# Making Sense of oscilloscope specifications

When looking for oscilloscopes, you often see the following specs:

• bandwidth [in Hz]
• sampling rate [in Samples/s]
• capture rate [in waveforms/s]
• record length [in pts/channel]

My question is: How are these specifications related?

Take for example the following comparison table from the company Siglent:

All these scopes have a sampling rate of 1GSamples/s, meaning they can take 1 billion measurements every second. From the Nyquist-Shannon Theorem these scopes should all be capable of accurately measuring frequencies up to 500MHz. But their bandwidth only allows for at most 200MHz, and only 100MHz for the scope in the centre row. How is this reconcilable?

• Also consider that 500MHz at 1Gs/s (2x) on a screen will be a nasty-looking squarewave, completely devoid of any detail whatsoever. For at least some detail, the sample rate needs to be 10x, and even then, that's only 10 data points on that 500MHz signal. May 13, 2020 at 16:41
• in addition to what rdtsc said, the bandwidth is for the whole scope. The sampling rate is typically a total sampling rate. So that 1GSa/s would be divided across 4 channels if you have all 4 going. So that would be 250GSa/s per channel. But each channel is still 200MHz capable. May 13, 2020 at 17:22

Sample rate is NOT synonymous bandwidth. That's why they are two different specifications and not one.

Sample rate is just how many readings per second you can generate and record in the digital domain. It has nothing to do with whether the high frequency components in the analog domain are actually able to get all the way from the input of the front-end amplifier all the way to the output of the ADC.

Just because you can sample at a very high rate doesn't mean very high frequencies are getting through the front-end amp or ADC. In the same way that sticking a lowpass filter at the front of your ADC doesn't suddenly reduce it's sampling rate.

In order to make proper use of your analog bandwidth, your digital sample rate needs to be higher than the Nyquist frequency in most cases. One exception is if you know you have a periodic signal then you can undersample and reconstruct the waveform from multiple passes.

It's important to remember that where something has several specifications, it's the tightest specification that controls the performance.

If an ADC has a 1 GSa/s sampling rate, then it will limit the bandwidth to at most the Nyquist limit of 500 MHz.

However, if it has a 100 MHz bandwidth amplifier or filter ahead of it, then that will limit the bandwidth.

Your assumption is wrong. 1GHz sampling rate does not mean it can accurately show any waveform containing up to 500 MHz of bandwidth. Sure, the ADC can take 1 Gigasamples per second. But the analog input stage may only be good for up to 100-200 MHz of bandwidth. I'd expect a 1 GSPS scope to be specified to 200 or 250 MHz of bandwidth tops.