Digital signals are very precisely engineered tiny pieces of technology sent in a rude world. How resilient they are is discussed here.

Noise has a broad band component. It can also show bursts due to lightnings  (QRN) or static activity at the antenna (snow or rain) and well defined signals in the frequency domain : carriers radiated by transmitters, power lines, computers treating the signal, household appliances, the neighbour's sheep fence ...

Except in very quite location, man made noise is the main steady component of broadband noise.

Under 10 MHz atmospheric noise can dominate.  It's due to the averaging of lightnings. It decreases with increasing frequency and is lower in the cold season than in summer  (contacting stations in the other hemisphere on 160 or 80 m requires therefore exceptionnally quite conditions).

Over 10 Mhz, man made and galactic noise predominate.

Digital modes ability to transmit information under very low S/N ratios make them fascinating.

These ratios refer to a 3 kHz bandwith for a 2% character error rate. For comparaisons with other data substract 34 dB and add 10log(Bandwith).

(source :KB9II http://my.core.com/~jematz/digicomp.html for CW and SSB, own tests with a path simulator for the digital modes).


Tuning becomes tricky with very weak signals. Multipsk tuning aid for Olivia and Domino's design make them much easier then mfsk under such circumstances. 

MFSK16 at -14 dB S/N How to tune it ?

Here are the examples illustrating the above mentionned limits.

RTTY at -7dB S/N

Feld Hell at -11 dB S/N

BPSK31 at -12 dB S/N

Lightnings cause the background white noise. Impacts are not not evenly distributed on the earth and their number is of course season's dependent. Lightnings bursts cover entire pieces of signal and can  affect  the sensitivity of the receiver which needs some time to recover.

I'm not aware of empirical data reguarding robustness comparaisons of amateur digital modes in that respect.

Two sites provide extraordinary pictures of the current lightning activity. This will help you to assess the level of burst to be expected
http://flash.ess.washington.edu/TOGA_network_global_maps.htm and http://www.wetterzentrale.de/topkarten/fsbeobl.html
for Europe.

This should help assessing the chances of lower band intercontinental contacts.

Local noise entering the receiving chain is a plague. Even in my rather remote location I'm affected by intermittent unknown sources. Don't expect your preferred mode to cope with this kind of problems which can take very tricky forms. Catching the culprit is a detective's job. Take it as it is : a game. The litterature and signal examples are supposed to help.

Broader modes including some redundancy resist better to QRM : Feld Hell accept a  PSK QSO within its bandwith as do Olivia with broader Pactor or RTTY modes to some extent. 

Due to BPSK31 very narrow bandwith, the probability to have an adverse signal landing nearly exactly on the frequency is modest. However, when bands have little or no silence zone (20 m at day and 80 during evening hours) they become crowded with strong (when not overmodulated) signals and sorting out weak signals become nearly impossible.

Multipath and doppler effects result from complex ionospheric refraction involving concentration irregularities and  movements in the ionized layers.

Signals arriving at the receiver having different times of flight will recombine but the interference can be destructive,  bits shape and identity can be modified seriously affecting the error rate. 

Doppler spread tends to broaden the signal's spectrum, averaging the signal's energy on a wider band, introduces phase modulation as well as amplitude variations. Doppler strongly affect paths over the poles.

Signals travelling a long way are likely to be affected by multipath delays. This effect can also be serious in short distance contacts via the ionosphere (so called Near Vertical Incidence NVIS) particularly if the critical and transmssion frequencies are similar : the ionospheric mirror let waves penetrate it where they travel to much lower velocities and show considerable delays differences when they eventually emerge.

Typical time delays are between 1 to 2 ms in mid latitudes (up to for NVIS).

Doppler spread can be as high as 50 Hz on polar paths.

The diiferent modes show quite different performances according to the ionospheric conditions.

This graph illustrates MFSK and Olivia excellent performances in the standart fading channel typical of long haul contacts. These modes give a 2 % Character Error Rate (required level for acceptable QSO's) with SNR 15 to 20 dB lower than than the required levels for PSK or RTTY  in similar conditions. More detailed results should be published soon.

 

Character Error Rate versus Signal to Noise Ratio in Fading channel conditions