you might try an absorptive filter for harmonics as described in QST
11/1968 which is constituted by a highpass filter terminated in 50 Ohm,
with its input parallel to the LPF, both having the same corner frequency.
Thus *all* frequencies produced by the PA see 50 Ohm.
In German we call this "Weichenfilter".
>From: [email protected]
>[mailto:[email protected]] On Behalf Of Paul Nicholson
>Sent: Wednesday, March 02, 2016 12:45 PM
>To: [email protected]
>Subject: Re: LF: MF: EbNaut
>I wrote the other day about seeing a transient that I thought
>was bouncing off the primary center.
>Now after dismantling and measuring a few things I find
>that transient was reflecting from the LPF which follows the
>transformer. The transient width was about the same as the
>FET switching time and it came back to the drain of the off
>transistor at a voltage many times higher than the supply.
>Pinging the transformer itself I can see no sign of any
>transmission line behaviour up to 50MHz, the turns are just
>too tightly coupled. In fact the coax cable with outer as
>primary and inner as secondary has a very tight coupling.
>351uH per winding and 1uH leakage inductance. I wont be
>using coax again though, the inability to tap the secondary is
>a severe limitation.
>So I learn that it's no good just looking at the bandpass
>and bandstop of the LPF, one might have to consider transient
>reflections because they end up back at the drain. Almost all
>the departure from idealness of the drain waveform was down
>to the LPF and not the transformer. I still haven't accounted
>for the excessive heat of the windings when the primary center
>is AC grounded.
>The divided down 477.69kHz multiplier output was compared
>with a GPSDO reference and the RMS jitter was less than 2
>degrees over timescales up to 60 minutes. Here is a plot of
>the 2-sample RMS jitter over a run of eight hours
>The plot means for example, that the average difference
>between two samples of the phase taken 512 seconds apart is
>1.04 degrees. The plotted phase is scaled back up to the
>tx frequency. The jitter at the soundcard output of EbSynth
>is one ninth of the plotted value.
>This pretty much convinces me that the spreading seen by Markus
>was unlikely to caused by my tx. Some doubt remains because
>I'm using the same GPSDO (Trimble) to reference both EbSynth
>and the phase comparison. Also the PA and antenna were not
>in the loop.
>The phase comparison is done using vtsid from vlfrx-tools,
>sampling the phase 100 times per second.