This JavaScript calculates estimated median values for external radio noise based on the spherical harmonic coefficient model of CCIR Report 322-3 and ITU-R P.372, the calculation is according to NTIA Report 87-212 p.11 ff. It can be expected that in 50% of all measurements the noise level is lower than or equal to and in 50% higher than or equal to the estimated median values, and that furthermore in 80% of all measurements the downward or upward deviation from the median values is not higher than Dl or Du.

It is important to note that these estimates apply only to background noise, which is white noise generated by many distributed distant sources. In contrast, superimposed impulse noise which sounds crackling or "corny" or rapid amplitude changes are generated by local sources like passing thunderstorms, rain or snowfall or electrical appliances in the neighborhood. A noise level significantly above the predicted values can be an indicator for such local sources. It is a fact that vertically polarized antennas pick up more of this kind of noise than horizontally polarized antennas. The reason is that the conductive earth more or less short-circuits horizontal electric fields and therefore ground-wave attenuation is much higher for horizontally than for vertically polarized noise. However, this does not apply to sky-wave but only to ground-wave propagated noise which comes from sources within a radius of typically a few hundred meters around the receiving site. The noise generated by even closer sources within a distance of a few tens of meters is not propagated by radiation but electrically or magnetically coupled into the receiving antenna by the reactive near field.

ATMOSPHERIC NOISE

Atmospheric noise is caused by electrical discharges in the atmosphere due to e.g. thunderstorms, rain and snow, sandstorms. It can propagate over long distances on shortwave via the ionosphere and depends on location, frequency, time of day and season. In high latitudes, only relatively weak noise is to be expected, while in the tropical zone, especially during the rainy season, very strong atmospheric noise is generated.

In the years 1957 to 1966 (19th sunspot cycle maximum to minimum and beginning of the 20th cycle), as part of an international cooperative program of the URSI, the atmospheric noise between 13 KHz and 20 MHz and its statistical distribution was measured by a network of 16 stations equipped with the standardized radio noise recorders ARN-2 at the following locations:

Balboa, Panama Canal Zone
Bill, Wyoming, USA
Boulder, Colorado, USA
Byrd Station, Antarctica
Cook, Australia
Enkoping, Sweden
Front Royal, Virginia, USA
Ibadan, Nigeria
Kekaha, Hawaii, USA
New Delhi, India
Ohira, Japan
Pretoria, South Africa
Rabat, Morocco
San Jose, Brazil
Singapore
Thule, Greenland

This noise data was grouped into four seasons ...

Winter in the northern hemisphere =
Summer in the southern hemisphere =
December, January, February

Spring in the northern hemisphere =
Fall in the southern hemisphere =
March, April, May

Summer in the northern hemisphere =
Winter in the southern hemisphere =
June, July, August

Fall in the northern hemisphere =
Spring in the southern hemisphere =
September, October, November ...

within which there are six 4-hour blocks of local mean time ...

00:00 - 04:00
04:00 - 08:00
08:00 - 12:00
12:00 - 16:00
16:00 - 20:00
20:00 - 24:00

... and converted to a matched and loss-free short monopole antenna (length << 0.1 λ) over a perfectly conductive radial network, which to this day is the reference antenna in ITU publications. Finally, a spherical harmonic analysis was applied to this data, which resulted in a set of coefficients from which the median value of the noise figure Fam [dB(kTob)] of the reference antenna and its statistical variance can be calculated as a function of geographical latitude and longitude, season, time block and frequency.

As early as 1964, before the measurements were completed, the ITU published CCIR Report 322 "Worldwide Distribution and Characteristics of Atmospheric Radio Noise" with 24 (4 seasons x 6 time blocks) isoline maps and diagrams showing these values for any point on the earth's surface and any frequency between 10 KHz and 30 MHz. Lucas and Harper developed a numerical representation of Report 322 in 1965 based on the 1 MHz maps instead of the original data points from which the maps were created, which resulted in an average error of about 2 dB and a maximum error of about 10 dB compared to the maps. In 1970, Zacharisen and Jones developed numerical "maps" from the original data and in 1982 Sailors and Brown developed a simplified numerical model for minicomputers. This 1964 report only considered the data collected between July 1957 and October 1961, but it was reprinted in 1983 as CCIR Report 322-2 with revised text and under a different title, but with the same atmospheric noise data.

