Using the K4JCW and KD2EAT balloon transmissions to improve your WSPR station (noise) performance

This posting describes a way to gauge a 20m WSPR station’s noise floor and performance using transmissions from one of the high altitude balloons which carry WSPR around the world.

Recently by applying surface wave transmission line theory, I’ve been working to create efficient, small dipoles for the amateur LF through HF bands. Application of this theory to very small dipoles, even ones less than 1 meter tip-tip has provided very good results for amateur HF WSPR use.

The 20m antenna I have been using for several months is only 75 cm, about 30”, tip to tip. It’s presently only 7-8 meters (24’ ) above ground level at N6GN. Even so, by working to improve efficiency and to eliminate common mode current on the feedline, I have come up with a very good performer. It is an excellent receiving antenna and one that actually does pretty well on transmit as well.

As part of testing this antenna, I’ve had N6GN on 20m WSPR quite a lot. As has been reported and tried by many others, WSPR is a great tool for examining not just propagation but amateur antennas as well. It just so happened that during this time of testing, one of the several amateur high-altitude balloons carrying a WSPR beacon “blew” past my QTH, close enough to be complete line-of-sight (LOS) to my antenna. This occurred when “flight Q2” of the K4JCW 20m WSPR balloon (one that also signs WSPR transmissions with “Q” frames carrying precise altitude and 6 character grid square information) came onshore the US into California beginning it’s 20th lap around the globe. It came within about 40km (25 miles) of my QTH at an altitude of about 8 km (5 miles).

This turned out to be a fortuitous event for me. Because the balloon transmitter including altitude, and transmitter power and ERP were rather precisely known and reported and because it was completely line-of-sight to my station, I had the rare opportunity to characterize my antenna and noise level. This was so because by simply looking at information obtained from the WSPR spot and then calculating path loss, I was able to determine what power a reference antenna should receive. Knowing this power and by comparing with the reported S/N makes it is possible to obtain a fairly good measurement of my receive station. Furthermore, because this balloon has been “mobile” I can also compare my results with many other 20m WSPR stations around the world which have had the balloon fly into their LOS range.

To achieve this, I wrote a script that allowed me to mine the WSPR database. I looked at the two months of May and June 2017 to find other stations that had spotted the balloon and what sort of S/N they reported. These results rather confirmed other tests I have been making. What I now believe to be true is that most 20m WSPR stations in the world have receive performance which is severely degraded by local, non-propagated, noise sources. This causes them not to hear and decode as they otherwise might. Perhaps most importantly, I now believe there is something that can be done to change this situation.

Within amateur radio, it seems to have become popular to blame noise on “all those modern digital devices”; TV’s, switching power supplies, power lines and such are all identified as the cause of the decline of the amateur HF experience. The ARRL, RSGB and even the FCC in the US have begun to address this problem. Meanwhile the bands often don’t seem as good as many of us who operated 50 years ago. However most hams seem resigned to this and few people have tried to do much more about the problem than complain. I think that there is a great deal that can be done about it and I believe that I can demonstrate this.

Returning to the balloon flight that went by my QTH, here is the best S/N I was able to report into the WSPR database. This is from the “compressed dipole” at about 8 meters above the ground. For this spot, the balloon position was not optimum. I didn’t realize what was going on until it was past. The antenna happened to be turned so that the balloon was almost off the end of this horizontal dipole, not broadside to it. The antenna on the balloon is a vertical dipole so for this spot the two antennas were nearly cross-polarization. For ionospheric propagation we don’t worry about crosspolarization but for VHF and line-of-sight communications at HF polarization purity, axial ratio, is maintained so it can matter a great deal. Misalignment can produce 15-20 dB or more of attenuation. Even so here is my best spot:

Timestamp Call MHz SNR Drift Grid Pwr Reporter Rgrid km ax
2017-06-29 01:04  K4JCW 14.097101 +15 0 CM87 20 N6GN CM88OK 109 12

A “Q” WSPR transmission at near the same time gave me the information to calculate a six character gridsquare locator and the precise balloon altitude. This +15 S/N is quite a strong signal from a 20 milliwatt ERP transmission. It’s strong because it is LOS rather than via the ionosphere and it is also strong because after quite a lot of effort to reduce it, the noise level at N6GN is relatively low. To compare, the attached file shows the best spots from the last two months of 20m WSPR for the K4JCW/Q2 balloon. Analysis of the grid squares shows them all to be LOS as mine was.

The balloon was transmitting at +13 dBm and went nearest to N6GN on 1 July, 2017 about 40 km to the south. Since it was at about 8 km elevation, it was a little more than 10 degrees above horizon and above hills and local clutter. N6GN should have had a clear and unobstructed shot at it. If N6GN had been using an ideal full size, half-wave dipole oriented vertically then the approximate signal level received could be expected to have been approximately:

Tx_ERP – pathloss + receive_antenna_gain = +13 – (37 + 20log(14.1) + 20log(25)) + 2 = 13 -88 +2 = -73 dBm

This is precisely S9 on a calibrated receiver. But WSPR doesn’t provide absolute signal level reports, only S/N. The S/N of +15 that WSPR measured indicates that WSPR thought the noise in a 2.5 kHz SSB bandwidth was 15 dB less than the actual signal power received. Had the N6GN antenna been an ideal dipole, perfectly aligned, this would have indicated a noise floor of -88 dBm in 2.5 kHz or about -122 dBm/Hz.

