[Elecraft] K2 keying bandwidth significantly reduced

[Wayne Burdick]

I was going to wait for Eric to come back from vacation to collaborate with me
on this, but I had a brainstorm and decided to give it a try. Seems to work
well. We'll have to get a few builders to try the modifications, but here's a
brief summary for those who may be interested (results are discussed at the end):

Hardware changes:

1. U10A on the control board is reconfigured as a second-order low-pass filter
rather than a simple R-C shaping network. This requires one new capacitor and
three component value changes, including replacing D3 with a resistor. The
result is an approximately symmetrical, rounded trapezoidal waveform with about
4 ms rise and fall times. The original circuit had a characteristic R-C
exponential decay waveform on the falling edge, which was a principle cause of
the observed bandwidth.

2. A large capacitor is added from pin 2 of U8 on the Control board to ground.
(The MAX534's output buffers are stable with any capacitive load.) This reduces
the slew rate of the output buffer, further rounding the corners of the keying
envelope. The final rise/fall times are about 4-5 ms.

Firmware change:

A change in transmit signal sequencing is required to take advantage of the
slower rise/fall times. (If the modified hardware is used without the new
firmware, the waveform will be distorted and there will be no improvement in
keying bandwidth.)

Results:

Initial tests show at least a factor of two reduction in the bandwidth of keying
sidebands, using a method similar to that posted recently by Earl, K6SE. The
signal also sounds very clean (and looks very clean on the scope). Waveform
symmetry is preserved over the full power control range. I haven't done any
extended testing yet, i.e. using spectrogram. 

One other point of interest. The bandwidth-limiting technique suggested by
W8JI--routing the CW signal through a narrow-band filter--would be very
difficult to implement on the K2. The transmit signal path cannot be
conveniently be routed through the CW filter, even with the KSB2 option
installed. It would require 10 or so additional parts, and probably a couple of
coax jumpers, and could degrade the ultimate rejection of both the CW and KSB2
filters on receive. The solution I described here (turning the shaping network
in to a 2nd-order LPF) is the only simple method I have found during two days of
lost sleep and head-scratching. In fact it's probably similar to what W8JI ended
up doing on his other radios, although I couldn't find a theoretical description
of his actual modifications, and don't have the FT1000 schematics.

Assuming this change passes muster with our short list of high-power contesters,
we'll phase it into the K2 and offer some type of mod kit. It's likely that this
and a few other minor changes will be included in next K2 firmware release. 

At present there are two many unknowns for me to give you a date on this. Please
DO NOT call Elecraft about it--we'll announce it, as usual.

73,
Wayne
N6KR

-- 

http://www.elecraft.com