Later, data from the 16 original stations were available until 1966, and additional data were available for many years from 10 stations in the then USSR and from Thailand until 1968. The new locations were:

Laem Chabang, Thailand
Alma Ata, USSR
Ashkhabad, USSR
Irkutsk, USSR
Khabarovsk, USSR
Kiev, USSR
Moscow, USSR
Murmansk, USSR
Simferopol, USSR
Sverdlovsk, USSR
Tbilisi, USSR

This new data was analyzed, corrected, converted and added to the existing data by David Sailors and his colleagues at the Naval Ocean Systems Center (NOSC) in San Diego, California, and the Institute for Telecommunication Sciences (ITS) in Boulder, Colorado. The result was a new set of 13,056 coefficients as the basis for the CCIR Report 322-3 and the Recommendation ITU-R P.372 (Spaulding, Stewart: "Atmospheric Radio Noise: Worldwide Levels and Other Characteristics", NTIA Report 85-173, 1985). The numerical version of Report 322-3 is now exact, i.e. the numerical and graphical versions provide identical data for all parameters including the median external noise figure Fam.

MANMADE AND GALACTIC NOISE

Manmade noise is caused by electrical devices, equipment and networks. The NTIA Report 87-212 also provided an improved model for this noise component, which was developed by Spaulding and Disney in 1974 and is based primarily on measurements taken in the United States. The measurements for quiet rural areas were carried out worldwide. The model for manmade noise is based on the following equation:

Fam = c - d log(f)

with f = frequency [MHz] and the constants c and d from the following table:

Environmental category: c / d
*****
Business: 76.8 / 27.7
Inter-state highways: 73.0 / 27.7
Residential: 72.5 / 27.7
Parks and university campuses: 69.3 / 27.7
Rural: 67.2 / 27.7
Quiet rural: 53.6 / 28.6
Galactic noise: 52.0 / 23.0


Normally, only the categories business (city, industrial area), residential (residential area), rural (rural area) and quiet rural (quiet rural area, remote and uninhabited) are used. The analysis of the measured data has shown that the following statistical parameters provide acceptable results for all four categories and all frequencies, but are most suitable for the shortwave range:

Du = 9.7 dB
σ Du = 1.5 dB
Dl = 7.0 dB
σ Dl = 1.5 dB
σ Fam = 5.4 dB

While in the past ignition systems of motor vehicles and overhead power lines also for low-voltage contributed a lot to the artificial noise, today's ignition systems and the low-voltage lines that we have now mostly laid in the ground are no longer an issue. Instead, new technologies (e.g. switch-mode power supplies, PLC, power-saving and LED lights, solar systems) and the cumulative interference caused by the rapid increase in the number of devices in use is very problematic. In recent years, several studies have been conducted with contradictory results. A large-scale measurement campaign by the BNetzA between 2007 and 2009 at over 100 locations in Germany at frequencies of 5, 12 and 20 MHz even came to the conclusion that the artificial noise was usually lower than predicted by the model (Report ITU-R SM.2155, result on p.29 Table 4).

For the artificial noise, the GENOIS procedure in the NTIA report calculates with σ Fam = 3.0 dB instead of the correct value given in the report as 5.4 dB, in the old program NOIS1.FOR with Dl = 6.0 dB instead of the correct value given in the report as 7.0 dB and in the new program GH_NOISE again with all correct values according to the report.

Galactic noise originates in outer space and, due to ionospheric absorption, is only significant above approximately 10 MHz. The following noise parameters apply to this noise component:

Du = 9.7 dB
σ Du = 0.2 dB
Dl = 2.0 dB
σ Dl = 0.2 dB
σ Fam = 0.5 dB


External electromagnetic noise ("radio noise") is mainly composed of the three components atmospheric, artificial or manmade and galactic noise, which must be summed correctly.

