It's true that a circularly polarize antenna can mitigate spin fading in satellite communications. This works because a linear wave is always received by a circular antenna at a uniform 3 dB loss. They are a couple equivalent ways to explain it:
A linearly polarized wave is a superposition of left- and right-hand circularly polarized waves, in equal amplitude. Thus, any circularly polarized antenna will receive half the signal power, regardless of the polarization angle. Half power is a 3 dB loss.
A circularly polarized antenna consists of a horizontal component and a vertical component. Say the incoming wave is vertically polarized: it will be received by the vertical component of the receiving antenna but not the horizontal component (which will receive only noise), thus a 3 dB loss. Likewise if the incoming wave is horizontally polarized. And it can be shown with a bit of trigonometry that for polarization angles between horizontal and vertical, the coupling is still -3 dB.
Underpinning each of these explanations is the fact that any possible polarization of a wave propagating in a particular direction can be expressed as the superposition of two orthogonal polarizations of specified amplitude and phase. Horizontal/vertical is one possible orthogonal pair, though any two angles separated by 90 degrees will do. Left- and right-hand circular polarizations will also do. There are also orthogonal pairs of elliptical polarizations. In general, any two points on opposite sides of the Poincaré sphere will do.
Any possible receive antenna will have some polarization that can be represented by some point on the Poincaré sphere. As such, any incoming wave can be considered the superposition of two components:
- the polarization that matches the antenna, and
- the orthogonal point opposite the antenna's polarization.
At any given time, some percentage of the available signal power will fall into the first component, with the remainder in the second component. Polarization fading occurs when a large proportion falls into the second component rather than the first.
For example, if the receiving antenna is a horizontal dipole, then the orthogonal polarization is vertical. When the incoming wave is vertically polarized, all of the signal power falls into the orthogonal polarization which is not received by the antenna, and the received signal power is zero.
Likewise, if the receiving antenna is right-hand circularly polarized, then polarization fading will occur when the incoming wave is close to left-hand circular polarization.
Since any possible antenna will have an orthogonal polarization for which the antenna will receive no energy, the question becomes: is there something about circular polarizations which makes their orthogonal polarization less likely to be observed when receiving a transmission through an ionospheric channel?
In general: no. Circular polarization works to mitigate spin fading because the line-of-sight channel over which satellite communications happen preserves polarization. If the satellite transmits a vertical polarization, it will still be a vertical polarization when the wave arrives at the receiver. If the satellite transmits right-hand circular polarization, it will still be right-hand circular polarization at the receiver.
This is not generally true in an ionospheric channel. A single path through the ionosphere can display Faraday rotation, which rotates the polarization. But ionospheric paths often exhibit strong multipath propagation as the transmitted wave reflects off multiple layers of the ionosphere at different heights. Each possible path can exhibit a different phase delay and rotation.
This means an ionospheric channel, unlike a line-of-site channel, can take any polarization from the transmitter, and transform it to any other possible polarization by the time the wave reaches the receiver. For example, an ionospheric channel can transform a vertically polarized wave into a circularly polarized one by having two paths which differ in length by 90 degrees, with one having a 90 degree greater rotation than the other.
Not all paths are as bad as others: NVIS paths are relatively mild, and may preserve linear polarization well enough that circularly polarized antennas are still of some use to mitigate fading. Multipath interference becomes more severe as the path becomes more oblique, and as more skips are added.
The ionosphere is in motion, so the polarization of the wave when it arrives at the receiver changes over time. In general, the polarization of the received wave will be a random point on the Poincaré sphere: it could be vertical, horizontal, right-hand circular, left-hand circular, or any of the many possible elliptical polarizations that lie between these points on the Poincaré sphere.
Because the polarization of the received wave is random and time-variant, there is no possible single antenna which will mitigate polarization fading on ionospheric paths that exhibit significant multipath propagation.
There is however a solution: if the receiver is allowed two antennas of orthogonal polarization, there always exists some combination of these antennas that precisely matches the incoming wave. A simple approach is to switch between the antennas and select the one with the better signal-to-noise ratio: this limits polarization fading to no more than -3dB. Slightly better but more complex to implement is to combine the two antennas by dynamically determined complex coefficients which can effectively move the antenna to any point on the Poincaré sphere to match the polarization of the incoming wave: this eliminates polarization fading entirely (or at least to the extent the algorithm is accurate)
As has been rightly pointed out, this theoretical explanation doesn't matter if experimental evidence doesn't agree.
The most pro-circular evidence I could find is a paper by Drew Schoen O'Shaughnessy of the Worcester Polytechnic Institute. This paper states:
From the data, it is clear that in many of the test transmissions circular polarization yielded a maximum SNR improvment of approximately 3 dB over the strongest channel using linear polarization. A 3 dB improvement is a non-negligible increase in signal strength,which roughly translates to a doubling of the effective range of the HF link. Although the SNR degradation due to polarization fading could not be measured directly, the tests show that the SNR performance of this link did improve when circular polarization was in use. It is unlikely that the differences in SNR between linear and circular polarization can be attributed to something other than polarization fading, given that the only change between them was the polarization of the antenna.
Although right hand circular polarization was dominant in these tests, the usage of left hand circular polarization should also be considered. Given simple phasing methods, it is not difficult to construct a ground based antenna system that would allow for switching between RHCP and LHCP. If the radios in a hypothetical HF link are operated manually,the operators could choose a polarization based on a test transmission scheme similar to the one used in this experiment.
I'm a little skeptical of the recommendation. The data consist of a limited number of observations over a limited number of paths, and the researcher found "There were several tests where linear polarization was actually stronger than either circular polarization. Circular polarization is not a universal solution for improving signal to noise ratio." The paper does not appear to be peer reviewed, and there's no analysis of the statistical significance of the results.
I also found Circular Polarization – Is It Worth the Effort? by Carl Luetzelschwab K9LA, which mostly includes anecdotal evidence such as:
I also talked to Woody WW1WW. He has a 6-element crossed-Yagi array at 50 feet on 10-Meters. He brings equal length coaxes from each Yagi into the shack, and can lead or lag either one by 90oto achieve left-hand or right-hand circular polarization. When I talked to him in early 2013,this antenna had been up for only a year and 10-Meter propagation was not great. Thus his observations were somewhat limited. He said he had observed up to 30 dB difference between left-handand right-handpolarization under fading conditions, and it does not seem predictable. He did not offer any comments on the duration of openings or on fading.
As for ‘cons’ with circular polarization, the last paragraph highlights the fact that you need to be able to select either left-hand or right-hand circular polarization (which means only one of the characteristic waves may be propagating at any given time)
I suspect switching between vertical and horizontal polarization works just as well as LHCP/RHCP, but given the antenna in either case is the same there's little practical reason to do one or the other. The author of this paper at least seems to have reached the same conclusion as I have: the ability to switch polarization is necessary, since circularly polarized antennas are still subject to fading.
I am finding indirect evidence that some ionospheric paths, especially NVIS paths maintain polarization well enough that circularly polarized antennas may be applicable to reducing fading. Mostly this comes from research on HF MIMO techniques, such as this and this, which find an NVIS path maintains orthogonality well enough to support some MIMO capacity gains. This implies that the polarization is not entirely random.
But as the path becomes more oblique things get worse, especially when the path involves multiple hops. For example HF Skywave Polarized MIMO Channels with Oblique One-Hop Paths by Umaisaroh Umaisaroh, Gamantyo Hendrantoro, *, and Varuliantor Dear, states:
If an implementation over a local area of up to a 200 km radius is desired, NVIS propagation with almost 90◦ elevation angles can be exploited [12]. However, for wider coverage area, the path of the radio waves should take a slanted elevation angle. In a single-hop radio wave transmission through the ionosphere, the elevation angle varies with the ground range between the transmitter and receiver. As the paths become oblique, the orthogonally polarized horizontal antennas do not appear as orthogonal as in NVIS case for the departing and arriving waves, which cause differences in MIMO capacity. This paper shows that oblique paths reduce the capacity gain relative to the SISO capacity due to the decreasing orthogonality between the cross-dipoles.
I can find very little research on MIMO over single-hop paths, and none on multi-hop paths. I suspect it's because the polarization is so well randomized that no useful orthogonality is preserved.