Is there a workaround to push autoscale of the waterfall deeper into blue?

The waterfall is a nice tool for identifying both signals and noise but the autoscale function tends to waste the best part of it, in my opinion. Virtually every time I hit 'autoscale' I end up also hitting 6-8 times to get the interesting region near the noise to have better visual identity. The default autoscale setting don't make good use of this IMO.

I'm unclear where this issue is best addressed, possibly in WebRX, but barring a systemic fix I'd like to know if there is a workaround that will achieve these results without user involvement every time a display is autoscaled.
Can someone tell me?

Glenn n6gn


  • The autoscale algorithm is very simple and has fixed parameters that probably need to be controlled by sliders in a "colormap" extension panel. Such a thing has been on my wishlist for a very long time.

    When you click autoscale the bins from one waterfall FFT are sorted by their dBm value. The 5th percentile value defines the "noise" level. The 98th the "signal" level. 30 dB is added to the signal level to set the WF max level. 10 dB is subtracted from the noise level to set the WF min level. All four of those values need to be adjustable.
  • The one useful thing about the current as-implemented values is that they are consistent across all of the installed and up to date Kiwis in the wild. That allow a few projects to log on and get SNR values for various bands and display on a webpage like

    I suspect that the SNR values obtained by the WSPRdaemon project are more accurate in reality and more useful to the end user given the values can be gathered and graphed locally, I'd happily lose the global consistency in favour of tunable waterfall auto scaling functionality.
  • You have a point about comparisons among different receivers but in examining the results I find the SNR algorithm to be a very poor reflector of station performance. The best WSPR stations with Kiwis don't necessarily rank well at all. A station with black-out spectrum is far above better stations showing the whole spectrum. At one time the 'winner' was a station with a big power supply oscillation or SMPS spectrum. In short, I ignore the SNR ranking as I have not found it useful.
    Given the trade-off, I'd prefer John's suggestion every time.
  • the linkfanel scheme shows little relevance to real usability from my experience.
  • Note that is using kiwiclient/kiwiwfrecorder (or some variant of that code) to obtain raw waterfall FFT samples as the input to its ranking algorithm. This is not influenced by what I mentioned above which is only what the Kiwi UI does to similar raw sample input. So any changes there would not alter the relative results of linkfanel measurements.
  • A year or so ago, I went down the rabbit hole of trying to come up with an automated measurement of the SNR/"quality" of online KiwiSDRs. It's one of those cases where you can look at a waterfall and pretty quickly decide how "good" one is, but very difficult to do via an algorithm. Plus a Kiwi good one one band may be poor on another. I presently have two online, one is on a 120 ft T2FD with good coverage overall. The other is on a 43/48 meterband dipole so it has somewhat better performance in the 6-7 MHz range, but not so good elsewhere. I'm not sure if I'll stick with that antenna or try another.

    Over time you learn which Kiwis are good/useful, and which ones, well, aren't. :)
  • Chris,
    "very difficult to do via an algorithm" is certainly true! In essence it is this algorithm , this model or theory, I have been trying to achieve during the last few years as I've been working with broadband antenna solutions for the Kiwi. If we had systems each capable of comparing to a common external source, measuring the excess noise coming from the galactic center say, then we might be able to compare our systems but of course this is too far below most Kiwi/antenna systems (but not all!) at this point.

    Next to that solution might be every Kiwi owner stating the system's noise floor referenced to the thermal noise in the radiation resistance of the actual antenna, not the real part of the feed point impedance and not the ingress arriving by other paths. This kind of metric is perhaps possible but almost certainly beyond the scope of most Kiwi owners since it involves a knowledge and complex interpretation of actual antenna, efficiency, polarization, near-field 'ground', far-field 'ground', take off angle ...". To my understanding this is what the ITU tried to do in preparing and measuring noise as reported in Rec. ITU-R P.372-8. Getting diverse Kiwi systems all meaningfully referenced to this seems unlikely due to the complexity of usefully integrating both theory and measurement.

    Even so the Kiwi is such a compelling tool for learning about, measuring and improving radio and site performance in general that it is hard for me not to keep trying to do better but I think it very unlikely that a single metric or ranking such as we have now is going to be useful any time soon, if ever.
  • Regarding current noise levels an interesting report describing measurements and methodology of man-made noise at 50 amateur radio locations in the Netherlands ranging from from rural to city environments can be found in the reference below. These measurements are likely representative for other areas in the world with similar technological development.

    They conclude that the noise ?oor is statistically signi?cantly higher than would be expected from the latest ITU-R noise ?oor data and that similar noise floor measurements should be repeated frequently to stay in line with the technological changes.

  • Thanks for sharing.
    I'm concerned that most measurements at amateur locations may not be of propagated noise but rather near-field or common mode ingress to the measuring system. I don't have much to back this up except a few tentative measurements from quiet Kiwi sites such as KPH and a some amateur radio astronomy work locally. These data seem to indicate that, if anything, the ITU numbers are too high.
    For example it appears possible to detect and measure galactic noise at an amateur location near 20 MHz. These levels can serve to qualify the system and thereby to validate a noise measurements, at least in the nearby spectrum. On systems that can achieve this measurement, the ITU numbers generally seem pessimistically noisy.
    I think more investigation is needed.
  • Hi Glenn,

    Near-field or common mode ingress could have occurred although I am pretty sure the folks making the measurements were aware of this possibility and took utmost care to avoid that.

    However only 3 sites classified as quiet rural were used in this study. The definition used for that was no residences or infrastructures within 1.5 km radius.

    The Kiwi and other stations nowadays able to receive galactic noise around 20 Mhz will probably fall in that category as well and perhaps the study could have been more comprehensive adding more locations.

    For the rural quiet areas that could be a problem in a small and densely populated country as Holland...

  • my professional experience supports Ben's graph. I work with HF as a day job and any field work we've done the last 20 years has shown a steady rise in noise.
  • I don't doubt the experience of increased noise, I've experienced that as well. I'm just not confident that it is a far-field increase rather than near-field and CM. I suppose a way to separate these is to show, or not, measured galactic noise relatively close to a city or residential center. This requires a well-matched antenna system (well matched to radiation resistance and in absence of ground loss) along with low near-field and CM ingress.
    I may be able to test this at home and nearby where I do have available locations that meet the 1.5 km criterion. However it probably requires a portable 20 MHz vertically polarized dipole which I haven't yet constructed. I tend to want a dipole because the symmetry allows verifying that CM is not significant. The location can be chosen to rule out near-field except for the measuring device.

    [sorry for the topic drift in this thread!]
Sign In or Register to comment.