Dots ModeNot every DSO has it, but Dots display mode is essential whenever the DSO gets near the limits of the sample theorem, hence signal reconstruction – or even acquisition itself – appears flawed.
As an example, consider a 12 MHz square wave with quite moderate 3 ns rise time, hence a perfectly adequate signal to a 200 MHz oscilloscope like the SDS824X HD.
SDS824X HD_Square_12MHz_3ns_1GSa_Vect
The screenshot above shows the standard use case: Auto memory with at least 10 Mpts max. record length, Sin(x)/x reconstruction, Vector display mode, no Color grading, Persistence off.
In 2 channel mode, where we get a sample rate of 1 GSa/s without aggressive AA-filter, the waveform looks pretty good. With these settings, it would be pretty hard to provoke major reconstruction errors or even aliasing on a deep memory DSO like the SDS824X HD. Yet there might still be situations where we can’t get a sufficient real time sample rate. To demonstrate this, we can use the Constant Sample Rate setting instead of Auto Memory.
As a first step, let’s reduce the sample rate to 250 MSa/s, which many would consider still adequate for a 12 MHz signal:
SDS824X HD_Square_12MHz_3ns_250MSa_Vect
Even though the fundamental frequency of the signal is just 12 MHz and the Nyquist frequency (125 MHz) is more than ten times higher, we still get to see massive reconstruction artefacts and aliasing already. So much for the sometimes mentioned “rule of thumb” which suggests that a bandwidth five times the repetition frequency of a square wave would be adequate…
Let’s take this one step further and set the sample rate to 100 MSa/s:
SDS824X HD_Square_12MHz_3ns_100MSa_Vect
With the previous settings, we still got something remotely similar to a square wave. We go one step further and reduce the sample rate to 50 MSa/s:
SDS824X HD_Square_12MHz_3ns_50MSa_Vect
Now we finally got a pure sine wave with lots of amplitude modulation and jitter – certainly not a very good representation of the original waveform anymore. We still want to take it to the extreme and reduce the sample rate even further to 20 MSa/s, thus violating Nyquist even for the fundamental frequency:
SDS824X HD_Square_12MHz_3ns_20MSa_Vect
This last screenshot needs not be commented, as it speaks for itself – except for the fact, that the SDS824X HD won’t let us use the original time base of 20 ns/div with such a low sample rate anymore. As a consequence, the DSO has automatically switched to 50 ns/div.
Anyway, this is not the end – after all we’ve got the Dots display mode up our sleeves:
SDS824X HD_Square_12MHz_3ns_20MSa_Dots
Yes, with only 1 point per division (10 points for the whole record!), there is no contiguous trace and the rendering is a bit dim. Yet nothing that could not be improved by a little Persistence time:
SDS824X HD_Square_12MHz_3ns_20MSa_Dots_P1
What we get now is a perfect visual representation of the original signal – within the bounds of the 244 MHz bandwidth, that is – despite the effective sample rate of only 20 MSa/s.