[aprssig] 9600 baud capable radios

[Henk de Groot]

Robert Bruninga schreef:
> Most new radios advertise 9600 baud "auxilliary" ports
> but I remember QST reviewed many of them and said
> most of them were not good performers with an external
> 9600 baud TNC.
> 
> Does anyone know what is the bottom line?  If I want
> to buy a CHEAP radio but then add an external KPC-9612+
> 9600 baud TNC, what radio is known to perform *best* in this 
> application.

Well, this is easy!

Most VCO's in modern radio's are PLL controlled. So the output frequency 
is devided and the output is compared to a stable crystal oscillator. When 
the VCO drifts away the PLL changes the voltage to keep VCO on the 
frequency. The voltage is not applied directly but through a low-pass 
filter so it can only change smoothly.

Now when you FM or PM modulate the VCO then the VCO obviously changes 
frequency and the PLL will output a voltage change in an attempt to keep 
the VCO on one frequency. The output of the PLL will be passed to the low 
pass filter and the output of the filter will be the avarage of the PLL 
output and therefore the VCO will be modulated but stay on the center 
frequency.

Now the above only works if the modulation is much higher in frequency 
than the passband of the low pass filter. If you apply a very low 
frequency of say 1 Hz then the PLL will follow this change easily and you 
will not see the VCO change frequency at all (or more acurate, you will be 
able see it but it is very, very much supressed).

So for such a radio the modulation should be well outside the passband of 
the low pass filter. With 9600 baud almost all of the 8 kHz voice band is 
used and it contains frequency components as low as 10 Hz. Now 10 Hz is so 
low that the PLL tries to compensate. So the PLL constantly changes the 
center frequency since the PLL sees the 10 Hz components in the signal as 
frequency drift.

A good 9600 baud G3RUH signal requires a flat envelope from 10 Hz to 8 
kHz. The supression of low frequencies distorts the signal. Result is that 
zero-crossing (needed for clock recovery) becomes less defined and the 
response of one bit has influence on the signal in the following bits 
(inter symbol interferrence). This reduces the detection marign. Remember 
that no radio has a perfectly flat envelope and this distortion makes it 
worse. When you also add noice then the signal cannot be decoded anymore 
when the RF signal becomes weaker, this happens much sooner than it needs 
to be.

Hearing this you may be thinking about lowering the passband of low-pass 
filter in the PLL. This means you have to use a much higher quality VCO 
and redisign the PLL. But remember that when changing from RX to TX you 
have to shift the VCO as well. Something like this will make the RX/TX 
changeover very slow! So this is not a feasible alternative.

By the way, the slow RX/TX changeover time is an additional problem for 
PLL boxes! Is't a waste to have 9600 baud and then spend most of the time 
on long TX delays.

This makes a normal PLL box unsuitable for 9600 baud, at least when using 
a signal according to G3RUH. Now there are some ways to make a good 9600 
baud (or better) radio:

1) Use a radio with a crystal oscillator. So your old TR7200 from of 
arround 1980 might proof to be a very good 9600 radio. Remember we need a 
flat envelope from 10 Hz to 8 kHz so we have to bypass the audio amplifier 
with all its capacitors and modulate the VXO almost directly. Also the 
special TEKK data transmitters use this.

2) Use a PLL VCO and mix the output with an FM modulated VXO - so not 
modulate in the PLL. The output of the mixer is then your signal. This is 
the kind of radio I use. A telephone is full duplex and a modified Nokia 
car telephone is very suitable. It uses one PLL VCO simutaneously for 
transmission and reception, for transmission the signal is mixed whith an 
FM modulated VXO. RX/TX changeover is very fast since the PLL VCO doesn't 
change frequency. Some modern multiband radio's also have this, they are 
excellent for 9600 baud.

3) Use a specially designed PLL VCO, here you modulate the VCO but also 
the PLL's referency crystal! Now the PLL doesn't see the 10 Hz modulation 
as VCO drift since the reference is also changed - the PLL sees no change 
and therfore does not try to compensate! This design is used in the German 
T7F packet radio by DF2FQ.

Of course there are other ways and that is by using another type of 
modulation. There are schemes possible that do not have these low frequecy 
components. But G3RUH just wanted to fit the signal in the passband of a 
normal HAM radio, so manchestercoding etc could not be used.

I hope this wasn't too technical or too hard to understand. So look for a 
radio with one of the above 3 designs. Any "staight" design with direct 
modulation in the PLL loop will fail, unfortunatly this accounts for 99% 
of the Ham radios from Japan.

Kind regards,

Henk.