cable derating

how far do you allow a cable to run before you will derate the cable.

most CU's i have seen have the cables incoming all bunched and should as far as i can see technicially be derated but i guess you use some common sense here and say if its less than x then dont bother?

have i missed some guidance in the book or am i correct, if so how much do you allow and is this the only case where you dont do it?

There is no guidance as far as I know that says de-rating can be ignored for a small distance, but there is a train of thought that says that the grouping can be ignored for the small distance that cables enter a CU together or pass through joist holes.
It is up to use as the designer to establish what percentage of full capacity the cables are likely to be operating at, for how long and the ambient as well. If you can prove in your calculations that it can be ignored then all well and good.
Personally I would ensure the cables pass into the CU in smaller groups and provide adequate holes in joists.

It would make sense to derate bunched cables because the inner ones in a bundle would be more exposed to thermal effects than the outer ones, right? By being bunched, the reference method for the inner cables is (in essence) different to the outer ones. As an example, if cables are classified under the "clipped direct" method the separate distance between the cables has to be one cable width.

EDIT ---- I forgot that one of the derating factors is grouping (Cg) so ignore my comments above about changing the reference method.

As ricicle states, the onus is on you to prove why you did or didn't do something - you are the competent person after all. There must be a cable length under which you it would be pointless to apply any derating factors (because the overall numbers are too small) but I'm guessing this would probably be 1m or under.

Hope some of that was of benefit anyway - it is rather late

Neil