Hmmmm ........
Having just looked again at the above 2 papers (via links), they are extremely technical, so perhaps I can explain a bit better .......
Impulsive interference basically comes from electro-mechnical machinery, eg, on/off switches, central heating thermostats, electric motors, car iginition systems, and NOT from "electronic type" stuff (which creates a different type of "radio pollution"!). This gets "added on top" of the digital signal you're already receiving.
Typically, it will be a rapid series of sharp narrow spikes, lasting a short time (eg, from an switch operating), although that "pattern" can also be continuous (eg, a motor, where the commutator is a repetetive on/off switch ......).
Now, for those interseted in Laplace and Fourier transforms, a very narrow pulse - in the extreme limit, a "Dirac delta pulse" - has an infinite frequency spectrum, which is exactly the problem! While such a pulse is "happening", then the signal to noise ratio of a digital reception system is - effectively - very much worsened.
(It's "as if" your dish or aeiral suddenly gets a lot smaller - in bad cases completely disappearing - for the brief time that the impulse lasts!).
But, how this affects reception very much depends on the modulation system used, ie, exactly how the digital signal is being "delivered".
For example, the ATSC terrestrial system - used throughout North America - uses 8 level vsb, ie, each seperate and succesive symbol represents only 3 bits! Therefore, a "spike" added to this will - at worst - only corrupt a few consecutive symbols, and the error correction system should then compensate (except in the worst cases).
Similarly, for satellite (DVB-s1, using QPSK), each succesive symbol represents only 4 bits!
(In any case, all satellite is "fairly immune" to impulse interference, as the dish "sees" only a narrow beam, instead of the much wider area of a Yagi terrestrial aerial).
Unfortunately, the DTT cofdm system is completely different!
For example, with 2K cofdm, 2k (2048) symbols are all transmitted at the same time! And, even though a single interference "spike" lasts a shorter time than the cofdm symbol "duration", nevertheless that many of them are still affected. If the pulse is very high energy - which it often is, for nearby electrical machinery - then error correction wil be unable to cope.
(although the "impulse energy" MIGHT still be low enough for the error correction to "just about" cope, the cofdm receiver can't possibly react quickly enough to "flag" the various carriers for their different error rates!).
Also important is the Freeview receiver. For the 2k (or 8k!) carriers to remain orthogonal, the local oscillator must be very precise. Otherwise, they're no longer orthogonal, and the bit error rate starts going up (very rapidly!).
(For example, my -now antique- Setpal receiver often has an annoying 8 second delay when changing channels, because it's checking carrier frequency against received Viterbi error rate, but most DTT receivers don't bother doing this!).
When compressed video broadcasting via cofdm was first demonstrated as feasible - as part of the Spectre project, around 1992 I think - everyone got extremely excited, but nobody seems to have considered the practical effects of impulsive interference, this only "emerging" during the On Digital fiasco!
For example, see:
www.ebu.ch/trev_256-marti.pdf ,
http://www.dtg.org.uk/reference/tutorial_transmission.html .
(google searching for "spectre dtt cofdm", etc, will get more stuff!).
There's been a "tiny cover-up" - in that - while DTT style cofdm is a "good idea in general" - as regards multipath immunity, and the possibility of single frequency networks - nobody is admitting that they "forgot about" this particular problem!