There are alot of discussions about crossover steepness. But regardless of one prefers steep crossovers or more gradual ones people seems to treat the question the same no matter the intended frequency range. But I wonder if the potential drawbacks with at steep crossover really matter in high enough frequencies?
Outside the critical hearing range it shouldn't matter too much as long as there isn't to much cancelation going on. If there is a 360 degree offset between the drivers, who would notice anything at all at 4000Hz? Or am I missing anything?
And if I am correct in my assumption, at what frequency would you say the potential drawbacks of lets say a 4th order crossover starts to disapear?
Outside the critical hearing range it shouldn't matter too much as long as there isn't to much cancelation going on. If there is a 360 degree offset between the drivers, who would notice anything at all at 4000Hz? Or am I missing anything?
And if I am correct in my assumption, at what frequency would you say the potential drawbacks of lets say a 4th order crossover starts to disapear?
Last edited:
From your post I get the idea that you do not comprehend or understand about the phase and group delay implications of high-order crossovers. Unless you are implementing this crossover with an FIR filter for which you can independently choose the phase response, a high order crossover will disturb the phase response relatively far away from the crossover point as well as cause a significant peak in the group delay at the crossover point. Both of these are generally considered bad, and have been tested as such sonically.
Also, I would not consider a 4th order crossover to be "high order".
The other thing to note is that the order is not so much a problem as is whether the crossover filter contains a pole with a high Q, which is going to cause the peaking group delay. In general, the sharper the rollover in the frequency response from passband to transition band the higher Q will be the pole closest to the crossover frequency and the more the group delay. This is the case within the same "order" of crossover filter.
Also, I would not consider a 4th order crossover to be "high order".
The other thing to note is that the order is not so much a problem as is whether the crossover filter contains a pole with a high Q, which is going to cause the peaking group delay. In general, the sharper the rollover in the frequency response from passband to transition band the higher Q will be the pole closest to the crossover frequency and the more the group delay. This is the case within the same "order" of crossover filter.
Last edited:
Although the human hearing is the most sensitive for SPL around 3-4kHz, this does not necessarily mean that it's the most sensitive range for timing differences.
The best way to experience the effect of timing delay of crossovers is to compare the same DSP with IIR (minimum phase) then with FIR (linear phase) filters with the same slope, imo. Blindly of course.
The best way to experience the effect of timing delay of crossovers is to compare the same DSP with IIR (minimum phase) then with FIR (linear phase) filters with the same slope, imo. Blindly of course.
It depends whether you want to test the audible differences between FIR and IIR in general or just the group-delay properties. If you want to test the audibility of group-dely then you should always test with IIR and add the group-delay distortion by means of allpass filters because IIR and FIR differ in other properties than just group-delay distortion.
Regards
Charles
Hi Charles, my method for trying to test for audible differences from group delay has been to use FIR files that are identical, where the only variable is whether the crossovers are linear or minimum phase (IIR).
(Using complementary LRs)
Do you see any issues with that method?
As for hearing differences in FIR and IIR in general.....
I find this is really tricky because most IIR-only dsps have such a relatively low IIR filter count availability (compared to embedding nearly unlimited IIR filters in FIR).
I've found I have to constrain the FIR file to however many IIR filters the IIR-only dsp can manage, and copy them and the IIR crossovers into FIR.
Doing that, I've never heard a difference between IIR using FIR, and straight IIR.
Make sense?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.