and 15-MHz color components without creating problems. This avoids issues like frequency response roll-off and group-delay distortions, which may cause ringing. It also rejects the introduction of spurious components, such as hum or noise, that might affect an analog signal. This is critical because absolute fidelity is needed for the many passes that the signal must make through the cable during video production.
Another benefit is derived from designing equipment to connect using the standardized SMPTE 292M serial-digital HDTV signal format: The production studio gets the ability to distribute any one of at least 18 different high-resolution HDTV image formats. These include 50-Hz and 60-Hz interlaced and progressively scanned images, along with 24-frame/s video for movie-film-compatible production.
Today, there are two most widely used video-production formats providing live and prime-time HDTV pictures to the home: the 1080I, 1125-line interlaced and the 720P, 750-line non-interlaced or progressive formats. For example, NBC has been broadcasting The Tonight Show with Jay Leno in 1080I since last fall, while ABC has been doing Monday Night Football and the Super Bowl in 720P. Both are presented in a widescreen 16:9 format, making an impressive image compared to NTSC's simultaneously broadcasted 4:3 aspect ratio.
Distributing these high-pixel- and line-rate digital signals requires a sampling clock of 74.25 MHz. It's used to sample and digitize the camera's luminance component with 10-bit/pixel resolution. Each of two color-difference components of the camera's signal are sampled at half this rate, or 37.125 MHz with 10 bits/pixel. These two color-difference components are combined into a single 10-bit parallel data stream of 74.25 MHz.
The 74.25-MHz clock and both the 10-bit parallel luminance and 10-bit parallel color-difference signals are then serialized into the SMPTE 292M serial-digital signal (Fig. 1). That standard's resulting signal is an NRZI data stream at 1.485 Gbits/s. The signal launched onto the cable is an 800-mV p-p signal that must conform to specified amplitude, rise-time, and overshoot limitations.
To measure that signal, the design engineer has to use a 75-Ω termination at the oscilloscope input with return loss of better than 15 dB at 1.5 GHz. Because most high-bandwidth digital storage oscilloscopes (DSOs) have 50-Ω inputs, a 75-Ω to 50-Ω adapter is needed (Fig. 2).
For easy calculation, consider using a 75-Ω to 50-Ω adapter that has a calibrated attenuation recognized by the oscilloscope. Otherwise, the adapter will affect the display of the signal on the oscilloscope's screen. Then the designer will have to calculate the actual measurement values from what is displayed.
The HDTV signal has a sin(X)/X spectral distribution with a first zero or null bandwidth of 1.5 GHz and several significant spectral lobes at multiples of 1.5 GHz. This makes it seem like an oscilloscope with a bandwidth of much greater than 1.5 GHz is required to measure this signal. But many of the measurements can be done with a 2-GHz bandwidth model. Just take into account the scope's own bandwidth effect on the rise-time measurement. It's possible to correct or reduce that measured rise time by the amount it has been increased due to the measuring device's finite bandwidth. Also, to ensure that the waveform is not unduly affected by cable loss, the length of the cable should be less than 1 m.
It's easy to determine if an oscilloscope will accurately capture the HDTV uncompressed waveform. SMPTE 292M specifies the rise and fall times of the signal to be no greater than 270 ps as measured from the 20% to 80% points on the waveform. The engineer making the measurements needs to take into account the capability of the oscilloscope. In the case of a 2-GHz oscilloscope, the 10% to 90% rise time of the instrument can be closely approximated by the following:
TO = 0.35/BW
For the Tektronix TDS 794D, for example, TO = 0.35/(2 × 109) = 175 ps.
The 10% to 90% rise time of the signal source, TS, and the oscilloscope, TO, are related in such a way that the effect of the scope's rise time can be subtracted as follows:
TS = (TM2 − TO2)1/2
where TM equals the measured 10% to 90% rise time on the scope display.
