[RTTY] Re: Rtty and Ritty

[Kok Chen]

On Apr 20, 2004, at 9:33 AM, Bill Turner wrote:

>>  It does not surprise me that your KAM out performed the sound
>> card.
>
> It not only surprises me, I'm sure there is some configuration problem
> there.  MMTTY easily outperforms a KAM, as many on this reflector have
> pointed out.

I don't know about the 232, but the KAMs are pretty simple.  I own a 
couple of KAM Pluses together with the KAM'98 (not recommended, just a 
Linear Tech FSK chip inside :-).

The KAM Plus consists of an analog input buffer, the gain of the buffer 
determines if you are operating in "FM" or "AM" mode.  In FM mode, the 
signal is first clipped before passing to subsequent stages.

This buffer is followed by a band pass filter made up of 6 stages of 
2nd order switched capacitor filters (run of the mill National MF10).  
This is then followed by the Mark and Space tanks circuits, each one 
again only a 2nd order switched capacitor filters.  This is followed by 
a diode discriminator into a straight forward slicer.  That is followed 
by a low pass switched capacitor data filter, and then converted to 
pseudo RS-232C levels.

By comparison, a 256-tap FIR filter will easily outperform that cascade 
of bandpass filters.  With DSP, you can also implement much more 
sophisticated adaptive slicers, not to mention selectable matched 
filters for the Mark and Space channels (which I am guessing RITTY 
uses) for different band conditions, and possibly adaptive equalizers 
to counter flutter.

If the input A/D converter stage is decent, there is no reason a DSP 
implementation cannot beat a KAM with one hand tied behind its back.

I use a Timewave DSP-599zx as a remodulator ahead of my KAM and it 
outperforms the standalone KAM by a good margin.  The DSP-599zx is just 
a firmware DSP implementation of FSK demodulation, and most of your 
computers should have oodles more processor cycles than what is in the 
Timewave box (mine does even before I code for the Altivec vector 
processor in my machine).

If you are not getting better print from a DSP set up, do try and check 
up on your A/D converter settings.  Use a software "oscilloscope" or 
spectrum analyzer to  look at the A/D converter output and you should 
run it just below clipping with strong signals.  The 'scope might also 
show if you have other problems such as hum loops).  Any decent 
software FSK program probably comes with some tool you can use to 
adjust the A/D gain with.

If there are no other signals in the passband, bump the gain up on a 
weak signal so you are using as much of the A/D range as possible.  
This is where a 20- or 24-bit converter comes in, you don't have to 
manually ride the gain as often to optimize print on weak signals.  The 
599zx has an AGC stage with you can engage to compensate; but the 599's 
AGC has its own problem, I usually miss the first couple of characters 
when a loud station sends an exchange.

Nowadays, I use offboard USB A/D converters even with computers with 
built in 16-bit sound input and output.  You can buy cheap $35 USB 
converters that are good to 20-bits (like the iMic from from Griffin 
Technology), assuming your OS can handle the extra bits.  The MacOS X's 
sound system is 32-bit floating point.

I will try and dig out the ISBN of a DSP book which has a chapter on 
FSK demodulation, including a section on adaptive slicers.  I don't see 
it on my bookshelf right now, so it might be at my Oregon QTH; I am in 
California right now.

73
Chen, W7AY