Undersampling of Bandpass signals

Hi Karthik,

It's been a while since I've dealt with AM modulated signals - but in thinking about this, your resulting signal would be the 230 KHz carrier in a 10 KHz envelope.  To accurately capture the amplitude, your going to have to capture peaks of that 230KHz carrier aren't you?  I think you would have to have a determinate sampling instance as well that is synchronized to the carrier frequency.

Did you have any particular ADC part number in mind?  We could potentially do a little internal brainstorming on what the best options would be.

Hi Tom,

    Thanks a lot for the reply.

Tom Hendrick said:

our resulting signal would be the 230 KHz carrier in a 10 KHz envelope

  Yes. It will have 2 sidebands at 220 KHz and 240 KHz

Tom Hendrick said:

To accurately capture the amplitude, your going to have to capture peaks of that 230KHz carrier aren't you?  I think you would have to have a determinate sampling instance as well that is synchronized to the carrier frequency.

  I am little bit confused there. For a normal Nyquist/shanon sampling, it doesn't specify synchronous sampling ryt. Input signal sampled at any sampling frequency greater than twice the bandwidth could recreate the original signal.

 But in our case, since the carrier is generated digitally, we can have synchronous sampling also. If synchronous sampling is a requirement we cannot use sigma-delta ADCs. ryt. Moreover many SD ADCs doesnt have higher input analog bandwidth and has digital filters. Both will prevent undersampling I guess.

Tom Hendrick said:

Did you have any particular ADC part number in mind?

 Our dynamic range requirement is on higher side. So I was looking into ADS8900B. Its true that I can sample at 1MSPS with this ADC. I wanted to reduce the digital processing load due to FPGA timing resources.

  Thanks and Regards

      Karthik R

Hi Karthik,

In order to process your signal, you're going to have to demodulate the signal from the carrier. Just undersampling the waveform will not be sufficient to ensure you are capturing the peaks of the carrier signal unless you sample coherently with the original carrier.

The simplest way to do demodulate AM is with a peak detector circuit. This is commonly done with a diode and capacitor. If you are trying to maintain very good signal integrity though, then you may prefer an active approach using an op amp rectifier + low pass filter circuit. 

The most precise option giving the best SNR is to mix the signal with a local oscillator matching the frequency and phase of the carrier on the receive side and filter out the resulting high frequency side bands. This requires synchronization though because if the phase is not correct then the amplitude of the received signal will vary but the noise will not. 

Best,

Zak

Hi Zak,

   Thanks for the reply.

Zak Kaye said:

In order to process your signal, you're going to have to demodulate the signal from the carrier. Just undersampling the waveform will not be sufficient to ensure you are capturing the peaks of the carrier signal unless you sample coherently with the original carrier.

  Yes. Very true. Infact there will 2 demodulation. One is carrier demodulation and second amplitude demodulation. If we are using synchronized undersampling and BPF, carrier demodulation happens. My worry is whether this undersampling and BPF results in loss of information or get more corrupted by noise. After that I have planned signal amplitude demodulation in the digital domain itself.  

Zak Kaye said:

The simplest way to do demodulate AM is with a peak detector circuit. This is commonly done with a diode and capacitor. If you are trying to maintain very good signal integrity though, then you may prefer an active approach using an op amp rectifier + low pass filter circuit. 

  But both are in analog domain ryt. To avoid noise and temperature dependencies of analog elements, I prefer digital processing. Also anyways I need to digitize the signals.

Zak Kaye said:

The most precise option giving the best SNR is to mix the signal with a local oscillator matching the frequency and phase of the carrier on the receive side and filter out the resulting high frequency side bands. This requires synchronization though because if the phase is not correct then the amplitude of the received signal will vary but the noise will not.

  The carrier used for modulating the signal will be generated in the receiver side so phase and frequency issue will not come. But whether its better than digital demodulation and filtering ?

 Thanks and Regards

    Karthik R

Hi Karthik,

Just curious, but what sort of application is this?  

Hi Tom,

   Its for a MEMS actuator control electronics where noise is a bigger concern.

     Regards

        Karthik R