While virtually all TV stations have DTV transmission systems on the air, many station engineers have never performed detailed measurements on the effectiveness of those systems. While the transmitter may have passed a proof when the manufacturer installed it, how many RF engineers really know if their DTV RF system is operating properly?
I and Q modulation
This article will provide a basic understanding of some key parameters that can be checked to ensure a properly operating DTV transmission system. Armed with this background, an analog-based engineer can begin to better understand how those parameters can affect the quality of the broadcast signal and what to look for when making tests.
A key to understanding the 8-VSB is to remember that a single-sideband RF signal can be produced by summing two modulated RF carriers, one offset from the other by 90° (cosine, sine). The cosine carrier is modulated by the in-phase, or I signal (baseband data). The in-phase components are shifted by 90° to produce the quadrature phase, or Q signal, which is then used to modulate the sine carrier. The two carriers are combined, and one of the sidebands is eliminated through phase cancellation. The resultant signal vector therefore contains both I and Q components.
Constellation diagram
Figures 1 through 3 illustrate the development of a constellation diagram from the 8-VSB waveform. As the signal propagates through time, the position of the carrier vector moves continuously along the time axis. Simultaneously, the carrier vector rotates around the time axis at approximately the carrier frequency. (See Figure 2.) The tip of the vector traces out the instantaneous values of the I and Q components when measured along those respective axes.
In Figure 3, the RF signal is viewed directly perpendicular to the I and Q axes. In this view, it becomes clear how the 8-VSB carrier vector assumes constantly changing I and Q values. The tip of the carrier vector shown in the diagram represents the position of the signal at one instant in time.
Constellation units
The vertical lines on the diagram represent the eight amplitude levels in 8-VSB also known as constellation units. If a sample falls anywhere along one of these vertical lines, it indicates that the I (amplitude) component of the signal is equal to the corresponding 8-VSB symbol.
The position of the sample measured vertically along the symbol line indicates the value of the Q component. In 8-VSB, the Q component carries no data, but it does provide information regarding signal quality and transmission impairments.
In a perfect system, the sampled points would fall exactly on one of the eight symbol lines. However, propagation errors cause the sample points to fall to the left or right of the lines. As long as the samples are clustered close enough to the lines, the receiver can still distinguish the eight discrete symbol levels and recover the data. But if the spread of the samples grows too large, samples for one symbol level will cross over the threshold for the adjacent symbol level, resulting in a symbol error.
Error vector magnitude
Figures 4 , 5 and 6 are constellation diagrams of off-air signals. In Figure 4, the error vector magnitude (EVM) is about 3 percent. In Figure 5, the EVM has increased to approximately 5 percent and to about 10 percent in Figure 6. It's important to note that even with an EVM of 10 percent, the signal was still producing essentially perfect pictures due to the ability of the receiver's adaptive equalizer to correct for propagations effects.
The difference between a sampled symbol value and the theoretical ideal value can be represented by a vector diagram. The length of this error vector, or its magnitude, is the hypotenuse of the right triangle formed by the I error and the Q error, as shown in Figure 7. EVM is expressed as a percentage of the outer symbol levels.
Error ratio
Over a short period of time, there will be millions of sample points in a constellation display. It's neither practical nor useful to evaluate the error vector for individual samples, so an RMS value of the error vector magnitude is calculated for a large number of sample points and then periodically updated.
Because the original I and Q channel components are defined by the well-known narrowband filter pulse response, any calculated EVM or errors shown on the constellation display represent the effects of the transmission system on the 8-VSB signal. Therefore, when measured at the transmitter output, EVM represents a figure of merit for the overall transmitter plant.
A common way to look at overall system performance is to compare the desired signal level with background noise. This is familiar in the analog world as signal-to-noise ratio (SNR). The 8-VSB equivalent is modulation error ratio (MER).
In the 8-VSB world, a desired output signal looks a lot like random noise, so you might think that an SNR measurement would be invalid. The secret to measuring SNR in an 8-VSB system measurement is to compare desired noise-like symbol power with the noise power resulting from all other errors caused by various factors.
From the section on EVM above, the concept of the ideal theoretical symbol value can be used to calculate the noise-like power level that would be produced by a stream of undistorted symbols. This undistorted power value is then compared with the error noise power from all other sources. The error noise power can be derived from the RMS value of the squared I and Q errors. (See Figure 7.) Using the usual decibel equation, the MER value represents the ratio of the carrier power to the background noise power level.
The ATSC recommends a minimum MER of 27dB at the transmitter to ensure signal availability within the designated coverage area. MER values will decrease with distance from the transmitter due to propagation attenuation, multipath, noise and interference from other sources.
Eye patterns
An eye pattern is a display of a number of symbols overlaid on one another and synchronized to the start of a symbol time. Anyone who has worked in digital audio, video or data communications for the past few years has probably seen a two-level eye pattern. The 8-VSB eye pattern extends this to eight levels.
