Automatic rating as Good if CW data decoded even is this comes from local noise?

Regarding Observation 1194164

This observation was automatically marked as Good, which I assume is because it has ‘decoded’ two CW data fragments (“ETEH” and “KKAT”)… however I cant really see a straight line in the waterfall that would indicate a Doppler corrected signal, is this data just spurious noise which happens to make it through the decoder?

Also I see the noise pattern on the left around 145.850 MHz, which I assume to be terrestrial from the post Doppler correction output on the waterfall, but I am curious as to why this appears to pulse about once per second and wonder if this is overloading the RTL-SDR?

Note I have +10 dB LNA with 2m bandstop filter which is tuned to 137mhz, and looks to have about -4 dB loss at 146 Mhz but tapers off pretty quickly so shouldn’t be any significant amount of out of band signals hitting the LNA…

However the RF chain is not shielded beyond an Alu case for the SDR and ferrite rings on the 1m usb extension, could the SMA connection between the LNA and SDR could be a weak spot that is picking up local interference?

There might be a few little fragments of CW in there that are visible, but it’s surely a stretch.

It seems to pick up 2 or 3 CW data fragments from each pass of XW-2D:

But also from UKUBE-1, which transmits its CW beacon on 145.840 nearby to to XW-2D at 145.855 MHz:

By comapring the Doppler shifts from both sets of observations it appears the interference is terestial, and centered at 145.850 MHz…

I will have a go at locating the source with gqrx on a laptop, but in the mean time does anyone recognise the intereference pattern which as far as I can tell doesnt match anything in the sigidwiki?

Looking back to see if I could find any successful decodes of UKUBE-1, I see Observation 1136419 which was “Vetted good on 2019-10-26 03:46:24 automatically”…

Should these observations be automatically marked as good based on the CW decoder picking some random strings from a waterfall with absolutely no sign of any signal at all?

This is a known issue with the CW decoder.

BTW, I inadvertently vetted one of the observations above as “bad” (rogue mouse click), but vetted it back to “good” as I don’t want to interfere. Sorry about that.

ah thanks, I have done some reading up of old threads about both the CW decoder and how auto vetting was added to address some of the shortcomings of the current observation rating process… which appears to be a can of worms!