[RTTY] 200 hz filter on RTTY

Kok Chen chen at mac.com
Tue Jun 1 12:47:47 PDT 2010


On Jun 1, 2010, at 11:05 AM, Jeff Blaine AC0C wrote:

> The stronger the signal being heard, the less 
> reliance there is on the power contained in the sidebands.  So the 325 hz 
> number discussed earlier on the board some time back is I think inclusive of 
> the first 2 sidebands on either side.  Hope I have that right...

Let me see if I can make it a little cleared with the three "filter selection" (my excuse is English is not my first language, HI HI):


(1) To pass the first keying sideband completely, you need to pass 215 Hz worth of perfectly centered RTTY signal to the demodulator.  

When the SNR is good, you can get 100% print even if you don't pass the higher order terms of the keying sideband.  215 Hz is all the bandwidth you need (if you are perfectly centered on the signal) when the signal is loud.


(2) If the filter does not pass 215 Hz, then the demodulator will throw a finite amount of error even when you have extremely good SNR.

In this case, no matter how clean and clear the signal is, you will always see some error.  Notice you can even copy an RTTY signal, but the error rate will be quite high.


(3) Being able to include at least the third harmonic of the keying sidebands will help improve copy under marginal SNR conditions if there is no QRM within this wider passband.

Also, a matched filter will only help if you are right at the threshold of the bit error curve.  When the SNR is very high, a additional 1 dB to 2 dB of sensitivity improvement won't be noticed (no one can tell the difference between one error hit in 1000 characters and 2 error hits in 1000 characters).

When there is any selective fades or flutter, you will also not notice any improvement with a filter that is matched to the keying sidebands of the signal.  A matched filter can get you a dB or two improvement.  Take a look again at VE3NEA's curves http://www.dxatlas.com/RttyCompare/ .

Notice in the AWGN curve that 1 dB shift along the SNR axis can move the character error rate significantly, by a factor of more than 2 in error rate around the detection threshold.

Now look at the Selective Fading case... a 1 dB shift along the SNR axis does not improve the character error rate noticeably (unless you are doing a long term statistical study).

Also notice in all cases that once SNR is very good (the right hand side of the plots), a 1 dB change in SNR again won't make any difference in perceived error rate.

In practice...

215 Hz will be almost impossible to tune quickly without some software aid such as a crossed banana.  A 215 Hz only helps when there is QRM close by.  That is why most people don't recommend filters that are narrower than 250 Hz to 300 Hz.  

On the other side of the coin, a matched filter is only helpful when you are trying to dig out that really weak signal that you can't seem to copy even when bands are stable and there is no QRM.  A wider I.F. filter together with a matched filter in the software can help to dig out the weak DX operating split as long as no one is transmitting close to the DX' frequency.

Remember than filter width in all the 3 cases above can be achieved in either hardware or in software.  As long as the sound card is not clipping, there are advantages to using a wider I.F. filter and select the filter you need in software.

All that being said... an antenna with good directivity can easily give you many dB of better SNR; easily better than any demodulator tweeks that you can make.  If you have a SteppIR, it is more important to adjust for best directivity than for maximum gain (they often do, but don't necessarily happen at the same point).  With digital mode copying, it is all about SNR and "loudness" has nothing directly to do with how well you can copy.

When there is fading, diversity reception can also help tremendously with digital modes.  Just look at the difference in SNR threshold in the VE3NEA AWGN plot (no fading) with his flat fading plot (Rayleigh channel with no multipath) and his Selective fading plot (Rayleigh channel with two paths).

73
Chen, W7AY






More information about the RTTY mailing list