A CW carrier should result in a higher reading on the 'RMS' trace if it really was continuous, in that there would be no gaps in time for the algorithm to find the real noise level. But the FFT trace should continue to read the noise - because it finds the gaps in frequency space not occupied by the CW signal. When both wsprdaemon FFT and RMS estimates are raised the implication is a broadband (well, at least over 320Hz) continuous source.
What I've found helpful in chasing down these sorts of events seen on the wsprdaemon graphs is to automatically take screenshots of a Kiwi waterfall channel with the WSPR band at the centre, e.g. using scrot on a Pi, and then pull together into a movie file, e.g. using ImageJ. Easy then to see a day's comings and goings in a few seconds. I've put an example from KPH on 630m at
Cheers Gwyn Great idea, I have a few ways to screen grab and would be able to combine a few useful windows..
--Later to avoid hijacking the thread any more--
Spurs went away on checking one loop control/power injector. Swapped to wireless bridge anyway at 7PM. I like sheilded CAT cable but I need to solder more wires to them and restrain those with some better method. Shielded CAT RJ45 connectors, in my use, seem to need a little movement every few months to keep a good shield.
I recently did some work at the Northern Utah WebSDR to fix a signal path issue on the KiwiSDR that had been excessively attenuating signal below 1 MHz and have consequently restored its ability to hear the band's noise floor on 630 meters (2200 still has some issues - will work on that later.)
In doing a bit a research I noted that according to the ITU and similar noise plots for a "quiet" site, I should expect approximately a 62dB noise figure at 630 meters, which correlates to an approximate -112dBm in a 1 Hz bandwith, getting much "worse" at even lower frequencies. This -112dBm figure would be for a very quiet location and daytime conditions, so the actual level will likely be higher than this.
The graphs at wsprdaemon.org, however, do not seem to go above -110 dBm/1Hz which, if one were able to do a proper calibration of actual noise levels, be wholly out of range of this graph on 630, and even farther off on 2200 - which means that the default graphs would be incapable of showing data from a properly calibrated receiver on these bands.
Am I missing something here, or is that essentially correct?
On a related question: How does one get rid of "old", unused - and blank - graphs from virtual receivers that are no longer in use?
Clint, I'm not sure about the actual vs. ITU at the low end. I know that ITU seems 10 dB high at 20 MHz, per University of New Mexico LWA measurements and others (even KPH) but I don't know about the LF end except that it seems pretty easy to get parity among disparate antennas/systems down there so it would seem that everyone is DX noise limited and perhaps 'quiet sites' aren't as big a deal (as long as CM near-field and other ingress is under control) To get rid of empty graphs, just blow away the directory for that band under the receiver of interest. The empty graph will no longer be plotted.
Clint, You've raised an important issue that we should look at. Those fixed scales have been there from the very early work Rob and I did on graphing the data, from before Rob added the option for band-by-band calibration. The simplest response would be to increase the upper limit of the graphs for 630m and 2200m. However, your current results, and the fact that the 630m and 2200m for most reporters on wsprnet.org are on-scale, does suggest we should encourage those reporting noise results to use realistic estimates of band-by-band corrections. In that case, the lower limit of the 630m and 2200m graphs could be increased too. If the options in wsprdaemon.conf are set, then data are also reported to an Influx database, and can be seen at this Grafana site - the link is for your recent 630m data - which very much reinforces your point.
Thank you all for the answers - I hadn't really dug around in the directories to find "unused" graph data (even the "old" band data files seem to have had been accessed by the system recently) - but this just goes to prove the old adage: "It's easy, if you know how!"
As for the absolute levels, taking 20 meters as an example, I find that the actual receive system noise floor (calibrated using a known-accurate directional coupler inline with the antenna along with a known-accurate signal generator) - which is sensitivity-limited by ionospheric noise - comes out within a couple of dB of the ITU "Quiet rural" graph at 14 MHz - this using the known ratings of the TCI-530 antenna and a coax run that has less than 1dB of loss.
I'm also using the TCI-530 for both 160 and 630 meters - both being frequencies below its design range of 3-30 MHz - and I know that the apparent gain at 630 is quite low compared to, say, 80 meters as I do not have an available NEC model for the antenna to attempt any estimation. Trying to find an absolute reading that is meaningful is a bit problematic, being subject to guesses and assumptions that may not be entirely valid. For 2200 meters, a shielded loop is being used as the output from the TCI-530 drops sharply below 400 kHz, but there's something in the system (probably an issue with the loop's amplifier coupling) that is causing it to be badly "gain starved" - a problem that will have to wait at least a little while to be addressed.
I just saw the Grafana chart for the first time, Gwyn mentioned above. I am sorry for my all but really calibrated noise charts! Being back after some 'pause', due to health problems, I need to start things up again - therefore I was experimenting a lot, trying to get rid of some local noise sources and switching antennas a lot also... so the noise graphs are rather meaningless from the past couple of weeks..
To get things better: How should I "calibrate" the noise absolute reading in the graphs with a simple system of let's say a beverage on 160m ? From 4Nec2 (NEC2 engine) such a beverage would have a gain of -14dbi - additional 3db loss from the feedsystem let's say and without any pre-amp directly to the Kiwi
What adjustments with the DEFAULT parameter I should do to get noise graphs with that Beverage antenna which can be compared with Clints noise graphs from Northern Utah antenna site ?
For a super quiet site such as Northern Utah site (quiet rural) it means: When I consider -14dbi the correct sensitivity for such a Beverage ant (170m at 1.5m over average ground) plus the 3db feed loss, then I can expect about -117dbm in 1kHz bandwidth at a very quiet qth at daylight hours. (quiet rural acc. to ITU-R P.372-12)
I checked my Beverage the other day. WD-noise plot with this Beverage connected shows about 8-10db higher figure that quit rural .... but how noisy is it really according to ITU ? The big question is, if the -14dBi for the 170m Beverage wire is correct ! So, would I need to correct anything with the DEFAULT setting? For the moment there is NO DEFAULT setting configured at all.
