Usually the confidence you have in your measurements is directly proportional to the cost of your instrument. Take care.
In debugging EMI problems you need instrumentation: scopes, spectrum analyzers, network analyzers, probes, etc.
And, measuring signals (correctly) is not trivial and easy. You need to understand the limitations of your instruments and the correct setup to make your measurement. Perhaps you will need to fail in some important project to really understand how important this idea is (as I failed in the past).
One example of those limitations is bandwidth and surely you know that a low bandwidth instrument offers you a non-complete waveform of the signal of interest (Figure 1).
Figure 1: A 10MHz clock signal with a 200MHz (top) and 8MHz (bottom) measurement setups.
Note how, with a low bandwidth setup (i.e. scope bandwidth, probe bandwidth, etc.), you will miss the high frequency part of your signal. If you are debugging EMI for example, you will not be able to locate a ringing that is creating some kind of conducted or radiated problem. The ringing is not there in your screen so there is no ringing for you. Additionally, some parameters as rise and fall times will increase in a low bandwidth system.
But, another parameter not so typical to consider is resolution bandwidth (RBW) when you are measuring the spectrum of your signal. For EMI debugging RBW can be decreased to analyze emissions so we can identify the frequencies with more RBW under some kind of “envelope” in emissions results.
The effect of RBW can be seen in Figure 2.
Figure 2: A 24MHz clock in time and frequency domains using 200kHz (left) and 2kHz (right) RBW setups.
Note that reducing RBW, the noise floor will decrease so some signals appear clearly on screen but, considering a typical spectrum analyzer of FFT instrument, sweep time (processing time) will increase for the same span (range of frequencies you are analyzing on screen).
At the same time, note amplitudes of harmonics are not changed for the two RBW values. Be careful about this: if the bandwidth of the signal is more that the RBW of your instrument, the amplitudes of the signals can change. That is because in EMC, the parameter RBW is usually fixed by regulations
(ie. 200Hz, 9kHz, 120kHz, etc.) so we can compare with the same limits when testing for compliance (all the people measuring in the same conditions).
Finally, you must take care about overloading. This is especially important if the signal you are measuring is changing in amplitude. One example of this situation is when you are measuring with near field probes. The amplitude of the signal can increase or reduce as your near field probe is close or far to the circuit.
For a good measurement, it is usually recommended to use the full screen range of the scope for measurement. But, the problem appears if, accidentally or unintentionally, the time domain signal increases amplitude (i.e. you put the near field probe very close to the circuit) and you do not change the vertical scale (V/DIV) to accommodate the new situation.
In Figure 3 you can see how the frequency spectrum for the same time domain signal is using the same RBW.
Figure 3: A 24MHz clock in time and frequency domains without (left) and with (right) overloading.
Left traces show the time domain and frequency domain waveforms for your signal. Note that, if the signal goes “out of the screen” in time domain (right), you will see how harmonics of the signal increase in amplitude (i.e. the relative difference of amplitudes of your signal increase).
The name for this effect is overloading and in practice it is similar to the overloading effect you obtain in spectrum analyzers when an input signal is too big. In spectrum analyzers a typical test to check if the amplitudes of the signals on screen are real or are obtained through overloading is to introduce some kind of attenuation (i.e. 3dB). If all the amplitudes of the signal decrease in the same value, the input was not overloaded. If amplitudes of high frequency signals are reduced in more quantity than low frequency harmonics (or some of those signals disappear), you were overloading the input circuits.
My final advice: when measuring FFT with your scope try to see time domain so you can be sure the signal is not too big in amplitude.