As a sanity check, a matched dummy load at room temperature at the receiver can be expected to deliver -174 dBm/Hz of noise. Thus, with -122 dBm as a worst case the N6GN noise figure would have been about -122-(-174) = 52 dB! In actuality I believe it is not this high, it is lower by the amount the test antenna and alignment was different from the ideal.

As I measure and estimate from other methods the antenna efficiency, which is essentially the attenuation relative to a matched dipole, was probably somewhere around 10 dB. Mismatched polarization and the off-end pointing of the antenna may have contributed another 10 dB, I really don’t know exactly. If this was the case, it would indicate a noise floor of about -122 - 20 = -142 dBm/Hz.

N6GN uses an Apache Labs Angelia board SDR. It has been calibrated with a calibrated signal generator. The measured noise floor at N6GN varies somewhat but at the time the balloon was spotted, it was reading approximately -112 dBm in a 1 kHz bandwidth which equates to -142 dBm/Hz so these two measurement methods, balloon and calibrated receiver, agree well.

As a result of these measurements and estimates, it is reasonable to suppose that the external noise floor, the signal level that would be received by a full sized, 100% efficient dipole would be different by the efficiency or on the order of -132 dBm/Hz or 42 dB above KTB, the noise power in a resistor. For our amateur stations, this noise floor value is a very important number to know since it gives a measurement of the ultimate sensitivity of the station and location.

This number describes the performance which determines the maximum S/N that a given antenna can deliver. Anything done to improve it improves all two way communications. Factors that worsen this number include radiated QRN or common mode noise current on the antenna feedline reaching the receiver hurt communications. We should all know and care a lot about what this number is for the various bands and antennas we use in amateur radio since it can easily overwhelm the advantage (or disadvantage) that antenna gain, front/back ratio and even height might determine.

How does measured noise compare with theory?

Reference books generally give a range of noise levels that depends upon the environment. ITUR P.372-9 defines the external noise factor Fa at frequency f for a short vertical antenna over a perfectly conducting ground plane. This parameter is related to rms noise field strength En along the antenna by

En = Fa + 20 log(f) + B - 95.5

where En is in dB (relative to 1 µV/m), frequency f is expressed in MHz, and the receiver bandwidth is B in dB relative to 1 Hz. A range of noise level is expected, depending upon the environment. At 20m these go from about .1 for “quiet rural” to about 10 microvolts/meter for “business”.

Calculating by using the “rural” level of 1 microvolt/meter for 14.1 MHz gives the noise factor. This is about -174 + 18.6 = -155.4 dBm in 1 Hz or -121.4 dBm in an SSB bandwidth. A calibrated S meter should read 42 dB (7 S units) below S9 or S2.

The estimated noise floor from the balloon measurement at N6GN is therefore about 13 dB greater than this “rural” level. But even with this degradation, the N6GN S/N spot of the K4JCW balloon was no worse than second place in the world! In examining the first space spot, I have reason to suspect that there is an error in the reporting. I’ll mention why I think this later. All this leads me to suspect that almost all stations on 20m WSPR and likely all amateur stations on 20m are suffering from elevated noise levels. This is bad news in one sense BUT it also offers the hope that some or many may be able to reduce the noise as I did.

I think that this is a widespread problem.

From these balloon measurements and from comparing a lot of WSPRnet database spots and watching hearing/heard ratios on 20m WSPR carefully for several months it has become my opinion that the majority of 20m WSPR stations are being significantly compromised, in some cases severely compromised, by elevated noise floors. Furthermore, it is becoming apparent that many of these cases are NOT due to radiated QRN as “common wisdom” seems to indicate. This, in my opinion, is actually very good news since it suggests that there are things we can do to greatly improve our station performance and amateur radio experience. The first problem to overcome is that of unawareness – if we don’t realize we have a problem and if we don’t realize there is something we may be able to do to fix it, the situation will stay much the same.

I hope that what I’ve presented this far has given some credible evidence that there is a widespread problem. The availability of known, mobile signal sources floating within measurement range of many stations can be a great boon to understanding our station’s performance, improving it and validating improvements. If you are able, I encourage you to watch the K4JCW/KD2EAT balloon paths and take any chance you have to “measure” them using WSPR. From that result and minimal knowledge of your station, I believe you can quantify how bad a problem you may have with noise. Armed with this information, you may be encouraged to take the next step – I hope to follow this article with another that details some methods for identifying noise mechanisms and for greatly reducing receiver noise levels.

How to find the balloon.

The easiest way to find the balloon is just to set the WSPRnet map page to show you spots of K4JCW. If you find any, look at the reported grid square for an idea of the balloon location. If and when you find it nearing your QTH, take the next step and look for related “Q” transmissions in the two minute period preceding it. The 4th and 5th characters of the call sign give the 5th and 6th characters of the grid square. For details please see this document.

