I became aware today through fred harris' excellent presentation at the DSP Online Conference (https://www.dsponlineconference.com/) of the harris-Moerder pulse shaping filter which was published 15 years ago: http://eon.sdsu.edu/~seshagir/SDR05.pdf

This filter reportedly results in an order of magnitude lower EVM for the same number of taps, and importantly significantly reduced rejection of adjacent channels, but to my understanding isn't commonly used since the Root-Raised Cosine filter is so baked into our standards.

Is anyone aware of actual use of this alternate pulse shaping filter and does anyone have further experience with it? Are there any other reasons, now 15 years later, that this hasn't been more widely adopted?

  • 3
    $\begingroup$ Dan, I wish I had an answer. My best guess is that, of all the "decent" pulse shapes, the raised cosine, while far from the best, is simply good enough. That and years of inertia has prevented industrial innovation in this field, as far as I'm aware. $\endgroup$
    – MBaz
    Sep 24 '20 at 22:27
  • 1
    $\begingroup$ I don't have an answer either but was similarly intrigued when I came across the same filter when reading some of fred harris's old papers (he'll be glad we're using the appropriate case). Part of it is surely existing hardware and systems that use rrc so future hardware does the same to stay compatible and follow what works. I'm sure eventually it will show up in open-source works but I know a lot of times what gets implemented comes down to what textbook ref you follow. I do think gnuradio and the like would be great places to start putting in these uncommon ideas to experiment with. $\endgroup$
    – user67081
    Sep 25 '20 at 0:17
  • $\begingroup$ @user67081 That's a great comment - agreed GNU radio would be a perfect place for it if not already there (assuming there is not some compelling technical reason). And to MBaz's point it would be good to quantify at the technical cross-over where it's use is compelling. I suspect the adjacent channel benefits would preface EVM given spectrum utilization concerns but haven't dug into the numbers. If power and achievable phase noise wasn't an issue then this could enable higher order modulations, perhaps applicable to DOCSIS (not very familiar). $\endgroup$ Sep 25 '20 at 1:38
  • 1
    $\begingroup$ Liguid-dsp implements: void liquid_firdes_hM3(unsigned int _k, unsigned int _m, float _beta, float _dt, float * _h) ..... github.com/jgaeddert/liquid-dsp/blob/master/src/filter/src/… $\endgroup$
    – Juha P
    Sep 25 '20 at 19:57
  • 1
    $\begingroup$ OK, made it an answer. $\endgroup$
    – Juha P
    Sep 30 '20 at 17:17

Liguid-dsp implements root-Nyquist harris-Moerder (hM-3) filter using Parks-McClellan algorithm as :

void liquid_firdes_hM3(unsigned int _k, unsigned int _m, float _beta, float _dt, float * _h)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.