• State
  • Locked Locked
  • Replies 11 replies
  • Subscribers 46 subscribers
  • Views 2299 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.

ADS1158: Anti-Aliasing Filter Design

tedex45
Prodigy 190 points
Part Number: ADS1158

Hi, I am looking at sampling 16 analog inputs using the ADS1158. Am I correct in saying that I would only need one anti-aliasing filter placed between the MUXOUTP pins and the AINP pins (AINN and MUXOUTN connected to GND) since the inputs are multiplexed. 

Also, is there by any chance a newer ADC you would recommend that supports 16-bit resolution with 16 channels as the ADS1158 seems outdated. 

Thank you in advance! 

  • 0
    TI__Mastermind 44550 points
    Hi tedex45,

    Yes, you can place place your anti-aliasing filter between the MUX output and ADC input. The MUX outputs are pinned out for that purpose so that a single filter or buffer can be used for all inputs.

    The ADS1x58 is the only 16 channel ADC we have. You could look at putting two 8 channel ADCs together if needed. Here is a list of other 16-bit high channel count ADCs: www.ti.com/.../precision-adcs-products.page

    Best regards,
    Chris
  • 0 in reply to Christopher Hall
    Prodigy 190 points

    Thank you for the detailed reply. Thus, when using all 16 channels in auto-scan mode, I will have a maximum of 1.5ksps/channel so my maximum signal bandwidth would be 750Hz. Thus, my anti-aliasing filter would need to cut off everything over 750Hz. Is that right? 

    Also can I leave PWDN and RESET in the ADS1158 unconnected (floating).

  • 0 in reply to tedex45
    TI__Mastermind 44550 points

    Hi tedex45,

    The maximum data rate in auto-scan mode is 23.739 kSPS. If you use all 16 inputs, then each channel will have an effective throughput of about 1.5kSPS (23.739 kSPS / 16) as you mentioned.

    For anti-aliasing, you don't actually need to have that low of a cut-off frequency. The ADS1158 has a built in digital low-pass filter. This filter will be fairly effective at preventing aliasing at all frequencies, except near multiples of modulators's sampling frequency. Therefore, your external anti-aliasing filter really only needs to have significant attenuation near fCLK/2, as shown below:

    If unused, /PWDN and /RESET should be tied to DVDD.

    Best regards,
    Chris

  • 0 in reply to Christopher Hall
    Prodigy 190 points

    Thank you very much. That explains a lot. So just to check if I understand correctly, f_clk/2 = 8MHz so I can simply use a RC filter with a cut-off frequency at 8kHz for example and I would then get -60dB attenuation at f_clk/2. Is that correct? 

    Also, with all 16 channels and thus a throughput of 1.5ksps, I am guessing my signal bandwidth is not restricted to the output data rate. Thus, I could have a signal bandwidth up to 8kHz (using the above filter). 

  • 0 in reply to tedex45
    TI__Mastermind 44550 points
    Hi tedex45,

    Yes, that is correct for the anti-aliasing filter attenuation.

    You actually bring up a good point about the signal bandwidth...if you were only sampling 1 or 2 channels, then the data throughput would be fast enough to get the full 8 kHz signal bandwidth. However, with the effective 1.5kSPS throughput from multiplexing 16 channels, your noise bandwidth is still 8 kHz but your signal bandwidth is limited to 750 Hz. The noise between 750 Hz and 8 kHz will alias back into your passband.

    ...It is as if you were sampling only one of your inputs at the 23. 739 kSPS data rate, but only keeping every 16th result. The digital filter is still allowing the full 8 kHz noise bandwidth through with each result, but by throwing away 15 out of 16 samples you lose a lot of information about the signal.

    So if you wanted to lower your anti-aliasing filter cutoff-frequency you could, and you probably would benefit a little bit by doing so. However, you'll likely run into a case of diminishing returns, where you'll need to make the filter more and more complex as the cutoff frequency gets smaller and smaller. Therefore, you'll have to decide what's good enough for your use case.

    Best regards,
    Chris
  • 0 in reply to Christopher Hall
    Prodigy 190 points
    Ahh I see, thats what I thought. Ignoring the additional filter complexity, would I need to design my AA filter with cut-off frequency at 750Hz then to achieve the best signal integrity. Then I would ensure no noise would alias back and I still have significant attenuation near the modulator frequency (8Mhz). As you mentioned before, the digital filter attenuates everything before so I wouldn't need any filtering around the 23.7kHz. Is that correct? Thank you again for all your detailed reply.
  • 0 in reply to tedex45
    TI__Mastermind 44550 points

    Hi tedex45,

    One problem with placing the RC cutoff frequency right at 750 Hz is that the RC filter will be attenuating signals in your passband, since it has a slow roll-off.

    In my opinion, unless you're expecting interfering signals (not just noise) within the 8 kHz bandwidth, then I don't think you'll gain much from lowering the anti-aliasing filter's cutoff frequency. It would be more effective to lower the data rate of the ADS1158 and take advantage of the digital filter, since the digital filter has a much steeper attenuation rolloff than the RC filter.

    Best regards,
    Chris

  • 0 in reply to Christopher Hall
    Prodigy 190 points

    Hi Chris,

    What if I use a 2nd order butterworts filter? I know its more complicated but then I wouldn't have any attenuation in my passband and still achieve the required attenuation. Would that work then? 

  • 0 in reply to tedex45
    TI__Mastermind 44550 points
    Hi tedex45,

    I don't see why you would need to build an active filter for anti-aliasing a delta-sigma ADC that oversamples the input. One of the main benefits of oversampling the input is that it relaxes the anti-aliasing filter requirement.

    You certainly can use an active filter if you think it is necessary; however, keep in mind that the components you use to build the filter could add additional errors and noise to your measurement.

    Best regards,
    Chris
  • 0 in reply to Christopher Hall
    Prodigy 190 points

    Just to clarify if I understood correctly. Sampling all 16 channels simultaneously achieves an effective throughput of 1.5kSps so for each channel, I would only be able to measure signals up to a bandwidth of 750Hz.

    However the data rate (f_DR) is still at 23.7kSps. This means that the digital filter removes signals only after 23.7kHz and thus leaves a noise bandwidth of 0 to 23.7KHz.

    If my anti-aliasing filter has a cut-off at 8kHz, it will solve the aliasing problem at 8Mhz (f_mod/2), but signals from 750Hz to 8Khz will alias back into my output as they are not removed by the digital filter and as they exceed my effective Nyquist frequency per channel of 750Hz.

    Is that correct?

  • 0 in reply to tedex45
    TI__Mastermind 44550 points

    Hi tedex45,

    That is correct!

    One thing to consider, if you don't need the bandwidth or data throughput, is to decease the data rate of the ADC. This would probably be one of the most effective ways to further reduce the noise bandwidth.

    Best regards,
    Chris