As I write this on July 8, 2017 a K4JCW balloon is working it’s way up the East coast of the US, in LOS range of many stations. Keep an eye out for one coming near you and see if you can spot it when it becomes LOS. That number and knowledge of your station may be able to lead you to a dramatically better and more enjoyable experience on HF. I hope it does.

An Aside: A WSJT-X S/N reporting anomaly

As shown from the table attached at the end of this posting, the highest S/N reported was from JH3APN. I haven’t been able to contact the operator but from other spots and data I suspect there may be an anomaly in this report.I’ve discovered that if one provides band-limited noise to WSJT-X, with bandwidths significantly smaller than the 200 Hz WSPR window, that the reported S/N is greatly in error.Once the noise is pre-filtered in this way, 10 to 20 dB of change is produced in the reported spot. I suspect this has to do with the internal algorithm used to measure and report S/N, but I haven’t examined the code so don’t know this for certain. Perhaps someone reading this can provide a better answer.

UPDATE: 27 January 2018!
There's a new worldwide winner in the spot-k4jcw-at-the highest SNR. PI4THT, the university of Twente club station and home of WebSDR has reported a bigger spot:
2018-01-25 13:56 QT2JBF 14.097106 +26 1 JO32 20 PI4THT JO32kf 28 192
This is probably because the balloon was less than 20 km distant, slant range, and the SDR's antenna was vertically/co-polarized with the balloon. The calculated noise floor for PI4THT is still on the order of -124 dBm/Hz and therefore not particularly low. Who will do better?

I suspect an error because other spots by JH3APN and the associated Hearing/Heard ratio do not seem to support a station with a phenomenally low noise floor. It may be that these numbers are correct but it also may be that they are skewed by pre-filtering of the noise presented to the WSPR decoder.

Whether or not this is a correct number, the fact that most stations in the world report numbers MUCH lower than these top few leads me further towards the conclusion that most stations have a higher noise floor than these top reporters.

Hearing/Heard Ratio on 20m WSPR

I also mined the WSPR database for another bit of information. I wanted to know, roughly, what the mix of unique transmit-only station call signs, unique receive-only callsigns and callsigns both transmitting and receiving. I knew that there were a lot of transmit-only beacons such as the U3s and PI3 and that often these are reporting considerably less than the 5 watt ‘nominal’ WSPR level.I filtered the data to remove the numerous “0” and “Q” balloon prefixes and analyzed the remainder. What I found for June 2017 was:

Total non-balloon calls spotted = 5771
Calls being spotted (transmitters) = 9597
Calls spotting others (receivers) = 2691
Calls both spotting and being spotted (both receive and transmit) = 1861

There were about twice as many transmit-only stations as there were receive-only stations. Only about one-third of the total callsigns spotted were from full transmit/receive WSPR stations.

This result leads one to think about another metric that is useful to reveal station noise levels. That metric is the Hearing/Heard_by ratio.If, on average, radio paths are reciprocal we can expect identical stations to hear each other just as well as they are heard if they are running the same antennas and ERP. We know that paths aren’t always reciprocal, after traversing the ionospheric plasma we think there are only circularly polarized O and X waves which don’t necessarily go the same places. The reverse path isn’t always the same but on average we might expect some sort of general parity. If every transceive station was running the same transmit power we might expect hearing/heard ratio to be about unity. Of course we know all of this is skewed. In general there are at least three times as many unique callsigns transmitted as there are receivers spotting them.If all transmitters were the same ERP we might expect hearing/heard ratio to be 3:1. Actually it should probably be somewhat less than this because many of these transmitters are running less than 1 watt.

By examining hearing/heard ratio data from a few stations which are both receiving and transmitting a metric for a “good” station may be obtained.This is not a precise measurement but generally a ratio of more than 2:1 should be possible. If one simply counts the number of lines on the WSPRnet map for a given callsign an idea may be gathered of that stations noise floor.

To get an idea of how your station compares to others’ compare your hearing/heard ratio with others in your general area. But remember, simply being as good as the average probably isn’t very good. With some further measurement, using calibrated receivers, spectrum analyzers or possible your spots of an LOS WSPR balloon, you may discover that your station has a lot of room for improvement.If possible I will write about some of the techniques I’ve found and used to reduce the noise at N6GN by almost 30 dB.

Here is a sampling from the 20m WSPR map late on a weekend afternoon:


The ratio is 28:13 or a little over 2:1. Many of the stations N6GN is spotting do not spot N6GN even though we all run similar ERP. These stations do not generally show as a high a Hearing/Heard ratio. I believe that the difference is largely noise floor. Before I made station improvements at N6GN, I had fewer Hearing spots and a much lower Hearing/Heard ratio, one more like the average 20m WSPR station. Since the improvements not only do I spot the fixed and known signal from the balloon among the best in the world I also am sometimes one of a very few stations hearing or spotting a DX station from my part of the world. All this is being done with only a very small and relatively inefficient dipole, not with a large high gain antenna. I think and hope that some of what I’ve learned and hope to write about soon may help others.

Glenn Elmore n6gn
contact information at the bottom of the N6GN pages at
July 2017

Image icon sn.jpg74.41 KB