It seems to me that you would want to add 'DEFAULT:17' to your config lines so that WD will add 17 dB to its noise measurements and thus compensate for the -17 dB Antenna Factor of your Beverage
I have not yet understood how we measure the gain or loss between the incoming e-field and the output of the feed line from the antenna, so I treat that Antenna Factor as 0 dB at KPH and AI6VN/KH6. However at KPH I have used a $50 NanoVNA (http://nanovna.com/) to measure the gain/loss from the antenna connector to the SMA input of the Kiwis and subtract the measured gain:
Kiwi72 is fed by a Marconi T through a 2-way splitter which introduces about 4 dB loss, so the config line DEFAULT:4 adds 4 dB to compensate for that loss. Kiwis 73-77 are fed by a TCI-530 antenna and the RF passes through a AM blocking filter => 23 dB gain LNA => 30 MHz LPF => 8-way splitter => Kiwis. The net gain is about 10 dB All those RF devices introduce frequency dependent gain of about +10 dB and to correct for that gain I configure those Kiwis for -10 dB. Kiwi78 adds a 10 dB gain LNA and 10 MHz HPF between the output of the 8-way and the RF input of the Kiwi78. That results in a net gain of about 20 dB on Kiwi78
Since the NanoVNA gave me gains for each of the WSPR bands, I specify exact values for each of them, although the uncertainty of the Antenna Factor almost certainly swamps the precision of those values.
Hi Rob, thanks for detailed information - very clear. Are KPH_HF_77 or KPH_HF_78 really the same receiver streams as on kphsdr.com:8077 and kphsdr.com:8088 ?
I connected today at about 17h30 utc, your local morning at 9h30 local time - so well local daylight. At that time, 30m and 20m is open - I can hear lots of westcoast stations in CW and WSPR spots also. TCI 520 should be a good antenna also (TCI 520 is specified with +5dbi gain at 17deg elevation angle on 20MHz)... At least much better than my small dipole...hi
But now I am really confused, that even in the 19m broadcast band none of the kphsdr.com kiwis shows strong broadcasters and the band is almost closed... At the same time, I am spotting lots of west-coast WSPR mini-stations with my simple 2x3m vertical dipole.... so band is open!
Considering your pre-amp gain of net +10db gain on top of the +5dbi antenna gain TCI is specifies for their model TCI 520, it seems rather strange that with an open 19m band (at least towards Europe) the strongest broadcaster is only between S7 and S9 on 19m band. (I just checked - no signals at all above 15Mhz!!)
If I can hear lots of loud W6/W7s at the same time on 20m, something must be wrong!
May be the kphsdr.com:8077 is NOT the same signal than KPH_HF_77 on WD-kiwi ? Or somewhere is a "lost" 30db attenuator, forgotten in the distribution of kphsdr.com:8077 or the antenna is broken ???
If you compare at home the absolute signal strength of the strongest broadcaster in 19m band with the kph kiwi signal strength - is there any major difference ? Considering the ant gain and preamp of kph system ...
I try to understand why even the strongest broadcasters are not much stronger than S9... max S9+20db... kph is not the Southpole nor deeply in aurora zone...
For comparison: At our local morning local the many far east local stations in local language for domestic service are already S9+ here in Belgium in 15Mhz and 17Mhz band and far east overseas service in morning time S9+60db on the Asia Beverage (which is most likely lower gain than +5dbi on 15Mhz at similar elevation angle of 17deg) and Beverage has NO pre-amp at all... (which results in at least 15db less net gain than KPH WD-Kiwi input) Unfortunately that is the reason, I cannot use the Asia-Beverage on the kiwi in the morning without additional -10db att. at the kiwi input. However with the attenuator in line, the bandnoise of the beverage wire on the high bands is far below the kiwi noisefloor - so many weak wspr signals get lost.
P.S.: "I can hear lots of westcoast stations in CW and WSPR spots also...." = I mean, I can hear these stations on my radio in Belgium with my little vertical dipole !!
Yes, all the KPH Kiwis 73 through 78 are fed by the same TCI-530 antenna, while Kiwi78 gets an extra 10 dB RF gain on its input.
You can see the WD listeners listed on Kiwis 72,77, 78 in the Kiwi's 'USERS' window, so you can tell that you are listening on the same Kiwis which are posting as "KPH". Today I am actively debugging WD v2.7a, so the WD listeners may go away briefly through the next few days.
Every Saturday afternoon (i.e. as I write this post) KPH is open for visitors and the fluorescent ceiling lights installed by our landlord the US National Park Service introduce a lot of broadband RFI above 10 MHz. Unfortunately we can't change those lights, so KPH will probably suffer from 6 hours per week of that RFI.
You might find it useful to compare Kiwi77 and Kiwi78 against Kiwi72 which is fed by a Marconi T with no LNA and only a 2-way passive splitter and AM band blocking filter in the path to Kiwi72. It has surprised me to see listeners using Kiwi72 at frequencies well above 3 Mhz where one would expect the TCI-530 to provide better signal levels. Those listeners and your observations suggest that some or many skywave signals (i.e. from far away) are not well captured by the TCI-530, an antenna which was chosen to optimize communications from 0-500 KM.
There is a currently broken TCI-540 (its signals give -10 dB SNR versus the 530) antenna at KPH and a (I think funded $50,000) project to repair it this coming summer when the marsh land under the TCI-540 dries out and 30+M high cranes can access the antenna. If the 530 is truly doing that poorly at long distances, then we can expect a real jump in KPH spot counts later this year if/when the 540 becomes available.
Hi Rob, I checked again the signal levels of KPH TCI530 antenna at about 8h...8h30 utc in comparison with other TCI 530 from a different location, but close enough so you may compare. I think just south of San Fransisco, west of San Jose, there is another ex-coastal station using TCI-530 with a kiwi-sdr
My believe now is, that TCI530 is a poor performer for low angle radiation on the higher bands (20m and up) My feeling is also, that +5dbi gain at 17deg elevation is probably too optimistic. (TCI info from data sheet of TCI 530) A horizontal dipole at half wavelength height has already +8dbi due to ground reflection.... A larger Yagi quickly has more than 15dbi at much lower elevation, especially when mounted at 1 wavelength height or higher.
My active vertical dipole is placed on a very large radial system once build for my 40m transmit array - so vertical radiation angle may be much lower than kph TCI 530 antenna (?!)
So my conclusion now: there is nothing fundamental wrong with kph set-up... just the TCI-530 is never build for DX-work on the high shortwave bands so antenna doesn't work especially well for that purpose... with really no good propagation on the high bands.
And congratulation: The former overloading of the Marconi vertical is almost completly gone - excellent reception now on 472kHz and NDB bands...! And the noise plots show the exceptional low-noise of kph antenna site.... really amazing. KPH is THE reference for low noise qth's to compare with, for sure! That is why it is interesting to achieve really absolute level calibrated noise graphs to the best extend... My noise graphs are far from calibrated, but I am working on it...hi
The TCI 530 is sold as an antenna that is, quote:-
"designed specifically to support skywave communications at short (0500 km) ranges. The 530 directs energy overhead to provide optimal support for shortrange circuits."
However they go on to say.
"In addition, at the higher frequencies used for longer ranges, the 530 radiates significant energy at lower angles for medium- and long range circuits."
(I think the marketing people got that bit added as an afterthought, as they sound much less 'enthusiastic' about that part of the specification :-)
The polar diagrams show something more like the truth, especially above about 20MHz where there is an increasing amount of upward radiation and not much below about 15-20 degrees.
I also note that they don't really mention a distance much greater than 1000 miles in their HF gain comparison charts at the end, I wonder why :-)
This problem with the pattern is quite common with broadband antennas of this type, as the frequency increases the pattern breaks up into more lobes, which tend to tilt upward.
KPH is a maritime communications station and the TCI-530 is the appropriate antenna for that application of contacting nearby ships at sea. There are a couple of unused V beams at KPH which I want to test, but they are aimed at Asia where there are probably few WSPR stations.
The currently broken TCI-540 was installed at the adjacent site KMI for use in voice (phone call) links to Asia where its low angle gain was appropriate, and once repaired it should be a stellar performer.
Since the KPH QTH was selected by Dr Beverage and many of his eponymous antennae were originally installed there, I have hoped that we might install one or more of them. But the strictures of the "Historic Radio Station" and other constraints seem to make doing that unlikely.
Rob - to justify: It is not THAT bad...hi During your morning time I listened again on topband ( during 160m contest) and compared your antennas: Marconi vertical, TCI 530 and also the 8078 port to boost 15Mhz and up a bit... All antennas have about the same sensitivity on 160m - will say, that stations have similar peak signal strength... astonishing! Is the port 8078 really highpass filtered ? No difference to 8077, really...
I listened to VK6LD on topband at about 6h30 your local time - great signal - a bit stronger on the vertical, but not much... perhaps about 3db .. Also on 40m, I heard EW7LO (CQing in CW) on the TCI 530 at kph-sdr and spotted him on the DX-cluster, mentioning he is heard at kph SDR site... great signal also...
So it is not that bad. Signal/Noise ratio at least is great. With the TCI 540 it will be even better as it better matches our hamradio preference to have low angle radiators on the high bands...
At KPH I fabricated a simple RC HPF in front of Kiwi78: 100pF in series with the feed results in approximately a 15 MHz HPF with only 6 dB per octave below that. But that attenuation is enough to almost eliminate the overload conditions which Kiwi78 suffered without it. That extra 10 dB gain on 15-30 MHz helps improve the SNRs 1-3 dB on 20M-10M and increase the spot count by perhaps 10 signals per day. I had the 10 dB LNA so I added it in, but I wouldn't buy an extra LNA just for those bands.
I need that 23 dB LNA gain at KPH in part to compensate for the losses in filters and the 8 way splitter. But even with the 530 connected directly to a Kiwi, that site is so quiet that the 20 dB noise figure of the Kiwi masked the atmospheric noise delivered by the 530 on 10 Mhz and above.
So whatever your antenna system, I would verify that your background noise goes up by 3 dB or more when you replace a 50 ohm terminator at the antenna end of your feed line. If not, then you may benefit from an LNA ahead of the Kiwi. Of course there are challenges keeping the LNA from introducing common mode to differential RFI. But that discussion is not Kiwi specific.
ok Rob, understood. Back to noise graphs: Is there a possibility (parameter in .conf file), so that noise graphs are only uploaded for specific bands, not all bands the kiwi configuration is set-up for, to listen.
So for example: Kiwi listens to wspr signal on 80m and 40m, but noise graph is only uploaded for 80m
Background: I want to examine the noise profile of a particular band in more detail. When I ask WD to make the noise graphs, the actual picture is not very detailed any more, because so many different kiwi channels need to be written on one page. That results in too much compression/loss in the y-axis of the graph... So writing only the noise plot of the few bands of interest will increase the visual resolution...
May be this question has been asked before, but at least, I cannot find it
Ulli ... I see you are posting your noise data to the wsprdaemon Influx database. So you can easily explore using the Grafana web tool, here are your recent 40m data for example:
many thanks for setting this up, Gwyn - great tool! However I have done nothing to get my data in this database... I would like to thank the person who forwards my data to this database (great service...hi) ... the only what I did, is post the noise data by configure "YES" in conf. file of WD 2.6f
One question: Does this Grafana tool real-time ? Because just about 11h30..12h today it stops showing my data from "graphs.wsprdaemon.org/ON5KQ/" (browser URL) Or does Grafana only works experimentally/occasionally ?
For better understanding, I renamed my Kiwi's in .Conf-file in WD and put antenna-name instead of KIWI_1. For KIWI_0 currently not the usual antenna, as the longwire is broken...
Ulli ... I should start by saying that the hard work with Influx and Grafana was done by Rob, AI6VN, and especially his friend Tommy - I just wrote the user guide.
Yesterday afternoon, I placed my Northwest antenna new - it was broken earlier. As I already had a 170m Beverage to the east (South Asia, Western Australia) in place, I wanted to have something similar to Northwest also. The Northwest antenna is a little higher (about 3m on average), the east antenna is only 1.5m high. Both antennas are free from any disturbing metal sourrounding on a clear field. You can see some foto's of the sourrounding on my qrz.com page.
The setup now is not so much to spot as many stations as possible, but rather to get a better idea of noise (especially man-made noise from the neighborhood). This morning I calculated the system gain/loss figures. I am not using any pre-amplifiers, but a carefully designed feed system with imp-transformers (compensated for reactances towards higher frequencies). So there is almost no mismatch which may introduce additional losses not included in the conf. file from 1Mhz to 30Mhz (SWR below 1:1.5)
Now the big question: NEC2 shows unbelievable gain for the beverages on the higher frequencies! I wonder, if this is really true, as ground losses, even at 3m high for the NW antenna might build up and destroy the (theoretical high gain for a 10 wavelength wire)... A gain figure of +7dbi on 10m for a beverage sounds very optimistic... but the actual conf file stick to the figures NEC2 offers me, as I do not have any other data so far...
The next step will be, to put my 2x1m vertical dipole in the middle of the field and compare real groundwave signal strength against the beverages on 1km distance or so...
I calculated -14dbi on 160m and -8dbi on 80m, which is less output than what John, ON4UN mentions in his lowband DXing book, but I think it is realistic... What I question is the strong increase of gain towards higher frequencies... up to +7dbi on 10m
Does anyone has some more data, than Nec2 figures for Beverage antennas ? Perhaps NEC4 calculations ?
As a result (taking the Nec2 gain figures including all distribution losses) I still end up with total system gain on high bands with the Beverages. Therefore my current Noisegraphs (http://graphs.wsprdaemon.org/ON5KQ/) show unbelievable low noise on the high bands... In fact, even with the calculated gain on the high bands from the very long wires, I better use pre-amps - at least for 20m and up (don't have any good design made yet)
What do you think the real gain of a 170m Beverage on 15m or 10m will be over average ground and properly terminated ? Do you believe it has any (positive) gain ? Or even loss ?
This is the best site I have found with information about Beverages: http://w8ji.com/beverages.htm I have been surprised about the > 10 Mhz performance of my 100M Beverage in Maui, but I put an AM blocking filter and a +23 dB gain LNA in front of the Kiwis.
Rob, I will setup the 2x1m measurement active dipole soon - the efficiency should be close to the value LZ1AQ calculated (about 2db/m from 350kHz to 30Mhz within 3db). Then I will compare my much longer 2x3m dipole with the smaller one on exact the same terrain/location. Using the Network analyser, I can measure the differences precise enough. As a result, my guess is, that the usual NEC2 simulations of Beverage antennas are not correct for frequencies >7Mhz... we will see..
My actual noise graphs from later than 12h utc of today show the real noise levels under the assumption the NEC2 model is correct... (using the DEFAULT paramenter) I will correct the noise graph as soon as I have newer/better data
Unfortunately the figures of 630/2200 will not be plotted here (http://graphs.wsprdaemon.org/ON5KQ/) as they values are arround -110...-100dbm/Hz... (out of graph level) They can be seen on the very nice Grafana site ..
Made some changes to the calibration adjustment parameter DEFAULT in conf.file.... There was a mistake in the formula of the spreadsheet. So the cable losses of the longwires were not included. I need up to 200m cables - so although on low frequencies the difference is not that big, on 10m the difference is very significant. Changed that now...
Ethernet speed change from 100Mb/s to 10Mb/s made a big change! In my installation the change to 10Mb/s reduces the noise in the spectrum above 10Mhz to almost not visible levels in the waterfall. Wow! The result is, that many weak signals are hidden in the rx-noise even in 30m band and defintely above. I tested shortly with a very sensitive other receiver and were surprised to see so many signals popping out of the noise, with additional 10-15db additional gain.... however it will be a challenge to achieve an MDS of -140dbm without bandpass filters... let's see...
A question: Will it be reliable for wsprdaemon software to run, if the kiwis are set to lower 10Mb/s Ethernet speed ? Otherwise I need to switch to Wifi connection and I saw the long discussion in the other thread running a simple stick... I am afraid of my non-existing Linux/code knowledge... To switch over to Wifi mode would need a ready-to-go-out-of-the-box solution, but I wonder, if this exist...
I am very happy with the results of the long wires now, with the low ethernet speed... much better with much lower noise. Now I must build pre-amps... Until now, I didn't use pre-amps with the wires... (except my active vertical dipole, which is not in use at the moment)
One more question to the noise graph: is it possible to change the vertical scale of the noise graph for 2200 and 630m ? The real noise it at 90...100db/Hz region - so it is invisible now (only on the Grafana site). It would be great to change the vertical scale for the VLF area so realistic figures can be shown...
You commented that "What I question is the strong increase of gain towards higher frequencies... up to +7dbi on 10m "
Yes that seems sensible to me.
As the wire becomes progressively 'longer' in terms of wavelength, the radiation pattern side-lobes tend to become narrower, tilt and follow along the direction of the wire, producing a very high gain but narrow beam.
I use a DX Engineering RPA-1 pre-amp clone, followed by an amplitude / frequency slope corrector (see previous threads on the subject) and broadcast band notches in order to improve the KiWi NF on the HF bands, whilst decreasing the overall system gain on the LF bands in order to prevent ADC overload.
Yesterday I started beta testing V2.7 on many of the sites which contribute to the Influx database. That new SW by default uses an new and much more CPU efficient way of calculating the FFT noise, but apparently it fails to upload that C2_fft value to the Influx DB. That problem only appeared when we changed Grafana to display C2 yesterday afternoon. I will upload the fixed 2.7 later today after which those C2 values should be present. It is possible to edit the Grafana page to customize the printouts, but I don't have time to find a link to those instructions this morning.
Ulli, Another advantage of the preamp, beyond simply improving the kiwi's noise floor, is that it pushes common-mode ingress occurring after the preamp further down by the amount of the gain. A disadvantage is that it may push the kiwi closer to OV. Optimizing a kiwi is a balancing act. It's likely that the LAN noise you see at higher speeds is due to CM current sourced by the LAN cable, continuing through the finite balance of the LAN input transformers, going through the kiwi (ground plane) itself and sinking out through the SMA/feedline. It doesn't take a lot of CM current or terribly low CM impedance to achieve enough noise at the kiwi preamp/adc input to exceed KTB. At 20-30 MHz propagated&galactic noise is very likely only ~10 dB or so above KTB, even lower if antenna efficiency is low (antenna unmatched, too small, etc)
Comments
What I've found helpful in chasing down these sorts of events seen on the wsprdaemon graphs is to automatically take screenshots of a Kiwi waterfall channel with the WSPR band at the centre, e.g. using scrot on a Pi, and then pull together into a movie file, e.g. using ImageJ. Easy then to see a day's comings and goings in a few seconds. I've put an example from KPH on 630m at
https://www.dropbox.com/s/ib7juls04hbxl70/KPH_630_10Jun2019_copy.avi?dl=0
where you can see the time in the clock top right and just about see the S meter as well.
Gwyn G3ZIL
Great idea, I have a few ways to screen grab and would be able to combine a few useful windows..
--Later to avoid hijacking the thread any more--
Spurs went away on checking one loop control/power injector. Swapped to wireless bridge anyway at 7PM.
I like sheilded CAT cable but I need to solder more wires to them and restrain those with some better method.
Shielded CAT RJ45 connectors, in my use, seem to need a little movement every few months to keep a good shield.
In doing a bit a research I noted that according to the ITU and similar noise plots for a "quiet" site, I should expect approximately a 62dB noise figure at 630 meters, which correlates to an approximate -112dBm in a 1 Hz bandwith, getting much "worse" at even lower frequencies. This -112dBm figure would be for a very quiet location and daytime conditions, so the actual level will likely be higher than this.
The graphs at wsprdaemon.org, however, do not seem to go above -110 dBm/1Hz which, if one were able to do a proper calibration of actual noise levels, be wholly out of range of this graph on 630, and even farther off on 2200 - which means that the default graphs would be incapable of showing data from a properly calibrated receiver on these bands.
Am I missing something here, or is that essentially correct?
On a related question: How does one get rid of "old", unused - and blank - graphs from virtual receivers that are no longer in use?
Clint
KA7OEI
I'm not sure about the actual vs. ITU at the low end. I know that ITU seems 10 dB high at 20 MHz, per University of New Mexico LWA measurements and others (even KPH) but I don't know about the LF end except that it seems pretty easy to get parity among disparate antennas/systems down there so it would seem that everyone is DX noise limited and perhaps 'quiet sites' aren't as big a deal (as long as CM near-field and other ingress is under control)
To get rid of empty graphs, just blow away the directory for that band under the receiver of interest. The empty graph will no longer be plotted.
You've raised an important issue that we should look at. Those fixed scales have been there from the very early work Rob and I did on graphing the data, from before Rob added the option for band-by-band calibration.
The simplest response would be to increase the upper limit of the graphs for 630m and 2200m. However, your current results, and the fact that the 630m and 2200m for most reporters on wsprnet.org are on-scale, does suggest we should encourage those reporting noise results to use realistic estimates of band-by-band corrections. In that case, the lower limit of the 630m and 2200m graphs could be increased too.
If the options in wsprdaemon.conf are set, then data are also reported to an Influx database, and can be seen at this Grafana site - the link is for your recent 630m data - which very much reinforces your point.
Gwyn G3ZIL
As for the absolute levels, taking 20 meters as an example, I find that the actual receive system noise floor (calibrated using a known-accurate directional coupler inline with the antenna along with a known-accurate signal generator) - which is sensitivity-limited by ionospheric noise - comes out within a couple of dB of the ITU "Quiet rural" graph at 14 MHz - this using the known ratings of the TCI-530 antenna and a coax run that has less than 1dB of loss.
I'm also using the TCI-530 for both 160 and 630 meters - both being frequencies below its design range of 3-30 MHz - and I know that the apparent gain at 630 is quite low compared to, say, 80 meters as I do not have an available NEC model for the antenna to attempt any estimation. Trying to find an absolute reading that is meaningful is a bit problematic, being subject to guesses and assumptions that may not be entirely valid. For 2200 meters, a shielded loop is being used as the output from the TCI-530 drops sharply below 400 kHz, but there's something in the system (probably an issue with the loop's amplifier coupling) that is causing it to be badly "gain starved" - a problem that will have to wait at least a little while to be addressed.
73,
Clint
KA7OEI
To get things better:
How should I "calibrate" the noise absolute reading in the graphs with a simple system of let's say a beverage on 160m ?
From 4Nec2 (NEC2 engine) such a beverage would have a gain of -14dbi - additional 3db loss from the feedsystem let's say and without any pre-amp directly to the Kiwi
What adjustments with the DEFAULT parameter I should do to get noise graphs with that Beverage antenna which can be compared with Clints noise graphs from Northern Utah antenna site ?
For a super quiet site such as Northern Utah site (quiet rural) it means:
When I consider -14dbi the correct sensitivity for such a Beverage ant (170m at 1.5m over average ground) plus the 3db feed loss, then I can expect about -117dbm in 1kHz bandwidth at a very quiet qth at daylight hours. (quiet rural acc. to ITU-R P.372-12)
I checked my Beverage the other day. WD-noise plot with this Beverage connected shows about 8-10db higher figure that quit rural .... but how noisy is it really according to ITU ? The big question is, if the -14dBi for the 170m Beverage wire is correct !
So, would I need to correct anything with the DEFAULT setting? For the moment there is NO DEFAULT setting configured at all.
Any suggestion ?
Ulli, ON5KQ
I have not yet understood how we measure the gain or loss between the incoming e-field and the output of the feed line from the antenna, so I treat that Antenna Factor as 0 dB at KPH and AI6VN/KH6.
However at KPH I have used a $50 NanoVNA (http://nanovna.com/) to measure the gain/loss from the antenna connector to the SMA input of the Kiwis and subtract the measured gain:
Kiwi72 is fed by a Marconi T through a 2-way splitter which introduces about 4 dB loss, so the config line DEFAULT:4 adds 4 dB to compensate for that loss.
Kiwis 73-77 are fed by a TCI-530 antenna and the RF passes through a AM blocking filter => 23 dB gain LNA => 30 MHz LPF => 8-way splitter => Kiwis. The net gain is about 10 dB
All those RF devices introduce frequency dependent gain of about +10 dB and to correct for that gain I configure those Kiwis for -10 dB.
Kiwi78 adds a 10 dB gain LNA and 10 MHz HPF between the output of the 8-way and the RF input of the Kiwi78. That results in a net gain of about 20 dB on Kiwi78
Since the NanoVNA gave me gains for each of the WSPR bands, I specify exact values for each of them, although the uncertainty of the Antenna Factor almost certainly swamps the precision of those values.
thanks for detailed information - very clear.
Are KPH_HF_77 or KPH_HF_78 really the same receiver streams as on kphsdr.com:8077 and kphsdr.com:8088 ?
I connected today at about 17h30 utc, your local morning at 9h30 local time - so well local daylight.
At that time, 30m and 20m is open - I can hear lots of westcoast stations in CW and WSPR spots also. TCI 520 should be a good antenna also (TCI 520 is specified with +5dbi gain at 17deg elevation angle on 20MHz)...
At least much better than my small dipole...hi
But now I am really confused, that even in the 19m broadcast band none of the kphsdr.com kiwis shows strong broadcasters and the band is almost closed...
At the same time, I am spotting lots of west-coast WSPR mini-stations with my simple 2x3m vertical dipole.... so band is open!
Considering your pre-amp gain of net +10db gain on top of the +5dbi antenna gain TCI is specifies for their model TCI 520, it seems rather strange that with an open 19m band (at least towards Europe) the strongest broadcaster is only between S7 and S9 on 19m band. (I just checked - no signals at all above 15Mhz!!)
If I can hear lots of loud W6/W7s at the same time on 20m, something must be wrong!
May be the kphsdr.com:8077 is NOT the same signal than KPH_HF_77 on WD-kiwi ?
Or somewhere is a "lost" 30db attenuator, forgotten in the distribution of kphsdr.com:8077 or the antenna is broken ???
If you compare at home the absolute signal strength of the strongest broadcaster in 19m band with the kph kiwi signal strength - is there any major difference ? Considering the ant gain and preamp of kph system ...
I try to understand why even the strongest broadcasters are not much stronger than S9... max S9+20db... kph is not the Southpole nor deeply in aurora zone...
For comparison:
At our local morning local the many far east local stations in local language for domestic service are already S9+ here in Belgium in 15Mhz and 17Mhz band and far east overseas service in morning time S9+60db on the Asia Beverage (which is most likely lower gain than +5dbi on 15Mhz at similar elevation angle of 17deg) and Beverage has NO pre-amp at all... (which results in at least 15db less net gain than KPH WD-Kiwi input)
Unfortunately that is the reason, I cannot use the Asia-Beverage on the kiwi in the morning without additional -10db att. at the kiwi input. However with the attenuator in line, the bandnoise of the beverage wire on the high bands is far below the kiwi noisefloor - so many weak wspr signals get lost.
Ulli
Thanks for your KPH report.
Yes, all the KPH Kiwis 73 through 78 are fed by the same TCI-530 antenna, while Kiwi78 gets an extra 10 dB RF gain on its input.
You can see the WD listeners listed on Kiwis 72,77, 78 in the Kiwi's 'USERS' window, so you can tell that you are listening on the same Kiwis which are posting as "KPH". Today I am actively debugging WD v2.7a, so the WD listeners may go away briefly through the next few days.
Every Saturday afternoon (i.e. as I write this post) KPH is open for visitors and the fluorescent ceiling lights installed by our landlord the US National Park Service introduce a lot of broadband RFI above 10 MHz. Unfortunately we can't change those lights, so KPH will probably suffer from 6 hours per week of that RFI.
You might find it useful to compare Kiwi77 and Kiwi78 against Kiwi72 which is fed by a Marconi T with no LNA and only a 2-way passive splitter and AM band blocking filter in the path to Kiwi72. It has surprised me to see listeners using Kiwi72 at frequencies well above 3 Mhz where one would expect the TCI-530 to provide better signal levels. Those listeners and your observations suggest that some or many skywave signals (i.e. from far away) are not well captured by the TCI-530, an antenna which was chosen to optimize communications from 0-500 KM.
There is a currently broken TCI-540 (its signals give -10 dB SNR versus the 530) antenna at KPH and a (I think funded $50,000) project to repair it this coming summer when the marsh land under the TCI-540 dries out and 30+M high cranes can access the antenna. If the 530 is truly doing that poorly at long distances, then we can expect a real jump in KPH spot counts later this year if/when the 540 becomes available.
Rob
I checked again the signal levels of KPH TCI530 antenna at about 8h...8h30 utc in comparison with other TCI 530 from a different location, but close enough so you may compare.
I think just south of San Fransisco, west of San Jose, there is another ex-coastal station using TCI-530 with a kiwi-sdr
My believe now is, that TCI530 is a poor performer for low angle radiation on the higher bands (20m and up)
My feeling is also, that +5dbi gain at 17deg elevation is probably too optimistic. (TCI info from data sheet of TCI 530)
A horizontal dipole at half wavelength height has already +8dbi due to ground reflection.... A larger Yagi quickly has more than 15dbi at much lower elevation, especially when mounted at 1 wavelength height or higher.
My active vertical dipole is placed on a very large radial system once build for my 40m transmit array - so vertical radiation angle may be much lower than kph TCI 530 antenna (?!)
So my conclusion now: there is nothing fundamental wrong with kph set-up... just the TCI-530 is never build for DX-work on the high shortwave bands so antenna doesn't work especially well for that purpose... with really no good propagation on the high bands.
And congratulation: The former overloading of the Marconi vertical is almost completly gone - excellent reception now on 472kHz and NDB bands...! And the noise plots show the exceptional low-noise of kph antenna site.... really amazing.
KPH is THE reference for low noise qth's to compare with, for sure! That is why it is interesting to achieve really absolute level calibrated noise graphs to the best extend...
My noise graphs are far from calibrated, but I am working on it...hi
Ulli, ON5KQ
The TCI 530 is sold as an antenna that is, quote:-
"designed specifically to support skywave communications at short (0500 km) ranges. The 530 directs energy
overhead to provide optimal support for shortrange circuits."
However they go on to say.
"In addition, at the higher frequencies used for longer ranges, the 530 radiates significant energy at lower angles for medium- and long range circuits."
(I think the marketing people got that bit added as an afterthought, as they sound much less 'enthusiastic' about that part of the specification :-)
https://www.antenna.be/tci-530.pdf
The polar diagrams show something more like the truth, especially above about 20MHz where there is an increasing amount of upward radiation and not much below about 15-20 degrees.
I also note that they don't really mention a distance much greater than 1000 miles in their HF gain comparison charts at the end, I wonder why :-)
Here's a hint https://www.researchgate.net/profile/Ben_Witvliet/publication/263054262/figure/fig6/AS:296376480092160@1447673067032/Relationship-between-elevation-angle-and-distance-for-several-frequencies-Simulations.png
This problem with the pattern is quite common with broadband antennas of this type, as the frequency increases the pattern breaks up into more lobes, which tend to tilt upward.
Regards,
Martin - G8JNJ
Thanks for those great observations.
KPH is a maritime communications station and the TCI-530 is the appropriate antenna for that application of contacting nearby ships at sea. There are a couple of unused V beams at KPH which I want to test, but they are aimed at Asia where there are probably few WSPR stations.
The currently broken TCI-540 was installed at the adjacent site KMI for use in voice (phone call) links to Asia where its low angle gain was appropriate, and once repaired it should be a stellar performer.
Since the KPH QTH was selected by Dr Beverage and many of his eponymous antennae were originally installed there, I have hoped that we might install one or more of them. But the strictures of the "Historic Radio Station" and other constraints seem to make doing that unlikely.
Rob
During your morning time I listened again on topband ( during 160m contest) and compared your antennas:
Marconi vertical, TCI 530 and also the 8078 port to boost 15Mhz and up a bit...
All antennas have about the same sensitivity on 160m - will say, that stations have similar peak signal strength... astonishing! Is the port 8078 really highpass filtered ? No difference to 8077, really...
I listened to VK6LD on topband at about 6h30 your local time - great signal - a bit stronger on the vertical, but not much... perhaps about 3db ..
Also on 40m, I heard EW7LO (CQing in CW) on the TCI 530 at kph-sdr and spotted him on the DX-cluster, mentioning he is heard at kph SDR site... great signal also...
So it is not that bad.
Signal/Noise ratio at least is great.
With the TCI 540 it will be even better as it better matches our hamradio preference to have low angle radiators on the high bands...
Ulli, ON5KQ
I need that 23 dB LNA gain at KPH in part to compensate for the losses in filters and the 8 way splitter. But even with the 530 connected directly to a Kiwi, that site is so quiet that the 20 dB noise figure of the Kiwi masked the atmospheric noise delivered by the 530 on 10 Mhz and above.
So whatever your antenna system, I would verify that your background noise goes up by 3 dB or more when you replace a 50 ohm terminator at the antenna end of your feed line. If not, then you may benefit from an LNA ahead of the Kiwi. Of course there are challenges keeping the LNA from introducing common mode to differential RFI. But that discussion is not Kiwi specific.
Back to noise graphs:
Is there a possibility (parameter in .conf file), so that noise graphs are only uploaded for specific bands, not all bands the kiwi configuration is set-up for, to listen.
So for example:
Kiwi listens to wspr signal on 80m and 40m, but noise graph is only uploaded for 80m
Background: I want to examine the noise profile of a particular band in more detail. When I ask WD to make the noise graphs, the actual picture is not very detailed any more, because so many different kiwi channels need to be written on one page. That results in too much compression/loss in the y-axis of the graph...
So writing only the noise plot of the few bands of interest will increase the visual resolution...
May be this question has been asked before, but at least, I cannot find it
Thanks for info.
Ulli, ON5KQ
So you can easily explore using the Grafana web tool, here are your recent 40m data for example:
https://grafana.int8.com/d/3dcOdAmWk/calibrated-noise-dashboard?orgId=1&from=1579955777194&to=1580153600394&var-site=ON5KQ&var-receiver=KIWI_0&var-receiver=KIWI_1&var-band=40
I have written an introduction to using Grafana and Influx for wsprdaemon noise, pdf file is attached.
best wishes
Gwyn G3ZIL
Attachments:
https://forum.kiwisdr.com/uploads/Uploader/8b/ed4f6a6afda009bfe880e736d99b41.pdf
However I have done nothing to get my data in this database... I would like to thank the person who forwards my data to this database (great service...hi) ... the only what I did, is post the noise data by configure "YES" in conf. file of WD 2.6f
One question:
Does this Grafana tool real-time ?
Because just about 11h30..12h today it stops showing my data from "graphs.wsprdaemon.org/ON5KQ/" (browser URL)
Or does Grafana only works experimentally/occasionally ?
For better understanding, I renamed my Kiwi's in .Conf-file in WD and put antenna-name instead of KIWI_1.
For KIWI_0 currently not the usual antenna, as the longwire is broken...
Ulli, ON5KQ
Yes, the Grafana tool is very near real time, and has been solid since about May 2019. By setting YES to post the noise data you contribute to this database. I can see you renamed some of your receivers, and so this morning's plot has several names, see
https://grafana.int8.com/d/3dcOdAmWk/calibrated-noise-dashboard?orgId=1&from=1580195248033&to=1580218093625&var-site=ON5KQ&var-receiver=All&var-band=40
For values that are very close together the dots may overlap, but you can zoom in as described in my notes.
best wishes
Gwyn G3ZIL
As I already had a 170m Beverage to the east (South Asia, Western Australia) in place, I wanted to have something similar to Northwest also.
The Northwest antenna is a little higher (about 3m on average), the east antenna is only 1.5m high.
Both antennas are free from any disturbing metal sourrounding on a clear field. You can see some foto's of the sourrounding on my qrz.com page.
The setup now is not so much to spot as many stations as possible, but rather to get a better idea of noise (especially man-made noise from the neighborhood). This morning I calculated the system gain/loss figures. I am not using any pre-amplifiers, but a carefully designed feed system with imp-transformers (compensated for reactances towards higher frequencies). So there is almost no mismatch which may introduce additional losses not included in the conf. file from 1Mhz to 30Mhz (SWR below 1:1.5)
Now the big question:
NEC2 shows unbelievable gain for the beverages on the higher frequencies! I wonder, if this is really true, as ground losses, even at 3m high for the NW antenna might build up and destroy the (theoretical high gain for a 10 wavelength wire)...
A gain figure of +7dbi on 10m for a beverage sounds very optimistic... but the actual conf file stick to the figures NEC2 offers me, as I do not have any other data so far...
The next step will be, to put my 2x1m vertical dipole in the middle of the field and compare real groundwave signal strength against the beverages on 1km distance or so...
I calculated -14dbi on 160m and -8dbi on 80m, which is less output than what John, ON4UN mentions in his lowband DXing book, but I think it is realistic... What I question is the strong increase of gain towards higher frequencies... up to +7dbi on 10m
Does anyone has some more data, than Nec2 figures for Beverage antennas ? Perhaps NEC4 calculations ?
As a result (taking the Nec2 gain figures including all distribution losses) I still end up with total system gain on high bands with the Beverages. Therefore my current Noisegraphs (http://graphs.wsprdaemon.org/ON5KQ/) show unbelievable low noise on the high bands...
In fact, even with the calculated gain on the high bands from the very long wires, I better use pre-amps - at least for 20m and up (don't have any good design made yet)
What do you think the real gain of a 170m Beverage on 15m or 10m will be over average ground and properly terminated ?
Do you believe it has any (positive) gain ? Or even loss ?
Regards,
Ulli, ON5KQ
I have been surprised about the > 10 Mhz performance of my 100M Beverage in Maui, but I put an AM blocking filter and a +23 dB gain LNA in front of the Kiwis.
As a result, my guess is, that the usual NEC2 simulations of Beverage antennas are not correct for frequencies >7Mhz... we will see..
My actual noise graphs from later than 12h utc of today show the real noise levels under the assumption the NEC2 model is correct... (using the DEFAULT paramenter)
I will correct the noise graph as soon as I have newer/better data
Unfortunately the figures of 630/2200 will not be plotted here (http://graphs.wsprdaemon.org/ON5KQ/) as they values are arround -110...-100dbm/Hz... (out of graph level)
They can be seen on the very nice Grafana site ..
Ulli, ON5KQ
There was a mistake in the formula of the spreadsheet. So the cable losses of the longwires were not included. I need up to 200m cables - so although on low frequencies the difference is not that big, on 10m the difference is very significant.
Changed that now...
Ethernet speed change from 100Mb/s to 10Mb/s made a big change!
In my installation the change to 10Mb/s reduces the noise in the spectrum above 10Mhz to almost not visible levels in the waterfall. Wow!
The result is, that many weak signals are hidden in the rx-noise even in 30m band and defintely above. I tested shortly with a very sensitive other receiver and were surprised to see so many signals popping out of the noise, with additional 10-15db additional gain.... however it will be a challenge to achieve an MDS of -140dbm without bandpass filters... let's see...
A question:
Will it be reliable for wsprdaemon software to run, if the kiwis are set to lower 10Mb/s Ethernet speed ?
Otherwise I need to switch to Wifi connection and I saw the long discussion in the other thread running a simple stick... I am afraid of my non-existing Linux/code knowledge... To switch over to Wifi mode would need a ready-to-go-out-of-the-box solution, but I wonder, if this exist...
I am very happy with the results of the long wires now, with the low ethernet speed... much better with much lower noise.
Now I must build pre-amps...
Until now, I didn't use pre-amps with the wires... (except my active vertical dipole, which is not in use at the moment)
One more question to the noise graph:
is it possible to change the vertical scale of the noise graph for 2200 and 630m ?
The real noise it at 90...100db/Hz region - so it is invisible now (only on the Grafana site). It would be great to change the vertical scale for the VLF area so realistic figures can be shown...
Ulli, ON5KQ
You commented that "What I question is the strong increase of gain towards higher frequencies... up to +7dbi on 10m "
Yes that seems sensible to me.
As the wire becomes progressively 'longer' in terms of wavelength, the radiation pattern side-lobes tend to become narrower, tilt and follow along the direction of the wire, producing a very high gain but narrow beam.
I use a DX Engineering RPA-1 pre-amp clone, followed by an amplitude / frequency slope corrector (see previous threads on the subject) and broadcast band notches in order to improve the KiWi NF on the HF bands, whilst decreasing the overall system gain on the LF bands in order to prevent ADC overload.
Regards,
Martin - G8JNJ
I will make a pre-amp and check, what equalization is neccessary to keep the Kiwi-rx well below clip-level.
By the way:
What happened to the FFT noise plots on the Grafana-website? Since this morning, they do not appear anymore...
Ulli, ON5KQ
Another advantage of the preamp, beyond simply improving the kiwi's noise floor, is that it pushes common-mode ingress occurring after the preamp further down by the amount of the gain. A disadvantage is that it may push the kiwi closer to OV. Optimizing a kiwi is a balancing act. It's likely that the LAN noise you see at higher speeds is due to CM current sourced by the LAN cable, continuing through the finite balance of the LAN input transformers, going through the kiwi (ground plane) itself and sinking out through the SMA/feedline.
It doesn't take a lot of CM current or terribly low CM impedance to achieve enough noise at the kiwi preamp/adc input to exceed KTB. At 20-30 MHz propagated&galactic noise is very likely only ~10 dB or so above KTB, even lower if antenna efficiency is low (antenna unmatched, too small, etc)