[RTTY] Re: external DSP audio filters

[Kok Chen]

On Oct 2, 2004, at 5:54 AM, Ian White, G3SEK wrote:
> Or do the outboard DSP boxes really perform so much better than the 
> DSP options in the radios and in MMTTY?

IMHO, they don't.  It depends on the hardware, and as they say, "the 
rest is just software/firmware."

On the hardware end, you need to look out for the A/D converters that 
are in the DSP boxes and how well the box is shielded, whether they 
have balanced inputs, what is the dynamic range (not just how many bits 
it has, but the true dynamic range) and how clean the power supply is, 
etc.

The same exact thing can be said of the "sound card" that you are using 
with computer based DSP programs.

A good DSP box will beat a lousy sound card, and a good sound card will 
beat a lousy DSP box.  The rest truly is just software/firmware, and 
amateurs can crank out equally good algorithms as commercial companies 
do.  Often better, since presumably they do it out of the love of it.

A good filter algorithm cannot make up for a crummy A/D converter.   
Analog garbage in, digital garbage out.

Naturally, the software in the computer or the firmware in the DSP 
boxes also have to also be able to handle the dynamic range of a good 
A/D converter -- using a 105 dB dynamic range A/D converter with 16-bit 
processing is a waste, for example.  Modern operating systems have 
floating point audio chains that accept 24-bit A/D converter inputs and 
enough horsepower to use it.

> As far as I'm aware, the twin peak filter is the only feature 
> mentioned that the MP doesn't have at all (don't know about the 
> Pro-II) but both bandpass and twin-peak filtering are also available 
> in MMTTY.

At 170 Hz shift and 45 baud data, properly designed set of matched 
filters for Mark and Space tones will implicitly create a "double peak" 
response.

I usually don't specifically design a double peak filter ahead of the 
demodulator, they only screw up the matched filtering, and you would 
have to compensate for the envelope waveform if you already have an 
ideal matched filter.   Instead, I use a symmetrical flat FIR bandpass 
for the RTTY signal for the "roofing filter" of the RTTY signal that is 
about 350 Hz but with very steep skirts.  I then let proper matched 
filters do their stuff; I use separate matched filters for Mark and 
Space ahead of the adaptive slicer.  The resulting system response is a 
double peak for the design data rate.

That said, all else being equal, for noise reasons I prefer that my A/D 
converter is outside of the computer.  I write my own DSP software, but 
I use external USB-based A/D converter boxes instead of PCI sound 
cards.  I use a handful of Griffin iMic converters, an M-Audio Transit 
and an M-Audio Quattro; these are 20- to 24-bit converters of different 
ages and dynamic ranges.

73
Chen, W7AY