wsprdaemon noise graphs



  • Hi Gwyn, I read your paper about the Noise measurements from researchgate with great interest... excellent writing!
    I am sorry, I didn't mention the sources of the pictures above:
    SDR picture is a screenshot of Simon's SDR-console V 3.0.14
    WX-map was from

    I am fiddling with the wsprdaemon DEFAULT- Noise parameter. Now I have set de parameter 10db up in level. So instead -10db, most bands are just arround 0 : "Default:10:0"... means, that 10m band does get 0db adjustment from what the S-meter reading actually is.
    The S-meter parameter in the kiwi-admin page is at -16 which gives me correct S-meter value for S9=-73dbm. Only 12m and 15m had to be adjusted as S-meter value was 3db too much (same as already found out by N6GN). I adjusted this with the DEFAULT parameter in WD-Conf file.
    Is my understanding correct how to calibrate the noise plots correctly ?

    Ulli, ON5KQ
  • edited October 2019
    Hello Ulli,

    I wrote the following before I saw Glenn's response, but I think my text complements his and Gwyn's.

    WD's noise level measurement levels are independent of those displayed by the Kiwi's waterfall and S-meter, although Gwyn, Glenn and I have gone to some trouble to correlate them.

    The Kiwi's levels are modified by the  two calibration fields of the admin/Config page. We have found that setting those to -16 results in S-meter and WF levels which match the single tone level we inject into the RF SMA. If WD.conf has no LEVEL_ADJUST field for a receiver or it is set to DEFAULT:0, then the WD levels should match those of a Kiwi set to that -16. If your RF distribution chain modifies levels, then you should change those calibrations appropriately.

    However the Kiwi's sensitivity varies with frequency with a +3 dB peak about 25 MHz and adjustments for that are built in to WD.
    In addition many sites (like KPH) have RF distribution chains feeding the Kiwi RF SMA which modify the relationship between the antenna feedline RF level and the RF delivered to the Kiwi. For example, at KPH there is a 23 dB gain DXE/Clifton LNA and a number of filters and passive splitters in that chain. In such cases adding one or more values to the (optional) LEVEL_ADJUST field gives WD the information it needs to adjust the noise levels it measures. For example, if there were a 10 dB gain LNA, then DEFAULT:-10 would subtract 10 dB from measurements on all bands. Adding BAND:level,... further refines the adjustments when you have such information. At KPH I used a $50 NanoVNA to determine the gain/loss for all bands in the system:

    "KPH_HF_77 KPH CM88mc foobar DEFAULT:-4.65,2200:,630:-3.11,160:-3.11,80:-3.11,60:-4.02,40:-4.42,30:-4.66,20:-4.65,17:-4.53,15:-4.43,12:-4.34,10:-4.31"

    You can see that at KPH the noise level on the higher bands is below -161/1 hz. But obtaining those levels requires a quiet site and a preamp in front of the Kiwi, since it has a poor noise figure. The -122 dB in 3 Khz I measure on a terminated Kiwi corresponds to about -156 dB/1 hz, which is 18 dB above -174 dB KTB and well above the -160 dB measured on 12M and 10M at KPH.

  • Hello Ulli

    Thank you for your nice comment on our detailed paper on ResearchGate, and for the sources of the two figures.
    Glenn and Rob have answered your questions - they have worked on those aspects and so are the best sources.

    Gwyn G3ZIL
  • Many thanks for all the answers.
    I deleted the old noise folders now, because obviously they do not show the correct levels, as I didn't understand the Default terminology correctly.
    Now to begin again, I just don't set any Default parameter - let's see, what I get.
    The S-meter in Kiwi-Admin is correctly set, as I have checked with the lab-generator

    It is just a matter of setting the code-words in conf. correctly - I don't understand anything of computer code ... therefore I am likely to be stupid.

    About ITU and field strength measurements, I used LZ1AQ's Excel spreadsheet with one of the loop designs. I found the calculation not far off from reality, when compared with a calibrated system in the EMC lab.
    Check Chavdar's (LZ1AQ) website here:
    "spreadsheet to calculate parameters...."

    To use a monopole is not very wise, as outside the EMC lab there are too many errors from other metal structures easily introduced - vertical monopoles are very sensitive against reflections from sourroundings.
    (magnetic) small loop antennas are much more useful as they are not that sensitive in real world (outside) areas...

    I found my qth, when measured 2 years ago about in the middle of rural and residential according to ITU recommendation
    At the IARU Region 1 website is a nice tool:

    Ulli, ON5KQ
  • The WD-Kiwi is now running without any Default parameter in the conf. file. But note I am using an active vertical. I adjusted the total antenna plus amplifier signal level as such, that the strongest stations with good propagation just NOT overdrive the kiwi-SDR. The pre-amp has very high gain so the max level is defined by the clipping of the Kiwi.
    As such we do not know which field strength causes these levels at the input of the Kiwi, as we do not know the frequency dependent antenna factor.
    But at least operating the kiwi like this gives me max dynamic range, although it is not enough and 10m is still too quit, if my neighbor switches off his computers in the office... (difference of 10db in noise level)

    I have just taken down the magnetic loops - so calibrated field strength measurements are for spring 2020,,,;-)

    When we see the noise with the antenna 30m from the houses -- how much it will be reduced, when I put the antenna 10x the distance (D) away from the houses (= noise sources).
    Can one assume Noisepower = 1/D or even 1/(DxD ) ?

    I have never really tested it on shortwave bands... (partly still nearfield)

    Ulli, ON5KQ
  • Ulli

    I have used a KiwiSDR in a backpack with an active monopole vertical and a slightly different processing path to wsprdaemon to measure and plot local noise. I've attached an example graph to show how one source of local noise (on 40m) decays with distance from the source.

    Gwyn G3ZIL

  • Ulli,
    You don't necessarily need to wait until spring. The characteristics of a short dipole are known and so are those of a short monopole. If you can characterize your preamp, its input impedance as well as its voltage gain then you can obtain a conversion from e-field to power at the (50 ohm?) output and thereby calibrate the entire system. Knowing the input impedance of the preamp may be the most difficult part but a good estimate should be possible.

    Another alternative is to measure with a known antenna, say a matched vertical dipole and then compare the active antenna levels with what you get when they are in the same location.

    Small antennas such as loops and short dipoles situated away from earth can be somewhat easier to analyze. I've seen LZ1AQ's analyses as well as Owen Duffy's. Even a monopole with nothing in the near field and well controlled/behaved image plane (ground, radial system,...) can be tractable, though the situation in the far field can make the pattern uncertain thus the excess noise reports as in Rec. ITU-R P.372-8 are subject to interpretation. Still, we try to get closer to a common measurement environment, to get from e-field to kiwi response better, so that we can see and perhaps improve our receive situation to approach a propagated and/or galactic noise limit. We may not get there but we try!
  • Glenn, I have now a thin piece of wire (approx. AWG 18) - 6.3m long attached to an old fishingpole - this is my vertical radiator. There are some 20 radials about just 5-6m long.
    I am feeding this directly with a high impedance pre-amplifier (in fact LZ1AQ's AAA-1C in dipole mode).
    I can feed this vertical passive as a matched 5/8 wavelength vertical and see, what the kiwi-S-meter will tell me.
    The vertical is totally free - nothing out of metal in the neighborhood about 20m in all directions.
    So, if we assume, the vertical has indeed an undistorted omni-directional pattern - how much losses do I account for the lossy ground ?

    NEC2 cannot model radials nor estimated ground losses correctly.

    Is there anyone who can advice what antenna losses I should acount for ?

    I cannot believe the 1dbi net gain from this antenna construction , which 4NEC2 (NEC2 engine) tells me....
    Does anyone know, what the gain of a typical quaterwave vertical with modest radial network on avarage ground really has - I believe no better than -5...-3dbi... probably even worse...

    I also have put a 14Mhz horizontal dipole well matched at 8m height between the trees - just for comparison on 20m... it produces S-meter readout on a free frequency of approx -115dbm in 1kHz bandwidth - so -145dBc... very quit antenna !
    I don't have a resonant 20m vertical dipole, but will build one (K9YC style, top at 16m height) - then I wonder how much higher the noise is and how much lower the signals...;-(
    Anyone some prediction in this comparison ?

    Thanks for the chart - if this values of would be measured with a 0dbi antenna - rather strong noise - that would certainly interfere.
    How much went the noise down 100-200m away ?
    I am planning to put my vertical out in the field some 200-300m far away from any house or metal structure...

    However with the wx raining cats and dogs, the fields are almost under water... so I need to wait ...

    Ulli, ON5KQ
  • As a first approximation, presume that the ground losses are zero. This isn't quite correct but other issues are bigger, mainly mismatch loss at most frequencies for a 6m vertical radiator. Its 'radiation resistance' only rotates to a good match (~37 ohms + ground resistance) at ~12 MHz. Everywhere else the mismatch loss will dwarf uncertainties about ground/radial effects. Thus just assume it's the same as the ITU antenna and get data at 12 MHz.

    What you will likely find is that your noise floor at 12 MHz is too high, much above the approximately -144 dBm/1-Hz (-110 dBm displayed by kiwi S-meter in USB bandwidth). This is *likely* the case because CM currents on your feedline due to imbalance, near-field noise sources and actual noise current in your earth (imperfectly isolated by any radial system) will cause it to be higher. The question is how much higher and what is the nature/cause of that lift. The mechanisms which cause it will likely apply to other frequencies but where the antenna requires special correction in order to obtain a calibrated measurement.

    You can use the NEC2 modeled data to obtain S-Parameters which can then be put into a modeling tool such as QUCS. This will give you an idea of the correction necessary, away from 12 MHz, should you want it. It will also show you that at some frequencies the mismatch/correction will be so bad that the kiwi will not be sensitive enough. That's just the way it is with a resonant antenna and a 50 ohm system.

    Once you determine your noise floor at 12 and have a look at the amount and nature of the increased noise, you can start to work on removing the coupling mechanisms in order to improve your system noise floor.
  • Hi Glenn,
    Taking S-parameters and using QUCS is exactly what I was planning to do...
    However wx outside is that terrible, that I better read a good book inside, rather than open the door and work in de mud of the fields...hi
    So - just wait a while, until wx condx get better... from Sunday they report wx improvement with dry periods...

    Ulli, ON5KQ
  • edited October 2019
    Measuring it when you can is great but even with a crude (optimistic model) you can see the general problem. It's not possible to match to the radiation resistance of a monopole far from resonance, this means that there is effectively (room temperature) loss between the source of the signal (desired noise floor) and the kiwi. Though the situation with real ground is probably worse, here's how 4NEC2/QUCS sees the mismatch loss of a 6.3m monopole over perfect ground:

    You can see that no correction is needed at resonance, there isn't any degradation of system noise floor mismatch, but things aren't good very far from there. Even with a low noise preamp having significant gain to swamp kiwi noise, between the monopole and the kiwi the antenna factor makes it difficult-to-impossible to get down to the ITU noise floor over very much of the range.
    To make matters worse, per other radio astronomy measurements the ITU numbers are likely 10 dB too high in upper HF so even a noiseless (perfect) LNA couldn't accomplish this. The noise associated with the mismatch loss pushes the attainable floor above actual propagated and galactic noise values.

    But this doesn't keep you from learning a lot that may help improve your system. Just measuring where the antenna is 'good' should let you start to identify noise coupling mechanisms, improve baluns, reduce CM current, position the antenna&feedline away from bad near-field sources and possibly improve grounding and radial systems. Those changes may help performance everywhere in the spectrum.
  • Glenn,
    it is clear, that loss with a given very short radiator will increase dramatically away from the matched situation of the resonant radiator.
    However in practice towards lower frequencies the noise levels increase also dramatically - what we need to achieve is, that the natuaral noise level (galactic noise) can be heard - or at least the system is that sensitive, that we can hear the actual noise at the antenna position and not any conducted noise sources coupling in feedlines, etc... (ideal isolation against "parasitic antennas")
    So, obviously we should do everything we can do and think of, to not pickup secondary 'parasitic antenna' voltages, such as coax cables, groundloops, etc.
    Then we should hear a base band-noise level, which is everywhere from 10kHz - 30Mhz just above the receiver noise level and which is constant with frequency. So, while field strength of actual noise increase towards low frequency in the same way the eefficiency of our antenna may decrease and still don't achieve the same S/N ratio.
    Keeping over the whole frequency the noise constant just above the rx noisefloor preserves max dynamic range of the rx-hardware...

    The problem is, that we still don't know the field strength of what the antenna receives...

    At the moment I do everything I can think of to:
    - decrease/surpress as much as possible any pick-up of noise sources, which are not coming from the antenna itself
    - after that adjust the gain of pre-amp, so that noiselevel is just above the receiver noisefloor - so any tiny voltage change will be immediately be recognized by the receiver and not hidden in rx-noise, nor pre-amp noise..

    I think this means, that there should be a constant antenna factor = Fieldstrength/ Voltage level
    So while field strength rise towards low frequencies, gain of antenna falls - as a result the voltage at 50Ohm input keeps constant

    LZ1AQ says that from SPICE calculation his pre-amp with a 2x1m dipole has antenna factor of 2db and is almost constant with such antenna.
    I will build such vertical dipole and compare the signal output with my current antenna...

    Need to think about how the dbm values of wsprdaemon 2.5a measurements (noise chart) can be linked correctly to real field strength for my qth...

    Ulli, ON5KQ
  • To support my experiments, I have now a 2nd kiwi on my table
    The Noise-plots from WD are as follows:
    - Kiwi_0 has a 3m monopole (GP with 20 short radials), LZ1AQ pre-amp in dipole mode. I have adjusted the Noise-plot taking some net-gain (including all feedline loss) into account: 20m 2db, 17m 4db, 15m 5db, 12m 8db , 10m 9db gain
    If this gain levels are correct, I need to measure, for the moment these are assumptions based on earlier experiments.
    - Kiwi_1 is connected with a well matched horizontal 20m dipole - 8m height, very clear sourrounding over very good ground (10km copper wire radial system), low loss feed system - no adjustment for levels in WD (DEFAULT: 0)

    No adjustment of DEFAULT parameter below 14Mhz for KIWI_0 with the short monopole = DEFAULT:0

    Ulli, ON5KQ
  • Ulli,
    A great many things to be said about what you wrote, perhaps too much and annoying to this forum. Let me just say a couple of things.

    For constant ERP from a distant source, available power follows a perfectly-matched antenna's aperture. At low frequencies an antenna has a great deal more aperture than at higher frequencies because lambda^2 is falling at 20 dB/decade. Constant ERP from a distant source produces decreasing power in the matched load, as frequency is increased. This is another way of viewing so called "path loss". Our goal is always to have the signal/noise from the DX source, arriving at some angle and from direction of interest to dominate at the Kiwi detectors.

    That available power appears as though it were sourced by a 'radiation resistance' which changes as antenna electrical dimension squared. The available open circuit voltage (as from an 'e-field probe') is twice that of the power matched voltage but both get small as an antenna gets short. Even though a given antenna is getting 'longer' as frequency increases, the ITU floor is falling faster. The result is that the upper end of the Kiwi range can be the hardest to solve for a broadband structure. Up at 30 MHz it takes almost a full half-wave dipole to get enough signal to dominate the system but an antenna of that sort is too high-Q to operate over more than a few percent of center frequency.

    Because the target DX noise floor has greater negative slope with frequency than the mismatch loss of a short antenna, the ITU limit can be met at the low end. This is why at LF, a short whip can be DX-lightning-noise limited even though the antenna match is so poor.
    This is why I think that only a hybrid approach can work if reasonably sized antennas are to be used . A short antenna can achieve the (high) ITU floor levels from VLF - mid-HF if coupled to a very high Z 'matching' preamp/feedline which transforms the (open circuit) antenna voltage present down to 50 ohms. At least it can if those circuits can operate within the typical very large local field strength which can easily generate distortion terms which then act like an increase in noise floor. But at upper-HF a short antenna of very much less than a half wave (or quarter wave monopole reflected by an image ground/plane) can't deliver enough signal from its low source resistance into a *broadband* match such that the resulting system noise floor is DX-noise dominated.
    If one can combine a short high-Z antenna follower for VLF-mid_HF coverage with a passive broadband structure such as a biconical then it may be possible to cover the entire Kiwi range. I presented mismatch loss for a 12-30 MHz antenna of that sort in another thread on this forum.

    Every practical solution does have the exact problem you describe. In solving the get-access-to-whatever-signal-is-on-the-antenna-being-used problem one has to simultaneously avoid ingress and IMD sources down to that same (very small) level. The required dynamic range is pretty staggering as is the symmetry/balance necessary to avoid these interferers.

    LZ1AQ has addressed some of these same fundamentals in his loop antenna discussions. A small antenna, loop or dipole, can't get broadband ITU coverage with currently known architectures and devices and available (room temperature) conductors and dielectric materials. He mentions that his loop antennas don't get down to propagated (and thus galactic) noise floors at 10m. They may work fine at .1 - 10 MHz though.

    Any antenna has nearly the same SNR as any other modulo its directivity . Transferring that SNR to the 50 ohm input of a Kiwi while not incurring noise/IMD ingress and not overloading the ADC on strong local fields is the trick!

    I'll stop ranting now, I feel like I'm filling the forum with arcane details not of interest to most.

    Glenn n6gn
  • Thanks Glenn... I'm sorry, for too much noise in this forum thread...
    Ulli, ON5KQ
  • While this antenna discussion is not directly related to WD, the topic is extremely important for users of WD. Many Kiwi users are suffering from RFI levels 20-30 dB greater than they could achieve with attention to antennas and especially common mode noise introduced by feedline/power supplies/GPS and other wired connections to their systems. In my opinion this subject deserves much more discussion, but perhaps it should move to its own discussion thread.
  • I'm with Rob on this, fascinating, informative worthy of its own thread.
    I've come to the induced noise subject from a non academic base with just observations and experiments and have almost enjoyed the forced education.
    WD has allowed insight I'd not have gained through other means as I simply didn't see the point so would not have assembled the kit.

    I'm going to stop WSPR for a while as I've learnt digital "in-the-noise" reception is not the same as a one that is enjoyable to the ear.
    I have ended up having three Kiwi's on that I hesitate to do much with as each change, or antenna test seemed to negatively affect WSPR spots, I'd like to just play radios and antennas for a bit.

    Another (pointless) observation I was slow to make, is the capture affect of WSPR users, if I set up a station in a similar geographical location to another user two things happen.
    1. I make spots
    2. I reduce other stations uniques
    I could not work out how my station being on seemed to coincide with a perfomance drop of some other UK users (on the unique count), as I said I'm a bit slow.
  • I have 3 kiwi, all currently on the same antenna. 2 are used 7 ch. each for wspr and one is a 3X20KHz public unit, for 5 public channels. I also use #3 on alternate antenna schemes for experimentation at times.
  • Your spots will have no effect on other stations.
    Uniques are the number of different call signs you have detected and has no direct relationship with reports from other WSPR listeners
  • Please continue these discussions, if necessary on a new thread. They are most informative.
  • Agreed , please continue :-)


    Martin - G8JNJ
  • jksjks
    edited October 2019
    This is the new thread. It was split from the previous "wsprdaemon version 2.3 (latest 2.5a) - A Raspberry Pi WSPR decoding service" thread which had 43 posts. Please continue the wsprdaemon noise graph discussions here (and there for non-noise graph issues).
  • Hi Rob,

    I'm now trying to calibrate my noise measurements in order to compensate for antenna gain and pre-amp / filter gain loss.

    I've been trying to understand the setup in wsprdaemon and the example you gave earlier in this thread - but I'm getting confused.

    "For example, if there were a 10 dB gain LNA, then DEFAULT:-10 would subtract 10 dB from measurements on all bands. Adding BAND:level,... further refines the adjustments when you have such information. At KPH I used a $50 NanoVNA to determine the gain/loss for all bands in the system:

    "KPH_HF_77 KPH CM88mc foobar DEFAULT:-4.65,2200:,630:-3.11,160:-3.11,80:-3.11,60:-4.02,40:-4.42,30:-4.66,20:-4.65,17:-4.53,15:-4.43,12:-4.34,10:-4.31"

    For example I have got a TC2M broadband vertical followed by a 10dB pre-amp.

    The modeled antenna gain for my configuration is:-

    2200 -63dBi
    630 -42dBi
    160 -22dBi
    80 -13dBi
    60 -8.1dBi
    40 -4.7dBi
    30 -2.9dBi
    20 -3.3dBi
    17 -2.2dBi
    15 +1.5dBi
    12 +1.3dBi
    10 +0.4dBi

    So should my config be:-


    Or am I misunderstanding the process ?


    Martin - G8JNJ
  • Also:
    I do not understand the syntax with the DEFAULT parameter..
    What is wrong with this syntax:
    "KIWI_0 ON5KQ JO10os *some password* DEFAULT:0,20:-2, 17:-5, 15:-6, 12:-8, 10:-9"
    "KIWI_1 ON5KQ JO10os *some password* DEFAULT:0,160:9, 60:-4.8, 60eu:-4.8, 40:-7.8, 30:-11.3, 20:-14.6, 17:-18.1, 15:-18, 12:-20.8, 10:-21.4"


    I don't understand it...
    Some ideas:
    - 60eu not defined ?
    - positive numbers need to be +
    - decimals must be komma instead of points ???

    Help required, please.

    Thank you

    Ulli, ON5KQ
  • Hi Ulli,

    "- 60eu not defined ?" I don't think that matters as 60 is close enough to correct for 60eu - but it my matter if 60eu is defined somewhere in a lookup table.

    "- positive numbers need to be +" Again I don't think it should matter, if it's not -ve it should be taken as a positive value.

    "- decimals must be komma instead of points ???" I don't think so as the comma is being used as a delimiter so using it for the decimal would be confusing.

    You have got extra spaces between some values which may be a problem ?

    I changed my values today and it has re-scaled the graphs. Unfortunately the LF band levels are now higher than -120dBm/Hz so they are off the scale. I'd like to know if that can be altered to accommodation the bands with higher noise floors.


    Martin - G8JNJ
  • Martin,
    If I change the line, so I do NOT have any extra spaces, it doesn't change anything...
    whatever I input behind the DEFAULT in the line, it don't change the noise graphs.... I don't understand it?
    That's why I copy and paste the lines here... perhaps I'm blind....hi

    Ulli, ON5KQ
  • By the way - I deleted everything with DEFAULT and after in the lines... = no change
    So the function is disabled anyway - why?

    Ulli, ON5KQ
  • I think the whole file-system crashed on the VM-ware machine... I will make a complete new install, as I get errors now with files missing and files are written into the wrong folders etc... finish for today
    Ulli - ON5KQ
  • Hi Ulli,

    Where are you looking to see any changes you make ?

    noise_graphs.png in /var/www/html/


    Martin - G8JNJ
  • Martin - the system here is totally reset - VMware machine destroyed - I am in the process to build everything new...
    Old noise files are in

    I have the system partly reinstalled, but no noise graphs are working yet - need to do more installs...
    Ulli - ON5KQ
Sign In or Register to comment.