Even in digital transmission systems, the actual information is carried from point-to-point by analog voltage or current waveforms. Therefore, while these waveforms represent digital symbols, they are still subject to all limitations associated with analog transmission. Pulses get rounded by limited bandwidth, signals are attenuated by distance, atmospheric and manmade interference are encountered, and distortion from excessive levels may be present.
To assess the transmitted quality of a digital system quickly, look at the eye pattern of the symbols. At the symbol times, the amplitude of the displayed signal should be at or near the defined symbol amplitudes. The space between the defined symbol levels at the sample times (eyes), should be open, i.e., there should be no signal traces through those spaces. Between symbol times, the signal waveforms will be in transition between symbol values. The display will be filled in with traces from many symbols during this period.
As noise, distortion and other impairments affect the signal, some symbol traces will deviate from the defined symbol levels, reducing the amount of empty space in the eyes. At some point, the system can't distinguish the correct level of a symbol. The eye patterns will close, and an error will occur.
Pilot level
Figures 8, 9 and 10 show eye patterns corresponding to MER's of 30dB, 25dB and 20dB, respectively. Note that even with the apparently closed eyes in Figure 10, 8-VSB error correction techniques will still allow the recovery of the data and provide an essentially perfect picture.
In the 8-VSB system, a pilot signal is inserted by adding an offset of 1.25 constellation units to the eight defined symbol levels. This results in a DC offset of the baseband signal.
In single-sideband modulation, a DC offset at the input to the modulator results in a fixed amplitude component at the carrier frequency. The amount of DC offset directly determines the pilot amplitude. The 1.25 constellation unit offset results in a pilot level 11.62dB below the total average signal power in the DTV channel. For practical purposes, the pilot is independent of anything related to modulation. The pilot is simply used by the receiver as a first step in tuning.
Power levels
Measuring power levels on an 8-VSB RF signal is more complicated than for an NTSC transmitter. Segment sync pulses do not provide the same peak reference as NTSC video sync, so there is no convenient point on the 8-VSB RF waveform to use as a reference.
The peak envelope power of noise-like signals is essentially random. The noise-like characteristics of the 8-VSB signal mean that energy is spread relatively evenly across the 6MHz channel. As a result, to arrive at a single power measurement number, the energy must be integrated (added up) across the correct bandwidth.
The most accurate measurement of total average power is provided by a thermal (calorimetric) sensor. This is impractical for most situations, so stations often use a broadband, full-wave rectifier-based power meter calibrated to read true average power on digital signals. (Some rectifier-based power meters need a correction factor when measuring noise-like signals.) This measurement also assumes that only the desired station's signal is being measured; there are no significant adjacent channel components from other analog or DTV stations contained in the sample. One might be tempted to use the pilot level as a reference and infer the total average power from that point. Pilot level measured at the transmitter output can be an indicator of power level, but it is not a substitute for accurate power measurement. In the field, the pilot level can be affected by selective fading or cancellation and is therefore not a reliable indicator of received signal strength.
The IEEE Broadcast Technology Society RF Standards Committee G-2.2 is working on a draft standard for 8-VSB emissions mask compliance. It introduces the term dBDTV and refers to spectrum amplitude measurements made with 500kHz resolution bandwidths. 0dBDTV is defined as the total average power within the 6MHz bandwidth, including the pilot.
Figure 11 shows the full-service DTV emissions mask along with key reference points for an 8-VSB signal. Note the small circle at the center of the 0dB line on the diagram. This refers to the total average power within the 6MHz channel at the licensed transmitter power output.
DTV mask
The -10.63dB line refers to the flat-top portion of the 8-VSB spectrum, measured using a 500kHz bandwidth. To derive total average power from this, 12 500kHz bands must be integrated across the 6MHz channel, with appropriate corrections applied for pilot level.
Spectrum analyzers are typically used for these measurements. Newer spectrum analyzers provide highly desirable marker bandwidth power measurement capabilities that can automatically integrate power over various bandwidth settings. Proper use of the spectrum analyzer for mask compliance and other 8-VSB measurements requires skill and attention to details that are beyond the scope of this article. I am currently engaged in efforts to standardize the use of spectrum analyzers as part of the IEEE G-2.2 Committee efforts.
There are many details to consider when establishing a monitoring and measurement program for your DTV operations. Not all measurements need be done on a regular basis. However, engineers need to become just as familiar with testing digital signals as they are with analog NTSC measurements. These new tests aren't necessarily more difficult than analog ones, but they are different. A key distinction is that with digital transmission, viewers won't get a poor signal. They'll either get a good one — or none at all.
This article represents a brief introduction to several important measurements for 8-VSB signals. It is an abridged version of a presentation I made at NAB2005. A complete copy of this presentation is available in the 2005 Proceedings.
More to learn
John D. Freberg is president of The Freberg Engineering Company.
