I can replicate the problem. It doesn't change decoding, but it does change
the SNR to a false high number. Simply unchecking RefSpec and going to
Flatten, makes the problem go away. (It is most noticeable on HF , where I
run FT8 a;; the time.) I have not checked it for any other mode.
I'll triple check it this afternoon when I have time. I'm on q65 on 6m at
the moment.
73, N0AN
Hasan
On Sat, Oct 16, 2021 at 5:31 AM Bill Somerville via wsjt-devel <
wsjt-devel@lists.sourceforge.net> wrote:
> Hi Hasan and Rich,
>
> a couple of points worth considering:
>
> 1) the "Flatten" filter has no impact on decoding whatsoever, so having
> "Flatten" checked or not will not change SNR numbers reported,
>
> 2) many rigs have considerable roll-off in receiver gain towards the
> bottom end of the expected receiver pass-band, the K3 is one of those BTW.
> Consider the consequence of that if a reference spectrum correction is
> applied to the pass-band? Note the reference spectrum does have an impact
> on decoding as the correction *is* applied to the sample data feeding the
> decoder.
>
> In summary, the "Flatten" filter is great for improving the visibility of
> signals on the waterfall and 2D spectrum, *if* the waterfall & 2D spectrum
> frequency bounds are set to reasonable values. The reference spectrum
> correction was introduced to help with the wider bandwidth modes where an
> uneven receiver pass-band can cause different tone frequencies of the same
> received signal to be artificially attenuated and correcting that can aid
> decoding of weaker signals since the decoders assume the whole signal is
> received at the same constant amplitude that it was transmitted with.
>
> In general with FT8 and other narrow band modes there is a *small* chance
> that a reference spectrum correction may even out the tone amplitudes of a
> signal received at the steep roll-off point of your receiver, and help with
> decoding. But you must also understand the consequences of the extra
> non-linear gain applied to that signal, and those even further towards the
> edges of your receiver pass-band, by the reference spectrum correction.
> Note that you can view the reference spectrum captured curve in the 2D
> spectrum on the Wide Graph window, or as a small insert graph at the bottom
> of the Equalization Tools dialog ("Menu->Tools->Equalization tools
> ...").
>
> It is also worth noting that for MSK144 mode the phase accuracy of the
> receiver is far more important than the amplitude accuracy. When using
> MSK144 mode a phase correction should be considered rather than a reference
> spectrum correction if you receiver performance warrants it. Capturing a
> phase correction curve is somewhat more complex as it requires a known good
> MSK144 received signal rather than background noise as a reference baseline.
>
> 73
> Bill
> G4WJS.
>
> On 16/10/2021 10:46, Hasan N0AN via wsjt-devel wrote:
>
> The cause here for that was using Ref Spec instead of Flatten . I chased
> it for weeks, it has no effect on decoding, but produces silly high SNR if
> below 400 Hz or so. Simply uncheck Ref Spec in the Waterfall controls, and
> go back to "Flatten" and I bet the problem goes away.
> 73, N0AN
> Hasan
>
>
> On Sat, Oct 16, 2021 at 12:03 AM Black Michael via wsjt-devel <
> wsjt-devel@lists.sourceforge.net> wrote:
>
>> I can see the opposite possibility in this data. Looks like below 300
>> the levels are low, high from about 300-400, then pretty consistent
>> Red line here is 50-point moving average of 5000 points. Above 2200
>> we're getting into the region where bandpass limits might have an effect of
>> some sort (lack of noise or perhaps more likely to have higher power
>> transmitters?).
>>
>> [image: Inline image]
>>
>> Mike W9MDB
>>
>>
>>
>>
>> On Friday, October 15, 2021, 11:45:16 PM CDT, Rich - K1HTV via wsjt-devel
>> <wsjt-devel@lists.sourceforge.net> wrote:
>>
>>
>> Has anyone experienced exaggerated dB signal reports given by WSJT-X to
>> decoded FT8 signals of stations transmitting tone frequencies below 500 Hz?
>> I use a K3 with 2.8 KHz filter using WSJT-X Version 2.5.0 . I have also
>> seen this in earlier versions.
>>
>> When stations using low audio tones are decoded, the signal levels that
>> WSJT-X gives them far exceeds the levels given to signals with higher
>> frequency tones of equal intensity level observed visually on the wide
>> graph.
>>
>> I have run the “Measure Reference Spectrum” tool on a quiet frequency and
>> have the “Ref Spec” checkbox on the Wide Graph checked. This didn’t resolve
>> the noted anomaly. I first observed this on Foxes running 3 or 4 streams
>> using tones below 500 Hz with the lower tones appearing a few DB stronger
>> than the higher tones.
>>
>> Any idea as to why this may be happening.
>>
>> 73,
>>
>> Rich – K1HTV
>>
>
> _______________________________________________
> wsjt-devel mailing list
> wsjt-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/wsjt-devel
>
_______________________________________________
wsjt-devel mailing list
wsjt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
