[aprssig] Interesting Findings - 300 Baud AX.25 on VHF-FM APRS

[Stephen H. Smith]

On 4/2/2016 7:30 AM, Ev Tupis via aprssig wrote:
> In regard to the findings below...you said that you were surprised that the
> difference played out on FM as it does on SSB but didn't go on to say why you
> thought that it would not.
>
> It's going to be tough for anyone to offer comments until they know what you
> are unclear about, exactly. ;-)
>
>

This is due to my previous experience with the very non-linear curve of 
S/N-vs-signal-level on FM as you start falling off the fully-quieted "cliff" 
and quite abruptly start losing quieting over a very small change of RF level. 
I.e. decreasing the RF level by 1dB at the "cliff" doesn't decrease the S/N by 
1dB; it may decrease by 5 or 6 dB.

This is very different from the "gradually getting noisier" behavior one 
experiences on SSB as the signal drops. I.e. where a decrease of 1dB in RF 
level yields more-or-less proportionally a one dB reduction in S/N.  In my 
experience, the loss-of-quieting cliff far out-weighed any subtle differences 
due to audio modulation mode, bits-per-second, pre-emphasis or not, ultimate 
demodulator performance, etc.



A couple of years ago, I did a controlled "shoot-out" comparing various 
hardware and software TNCs.   The test setup was similar to the setup I 
described in the first post of this thread.   The difference in that test was 
that I used a high-quality digital off-the-air recording of Mic-E packet bursts 
from a Kenwood D700 to modulate the IFR's RF generator.   (It's actually one of 
the tracks on my test CD.)

I tested an MFJ-1270, a KPC3+, an original KAM (VHF port), an AEA PK-88, the 
AGW packet engine, MixW in packet mode, and the UZ7HO Soundmodem.  All of them 
failed within 1 dB RF level of each other except for the Soundmodem which was 
dramatically 4 dB better.  Later, I discovered that the Soundmodem had 
implemented a missing-bit recovery algorithm that the other hardware and 
software didn't.

All those previous tests had been done at 1200 baud only. At that time, any 
subtle differences in audio demodulator performance were overwhelmed by the 
rather abrupt loss-of-quieting from the decreasing carrier level.

(I later tried a direct audio-audio connection between the test signal source 
and the TNCs under test. Gradually decreasing the audio level did show a wide 
range of differences, but it's inconclusive since the optimum audio input level 
for the various devices may be sharply different.

Some day, I will repeat these tests, but mix varying amounts of white or pink 
noise into constant-level test audio, to truly characterize the different 
devices' performance with noisy signals, unbiased by the FM discriminator 
quieting threshold effect.

This time around, I was contrasting 1200 vs 300 baud on the same software (the 
UZ7HO  Soundmodem) and saw dramatic differences in the decode performance.  It 
turns out the Soundmodem has DSP audio bandpass filtering before the 
demodulator. The filter bandwidth automatically varies for the different baud 
rates. Note that these are pre-detection audio filters; not the post-detection 
low-pass filters with a cutoff just above the bit rate common in many hardware 
and software modems.

___________________________________________________
Stephen H. Smith    wa8lmf (at) aol.com
Skype:        WA8LMF
EchoLink:  Node #  14400  [Think bottom of the 2-meter band]
Home Page:          http://wa8lmf.net

  _______ Windows 10 Outrages! _______
    <http://WA8LMF.net/Windows10_Info>

Live Off-The-Air APRS Activity Maps
    <http://wa8lmf.net/map>

Long-Range APRS on 30 Meters HF
    <http://wa8lmf.net/aprs/HF_APRS_Notes.htm>