Anti-Aliasing Filter on Analog Discovery 2

[kilian987654321]

Is there an anti-aliasing filter on the Analog Discovery 2?

If so, how can I implement it?

If not, do you have any suggestions on how to avoid aliasing (Oversampling?)

[attila]

Hi @kilian987654321

There is no adjustable filter available in the device only the averaging of N samples = 100MHz ADC rate / sample rate
The scope input bandwidth +30MHz acts as an aliasing filter.

In the WaveForms application you can enable oversampling under the Time options and add Math filter channel like this:

[kilian987654321]

Thank you, this is very helpful. As a follow-up question, this also means that there is no anti-aliasing filter to filter out frequencies higher that the maximum bandwidth, correct (a 6MHz signal for example)? It's fine for me if not, just curious.

[attila]

Hi @kilian987654321

The scope input bandwidth is +30MHz, there is no other filter available in the device.

[TTL]

Dear,

 

I really surprise about the absence of adaptative anti-aliasing filter (depending on sample rate configuration).

Is there any Reason of it ?

 

The solution you give is not a really one because the spectrum is still altered in the low frequencies.

Just try to Watch the Spectrum of a 110kHz square wave… It is awfull… That mean, your device is unusable exept for special frequency (or use a large oversampling which limits the bandwidth of the Spectrum analyser).

[attila]

Hi @TTL

The 'Reason" for not having adjustable filter are the device size and price.
It would require additional components and larger processing unit to handle software filtering.

If you use a high order low pass filter Math channel, the low frequencies won't be affected.
You can select the 2nd device configuration, under Settings/ Device Manager, to double the scope buffer, to double the resolution.
For lower frequencies, below 1-2MHz, you can record more samples in the Scope.

Here you have shown the most extreme situation with a square waveform.
With typical sine waves modulated in a way or other, or distorted, you don't have that many significant harmonics which could cause alias.

Even with expensive tools I use to analyze the signals at different ranges, sampling frequencies.
This to identify which components in the view may be alias of a different frequency domain.

 

[Mike Simmonds]

When trying to measure white noise levels, I noticed that it was necessary to select a Sample Mode of "Average" rather than "Decimate" (under Channel Options) to avoid having spectra that were badly contaminated by aliased noise.

For example, I set up the following situation:  I fed C1 from the output of a low-noise 1000X preamp with 1MHz bandwidth and put a 10K ohm resistor on its input.  Instead of seeing the expected 14nV/RtHz, I saw about 40nV/RtHz when in Decimate mode.  I could fix this switching to Average Sample Mode.  My Analog Discovery 2 came with Decimate as the default mode.  When using Decimate Mode, I had to apply an analog LP filter set to about 30KHz in order to suppress all the aliasing.

Other users may want to verify that they are using Average Mode if they are having problems getting good spectra.

[attila]

Hi @Mike Simmonds

Yes, the Average sampling mode acts as a low pass filter.