Using NB during JT65 & JT9 operation

N4AB

When operating the digital modes with a 6700 and DAX to WSJT-X on the computer, I find that in my situation (noise floor at S5 to over S9 most of the time) using the NB gives me a much better Decode on the WSJT waterfall (and also seems tohelp the SSDR waterfall). On the attatched screen shot, at times 1738,1739 and 1740 I turned the NB off during the minute of decoding. Its a little hard to see on the JT65 signals, but during the NB off period the signals are greatly reduced, however it is easy to see the drop out on the JT9 signals. Over the last two years on a couple of occasions Flex has recommended not using the NB and NR functions for the digital modes. I would like to know how other people are setting up their radios for these modes. I use USB at 4kHz on all HF bands.

Bob-N4HY

Al:

I do not believe anyone that knew a lot about what is going on told you not to use NB with digital modes.  They probably did tell you not to use NR and the other one was almost surely ANF unless they made a mistake.  The latter two, NR and ANF, introduce phase distortions that will almost surely degrade performance.  The things you mentioned you observed in use of NB are good and desirable.  It also means there are no really strong signals nearby.  NB will work well for you under these conditions in the removal of impulse noise.  A strong CW signal nearby and you might not like it as much.  We are working on new code to help fix that now.  It has been announced that the next major release, 1.5,  will be a major reorganization and rethink of the entire noise and interference mitigation processing along with new features that are in that theme.  The goal is to include a new NB.

73's Bob N4HY

N4AB

Thanks for the response Bob. I can't remember exactly how it was phased but it might have been "do not use the DSP function" and I assumed (dirty word) that included the NB.

Thanks again, Al   N4AB

DrTeeth

From Tim W4TME: "Also turn off ALL DSP noise mitigation (NB); you should never use noise mitigation with digital modes". His emphasis, not mine.

Bob-N4HY

Again,  Tim and I have known each other, like each other,  fought a few decision making battles before.   When this is over, we will still be friends.  I disagree with Tim on this.

The thing that sold the superbly performing NB we have in the conditions that are correct for it,  were the use by W0VB doing JT65 and EME.  I wrote the algorithm myself and tuned it for Terry VB's operations.  Since I am the architect and original author,  I reserve the right to have an opinion on how it is or isn't used.

NR and ANF should never be used on digital signals.  You might use TNF.  NB can be used very effectively to pick weak signals out of a sea of mess from impulsive noise so long as the band you are in doesn't have,  contemporaneously,  large signals.

73s
Bob, N4HY

Bob-N4HY

I also believe the signal dependent mean algorithm, NB2, should always but always be set ON first and tuned for best performance without distortion by strong signals, and then, and only then, use of NB.  I will contact the current maintainers of the code to remind them that the correct order of operations in the code is backwards from the name: NB2 first,  NB second in time order of operations.

DrTeeth

Thanks for that Bob. Time to experiment.
Have a great weekend and Happy Easter
73 de Guy