[Elecraft] K3 Cw pileup problem: effects of Noise blanker?

[Guy Olinger, K2AV]

I'm really having a really hard time identifying with any of the AGC & NB 
complaints.

I used the second-to-last alpha test firmware (not latest) installed to my 
K3 in the recent ARRL DX CW on 40m from NY4A and was simply delighted. I can 
nearly or effectively remove key clicks with certain choices of  *digital 
only* NB coupled with same bandwidth narrow roofing and DSP.  The K3 front 
end plus these settings turned the usual 40m incoming cr*p into a zone where 
not copying a QRP signal was the exception as long as there wasn't a 
stronger station calling. At most there were a dozen or two signals all 
weekend calling that were not pulled out.

It was, frankly, astounding.  And *very* beneficial to score (1700 Q's on 
40m).

I did not experience AGC pileup distortion of any kind. With the key click 
nullification I was able to work 3/4 of the contest USING SLOW AGC, albeit 
set to its fastest decay. No theory here, just what happened.

The most exceptional reduction was one period, about two hours, where a 20+ 
over clicky EU station was right below me, and was so close I had to use 350 
width for running. With NB off had S6-8 key clicks, NB on had S1 S2 S3 band 
bottom.  I used slow AGC during this time and worked QRP stations and weak 
Russians through it.  Slow AGC was easiest to listen to (no idea how to 
express technically what I mean).

There used to be that "next unworkable layer", stations you could hear were 
there, but just could never pull out.  The unworkable layer was workable 
this contest. The end of workability occurred with the station dropping into 
absorbtion, as I now know is a very clean and transparent transition that 
doesn't sound like anything else on HF.  You could hear it happen.  I 
suspect it sounds like a vhf beacon signal dropping into solar noise though 
that is a very old memory for me.

We worked Russians on 40m at 1130Z when it was 1:30 or 2:30 in the afternoon 
their time, which were only a whisper above a remarkably constant and 
unjunky noise level.

More specifics on click removal:

The clicky station must be well down on the skirts of the roofing filter for 
the trick to work and the best results seem with DSP width set at the TRUE 
bandwidth of roofer.  Both my "250" 8 poles measure out ~ 330.  350/300 DSP 
width was best. Narrowing DSP below Roofer width seems to mildly worsen 
results. I presume getting the clicky station well down the roofer skirt 
makes sure that within the DSP IF the click appears as a detached click 
rather than a clickish aspect of discrete signal that is inside the DSP IF.

There is one necessary artifact to nullifying key clicks, and that is a 
discernable notch on a very weak signal, that with a little help from fast 
deep QSB can break a dash into two dots, or turn a dash into a dot.  I 
figured that out very early and made upward tweaks on my callsign 
correctness paranoia to compensate.  I will take that mild mental adjustment 
with a smile anyday over frequencies rendered useless for weak signal work 
by key clicks.

I cannot comment on use of 5 poles for this.  For running bandwidth reasons 
I use 400/"250" 8 pole roofers in both RX, set to kick in at 450/350 Hz 
width settings.

I did not use IF NB, or use NR at any point in the contest. I do not find NR 
at all useful on CW, confirming Lyle's statement that he did not see how it 
would be useful, everything else optimal.  Be careful that in messing with 
the AGC for these complaints I don't understand, you don't UNDO what I used 
last weekend.

73, and thanks for a real edge in the contest past.

Guy, K2AV

P.S.  Puhleeze, 10 Hz step for shift and width on CW.

----- Original Message ----- 
From: "ni0c" <ni0c at earthlink.net>
To: <elecraft at mailman.qth.net>
Sent: Friday, February 27, 2009 5:24 AM
Subject: [Elecraft] K3 Cw pileup problem: effects of Noise blanker?


> I've been thinking about the recently reported "cw pileup" problem,
> wondering why this apparently wasn't noticed before-- particularly by
> the VP6DX operators.

> My guess is that the VP6DX operators probably never needed the noise 
> blanker in their (presumably) quiet QTH.
>
> It may be that the solution to this problem may be different for those 
> using
> a single receiver K3 versus the dual receiver.  It may be that the 
> solution could be as simple as disabling the NB when switching from VFO A 
> to VFO B.
>