Last Spectrum persists when all signals removed [fixed in v1.279]

edited April 2019 in Problems Now Fixed
Z14 with min/max adjusted such that with no input much of band is below 'dark blue'. Connect antenna (or turn on antenna/preamp) for normal display, say with noise floor dark blue some places and peak signals all on screen. Then turn off input such that all falls below dark blue. Waterfall goes black but spectrum display does not disappear, last display persists indefinitely whether decay is set for fast, medium or slow. Quickly zooming in one step and back out washes out that old, erroneously persistent display.

This also seems to occur some times for signals very nearly off screen in the dark blue direction, e.g. with setting such that > 15 MHz is very low, dark blue noise floor, if a signal noise burst comes along, the display will capture it and never let go of it even if the actual noise floor reverts to previous low value.

This is a different problem from the 'long time to decay' one described in the bug:spectrum list.
This bug makes it difficult to study changes in system intended to improve noise floor since only the the waterfall captures the changes and 'spectrum anlaysis mode' isn't available.


  • jksjks
    edited April 2019
    Your text says Z14, but all your images are of Z0. For which do you think there is a problem? Or is it with both?

    I don't see any "indefinite" persistence of the spectrum display. Granted, in "spec slow" mode it takes many minutes for the spectrum to decay all the way down to the noise floor for a disconnected antenna. Even fast mode takes a good 30 seconds or so. But it does eventually happen. And in fast mode I observed a signal burst eventually decaying from the display. This is just the way the IIR filter works given the relatively slow update rate of the Kiwi waterfall. There is probably an argument for including a more traditional averaging filter.

    Any action you take that alters the spectrum significantly (e.g. zoom in then out) clears the spectrum buffer because you lose information needed for the IIR (or averaging) to continue.
  • I confess, I didn't wait > 15 minutes but effectively it is useless as a means of making low level comparisons. It's a peak detect for signals where the 'low' signal is off screen.
    I do not find it useful.
  • You sure it's decaying? T=1 hour:
  • There could easily be a bug that I simply haven't stumbled across yet (certain conditions required to trigger). But for me it works fine. I'll try and work on adding an averaging filter.
  • It may be that the symptom can be described as "no refresh of old data if new data is off screen".
  • jksjks
    edited April 2019
    v1.279 has improved spectrum filter options. In addition to the default IIR filter you can now select an MMA (modified moving average) and EMA (exponential moving average) filter or no filtering at all. There is also a new slider for the parameter associated with each filtering algorithm:
    IIR: param = gain coefficient 0..1 (slower..faster)
    MMA: param = averaging size in samples (1..64)
    EMA: param = decay period (averaging size) in samples (1..64)
    The URL parameter options sp= and spp= allow the filter and filter parameter to be preset in the URL. See

    I implemented a minimum decay rate for the IIR filter. For various reasons the filtering is implemented after the waterfall colormap scaling. And with a more restricted colormap range (i.e. WF min slider less negative so no-signal waterfall area is black) the IIR filter would not decay for small signals as was observed.

    There is also a new "GND" momentary button (i.e. click/touch and hold). This zeros (grounds) the waterfall/spectrum input data so you can see how the various spectrum filters decay under no-signal conditions.
  • Took me a moment to workout I should switch away from IIR before playing with the SP Parm slider to see large affects, once I did that loads to play with.
    Very useful for "test equipment" side of Kiwi use.
  • But there is a significant difference between an IIR of 0.2 (the default) and 1.0
  • I was looking at a decaying strong signal and counting seconds for it to fall between two 10db lines when GND used.
    Just tried it again, can see a difference in the running spectrum activity fine but that mode is (as previously described) the least responsive to a decay test.

    I was sort of trying to make sure others tried the new modes first, before playing with the GND and SP Parm options.
  • Hi John,

    That's a very useful improvement, so thanks for that :-)

    The MMA & EMA filters work the best for me with SP settings of around 5.

    Any chance of adding peak hold 'outline', something like this in SDR Console V3


    Martin - G8JNJ
  • This is an excellent enhancement
  • edited April 2019
    Hi John,

    Brilliant :)

    Makes spotting and determining the amplitude of fleeting signals much much easier.

    Thanks for this (and all the other stuff too).


    Martin - G8JNJ
  • With the new spectrum options I spent a fair amount of last evening trying to screen dump a fleeting burst from mains networking, failed over forty times to hit it.
    Today with peak capture enabled, so I didn't need to be frame accurate, I hit it third attempt.
Sign In or Register to comment.