2 ms·

Naively I would upsample by 4-8 by 0 stuffing and low pass filtering and then interpolating. That can't be that bad, can it?