v1.223: CIC filter compensation implemented, possible DRM benefits

jksjks
edited September 2018 in KiwiSDR Discussion
I should have done this long ago, but I finally figured out a simple way to implement CIC filter compensation by empirical measurement. The before and after pictures below show the result. No more 10 to 15 dB roll-off at the ends of a wide passband. This will improve wide-band AM frequency response and more importantly wide-band IQ mode uses. DRM in particular might have an easier time without the roll-off (see DRM PSD plots below).

IQ mode, 12 kHz passband, 10 dB/div:
(display is from the new audio FFT)
image


With CIC filter compensation. Note flat passband:
image


DRM PSD before and after compensation:
image


image

Attachments:
https://forum.kiwisdr.com/uploads/Uploader/08/6df1555724dd0ffc7e95695fdabe22.png
https://forum.kiwisdr.com/uploads/Uploader/a1/7bca33c3144de8b203b9f8a19d862b.png
https://forum.kiwisdr.com/uploads/Uploader/6b/95bbfc2e91e03bde34047052a51da9.png
https://forum.kiwisdr.com/uploads/Uploader/36/6de24562129d9d418a68613df820db.png
PowernumptyWB4klj

Comments

  • Thank you John, I will do some testing and advise.
    Cheers
    PK
  • Both charts were taken from same time from same programme source

    Cheers
    PK


    V1.222 FIR




    v1.223 CIC
  • "I should have done this long ago, but I finally figured out a simple way to implement CIC filter compensation by empirical measurement. The before and after pictures below show the result. No more 10 to 15 dB roll-off at the ends of a wide passband. This will improve wide-band AM frequency response and more importantly wide-band IQ mode uses. "

    Hi John,

    Could this also potentially affect the TDoA calibration ?

    Regards,

    Martin - G8JNJ
  • I don't think so, but that's really a question for Christoph. The roll-off is more significant for bandwidths between 10 - 12 kHz. The default IQ demod b/w is 10 kHz and hopefully people are adjusting it lower to just fit the spectrum b/w of the signal they are DFing.
  • I was just concerned in case it perhaps altered the filter group delay (or whatever the digital equivalent is) which may have altered the time delay at particular frequencies.

    Regards,

    Martin - G8JNJ
  • A possible change in the group delay would be identical for all used KiwiSDRs in a given TDoA analysis and therefore cancels out in the cross-correlations.
  • In v1.262 Christoph has come through with what looks to be proper CIC filter "droop" compensation. It has been tested and found not to cause the distortion to WSPR reception of my previous solution. Sincere thanks Christoph!
  • v1.262 seems to work well.

    Before looking at wide-band digital signals I recorded the OTHR signal from Cyprus and one CODAR signal in order to convince myself that the signal reconstruction out of three IQ datastreams works as expected, see
    https://hcab14.blogspot.com/2019/01/kiwisdr-iq-data-streams-with-2025-khz_29.html
    There you can find also an example of how to use kiwirecorder.py which might be helpful.
Sign In or Register to comment.