|Figure 1: The acquisition circuitry of a digital oscilloscope|
Digital bandwidth is largely a function of the instrument's sampling rate. While the analog bandwidth determines the highest frequency that can be fed into the oscilloscope without incurring losses, the digital bandwidth must be high enough to sample that input signal without errors caused by exceeding the Nyquist frequency.
The Nyquist limit says that analog signals sampled at greater than two times the highest frequency contained within the signal can be reconstructed with no loss of information. Thus, the highest frequency (digital bandwidth) that can be present in a digitized waveform is one half the sample rate. More practical limits are 1/3 to 1/10th of the sampling rate. Narrowband waveforms such as sine waves can be sampled at a lower rate; broadband signals such as pulses need to be sampled at higher rates. The consequences of sampling at too low a rate include aliasing and distortion.
|Figure 2: The frequency at which a sinusoid is attenuated to 70.7%|
of the initial amplitude is known as the -3 dB point.
So how much bandwidth should your oscilloscope have? For one thing, it should have enough bandwidth to keep up with high-frequency changes in the input signal. An inadequate amount of bandwidth will result in amplitude distortion and slower edges.
For example, in serial data applications, a rule of thumb is that the oscilloscope bandwidth should be at least five times the fundamental frequency of the signal being measured. A PCIe Gen 1 signal at 2.5 Gb/s has a fundamental frequency of 1.25 GHz. Thus, a 6-GHz oscilloscope would be required to accurately display the signal.