Topband: Tuned cw speaker

[Jim Brown]

Having worked in pro audio, I've dealt with digital for a decade or so 
longer there than at RF. It comes down to how hard the signal plus noise 
is hitting the digital system, and how close that signal is to the 
bottom of the tower of bits. It could be,

for example, that the analog RX had a better front end, and that the 
digital RX wasn't hitting the A/D converter hard enough.

I suspect that the advantage of JC's acoustic filter is that it has less 
phase shift associated with the skirts than electrical filters. The 
human ear does not like phase shift -- it's a form of distortion that 
degrades speech intelligibility. RTTY decoders also don't like it, and 
when I got back on the air 20 years ago and got interested in RTTY, I 
was telling top RTTY op W0YK that, and that IF should be run fairly wide 
to avoid that phase shift. He didn't believe me until G3YYD came up with 
his wonderful 2Tone decode and gave the same advice.

K1JT, the lead developer of WSJT modes, advises that the RX IF should be 
run wide open at 3 kHz, allowing WSJT decoders to do the filtering in 
the digital domain, and part of the secret sauce is the skill with which 
that is done. It's entirely possible that the guys writing the code at 
Flex don't get it.

I've never seen a Flex, but ten years ago, K6TD, a great engineer and 
great op, told me that the Flex 6500 he'd bought was the best radio he'd 
ever owned, but that the UI was awful, because the designers were 
programmers, not operators. Soon after, several top contesters, N6WM and 
K9CT among them, hooked up with them and within a few years the UI was 
something they love.

I've been using an Elecraft K4 for several years, and one of the great 
things about it is the performance of the filters. Elecraft's chief 
engineer, is an active CW op. AND both he and his partner, WA6HHQ, have 
always been VERY well connected with their user base. The only common 
knock with their first rig using digital components was that it didn't 
sound great. No surprise -- it was a fundamental limitation of the chips 
you could buy at a price hams could afford in 2007 when the K3 was 
designed. The K4, designed more than a decade later, sounds much sweeter.

And back in 2007, Wayne came up with what he called "sigmoidal" shaping 
of CW keying that redefined the state of the art in close-in CW TX 
bandwidth. Running K3s and amps with a pair of 3CX800A7s, close neighbor 
K6XX and I could operate within 500 Hz and hear the other only as a loud 
signal. 15 years later, that holds with K4s and KPA1500s. Wayne also 
took design steps that drastically reduced phase noise. One day back 15 
years, I accidentally ended up with both K3s on 20M, one feeding my 
SteppIR and the other a 3-el Yagi, both pointing about 70 degrees, 
making them approximately colinear by virtue of their locations. 60 kHz 
from each other, they didn't know the other was there, both running 
those Ten Tec amps legal limit.

When ARRL Labs tested the Flex 6500, it's CW bandwidth was pretty nasty 
(like the un-modified MPs). They subsequently implemented Wayne's 
method, and I've heard that the rigs are much cleaner.

What I'm getting at is the critical need for both skill AND 
understanding by the designers. It's more than just the technology! One 
of Wayne's early goofs was that the preamp on the original K3 wasn't 
nearly good enough above 15M. Within a year or two, there was an add-on 
that could go into an I/O patch point at the RX input, and the updated 
K3S ten years later had that preamp built in. And he was hip enough to 
have built the rig with that patch point!

73, Jim K9YC


On 2/7/2026 8:17 AM, Ron Spencer via Topband wrote:
> My take: for a SDR, if no signal was converted during the A to D stage, there will be no signal to "dig out" during the D to A stage. No amount of signal processing or noise reduction will help since there is nothing there. Nothing is buried in the noise floor.