How to Avoid Oscilloscope Aliasing Pitfalls for Accurate Measurements - Workbench Wednesdays

The Nyquist Theorem says that you must sample a signal at two times its fastest frequency, right? However, even if you do that, aliasing on an oscilloscope can still occur! Aliasing is when a fake signal shows up on the screen due to undersampling of the original signal. Aliasing can happen for several reasons, even on an oscilloscope with enough sample rate.
The most common cause is that the sample rate drops on longer time bases. In this video, learn from James what the REAL theorem says, what aliasing looks like, and how to avoid it in your work.
Discuss the episode and ask James questions on the element14 Community! bit.ly/3sS2PKN
James reviews the Korg NTS-2 oscilloscope kit: bit.ly/3NF5u0P
Episode 47: What Does Bandwidth Mean for Oscilloscopes? bit.ly/3xiP85O
Engage with the element14 presents team on the element14 Community - suggest builds, find project files and behind the scenes video: bit.ly/3tmdewv
Visit the element14 Community for more great activities and free hardware: bit.ly/3q6YMpu
Tech Spotlights: bit.ly/3qPrDhM
RoadTest and Reviews: bit.ly/3pV5Bux
Project14: bit.ly/31wbnJY
#0:00 Welcome to Workbench Wednesdays
#0:36 Bald Engineer's Sampling Explainorem
#2:02 Aliasing Examples
#4:16 Why Aliasing Can Happen
#7:06 Sample Rate
#antialiasing #oscilloscope

[@emailuser3869]

Excellent video!

[@element14presents]

Glad you liked it!

[@Darkipod]

Very good video on aliasing, I feel like most labs in school dont cover this topic appropriately!

[@bald_engineer]

I think most DSP classes covers the topic well. However, not everyone takes them. And while fundamentally they apply to how an oscilloscope (or any test tool really) captures data, there are practical differences. Oscilloscopes are really easy to use. However, they are also easy to configure wrong while showing convincingly good waveforms.

[@AdrianGonzalezOK]

@@bald_engineerI need to discuss a small point with you! How can I reach a possibility to sample a CW signal with smaller sampling than 2.5 Nyquist? Some systems declare they do and I do not understand how

[@bald_engineer]

@@AdrianGonzalezOK If the signal is repetitive and you do not care about transients, then you can sub-sample. It is how the early high-bandwidth digital oscilloscopes worked. (Including the HP I showed at the very end.) They were only really useful for clock signals or drawing eye diagrams of data signals.

[@MAYERMAKES]

i recently upgraded my scope, could not trust the old one anymore..it might just have had a to slow sample rate for what I measured. huh. would explain why the error was not rproduceable..it was just off.

[@HL65536]

Is there any chance of something showing as the wrong frequency on a purely analog scope? I saw a 200MHz wave (some ringing) on my 60MHz BW scope. Assuming attenuation like a standard r-c filter I calculated the signal to be ~4ish times larger amplitude in reality than displayed on the screen. How good are the chances of this being correct?

[@HL65536]

Is aliasing a practical problem if the analog bandwidth is significantly lower than the signal? E.g. with an analog BW of 50MHz and 500MSa/s? (Signal >250MHz)

[@bald_engineer]

It depends. The analog bandwidth is effectively a low-pass filter. Bandwidth is specified at the -3db point, like all first-order low pass filters. So if you put a 250 MHz sine wave into an oscilloscope with a 50 MHz bandwidth, you won't see (much of) anything past the pre-amp. You might see a very small ripple but it would probably be in the noise floor. (Also, I should point out that some digital oscilloscopes don't have a first-order response. When the gap between analog BW and sample rate is smaller, they tend to be much steeper as a way to prevent aliasing.) (Edited after I realized what you were asking.)

[@HL65536]

@@bald_engineer So, I had the idea of making a Raspberry Pi Pico Oscilloscope, so I overclocked its ADC to 2MSa/s. However, I was unsure at what value to set the analog bandwidth (via r-c filter at the adc). Then I had the idea that it should be possible to remove the effects of the r-c filter in software and set the analog bandwidth relatively low. But, as the filter is there in hardware, instead of sampling at discrete timeslots with the sample-and-hold capacitor disconnected in-between these timeslots, I would effectively average the voltage of the whole timespan, not just the sampling slot. Tests reveal that I can perfectly reproduce square waves (except the one sample where the voltage level switches, which is in-between the 2, level depending on the exact time where the level switched). I was just now thinking about if this was alias-proof. Thinking about it, it likely is not completely, but should perform better than without the hardware+software hack.

[@Clark-Mills]

Random thought... Add randomness / jitter to the sampling slice... eg +/- rnd(5%) and that should help alert to aliasing.

[@bald_engineer]

Interesting idea. However, scope manufacturers work very (very) hard to keep their sampling apertureas tight as possible. I don't think they would like the idea of adding any kind of jitter. That said, sequential (non-real time) sampling scopes, like the HP I showed at the end, purposely randomize their starting relative to a trigger to help reduce aliasing. Those style scopes significantly under sample relative to their analog bandwidth. (I cannot say for certain if the HP 54501 does, but I know other sampling architectures support it.)

[@MrWaalkman]

Nice scope. I'm jealous.