[Elecraft] K3 pwr out on digimodes

[Kok Chen]

On Jan 17, 2009, at 1/17    3:00 AM, hb9ari wrote:

> I can imagine that digimodes software generate  ± complex audio  
> signals,
> then we can adjust this audio level to allow TX chain, including  
> PA, to work in a
> linear zone.

Out of the common HF digital modes, I believe that BPSK31 and QPSK31  
are the most sensitive to nonlinearities.  RTTY, MFSK16, Olivia and  
Domino are M-ary FSK (M=2 for RTTY) and are less sensitive to  
nonlinearities since their crest factor is 1.  However, with poor  
linearity, you can no longer use envelope waveshaping to help reduce  
the FSK keying sidebands.

BPSK31 is differential phase shift modulation, with either a phase  
change of 180 degrees or no phase change, between successive bits.   
The envelope is also waveshaped so that during a non-zero phase  
transition, the amplitude is modulated in a sinusoidal manner.   The  
envelope modulation, what W4TV had called "predistortion," it is  
simply the waveshaping of the keyed signal with a sinudoidal window,  
easily achieved in the I and Q components of the PSK31 generator.

(A digression: a PSK31 signal can be easily generated by starting  
with baseband I and Q signals that swing between +1 and -1.   
Depending on whether you want 0 or 180 degrees (and the addition of  
90 and 270 degrees for QPSK31), you either keep I or Q, or both,  
unchanged and slew any component that changes between +1 and -1 with  
a half a period of a cosine function).

When no information is being transmitted in PSK31, it puts out the  
"idle" Varicode, which in PSK31 simply causes a 180 degree phase  
transition at every bit period.  The dibit representations in QPSK31  
are chosen so that an idle Varicode also puts out a 180-degree phase  
shift per dibit period.

When a PSK31 signals goes into idle, it therefore creates the  
familiar two tone signal, with only two sidebands (the center  
frequency is suppressed) that are 31.25 Hz apart. This causes a clean  
PSK31 to look like two parallel "rails" on a waterfall display.

Any deviation from a perfect sinusoidal modulating envelope will  
cause an idling PSK31 signal to develop higher order keying  
sidebands, just like what a two tone test signal does with voice mode  
SSB.  The keying sidebands at 2.f0-f1 and 2.f1-f0 are measured,  
compared with the "fundamentals" at f0 and f1 and this ratio is  
reported as the IMD number in PSK31.

An IMD somewhere around -28 dB is considered decent, with many good  
stations putting out signals that are cleaner than that.  People do  
differ in what they consider to be "good enough" until they get  
QRM'ed by a loud, but not so clean, signal :-).

With the 100 watts PA turned off, I have measured an IMD of -34 dB on  
my K3, running at 10 watts peak (average idle signal of 5 watts).

However, with the PA turned on and running above 25 watts average on  
idle (50 watts peak), the IMD from my K3 is only around a -24 dB.  
(IMHO, very mediocre.)  With the PA turned off, the K3's IMD also  
starts to rise slowly when power is decreased below 500 mW.

Please note that this was measured on my K3 that was built from a kit  
(SN 1432).  Your K3's performance may be different from mine.

For rough comparisons, my FT-1000MP running below 5 watts and with no  
ALC action whatsoever, is capable of putting out an IMD that is  
cleaner than -40 dB IMD both on the bench and with on-the-air tests.

I have also copied a local who was using QRP levels on his TS-2000  
with a tad better than -40 dB IMD.  The IMD of my FT-1000MP does  
degrade to -30 dB dB when I raise the average PSK31 power to 25  
watts, but still adequately clean.

A clipped modulating signal will create something like -11 dB IMD,  
and I have seen even be worse than that on the air -- I suspect that  
the signal was not just clipping but had overshoots and undershoots  
in the envelope.

73
Chen, W7AY