Two different limits
Bandwidth is an analog limit: the frequency up to which the signal reaches the converter without excessive attenuation. Sample rate is a digital limit: how many samples per second the converter takes. The two must be balanced. A very high bandwidth with insufficient sampling lets through frequencies the converter cannot represent well; a very high sample rate with reduced bandwidth wastes samples.
The Nyquist criterion
To represent a signal unambiguously, it must be sampled at more than twice its highest frequency component. Otherwise aliasing appears: fast components are interpreted as slow ones. But twice is the mathematical minimum, for a signal whose content is strictly band-limited, and it assumes ideal reconstruction.
Why more is needed in practice
- The input filter does not cut off abruptly and still lets some energy through beyond the nominal bandwidth.
- To see a waveform and not just its frequency, several samples per period of the relevant components are needed.
- Interpolation between points, whether linear or with a reconstruction function, improves the appearance but does not create information the samples do not contain; with few samples, it can introduce errors at the peaks.
This is why manufacturers design their instruments with a sample rate that is a multiple of the bandwidth, and that margin holds only as long as all channels and settings allow it: on some instruments, using more channels divides the rate available per channel.
The time base rules
The sample rate displayed depends on the horizontal scale and the available memory. If a slow time base is selected, the instrument lowers the sampling to fill the memory over the capture time, even though its maximum capability is much higher. A fast signal viewed with a slow time base can end up undersampled without the user noticing.
How to interpret the result
Always check the effective sample rate reported on the screen, not the maximum in the catalog. If it does not clearly exceed the signal, distrust the peaks and frequencies you see.