Digital RF bandpass filter for FM radio translators

Benjamin,

Thanks for your reply.  The desired 91.5 MHz signal is about 40 dB SNR, and the interfering 91.1 MHz signal is a few dB higher than that, say 43 dB SNR.  Our tuned-cavity bandpass filters only attenuate the interference by about 10 dB, but if we could attenuate it by 30 dB, I think that would be enough.  Our output bandwidth is 400 kHz, which puts us too close to the 91.1 MHz station's upper sideband.

I haven't been able to find commercial off-the-shelf digital filters like this, but with improvements in ADC performance and software-defined radio technology, I was wondering if it was feasible.

Doug

Hi, Douglas,

    It all comes down to dynamic range: in general the passive components gives you better dynamic range, subject to available Q of the components that define the filter skirts (selectivity).  You can get better selectivity with digital filters only if the analog path and ADC/DAC can support the dynamic range that you need.

   Before a blanket dismissal of the possibility, can you tell us the desired and interfering signal levels, the output BW, in-band SNR & output interferer level you are trying to achieve?  I have to admit that this is probably more of an academic curiosity than a practical solution.

Benjamin

Hi, Doug,

    The requirements may not be as stringent as I first envisioned!  A 12-bit RF ADC-DAC system should be able to provide 70+ dB of dynamic range at Nyquist, and much better once the BW is decimated down to 400 kHz.  So a line up as below may work:

    Cavity Filter - LNA - VGA - ADC - Digital Filter FPGA - DAC - Filter

where the Digital Filter block includes Decimation, FIR filtering, and up-sampling.  Please double check my assumptions below:

    Input power (within pass band of the filter) ~= -75 dBm (40 uV)

    Gain required ~= 1/(12*40u) = 2096 => 66 dB,

            to convert 40 uVrms of random signal to 1 Vpp for ADC

    Fsamp(input) = 250 MHz

    Decimation ratio = 32 => Fsamp(output) = 7.8 MHz, i.e. > 8x oversampling

Suitable ADI components would include:

  LNA - ADL5545 (among others) x 2

  VGA/Driver - ADL5101

  ADC - ADL9634-250 (12-bit 250 MS/s)

  FPGA - ADI does not make FPGA's, but some evb's have them

  DAC - AD9753 (12-bit 300 MS/s)

  Output filter - may need to be a cavity filter, or a dielectric resonator could work

The parts cost would be much higher than an analog implementation.  But since you are looking at small quantities (1 or 2 ?), development cost would outweigh parts cost.  The filter you need is probably a bandpass with 250kHz 0.5 dB passband and 33 dB suppression at 400 kHz.  You may want to repost this on the high-speed ADC forum, and see if they may have an evb that already has an appropriate FPGA on, or even a bandpass sigma-delta ADC that would do the job.   Digital filtering is outside the scope of this forum and my expertise.

Benjamin

Benjamin,

Thanks very much for your reply.  I tried responding the day before yesterday, but my message apparently didn't get through.

I'm not familiar with bandpass sigma-delta ADCs.  If a sigma-delta ADC has an integral bandpass filter, could one be connected directly to a DAC and bypass the digital FPGA filter altogether?

I'm also not familiar with dielectric resonator filters.  Can you recommend a source for a custom filter like this?

Doug

This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

Thank you,
EZ Admin