Not teaching anyone to suck eggs & I'm sure I've explained this better on previous posts in this locale* but it may be helpful to readers.
Re passive filters :
They all depend on the rolloff/response curve. The ones linked above will probably wipe 50% off a VDSL2 connection minimum but provided you aim for a rolloff (-3dB) at about 3MHz then it will have no adverse effect on ADSL of any type. You could aim a bit tighter if required, a rolloff @ 2.5MHz is unlikely to trouble any ADSL2+ connection other than the highest syncing ones.
The other problem is the type of passive filter - in general you get a fast rolloff (slope on graph would be a cliff) with a lot of ripple (not good) or a slow rolloff with a nice flat response curve. You can ameliorate this by increasing the number of poles in the filter which increases cost/size & probably has very limited effect in most instances. Its all a trade-off though.
Re active filter :
First time I came across this was working on Rapier B1MLI upgrades where basically the target radar returns were sorted for doppler shift through a load of notch filters (10kHz IIRC) implemented with op-amps.
Sounds totally insane these days but back then you couldn't do the DSP stuff now possible so it was just better to do this stuff in (analogue) hardware, then digitise it & put it into a 68020 cpu running at 25MHz to do the logic. As an OT point Rapier B2 (would slew vertically onto target automatically) had 32-bit FFT butterfly filters implemented in discrete logic - you've never seen so many 16-pin ICs in your life & the heat generated was incredible (was all TTL, some ECL
). I have huge respect for the man who designed those PECs (panel electronic circuit in milspeak).
Its perfectly possible to build a wee opamp (I loved ua741s
) circuit to cover the ADSL2+ frequencies and have it capable of only knocking out one or two bins if required. Loads of resources online. You'd probably need to have a look how to power it though. Line voltage might do, it isn't going to draw much power but the variations in voltage on POTS are unlikely to be helpful.
Real problem would be multiple noise spikes where I wouldn't be happy cascading the signal though multiple notch filters (another few opamps basically) because of gain & heat issues possibly distorting the ADSL2+ carrier. There's probably a way of sorting that but its getting towards the stage of reinventing the wheel because the road has potholes
I'd be inclined just to tune it as tight as possible either side of the spikes & accept the losses.
*something with asbokid maybe? There was a fairly extensive & technical discussion which will be in my posting history.
Edit - the post above refers to filtering the phoneline, not the mains power supply just in case there's any (probably short-lived) confusion