2
$\begingroup$

This is a waterfall made in gnuradio with a simple audio source->waterfall block at 12ksps. The filtering I'm talking about is therefore all on the Elecraft KX2.

Waterfall showing odd filter behavior

At the top of the waterfall the filter is pretty wide. But there's some unrelated signal (in red) in the top right. I want to change the filter to remove that.

I press "FIL" on the KX2, and reduce the bandwidth. You can see that the yellow portion becomes narrower for a short bit. When it goes below about 1.8KHz filter BW, it seems to create a band stop filter around 1.5KHz.

That doesn't seem right. I just want to make the filter narrower, not notch out the middle of the band.

At the bottom of the waterfall I reset the filter bandwidth to about maximum.

Is this a bug in the Elecraft KX2 DSP, or am I misunderstanding what I'm seeing?

If you're curious, this is me listening to 14.074 JS8, but there's some (I think FT4) high up in the band. It's not necessarily actually interfering with me being able to decode the signals below it, but I do want to understand the Elecraft filters.

So tl;dr question: Why the notch filter when the filter BW goes below 1.8KHz?

Edit: second screenshot shows that this notch stays about the same, as the BW continues getting narrower.

To be clear: This is not I/Q signalling, it's the upper sideband only. The X axis is correct, in that it starts from 0. I see the same in wsjtx/js8call.

Narrowing filter down to minimum

$\endgroup$
4
  • $\begingroup$ Are we just seeing the two sidebands? If you narrow it more, do they get thinner? $\endgroup$
    – aMike
    Commented Jul 9, 2023 at 13:18
  • $\begingroup$ No, this is just the (mono) audio spectrum. The waterfall, when given a float (not complex), has the option to show only the right half. This is that. So the axis at the bottom is correct. It goes from 0Hz to ~3KHz. It's not I/Q, and not duplicated (the unwanted signal is not mirrored on the left). $\endgroup$
    – Thomas
    Commented Jul 9, 2023 at 15:28
  • $\begingroup$ I see the same notch in the frequency (audio) spectrum and waterfall when I run wsjtx / js8call. $\endgroup$
    – Thomas
    Commented Jul 9, 2023 at 15:29
  • $\begingroup$ Sorry, didn't explicitly answer your second question. I have now added a second screenshot showing the BW going down to minimum. As you can see, the notch does not grow. $\endgroup$
    – Thomas
    Commented Jul 9, 2023 at 16:35

1 Answer 1

1
$\begingroup$

I got an answer from Elecraft directly. It's not the greatest answer, but:

That is a side effect of Weaver demodulation for SSB. Since that post is mentioning digital modes like JS8 and FT4, you should be using DATA A mode for digital modes like that. DATA A does not use Weaver demodulation, so it will not have that notch in the
passband.

I don't really understand why that narrowing the filter for SSB should create a notch at 1.5kHz, or why that would be desirable, but apparently that's what it does.

I've confirmed that this phenomenon does not happen when in DATA mode.

$\endgroup$
3
  • 1
    $\begingroup$ A clue from some course material on the Weaver SSB modulator (but demodulation is similar, only backwards): "Note that the second pair of quadrature multipliers should, ideally, be DC coupled. [...] In practice DC coupling is not often provided, since it produces DC-offset problems. As a result there can be a small gap in the message, as received, in the vicinity of (B/2) Hz." $\endgroup$ Commented Jul 19, 2023 at 6:26
  • 1
    $\begingroup$ So, I think, the depth of the notch at 1500Hz depends on your filter response at 0Hz. Turning down the top end of the filter shouldn't affect it, only raising the low cut. $\endgroup$ Commented Jul 19, 2023 at 6:28
  • $\begingroup$ Very interesting link, thank you. Though it doesn't make me able to answer its Q6 quite yet. $\endgroup$
    – Thomas
    Commented Jul 19, 2023 at 17:00

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .