M0AGX / LB9MG

Amateur radio and embedded systems

Oscilloscope bandwidth benchmark

What is the impact of oscilloscope bandwidth on the observed signals? Does a bandwidth of 25 MHz mean that a scope will display a 30 MHz signal only with a lower amplitude? Can oscilloscopes work around Nyquist with equivalent sampling? Is a scope four times the bandwidth worth four times the price? Let's have a look! (spoiler: it is complicated)

I tested the following Picoscopes:

  • Model 2205A (25 MHz BW, 8-bit)
  • Model 2406B (50 MHz BW, 8-bit)
  • Model 5244D (200 MHz BW, 16-bit)

The highest model has adjustable vertical resolution. I tested the 16-bit and 8-bit modes. I used an AD9834 direct digital synthesizer clocked at 75 MHz as the signal source. A DDS is basically a high-speed RF DAC + some logic that can generate sine waves. Because all control is fully digital it can generate programmable sine waves with sub-Hertz resolution from a fixed clock. The DDS is a good test subject for analysis because it is commonly used as a sine wave source, yet the output gets "funky" as the frequency approaches the Nyquist rate. All measurements were taken from the same circuit with the same 200 MHz probe at 10x setting.

1 MHz sine

1MHz_BW_25MHz_8bit_200MSPS 1MHz_BW_50MHz_8bit_1000MSPS 1MHz_BW_200MHz_16bit_62.5MSPS 1MHz_BW_200MHz_8bit_1000MSPS

All traces are pretty much the same. Very little to see here.

10 MHz sine

10MHz_BW_25MHz_8bit_200MSPS 10MHz_BW_50MHz_8bit_1000MSPS 10MHz_BW_200MHz_16bit_62.5MSPS 10MHz_BW_200MHz_8bit_1000MSPS

Things look more interesting now. The 1000 MSPS traces clearly show output steps of the DAC waveform, while the 16-bit 62.5 MSPS mode pretends that the signal is still a sine wave. The differences between the 50 MHz and 200 MHz scope are not that big.

30 MHz sine

30MHz_BW_25MHz_8bit_200MSPS 30MHz_BW_50MHz_8bit_1000MSPS 30MHz_BW_200MHz_16bit_62.5MSPS 30MHz_BW_200MHz_8bit_1000MSPS

All traces show significantly degraded output of the DDS. The trace with 25 MHz bandwidth shows lower apparent amplitude of the signal. 16-bit 62.5 MSPS trace is close to useless. The 50 MHz and 200 MHz scopes show similar traces that are far away from a pure sine wave.

Summary

My most surprising find is that the differences in 8-bit mode between scopes with 50 MHz and 200 MHz bandwidths are not that big (for a 30 MHz signal of course). The most important find is that the 16-bit high (vertical) resolution mode of the 200 MHz scope downgrades the sampling speed to 62.5 MSPS and makes it perform actually worse than the cheapest 25 MHz scope running at 200 MSPS (obviously as long as you do not need the improved vertical resolution). Samples per second seem the most significant factor affecting the temporal resolution and there is just no way around Nyquist. When buying a digital oscilloscope (without having any special requirements) I would first consider the MSPS and bandwidth and maybe later the memory depth (to capture longer traces).