[RTTY] LMS Notch filter

[Kok Chen]

On Feb 24, 2012, at 1:46 PM, Jeff Blaine wrote:

> Is there a benefit when using a twin peak filter (of whatever 
> implementation) - assuming there is only one signal within the passband?

An arbitrary twin peak filter can actually be detrimental if the software modem or TU already includes an optimal filter (a Matched Filter or Raised Cosine filter that are designed for the specific baud rate).

My own modus-operandi is to use as wide a receiver filter that I can get away with (i.e., without having QRM clipping the receiver chain and the sound card).  This presents a signal with the flattest response, and with the least group delay, to the modem.  

Then feed that wide signal to a software modem that is specifically designed for the baud rate which you are receiving.  

The Raised Cosine filter is the narrowest filter than can be used.  For a 45.45 baud signal, the Raised Cosine basically has -6 dB points at 23 Hz from the center of the mark and space tones -- for 170 Hz shift, you can think of this as an optimal "dual peak" filter that is 216 Hz wide, and a hole in the middle.  With overall -6 dB width of 216 Hz, and a very sharp skirt, falling to below -80 dB by around 260 Hz.  

Don't try this with any dual peak filter/I.F. filter combination that is 216 Hz wide, however.  The shape of a filter (plus the group delay) can cause much intersymbol interference for an arbitrary 216 Hz filter.  To achieve 216 Hz (and not suffer from intersymbol interference) the filters need to be a pair of perfect Raised Cosine shapes around the mark and space tones.  So, it looks like a dual peak filter, but is not any arbitrary dual peak filter, and you need to change the width of the Raised Cosine when you change baud rate.

As long as the QRM is not clipping the sound card, the raised cosine filter will completely reject any QRM (and noise) passband that rejects the QRM completely outside the 260 Hz passband.  If you receiving chain and sound card combination has enough dynamic range, just let the modem reject the QRM .

When there is no QRM, a Matched filter outperforms the Raised Cosine by just a tad, but it is also very wide.  For 45.45 baud, the -30 dB points is more than 500 Hz on each side of the Mark and Space tones.   At some loss in sensitivity relative to the Raised Cosine filter, you can "roof" the Matched Filter with an IF receive filter when QRM is present, but open the I.F. filter up where there is no QRM (weak DX working split, for example). 

Both Raised Cosine and Matched Filters are very difficult/expensive to implement in hardware.  That is why a third order Butterworth filter is used in the past to achieve optimal RTTY copy in the better TUs.  In software, both Matched Filters and Raised Cosine Filters are a matter of properly designing the coefficients for an FIR filter.

73
Chen, W7AY