I'll deny the part about oversampling. The filter "core" only really sees a normalized frequency, and oversampling will make the low normalized frequency even lower. This is make the problem worse!thevinn wrote: IIR filters with low cutoffs definitely start getting wacky especially as the order increases. I think the typical solution is "oversampling".
Can anyone confirm or deny this?
Anyway, this problem is dependent on the filter structure. Direct forms are probably some of the most sensitive (of the commonly used structures anyway). Like I said in the previous post, I've been meaning to write a tutorial on how to take an arbitrary set of direct form coeffs and map it to arbitrary filter structure (it's not exactly very hard in most cases).
Basic "general" outline is to define the filter in state-space form and take the inverse of (zI-A) and then multiply with input/output vectors, add direct path, and solve for whatever unknowns you have so that the response matches. In some cases (uum, ladders/lattices) there are also published formulas to do it directly (though the state-space approach works too).