Not all KPA100s are created equal

[DaveL G3TJP]

Mine, constructed to the latest Rev. D, seems to have had an uncommon problem that was easily traced to noise radiating from the high voltage bias oscillator and generator.

 

I was experiencing bands loaded with spurious signals, most of which were S5 or less, but some, notably 1875, 3560, 14068, 14224, 18087, 18165, 21229, and 28224 were way above S9+40 (USB, OP1 = 2.4KHz, DSP bypassed).  Other than that, the PA performed admirably and received good reports on CW and SSB.

 

Lifting the anode of D7, providing the -ve voltage to the MAX1406 reduced the largest of these spurii to around S4/5.   Cutting the HV bias track adjacent to cathode D1/C43 reduced the noise another S-point.

 

Advice and assistance from reflector contributors covered soldering, grounding and component problems.  All was eliminated using the "grim reaper" principle.  EVERY component in the HV bias supply, including Xtal, transistor and transformer were replaced with new spares purchased direct from Elecraft.  Components around the T/R switch were also replaced.  This made not the slightest improvement and the only remaining advice was to 1. Construct the HV supply externally in a screened box, 2. Mount the PA in a separate cabinet or 3. Test out the N0SS HV line filter.

 

1 and 2 were unacceptable, so the N0SS mod was tested out in an enhanced format.  Since most of the noise seemed to be radiating from the -ve supply to the MAX1406, that track was cut just after C36/D7 and a 1mH (1 millihenry) choke was inserted, followed by a 0.1uF cap. decoupling to a convenient point on the ground plane.  This reduced the strongest spurious signal to S5 and ensured the disappearance of all spurii originally below S5.  N0SS's original modification, kindly supplied to me by Gary AB7MY in PDF format required the addition of a .01uF 250v capacitor between junction C43/D1 and the ground plane closely followed by a track cut to the T/R switch, bridged with a 100uH RFC.  I also decided that C77 in the T/R switch could be paralleled with another .018uF 250v as I now had a spare, so it might as well be used.  This gave a further 1 to2 S-point reduction in the strongest spurii.   Experimenting with filtering of the 12v feedline (in the ribbon cable) from the control board, showed no reduction in noise level whatsoever so that was left alone.

 

Originally, my bias voltage was 148v but after installing the 1mH choke/.1uF HV line filter together with the additional capacitor in parallel with C77, it was down to 128v.  The -ve supply was originally -26v but after installing the N0SS mod, it was down to -21v.  As there was still plenty headroom with the voltages, I decided to reduce the oscillator's energy by replacing R4 (100k) with a 220k.  This reduced the HV bias voltage to 110v and the -ve supply to -18v but it also reduced the strongest spurii to S1 (in the case of 18087) and below S1 (although still perceivable) for the rest of 'em.  That was with the KPA resting on a box beside the open K2 and without the shield in place.  I anticipate further reduction in perceivable spurii when everything is reassembled and screwed up tight, but that won't be for a few more days as the K2 is scheduled for further (minor) surgery.



There's bound to be scope for further improvement, but it'll do for a start.  I hope that when the rig gets back on the air, these spurii will be below the noise floor.



73 all.  DaveL  G3TJP

 

 
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com