This requirement is sometimes misinterpreted due to the wording in the rules. Another way to state the measurement range is from 500 kHz below to 3.75 MHz above the visual carrier frequency.
When running this test using a multiburst test signal and a spectrum analyzer, its important that the multiburst
test signal packets be of equal width. This will produce equal amplitude pulses on the analyzer.
Some multiburst test signals have narrower packets at the higher frequencies. In that case, even though the amplitudes are constant when viewed on an oscilloscope or waveform monitor, a spectrum analyzer will indicate that the narrower packets are lower in amplitude.
Some systems run this test by demodulating the signal, then measure the flatness of the test signal with a waveform monitor or oscilloscope.
If you use that technique be aware that, technically, it doesn't quite follow the rules. The problem is that this process largely ignores frequencies below the visual carrier frequency. Therefore, you aren't checking from 500 kHz below the visual carrier to 3.75 MHz above the visual carrier.
Some systems prefer to use signals such as the Ghost Cancellation Reference signal (GCR) or line sweep for this test. They feel the "continuous" display of those signals allows them to see response variations that may be missed if a multiburst test signal is used. That's a possibility. However, the nice smooth display generated by those signals can cause a false sense of security. Here are some additional considerations:
1. The ability to see narrow dips or peaks in the frequency response is limited by the resolution bandwidth (RBW) filter being used in the analyzer. To demonstrate this, connect a line rate sweep signal directly to your spectrum analyzer and adjust the analyzer for 2 MHz center frequency, 500 kHz total span (50 kHz/div), 3 kHz RBW, 1 or 2 dB per division, and adjust the reference level to center the display. It will look something like this:
High level sweep signal.
2 MHz center, 50 kHz/div, 3 kHz RBW, 1 dB/div
The sweep signal actually consists of many "packets" of energy spaced 15.75 kHz. In fact, if your analyzer has a sufficiently narrow RBW filter, you can see that it also consists of packets spaced 30 Hz apart. This, of course, is due to the 15.75 kHz horizontal and 30 Hz frame rates of our television system. A simplified way to think of this is that the sweep signal is being 100% modulated by the horizontal and vertical sync pulses, resulting in 15.75 kHz and 30 Hz "sidebands". In fact, there are several other spacings depending on the nature of the signal.
The point is that the sweep signal has 15.75 kHz peaks and valleys, yet under normal in-channel response test conditions, they are not even visible! i.e. the nice smooth display produced by GCR or sweep signals isn't the whole picture.
In a similar fashion, if the peaks and valleys were spaced 100 kHz apart, you probably wouldn't be able to see them using the normal 300 kHz RBW filter. If the peaks and valleys were spaced 300 kHz apart, they would be visible, but the "peak to valley" measurement would not be very accurate (depending on the shape factor of the 300 kHz RBW filter being used).
If the peak or valley occurs near one of the standard multiburst frequencies, the test results using multiburst could be more accurate than the GCR or sweep result.
2. If multipath is present on the signal, there can be large differences between in channel response test results using GCR or sweep signals relative to that obtained using multiburst signals. Take a look at some off-air signals that have both GCR (usually on line 19) and multiburst test signals in their VITS line ups. The displayed frequency response depends on the type of test signal being used and the multipath delay(s) and reflection amplitudes.
3. An advantage of high level (100 IRE) multiburst test signals is that they can be used to provide non-disruptive in channel response tests using virutally any spectrum analyzer and some signal level meters. The 4100 & 4200 In-Service Test Processors provides this capability. See the technical paper "How to run non-disruptive tests even if your test gear doesn't do gated measurements" for more information.
Overall, there are advantages to using perceived "continuous spectrum" types of test signals for in-channel
response tests, but the advantages are much less significant than you might think.
How to run non-disruptive tests even if your test gear doesn't do gated measurements