Hello David,

First of all many thanks for the updated FLdigi test version!


Here are the results and observations on THOR25 in brief. Please refer to the attached updated spreadsheet for details, if necessary.


1. Raised Cosine Signal Shaping: Excellent!

- Tested and confirmed OK: no more clicking when signing on using a Shape factor of e.g. 100 or more, while the setting '0' reproduces the previous state (non-shaped).

- Not sure whether the removed end of signal shaping might be an issue, since it clearly creates a 'click'.
=> This could be avoided, if the same (user-set) yet reverfsed raised cosine shaping was applied there, too.


2. New Thor-25 DigiMode Definition: Major Improvement!

- Compatibility with previous alpha test Thor-25 DigiMode Definition: NO (just for information).

- RSID (TX and RX incl. the new secondary RsID code 691): OK.

- Speed: similar to previous Thor-25 test version. OK.

- File Audio Playback (original FLdigi created THOR-25 stream): Text reproduced 100% OK (Columns H+I in the spreadsheet referred to below). OK.

- The signal decoding bandwidth ranges about +/- 250 Hz around the true center frequency (+/- 130 Hz with 'Slow CPU'). OK.

- Not shown in the spreadsheet: the SNR is stuck around +5 dB (3 dB with 'Slow CPU') when decoding the original (perfect) audio.
=> This is certainly not realistic and should be corrected before a final release.


3. PathSim Mid-Latitude Disturbed NVIS AWGN SNR 0dB resulting audio: RX text with 4 line errors (the previous test version had produced 10 errors)!

* See sheet 'PathSim NVIS THOR25A v4.2.06.14' see RUN #1 Columns B+C, RUN #2 Columns D+E, RUN #3 Columns F+G (Slow CPU). All previous sheets are unchanged:

- I also attach a gently filtered flac audio file (smaller size but keeping the entire relevant bandwidth unchanged) just for your information.

- Initially, for 38 consecutive lines or about 4 minutes, the decode was perfect!

- Thereafter, in a few lines, there were 1 or 2 decoding errors, leaving part of the text scrambled.
Note that if a linefeed was missing (error at the end of a line), I have corrected it manually to avoid auto-marking all lines below as errors.

- A repeat using the exact same audio a 2nd time has reproduced the same fair result, interspersed with the few errors described above. Repeatability OK!

- When checking [x] Slow CPU, the error rate increased to 14. Hopefully this might provide a hint as to where the remaining decoding issues may come from...?


* See sheet 'SNR PathSim NVIS THOR25 v.06.14':

- The FLdigi FEC log shows, that the signal is rated between error-free, yet still quite frequently at 50%, and once at 75% error, thus still a bit variable.

- The average SNR on this disturbed signal is rated at -2.3 dB (unchanged), and it varies quite a bit along the way, yet probably in phase with the simulation conds.


4. Questions:

- To find out about the remaining decoding issues, and possible further improvements:
=> Is there a way (e.g. using the existing log protocol - upon request?) to continuously track / log the current de-facto decoder frequency offset, SNR, and FEC?
=> In other words, could these values be added to the daily log file tags, e.g.
current definition: RX 1500 : THOR25 (2024-12-09 18:16Z): <...RX TEXT...>
extended definition: RX 1500 (1497): THOR25 -2.3 dB 75% (2024-12-09 18:16Z): <...RX TEXT...>
=> I thought about maybe 'seeing' whether the decoder does stay stuck on the 'correct' signal center frequency, or whether (and how) it starts sliding sideways?
=> Since decoding works so well over long stretches, maybe this might explain the short dropouts, and a slower 'AFC' in the THOR sense might stabilize the result...?

- The AFC Button is still active. Better deactivate it or does it have any function in THOR? I have not seen a difference when activated.

- On a related note, for MFSK AFC, if possible, I'd prefer allowing an even higher maximum low-pass filter setting, so far it's limited at 200 (quite helpful!).


Bottom line, this has been yet another amazing leap ahead in the new THOR-25 DigiMode definition!


Thank you, David!


73s Tobias

.-.-.