After seeking the advice of other BITX20 V3 builders on the Yahoo forum Leonard, KC0WOX, suggested I check the impedance of the transformers T1 and T2 and L1 in the linear amplifier.
I measured T1 at 3.8 uH for each winding and Leonard suggested this was too low as was T2.
I re-wound T1 to be 38 uH and T2 is now 3.8 uH.
For L1 I put 40 turns on a small binocular core and this measured at 130 uH. Leonard suggested at least 40 uH so this should be OK.
But, even with these changes the linear still seamed to be self oscillating when keyed down and connected to an antenna but not on a load.
Just to see what would happen I tried the linear amp without the BITX20 V3 connected, instead driving the linear from my signal generator. No problems now into the load or either antenna.
Then I tried it with the BITX RX input disconnected at the main board – again no problems.
I then added a solder bridge from the normally open side of the K1 relay to ground so the input to the receiver is shorted to ground rather than open during transmit and things were a lot better. It still occasionally self oscillated but nowhere near as bad.
Note: This turned out to be a bad thing to do. See BITX update #23.
So I think my problem is TX RF getting back into the BITX V3 main board and creating an unwanted oscillation. This may because of my mounting method and cables between the linear and the main board. I have the main board mounted on wood and the linear board floating in space. The RF connections between the boards are done with shielded audio cable and are about 15 inches longer than need be.
What I will try next is mounting both boards on an aluminium chassis, with brass standoffs and use RG174 for RF connections between the boards.
This should improve the grounding of each board and the BITX20 V3 should then be better able to deal with stray RF.