>Alex's plots show that TrueTTY is 3 dB more sensitive than MMTTY when
>presented with an RTTY signal in the simple white noise case. But
>both demodulators require *much* more SNR when faced with the real
>world conditions, i.e., "Fig 17." It is probably because of this
>variability why most people don't notice that MMTTY is poorer than
>TrueTTY, even though TrueTTY beats MMTTY by an even bigger margin for
>the selective fade case.
Hi Chen
Many thanks for pointing out Alex's work, which for some reason I had missed...
Very interesting indeed...
I have not tried true-TTY yet, but having played a lot with pathSim and MMTTY, I
know that MMTTY sensitivity can be simply increased by turning the BPF
filter on (which unless mistake is turned off when using MMTTY defaults).
Further sensitivity improvement can be provided but playing with
filters parameters i.e using very sharp FIR filters and even a sharp notch
between the tones (to make a kind of twin-peak filter).
For example the "preset" copied below, will provide identical (even slightly
better) results than the ones plotted by Alex for trueTTY in an AWGN channel (
ie using Alex's nice tool: CER 0.02 for -8dB in AWGN)
But this is on the simulator... the BPF becomes almost useless, once considering
MMTTY connected to a receiver like the pro2 and using the RTTY filter (normal
the transceiver DSP does the job).
I also know, for having done quite a few contests with mixW and MMTTY in // that
mixW is an excellent tty decoder, but that MMTTY can sometimes do slightly
better, especialy when using the most appropriate preset... something which
cannot appear in Alex's plots, as apparently all tests were done using the
"standard RTTY" preset.
So definitely an interesting job, but I would wait a bit before considering
MMTTY as an obsolete decoder... Applying the full unfiltered 3khz BW audio at
the MMTTY input and not using the BPF, may not give a real idea of the real
MMTTY performance, and at least not a real idea of what's happening when the
modem is connected to a receiver, furthermore if it has its own DSP...
I am of course impatient to try TrueTTY, and take this opportunity to thanks
Alex for providing RttyCompare, a really useful tool.
Patrick
PS This MMTTY preset provides about 2.5dB sensitivity improvement (compaired to
VE3NEA figures http://www.dxatlas.com/rtty/ ) on the simulator (AWGN channel)
[Define6]
Name=hyper sensitive
Enabled=1
AFC=1
AFCFixShift=0
AFCSQ=32
AFCTime=8.000000e+00
AFCSweep=1.000000e+00
UOS=1
TxNet=1
TxDisRev=0
LimitDxGain=2.000000e+02
LimitOverSampling=0
LimitAGC=1
ATC=0
ATCTime=4
Majority=1
IgnoreFreamError=0
SQ=0
SQLevel=7.000000e+01
DefMarkFreq=2.125000e+03
DefShift=1.700000e+02
DefStopLen=0
OutputGain=2.457600e+04
TXLoop=1
TXBPFTAP=48
TXLPF=0
TXLPFFreq=1.000000e+02
TXWaitType=0
TXCharWait=0
TXDiddleWait=0
TXCharWaitDiddle=0
TXRandomDiddle=0
TXWaitTimerDiddle=0
Rev=0
ShiftFreq=1.680000e+02
RXBPF=1
RXBPFTAP=256
RXBPFAFC=1
RXBPFFW=5.000000e+01
RXlms=1
RXlmsDelay=0
RXlmsMU2=3.000000e-03
RXlmsGM=9.999000e-01
RXlmsAGC=0
RXlmsInv=0
RXlmsTAP=56
RXNotchTAP=512
RXlmsBPF=1
RXlmsType=1
RXlmsNotch=2211
RXlmsNotch2=0
RXlmsTwoNotch=0
DEMTYPE=1
BaudRate=4.545000e+01
TTYBitLen=5
TTYStopLen=4
TTYParity=0
SmoozType=0
SmoozOrder=5
SmoozIIR=6.000000e+01
Smooz=3.000000e+02
Tap=512
IIRBW=60
pllVcoGain=3.000000e+00
pllLoopOrder=2
pllLoopFC=5.000000e+01
pllOutOrder=4
pllOutFC=2.000000e+02
Diddle=2
TxFixShift=0
InvPTT=0
TXUOS=1
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|