[Top] [All Lists]

[TenTec] Pegasus/Jupiter keying - In search of Perfect CW :-)

To: <tentec@contesting.com>
Subject: [TenTec] Pegasus/Jupiter keying - In search of Perfect CW :-)
From: W8JI@contesting.com (Tom Rauch)
Date: Sun, 3 Feb 2002 19:45:17 -0500
Hi Lee,

I see what you are describing now, and it should be clean CW. 
Waveform is VERY important, more so than rise and fall in most  

I'll try to do a better job of explaining, because I still some of us 
assume we can hear a raised sine or filtered waveform as a "soft" 

> OK........getting back specifically to the keying characteristics of
> the PJ's,  perhaps I should have explained how I measured the rise and
> fall times.   I  used a long standing standard for measuring pulses of
> timing the slope  between 10% and 90% amplitude, and then I rounded it
> to  the nearest full  millisecond.   I did not think we were into
> splitting hairs here.   Beside, I  have not calibrated my 'scope
> against a traceable standard lately so I do not  know how the error it
> may be contributing.  

The problem isn't splitting hairs.

The problem is we are trying to verbally describe something that is 
almost impossible to look at and decide if it actually clicks. One 
difference is I use the starting and final points of both the rise time 
and the fall time because that is what concerns us in switching 
times and bandwidth, and you are using the 10% and 90% levels of 

I'd like to go back to the point I made earlier. We can not tolerate a 
fast turnover in slope at any point in the waveform. That would 
include points below 10% and above 90% amplitude!

We should suppress sidebands (clicks) outside the normal 
bandwidth of our receivers at least 50dB, and more if we are 
parking next to weak signals. The typical range of signals at my 
house measures about 60-70 dB, and at times it approaches 80dB.

You can't reach that level of attenuation with the type of waveform 
shown in the Handbook, because if you look at the waveform 
carefully you will see sharp slopes at the very start of the rise, and 
the very start of the fall. We have to be very fussy what we do in a 
radio, otherwise we will bother people. 

A good start is to use a 5mS (total) rise and 5ms or slightly longer 
(total) fall time, and if you use a sine-shape (error function or 
similar) you can get a bit faster and not have problems.  
Sharp slopes don't make a pleasant "sound", they cause clicks. 
They are too brief to make it through the filters of anything but a 
very wide receiver filter. We will never even hear those "clicks" 
unless we are tuned off to the side of the signal, and the only thing 
they do there is cause QRM!
What has been described here as "desirable" is actually the  
waveform Kevin describes as an exponential waveform and I 
describe as "single-pole R/C filtered" waveshapes. This is the 
shape provided by FT1000's as well as those old grid-block keyed 
rigs, it is the result of filtering the modulation (a square wave from 
the key) in a simple R/C circuit. The result (harmonics and all) then 
AM modulates the carrier.
> More importantly, the keyed CW pulse looks very much like the 
> "error  function" turn on described by Kevin Schmidt K9CF in his paper
> found at: 

The shape I call a "filtered dot" is nearly identical (nothing is 
perfect) to Kevin's error function. If it has the shape or a similar 
shape as Kevin describes it is also known as a "filtered dot". Again 
what we can not tolerate is a sharp bend or steep slope at any 
point in the waveform.

> Kevin goes on to say in that paper "...For a given rise time, the
> error function  shape for the rise and fall will attenuate unnecessary
> interference away from  the carrier frequency much better than
> exponential keying..." Hence, the PJ's, while having maybe a little
> less than some arbitrary ideal  rise time, have the ideal wave shape
> which leads to minimizing key clicks.    See the paper referenced for

Indeed! Also the sound will NOT be too soft. The point I try 
(unsuccessfully so far, hi hi) to make is when filtering is set for a 
reasonable bandwidth you get something similar to a raised-sine 
shape, a waveform without abrupt changes in slope at any point 
from zero to full power.

There is no possible way any of us can "hear" a few cycles of a 
500 Hz or 800Hz tone as a "soft" rise and fall, because the period 
is too short for the ear and brain to register. Even if your ear and 
brain worked that fast, the receiver would not permit it when using a 
normal CW filter.

If you want an example of how you "hear" a few cycles of a tone, 
tune in WWV and look at the seconds "tic" on a scope. The "tic" is 
a perfect sine-wave tone (I think it is a 1000Hz tone).

> a full discussion. One can confirm that the PJ's do not create
> excessive clicks by doing a  fairly simple test.  Connect the radio to
> a dummy load.  Loosley couple a  clean, quiet receiver like a
> CorsairII such that the keyed signal of the radio  under test produces
> a 10 to 20 over S9 signal.  Send a series of dots with  the
> transmitter being tested.  Switch in the narrowest filter in the
> receiver  and tune around the signal.  If there are any troublesome
> clicks, you will  hear them. I did just that a few minutes ago and
> found NO key clicks anywhere on the  band.   I did have some phase
> noise close in as expected, but no key clicks

Actually that is the best test possible, because you are looking at 
the signal the way the receiver and ear looks at the signal. My web 
site describes that method in my tests, except I used both a 
receiver and a peak detecting selective level meter with sample and 

Every radio should be tested that way.

It is possible to actually have a rise and fall and shape that looks 
perfect, and have bad clicks! It is not possible to have an 
excessively short rise and fall and not have excessive bandwidth, 

But now that The waveshape and times are described in more 
detail, the "PJ" certainly has the potential to be what for all 
purposes is click free...and with that rise and fall time, it should not 
sound too soft either.
73, Tom W8JI

<Prev in Thread] Current Thread [Next in Thread>