• State Resolved
  • Locked Locked
  • Replies 1 reply
  • Subscribers 46 subscribers
  • Views 233 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

When does an Antialiasing filter more harm than it does good?

Philip Egger
Prodigy 30 points
Other Parts Discussed in Thread: ADS8422

Let's imagine I'm digitalizing a 1MHz signal with the ADS8422 (4MSPS, 16-Bit).

Therefore, I need an Antialiasing filter with a Gain of 0dB @ 1 MHz and according to the dynamic range a Gain of -97dB @ 2 MHz.

A filter like this would be something like a 10th order Chebyshev filter and therefore introduce a quite significant amount of noise and error.

This might be an extreme example but it represents my question quite good:

Where is the point where an Antialiasing filter does more harm than it helps?

Is there a rule of thumb for what is the maximum usable order from a signal point of view?

Or is it really calculating the sweet spot between ADC errors getting larger and filter errors getting smaller with increasing ADC speed?

In addition, does the filter really have to block the full dynamic range of the ADC? If not, where is the sweet spot for that?

 

Thanks in advance.

 

Bets regards

  • 0
    TI__Genius 17737 points

    Philip,

    That is a great question. I’ll give you a brief summary of my thoughts and some links to collateral.

    1. To restate your problem: Your input signal range is 1Msps and the sampling rate is 4MHz. So, for your example, the Nyquist frequency is 2Msps. In this case any signal above 2MHz will alias.
    2. The next thing to think about is the magnitude of any input signal above 1Msps. Do you expect that a full scale 2MHz or greater signal may be applied to the input? If so, your assumption of requiring a very high order filter is true. In other words, you will need to attenuate the input signal by the dynamic range of the converter.
    3. A very high order filter is usually not practical. Besides the cost and PCB area the multiple stages will likely degrade the system performance. A better approach would be to choose a higher sampling rate converter or to limit your input frequency range further. For example, if you limit your input frequency range to 100kHz your filter order can be lower. As another option, if you increase the sampling rate to 8Msps your Nyquist will increase to 4Msps and reduce your filter order.
    4. I think a more common issue is that the input signal range is limited (e.g. 0 to 1MHz), and you want an anti-aliasing filter to eliminate noise pickup issues. In this case it is not correct to assume that the alias signal is full scale. In fact, there is no real way to know how large the interference signal will really be, but you can make an assumption. For example, you might assume that a noise pickup signal is will probably be less than 1mV. In this case you only need a 3rd order filter. The pocket reference calculator has a utility that lets you enter your system information and calculates the required filter order.
    5. Precision labs covers anti-aliasing filter design. I suggest you watch Aliasing and Anti-aliasing Filters and Hands-on Experiment – Aliasing and Anti-Aliasing Filters . The second video is a “hands-on” experiment. Even if you don’t have the hardware called out byt the video, I suggest watching as it covers a full design and shows measured results. This series also discusses using filterPro to design the filter.

    Let me know if you have further questions.

    Art