The FCC, in order to make a foolproof definition, made the out of channel emissions regulation somewhat hard to understand. To establish the reference value to measure to, the FCC specifies the total DTV power, including the pilot carrier. To make the sideband power measurement, the FCC specifies that a 500 kHz (equivalent) bandwidth be used to determine that the emissions be below the mask specified in the curve.

Clearly, a 500 kHz measurement bandwidth is unusable because it would obscure the necessary detail required at the channel edge. However, since the 8-VSB signal is very noise-like when you change the measurement bandwidth the measured amplitude of the signal changes (with the exception of the pilot amplitude, which remains constant with measurement bandwidth). This includes all of the DTV transmitter's out of channel emissions also. Therefore, to achieve a 500 kHz (equivalent) measurement bandwidth, one makes the measurement in a narrow enough bandwidth to see close-in detail (30 kHz is typical) and then scale each reading by the 10log the ratio of (500khz/measurement noise bandwidth) to display the reading in the correct bandwidth. The RFA300 performs this algorithm automatically so that the user does not have to perform any mental arithmetic.

If the amplitude vs. frequency response of the DTV signal is within a dB or so of the correct root raised-cos performance required of the transmitter (i.e. flat clear to the channel edges), it is possible to make the out of channel measurement with general purpose test equipment. (At least within the dynamic range limitations of the equipment. See below.)

To scale the amplitude of the measurement, we note that the noise bandwidth of the entire DTV transmission, excluding the pilot, is one-half of the symbol frequency or about 5.38 MHz. Now, 10log(5.38MHz/500kHz) = 10.3 dB. If the pilot signal is removed from the DTV transmission, the total power will fall by about 0.3 dB. Therefore, if we say that a total DTV signal has a power of 0.0 dB, removing the pilot will diminish it to -0.3 dB. Now, measuring that DTV signal in a 500 kHz bandwidth, one will determine its amplitude as -0.3 - 10.3 = -10.6 dB. That is the center part (sometimes called Bart's head) of a flat DTV signal will be -10.6 dB below the reference amplitude when shown on the FCC's scale. Since all non-pilot signals scale equally with measurement bandwidth, one can use, say a 30 kHz measurement bandwidth, to make all the measurements, and then scale all the measurements with the knowledge that Bart's head is -10.6 dB. For example, the breakout or shoulder amplitude at the channel edge should be 47-10.6 = 36.4 dB below Bart's head.

Keep in mind that as far as we know, the -110 dB limit of the FCC mask is beyond the ability of any currently manufactured test equipment to measure without aid of an external filter. The Tektronix RFA300 and Rohde &Schwarz FSIQ are the best instruments available today and they are able to measure to about -85 dB on the FCC's scale. This is limited by the instrument's inherent intermodulation performance. Because the signal is noise-like, it is not possible to narrow the bandwidth and lower the signal amplitudes to extend the dynamic range like it is with sine wave-type signals (as you narrow the bandwidth the noise-like signal falls in amplitude also).

Measurements can be made directly in one of two ways with external filters. The first is to use a bandstop filter between the measurement instrument and the signal. The bandstop should lower the amplitude of the central portion of Bart's head by perhaps 20 to 40 dB depending on the performance of the measurement instrument but have no effect on the sideband amplitude. To make the measurement, the filter must be switched to normalize on Bart's head and in to make the deep sideband measurements. The second technique is to make the measurements at the input of the transmitter's own channel filter. The measured amplitude is then diminished by the filter's response to infer the response of the combination of the transmitter and the filter. This method was illustrated by Bob Plonka of Harris at the 1998 NAB.

The RFA300 features an easily changeable measurement mask for all measurements. When making out of channel emissions measurements, it is very easy to modify the FCC's mask with the characteristics of the transmitter's channel filter. When thus modified, one can easily determine when the transmitter is meeting the FCC's specification.

We are all familiar with the fact that as you increase the input signal amplitude to the transmitter's power amplifier, at some point the amplifier compresses and its output amplitude no longer increases in proportion to its input amplitude. That is, the amplifier's gain falls. Amplitude error can be thought of as a plot of gain variation as a function of the signal's magnitude or amplitude. (Only the signal's magnitude is considered here. Phase is dealt with below.) Amplitude error is sometimes called AM to AM conversion. That is, small changes in (the correct) signal amplitude as caused by variations in its instantaneous output amplitude.

Looking at the amplitude error display we see that the vertical axis represents the amplitude or magnitude error, in dB measured between the actual signal amplitude and that of an ideal version of the signal (generated within the test equipment). The horizontal axis represents the instantaneous signal magnitude or amplitude as expressed in constellation units. The graph displays the instantaneous variations in gain (from nominal) as a function of signal amplitude. For instance; typically when the instantaneous amplitude is high (toward the right end of the graph) the gain compresses somewhat and we see the curve bend down a few dB. An ideal signal has a zero error from the origin to the extreme (to the right) peak signal amplitude. To compute the curve, the gain of the signal a few constellation units from the origin is considered to be the correct value and variations elsewhere are shown with reference to this nominal value.

The constellation unit deserves some comment. Analog TV engineers are familiar with measuring their TV signal in terms of IRE units where 140 IRE is the distance between sync-tip and peak white regardless whether the signal actually measures 1V or 10V on an oscilloscope. The IRE unit is used to measure the amplitude of signal details with respect to the signal itself rather than to an absolute voltage. Likewise, the 8-VSB signal is measured in constellation units, where one constellation unit is the distance between the origin and the first constellation state along the real axis (with the pilot offset removed) regardless of its absolute amplitude in volts. When you look at the constellation diagram you will note that the lines occur at odd integer values between -7 and +7, each separated by two constellation units from the next (again, with the pilot offset removed). Like an IRE unit, the constellation unit is used to measure the amplitude of details with respect to itself. The actual amplitude does not matter.

Phase error is a graph of the instantaneous error in phase between the measured and an ideal version of the signal. It is also plotted as a function against the signal's amplitude in constellation units. Phase error is very similar to the well-known ICPM or incidental carrier phase modulation observed in analog TV transmitters. Phase error, is sometimes called AM to PM; phase variations in the signal caused by instantaneous changes in its amplitude.

The non-linear displays are a major tool in identifying and correcting non-linear problems in transmitters. They are a RFA300 exclusive feature. All other types of measurements used to diagnose non-linear performance infer the cause by making indirect measurements of such parameters as out of channel emissions or peak to average ratio. Using these indirect measurement to infer the cause of the problem require more experience and skill than when using the RFA300 to give the operator a direct picture of the signal's non-linear characteristics. Using the RFA300, a user can more easily adjust the transmitter's linearity (e.g. by adjusting the amplitude and/or phase error compensation sub-system) to minimize any out of channel emissions and improve the transmitter's signal to noise (S/N).

Just as a note, one way of identifying if a transmitter's low S/N readings are being caused by linear or non-linear effects is to start with a constellation measurement. Note the S/N as the equalizer is turned on and off. If there is a major improvement in S/N when the equalizer is turned on (thus removing any effects of the system's unflatness and/or envelope delay) the problem is probably being caused by linear (envelope delay and/or flatness) errors. If the S/N remains about the same with the equalizer on or off, the poor S/N is probably being caused by non-linear effects. Note that if a transmitter does not display a S/N of greater than 32-33 dB with the equalizer ON, it probably will not meet the FCC's out of channel emissions mask requirement at the channel edges. Therefore, typically one will find that low S/N values will be caused by linear effects in transmitters that even come close to meeting the out of channel emissions mask. Excessive phase noise can also cause low S/N when the equalizer is on or off, but is not likely to be a problem in today's transmitters.