COMBINED NOISE

The normal distribution (Gaussian distribution, bell curve) is the most important probability distribution. Examples: intelligence, body height, income (if it is logarithmized !) are normally distributed. The "logarithmic normal distribution" (abbreviated "log-normal" distribution) describes the distribution of a random variable x if ln(x) or log(x) is normally distributed. Conversely, if y is normally distributed, then e^y or 10^y is log-normally distributed. While a normally distributed random variable can be understood as the sum of many independent random variables, a log-normally distributed random variable is the product of many random variables and therefore the log-normal distribution is the simplest distribution for multiplicative models. This includes atmospheric, galactic, and artificial noise, because their noise factors fa are log-normally distributed and thus their noise figures Fa = 10 log(fa) are normally distributed. Therefore, the mean value can be directly calculated from two noise figures Fa in dB and linear interpolation can be performed directly between them.

Spaulding & Stewart showed in NTIA Report 87-212 that the previously applied method of determining the distribution of the combination of the three noise components was not correct and developed a statistically more correct method. Simply adding up the average values for the atmospheric, artificial and galactic noise power into a resulting average value would be correct, but the model provides median values and adding these up into a resulting median value would produce an error. Therefore, the noise components must be combined using split log-normal distributions.

The strength of all three noise components decreases with increasing frequency. Atmospheric noise is usually stronger than artificial noise at electrically quiet locations at low latitudes below 20 MHz, while galactic noise can only become the dominant component at electrically very quiet, remote locations and at high latitudes.

The non-local background noise received by an antenna is generated by a very large number of sources and comes from a correspondingly large number of random directions. Therefore, it may be assumed that the gain of any lossless antenna with respect to noise is almost isotropic. It may further be assumed that the noise coming via the sky-wave from distant sources or via the line-of-sight from very nearby sources is randomly polarized, and that the noise coming via the ground-wave is predominantly vertically polarized. Therefore, any passive matched and lossless electric antenna will produce approximately the same noise power except for ground-wave noise, which is suppressed by horizontal antennas. In densely populated electrically noisy areas, more artificial noise will be received by vertical antennas via the ground-wave than in electrically quiet areas, where there is likely to be little difference between vertical and horizontal antennas.

The noise power supplied by an antenna is the sum of all the power received from individual sources with their own Poynting vectors and polarization planes and therefore cannot be calculated with the antenna gain over the field strength in the deterministic physical sense as we do for a singular radio signal. In order to create a basis that can be applied to all noise components and that also allows a comparison of their strength, in the relevant ITU documents of the ITU a noise power pn is assumed, which the matched lossless short monopole antenna over a perfectly conductive radial network delivers from external sources. An Ohmic resistor delivers the noise power kTob, regardless of its value. The ratio of these noise powers is the effective antenna noise factor fa:

fa = pn / (kTob) = Ta / To
Fa = 10 log(fa)

where:

fa = external noise factor
Fa = external noise figure 10 log(fa) [dB(kTob)]
pn = available noise power [W]
k = Boltzmann constant (1.387 x 10^-23 J/K)
b = effective noise bandwidth [Hz]
To = reference temperature (290 K / 17° C)
Ta = effective antenna temperature due to external noise [K]

so it is:

pn [W] = fa k To b = 10^Fa / 10 k To B

and with B = 10 log(b) and 10 log(kTo) = -204 dBW we get:

Pn [dBW] = Fa + B - 204

and the effective noise voltage Un at a receiver with the input resistance Ri Ω is obtained:

Un [Veff] = SQRT (pn Ri)

For the short vertical monopole above a perfect groundplane, the following applies:

En [dB(uVrms/m)] = Fa + 20 log(f) + B - 95.5

where:

En = electric noise field strength in the bandwidth b [dB(uVrms/m)]
f = center frequency [MHz]

fa and Fa are independent of the receive bandwidth b, because the noise power supplied by the antenna and that supplied by a resistor are both proportional to the bandwidth. By contrast, the noise power pn and noise field strength En are dependent on the bandwidth. The spectral noise power density is obtained for a bandwidth of 1 Hz. Background noise does not come from a single source and therefore does not correlate. Pn and En therefore increase with the measurement bandwidth by 10 log(b) or 3 dB when doubled. Most digital signals are similar to noise. Correlated signals, on the other hand, come from a single source and as long as the measurement bandwidth is smaller than the signal bandwidth, Pn and En increase with 20 log(b) or 6 dB when doubled.