To get a conservative estimate for the TSMPTE 20% to 80% rise time when measuring toward a maximum rise-time specification, divide the 10% to 90% value by 1.42. This provides some margin over a straight-line linear interpolation value. It also compensates for the typical rise-time shape of the band-limited signal. Consequently,
TM20-80 = TM/1.42
where TM is a 10% to 90% rise-time value.
These relationships are then combined. This provides a good approximation to the SMPTE 20% to 80% rise time as a function of the oscilloscope rise time, TO, and the measured 20% to 80% rise time, TM20-80, as follows:
TSMPTE = \[(1.42TM20-80)2 − TO2 \]1/2/1.42
or, after simplification,
Tsmpte = (Tm20-802 − 0.5To2)1/2
As an example, the measured value of 260 ps in the eye diagram of Figure 3 is plugged into the given equation:
TSMPTE = (TM20-802 − 0.5To2)1/2
= \[2602 − (0.5 ↔ 1752)\]1/2
= 229 ps
Similarly, the fall time as per SMPTE 292M can be measured. As in the previous case, use 175 ps for the value of TO because the oscilloscope is specified to have a 2-GHz bandwidth. Note that the given equation works well for oscilloscope bandwidths down to about 1 GHz. But the accuracy and validity of the approximations degenerate very quickly for lower-bandwidth oscilloscopes.
Comparing results with 5-GHz sampling scopes revealed that the 2-GHz-bandwidth scope measurement, when used in the given equation, is surprisingly accurate. At Tektronix, the view is that the 794D digital phosphor oscilloscope (DPO) makes it easier to see the trace profile for setting the cursors. This was discovered when the DPO was compared to a DSO with point-accumulation display.
So the bandwidth of the 2-GHz oscilloscope suffices for measuring the uncompressed HDTV signal. Now the amplitude of the signal can be measured. According to SMPTE 292M, that amplitude is required to be 800 mV within ±10%. This can be measured with the oscilloscope by using the amplitude cursors (Fig. 4).
The measurement of 852 mV demonstrates that this HDTV serial-digital source is just within the maximum limit of 880 mV. Note that the amplitude specification does not include the overshoot. It's intended to be the most probable values of the trace density's histogram, indicating the eye height that most commonly occurs.
Again, company opinion states that the DPO display makes it much easier to evaluate the distribution of the signal compared to traditional infinite-persistence and point-accumulate modes. With other DSO displays, it might be tempting to set the cursors a little farther apart because the trace density or histogram peak is not as well defined as it is in Figure 4. This could lead to a measurement that is out of spec when it is, in fact, just within the SMPTE specification limit of +10%.
SMPTE 292M specifies that the overshoot of the eye pattern be less than 10% of the measured amplitude. Figure 4 shows that this waveform has less than 10% as measured with a 2-GHz scope. However, it's not possible to confirm that the overshoot would still measure less than 10% on a much wider-bandwidth oscilloscope. As a practical matter, cable lengths of over 1 m restrict the signal bandwidth and, if properly terminated, would reduce the overshoot removed by the 2-GHz bandwidth limitation anyway.
When calculating overshoot, it's not the bandwidth but rather the termination return loss that is most important. Even on a short cable, this may cause reflections that are well below the 2-GHz bandwidth of the oscilloscope and easily measured. In other words, if a SMPTE 292M signal fails the 10% overshoot specification, it may just be due to the 75-Ω to 50-Ω pad's return-loss used in the measurement. This should be verified with a known, high-quality return-loss pad.
The jitter specification for the SMPTE 292M HDTV signal also must be calculated. To do this properly per the SMPTE recommended practice, use a bandwidth-controlled clock extractor to externally trigger the oscilloscope.
Using a general-purpose oscilloscope with sufficient bandwidth, the design engineer can easily test uncompressed HDTV digital outputs. Specifically, the rise time, fall time, and amplitude of these complex, high-speed digital signals can be captured and measured using a 2-GHz bandwidth, 50-Ω oscilloscope and a high-quality 75-Ω to 50-Ω adapter.