[aprssig] Interesting Preliminary Findings on Soundcard "Soft TNC" Shootout

Stephen H. Smith wa8lmf2 at aol.com
Mon Dec 26 03:44:21 EST 2011

I have started comparison testing the relative performance of three sound-card 
software packet TNCs that can be used with APRS programs as a replacement for 
hardware TNCs such as a KPC3 or TNC2.       The three "soft TNCs" I am testing are:

o      MixW 2.20 running in packet mode.

o     The most recent freeware version of the AGW Packet Engine (Ver 2010.414)

o     The new UZ7HO "soundmodem" beta Ver 0.41

The "transmitter" in my test setup consists of the signal generator of my 
IFR-1500 communications monitor.   Audio from a PC running UIview, the UZ7HO 
soundmodem and a Griffin Electronics "iMic" external USB sound system is fed 
into the external FM mod input of the IFR.   This is a very low-distortion 
modulator with a flat response from less than 5 Hz to over 100 KHz.  (I have 
actually fed the output of a homebrew stereo multiplex generator into this 
input with the IFR set to generate in the FM broadcast band,  and have exceeded 
the performance of commercial FM stereo broadcast transmitters.) The RF output 
of the IFR is variable with a digital attenuator in 1dB steps from 0 dBm to 
-127 dBm.

The receive setup for the first quickie tests was a Kenwood TM-G707 VHF/UHF FM 
transceiver.  The IFR generator is connected to the antenna input of the 
Kenwood, as it would be for receiver alignment or a sensitivity/SINAD test. I 
am taking audio from the 6-pin mini-DIN "data"/"packet" jack, passing it 
through one of the homebrew soundcard interfaces described on my web site, and 
into the line-level audio input of a Dell L-866r PC.    [The built-in sound 
system in this elderly 866 MHz Pentium III is not the typical low-performance 
brain-dead mostly-software-based AC'97 motherboard sound system found in 
virtually all current PCs.   Rather there is an actual HARDWARE-based 
Soundblaster AudioPCI 64 built into the motherboard; i.e. the same chipset and 
accurate crystal-controlled timebase used on many PCI-card add-on sound systems.]

The test consisted of repeatedly hitting "F9" in UIview to force a beacon 
transmission on demand, while gradually reducing the RF level of the generator 
in 1 dB steps, until the receive setup stopped consistently decoding the packet 

I ran one of the softmodems described above at a time on the same PC, trying 
them at both the standard 1200 baud  1200/2200 Hz tones  and with 300 baud 
2100/2300 Hz tones.    The RF levels were measured in dBm rather than 
microvolts since the IFR's attenuator is calibrated this way.      The 
correspondence, in 50-ohms systems, for those not familiar with dBm is:

-119 dBm     .252 uV
-120 dBm    .224 uV
-121 dBm    .200 uV
-122 dBm    .178 uV
-123 dBm    .159 uV
-124 dBm    .141 uV
-125 dBm    .126 uV

At -119 dBm, the TM-G707 has definitely lost full quieting on an unmodulated 
carrier, with soft hiss evident. 1 dB lower at -120 dBm, "popcorn" type noise 
starts appearing , along with the "fine-grained" hiss.   (The transition from 
fine hiss to "popcorn" corresponds closely to the "12 dB SINAD" sensitivity in 
most narrowband FM receivers.)

Note that the test setup has "flat" audio response on both transmit (no 
pre-emphasis in the IFR modulator), and on receive (no de-emphasis on the 
mini-DIN dataport output). As a result, more high-frequency hiss was present at 
low RF levels than a pre-empth/de-emph mic&speaker hookup would have.

The results were:

*** 1200 baud ***              Lowest level (dBm) for reliable copy; i.e. EVERY 
packet decoded
MixW 2.2                          -119 dBm
AGWpe                             -119 dBm
UZ7HO                              -120 dBm

*** 300 baud *** Lowest level (dBm) for reliable copy;i.e. EVERY packet decoded
MixW 2.20                         -120
AGWpe                             -121
UZ7HO                              -123

UZ7HO                              -124   (Still copying 90% of test packets!)

1)    Note that these 300 baud tests were conducted over FM; not SSB. 
However, the 300 baud mode shows a worthwhile improvement in weak signal 
performance over 1200 even on FM.   Conceivably this might be useful in some 
specialized low power application over a marginal path on FM.     Note UZ7HO is 
fully-tunable to any audio tone pair at either baud rate.  MixW is tunable at 
300 baud but fixed at 1200, while AGWpe is fixed at both baud rates (1200/2200 
at 1200 baud and 2100/2300 at 300 baud).

2)    The UZ7HO modem clearly outperforms the other two , especially at 300 baud.

3)    In actual off-the-air testing on 30 meters HF APRS (i.e. SSB with lots of 
non-gaussian noise),  with all three modems running at once off the same 
receiver (a Kenwood TS-690), I am seeing about 10 successful decodes for UZ7HO 
vs 7-8 for AGW vs about 1-2 for MixW.      UZ7HO has the capability of using 
multiple additional pairs of audio detectors (a.k.a. "receivers") spaced in 
intervals of 30 Hz to each side of the main channel.   These increase the CPU 
loading of the PC, but the UZ7HO modem then reliably and consistently copies 
off-frequency stations that the others completely miss, thus dramatically 
increasing it's edge over the other two.

4)    I intend, in the next week or so, to expand this series of experiments to 
include a variety of radios, and to include some hardware TNCs, probably a 
KPC3+, a KAM and a PK-232.

I will also experiment with mic jack/speaker audio hookups vs flat in/flat out. 
   (The IFR has a mic jack with typical 3dB/octave pre-emphasis speech input - 
I need to make up a weird cable to go from a PC's 3.5 stereo minijack to the 
IFR's oddball-spaced 6-pin full-sized twist-lock DIN mic jack.)



Stephen H. Smith    wa8lmf (at) aol.com
Skype:        WA8LMF
Home Page:          http://wa8lmf.net

*****  NEW   Precision Mapping 9 For UIview Released! *****

  Vista & Win7 Install Issues for UI-View and Precision Mapping

30-meter HF APRS over PSK63

"APRS 101"  Explanation of APRS Path Selection & Digipeating

More information about the aprssig mailing list