[amsat-bb] AFSK vs FSK

[Bob Bruninga <bruninga@xxxxxxxxxxxxx>]

Just to summarize and invite any indepth analysis from the experts...

We all accept that AFSK is inferior to other modulation methods, But
recent testing and mental gymnastics for satellite links to EXISTING
Mobile FM radios is leading me to a narrow, but contrary conclusion:

  1) 1200 baud AFSK in a 15 Khz FM channel appears to be 7 dB better than
     9600 FSK in the same channel bandwidth.

  2) 1200 baud AFSK in a 15 KHz channel is only about 2 dB worse than
     could be obtained even in a matched bandwidth 1200 baud FSK radio.

  3) Most on air 1200 baud AFSK signals are terrible implementations...
     (I only threw this in so that people dont try to correlate their)
     (existing on air terrestrial experience with this discussion....)

We welcome any of the experts to correct us if we have made any grievous
errors in the following assumptions.  Here are the data points we are
using:

1)  Assuming Kenwood with its all-in-one-integrated dual 1200/9600 baud
data radios did the best they could with their FM radios, testing has
consistently revealed on 4 different radios a measurable 7 dB difference
between 1200 baud AFSK and 9600 baud FSK.  THus, only 2 dB worse than
the theoretical difference between matched receivers for 1200 and 9600.

2)  In a fixed 16 KHz bandwidth channel, with simple FSK demods, there
will be no appreciable difference in performance with Baud rate.  IE, a
1200 baud FSK and 9600 baud FSK signal will be about the same since the
noise power is the same)... (again, assuming only simple FSK demod)

3) McDermott's book gives a good mathimatical analysis of an almost 4 dB
improvement of FM AFSK to AM AFSK.  If we assume that an AM AFSK signal is
placing 1/4 of the energy in a single sideband and that AM AFSK in a
single  sideband is equivalent to FSK, then one could say that SSB FSK
gives a 6 dB advantage over FM AFSK and so 6 minus 4 gives us again, a 2
dB performance loss of AFSK compared to FSK in matched bandwidth
receivers.
But we dont have matched bandwidth receivers.  We have 15 KHz.

4) Looking at it another way, we can consider the AFSK modulation is in
fact, spreading the 1200 baud signal bandwith to fill the full 15 KHz
available.  The FM is taking an approximate 3 KHz bandwidth voice
channel and spreading it to 15 KHz or a factor of 5.  If we think of this
as "spreading gain" then again, we can expect about a 7dB gain with the
1200 baud AFSK signal properly demodulated..

SO, all of these simplifications, assumptions, and tests seem to support
one another.  BUT we have one contradicting data point:

We have noticed subjectively that the 9600 baud downlink from the Pacsats
does occassionally decode on the Kenwood receivers almost 6 dB lower than
our terrestrial tests.  Some of this could be due to these reasons?

   1) bursts-vs-continuous data stream, PLL sync, TXD etc...
   2) 9600 PACSAT transmitter is XTAL instead of PLL (loop modulation)
   3) These are "occassional" decodes, so we are a few dB lower on the
      BER curve.

There are many AMSAT members who operate 9600 baud FSK for the pacsats
reliably and continuously for the last 10 years or so.  It would be nice
to be able to compare the Kenwood D700 and D7 9600 baud Demodulation
performance "against" someone who thinks they have an optimum 9600 baud
PACSAT receive capability.  This would tell us if this 7 dB is real, or
just Kenwoods implementation....

Our tests show that the Kenwoods decode each other reliably at about -113
dBm using 9600 baud.  Can anyone with an optimized PACSAT station observe
their miniumum decode level from another kenwood or a PACSAT and then
calibrate that signal level to dBm with a good quality instrument?

THanks.  Always hoping to learn...

de WB4APR, Bob




----
Via the amsat-bb mailing list at AMSAT.ORG courtesy of AMSAT-NA.
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org