[RTTY] RTTY K3 Roofing Filter

[Kok Chen]

On Oct 19, 2013, at 1:38 PM, John Reilly wrote:

> Is there a consensus on the best K3 roofing filter for RTTY? I currently only have a 250 Hz, which isn't very useful.

A modern RTTY demodulator only requires between 23 Hz to 45 Hz of bandwidth for 45.45 baud RTTY.

The RTTY demodulator has two of them, at 170 Hz from one another.  This means that a perfect bandpass RTTY filter will need to be no more than 215 Hz to 260 Hz wide.

However, to be this narrow, the filters need to have a very specific shape.  In practice, when you add a bandpass filter ahead of the modem is to use one that is perfectly flat and with no group delay to about 280 Hz.

This does not mean that you cannot go narrower -- if you are willing to allow the error rate to be double, for example, then you can even use a filter that is only flat up to 210 Hz wide.

However, notice that I mentioned flat and no group delay.  If the filter is not perfectly flat across said passband, or it has significant group delay, then the errors will also rise. High order crystal filters are notorious on both counts.

With something like a superhet, the other extreme is no good either.  If there is no QRM, you would want as wide a filter as possible.  But I as I pointed out in the following article regarding specifically the K3, the receiver (at least mine, serial number 01432) is very prone to overloading where there are more than one loud station under the roof.

http://www.w7ay.net/site/Technical/K3/Content/imd.html

While the front end of the K3 may have a dynamic range of 100 dB, the entire chain all the way to the K3's line output only has a dynamic range of no better than 80 dB.  The headphones output of the K3 actually fares better than the line output, but you need to be careful when you use those (watch out for noise and hum when the modem's ground is not at the same potential as the receiver's ground).

Because of the overloading problem, when there are lots of loud signals close by, a moderately wide filter will not work well.  The K3 line output will collapse even before a good sound card is overloaded (good 16 bit sound cards have around 95 dB of dynamic range).

If you are a DXer looking for weak signals with no QRM, go for something that is low order (to avoid group delays) if you have a choice, and something around 400 Hz or wider (I have 250, 400, 2.8k and 6k roofing filters in my K3).  Under flutter conditions, something wider than 500 Hz will be more optimal.  

When there is no QRM that can overload the receiver (or the sound card), just open the bandwidth to 2.4 kHz.  The modem will thank you.  When there is no QRM, the worst thing you can do to optimal filters like the ones found in 2Tone and fldigi is to ruin them with a narrow crystal filter.  

I also cannot overemphasize the need to widen your filters when there is flutter.  Stop listening to what your ears tell you.  Widen the filters and the modem will prefer them for flutter.

In a tight contest environment, where the line output of the K3 can be overloaded by adjacent QRM, something like 300 Hz might be better.  I.e., the ISI may end up causing less error than the mess from IMD.  Your error rate will likely be double of what you get from a wider filter when there is no QRM, but that is still better than the error rate going sky high due to overloading (or AGC pumping, if you use AGC).

(Because of all of these problems, I have abandoned superhets and gone to the DDC variety of SDR (a HPSDR Hermes) to get good true dynamic range all the way to the demodulator, with very flat, and wide linear phase filters all the way to the demodulator.  For RTTY, I am basically using a bandwidth of 96 kHz today, with no AGC into a modem that has a pair of Matched Filters.)

73
Chen, W7AY