A Tale of Two Digimodes A project to satisfy particular aims in the digimode domain G3XSD Feldhell# 3431
What were the aims? To find two modes capable of extremely good weak signal performance To find at least one mode where the duty cycle was low enough to allow 100 Watt usage from a nominally 100W tx To find at least one mode better than PSK31 and relatively competitive with JT65 but with none of the limitations e.g. fixed format, macro based exchanges, max freeform message 13 characters
Aims (contd) To find one mode that is human readable and not requiring a complex decoder hungry for CPU power That both modes should have enough adherents to allow a reasonable number of casual QSOs With suitable propagation conditions/freq choices, that properly DX communication should be possible
How to judge performance? Get consistency by using simulation (PathSim AE4JY) Use only one model for condx CCIR520-2 (set of conditions and modelling) At least 2 people have already done it! (Dave Freese (W1HKJ author of FLDigi) and Jaak Hohensee (ES1HJ). Rather than look at massed columns of figures we'll use Dave's approach. ref:http://w1hkj.com/fldigihelp-3.21/modes/compare.htm Jaak ref: http://kambing.ui.ac.id/onnopurbo/oraridiklat/teknik/psk31/doc s/ps_200810_pub.pdf
Mid Latitude disturbed NVIS 0dB S/N This is a near approximation of poor multi-path conditions, without static or other interference, but with white noise reducing the signal to 0dB S/N in 2.4kHz bandwidth- typical of 80m at night - W1HKJ From practical experience a QSO is viable if '% Copy' is on average greater than 90%, and impossible below 70%. - W1HKJ.
Errors vs Speed
Error efficiency
Bandwidth
Coding Gain
Weighting the criterion Bandwidth is not critical unless the mode cannot be used in the narrow band segments Errors vs speed may not be critical if CW equivalent speeds are OK as is Error/Efficiency Coding Gain -a measure of the typing speed compared with the symbol rate. Implies redundancy and redundancy=resilience What about computational speed and simplicity?
What were the modes of choice? One where simplicity/performance/resilience was best balanced Hellschreiber One that would produce the best % copy in the worst conditions Olivia
Hellschreiber The signal characteristics are similar to high speed CW (maybe why I like it?). Greatest redundancy of all modes tested: 49 symbols per character Relatively insensitive to poor signal quality or interference Uses the human brain's pattern recognition powers (just like CW..but vision instead of audio If you can see and you can read then OK no code to learn Insensitive to group path delays,polar flutter, and QRM/N whilst exact tuning and drift are not critical A Fuzzy Logic mode rather than purely digital
What's not to like? Narrow bandwidth ~300-400Hz and may be used in narrow band segments Compatible with narrow CW filters in the rig Very low duty cycle (min 6%-max 39%-average 22%) - Rig can run flat out Simple no special equipment required.can use the simplest CW rig no other digimode can do this! NO CODE TO LEARN!
What does it look like?
Noisy Signals?
How it works Vertical rasterisation sent sequentially No sync signals so scan is two lines wide* CW performance *outdated notion?
Many variants 3wpm/25wpm/50wpm 125wpm/225wpm/ OOK ASK. ON/OFF Amplitude Shift Keying 2-MSK. Minimum Shift Keying 2-FSK. 2 tones Frequency shift keying
Hell Club https://sites.google.com/site/feldhellclub/home/feld-hell-faq Awards Worked All States Frequencies Worked All Canada WARC'n Overtime DXCC Worked All Continents Worked All Europe SWL Awards 160 METERS 1.840 80 METERS 3.574 to 3.584 40 METERS 7.077 to 7.084 30 METERS 10.137 10.144 (Region I) 20 METERS 14.063 (PREFERRED), 14.073 (Note 1) 17 METERS 18.104 Contests Contests General Sprint Rules Sprint FAQ 2015 Sprint Results 2016 Sprints 15 METERS 21.074 12 METERS 24.924 10 METERS 28.074 6 METERS 50.286
OLIVIA The diametric opposite A more complex encoding is hard but not impossible to achieve! But can give 100% copy when s/n floor is below -14dB Revisit W1HKJ measures Many variants from 64/2000 32/1000 down to 4/250 Works at so low signal levels that VFO searching is useless and a system of 'voluntary channelisation' is used. Also copies well under QSB/QRN/Flutter(polar path) and Auroral conditions Relatively comparable with JT65 BUT ragchew capable!
The Big Idea
16/500 and Contestia
16/500 and Contestia (contd)
The encode process
FEC 1 7/64 block code 7 bit ASCII to 64 bit Walsh function 128 possible 64 bit words ( from a choice of 18,446,744,073,709,551,616) Possible to find 128 unique words easily differentiated from each other with a maximum given number of errors! (how many errors before one is mistaken for another?) If any combination of error bits is to be corrected to the right answer typically this is 2 or 3 max. ( for this redundancy) Many algorithms: BCH ; Reed Solomon ; Hamming ; WalshHadamard.
Q.What if burst noise occurs? A. Interleaving
Scrambling? To avoid rx false lock up on certain patterns e.g.make patterns seem more random and with less self correlation Each 64 bit word is scrambled with: 0xE257E6D0291574EC With end-around carry a 13 bit shift as each sequential 64 bit vector is scrambled with the pseudo random pattern
FEC 2 Now the block is 64 vertical X 5 horizontal each 5 bit entity ( now called a symbol) is signalled as one of 32 tones An FFT decoder will identify the tone and as ONLY ONE tone is present (2.5/sec spaced by 31.25Hz and shaped to efficiently contain the energy within a single tone BW) The demodulator has a resolution of 15Hz. The modulator allocates the tones using Gray Code Gray Mapping is always optimal in a sense that it minimizes the number of bits that can be demodulated with error when using Maximum Likelihood Demodulation (or Minimum Distance).
Olivia FEC1 decode
Better than other modes? Research published by DK8OK with cooperation by Dave W1HKJ using Pathsim by Moe AE4JY -Using Watterson's Ionospheric channel modelling and CCIR 520-2 conditions definitions see: http://www.moetronix.com/ae4jy/files/pathsimtech100.pdf It depends what you want..high information content messages or high speed transmission or minimal bandwidth or what? The results for many modes follow:
Green=good copy,yellow=acceptable,red=unusable
Green=up to 10% CER,Yellow=11-30%CER,Red =>30% CER Green=good copy,yellow=difficult copy,red= impossible copy
Contestia=Olivia adjusted to run 2X faster in ½ the BW
QSO's about?
What about G3XSD?
Further Tests? Extended daily tests for 7 days G3XSD received at 100% copyable level for each day, every day in VK (Nov/Dec 1400 UT PSKReporter) ZL2SEA HF Echolink still usable but direct link preferred Practical tests to be done : 8/250 lowest s/n -14dB slow 14.5wpm 16/500 lowest s/n -13dB medium 21.5 wpm 32/1000 lowest s/n -12dB faster 25 wpm What's best for the context of the experiments?
M0GAM - The other end
Where are they?
Live trials begin soon.m0gam, Avril, in blue...