I have a few questions that I am hoping someone with more experience can help me with. I am building a two-way system and would like to use a digital crossover for flexibility to change filter characteristics (such as a minidsp nanodigi). The crossover will be between 300-500Hz. Here are my questions:
1. Is there a preference between using steep filters vs blending HF/LF drivers together with low order crossovers? Or is there a situation when using steep filters is very beneficial. I believe an IIR filter is an equivalent biquad implementation. Is there a specific filter IIR type (e.g. butterworth, linkwitz-riley, etc) that people use most and why?
2. I have been reading a bit on FIR vs IIR filters. The main difference I have focused on being discussions on the benefit of linear phase (although I have read very little on the subjective differences of IIR vs FIR filters. From what I've gleaned (assuming you have the DSP horsepower) FIR filters can be very steep and linear phase but can have pre/post ringing. It appears to me that a lot of the room correction software does the same thing (please correct me if I am wrong) but they are making adjustments to phase, amplitude, time, etc at the listening position taking into account reflections and room interactions. Obviously, the microphone is also picking up an crossover discontinuities. Lets assume that I am going to use a room eq product such as Dirac Live. Would there be any benefit in building an FIR filter for my two-way vs an IIR filter when correction is going to be made anyway by a room eq product.
Thank you very much. Any input would be much appreciated.
Regards,
gwho
1. Is there a preference between using steep filters vs blending HF/LF drivers together with low order crossovers? Or is there a situation when using steep filters is very beneficial. I believe an IIR filter is an equivalent biquad implementation. Is there a specific filter IIR type (e.g. butterworth, linkwitz-riley, etc) that people use most and why?
2. I have been reading a bit on FIR vs IIR filters. The main difference I have focused on being discussions on the benefit of linear phase (although I have read very little on the subjective differences of IIR vs FIR filters. From what I've gleaned (assuming you have the DSP horsepower) FIR filters can be very steep and linear phase but can have pre/post ringing. It appears to me that a lot of the room correction software does the same thing (please correct me if I am wrong) but they are making adjustments to phase, amplitude, time, etc at the listening position taking into account reflections and room interactions. Obviously, the microphone is also picking up an crossover discontinuities. Lets assume that I am going to use a room eq product such as Dirac Live. Would there be any benefit in building an FIR filter for my two-way vs an IIR filter when correction is going to be made anyway by a room eq product.
Thank you very much. Any input would be much appreciated.
Regards,
gwho
If your crossover point is as low as 3-500Hz, and your driver-driver c-c distance is reasonably low, then I'd say you're fine with anything above 2nd order - by acoustic reasons.
One thing to consider here though is the absolute power/displacement demands you put on the higher range driver. 2nd order highpass means constant displacement, i.e that the driver will move exactly as far in/out at 30Hz as at 300Hz. This does induce some modulation (I know, that's a question contended, but let's at least say it moves the driver out from ideal working position without adding anything of value to the sound power).
One thing I noticed very clearly from a purely non-scientific PoV was that my setup (three-way digital) gained the most from phase-matching the drivers in the x-over regions. That is; making sure that you have perfect addition at all points. For me, that meant using crossvolver - one FIR impulse correction per working band. I did this to ensure both that I didn't needlessly overwork the drivers (making drivers work out-of-phase is the same thing as forcing them to work harder without actually making more sound), and to minimize lobing.
This had a much bigger impact on the way I perceive the long-term sound of my system than any other configuration with digital filters before that.
I used a combination of automated and manual tweaking of the impulses by specifying the responses I wanted in DRC, and then tweaking the measured results of that in rePhase.
Another thing to note regarding phase:
There's mainly two ways to do that. The first is that you use "normal" linear phase response for the x-over filter and make sure that you have good phase matching there (2nd order Butterworth acoustic response or any other well integrated model), and then making the phase more linear as an added (or actually prepended...) step.
The other is to make the responses for each band as linear as you want it directly, but that does however necessitate having one FIR filter per band.
My system is entirely HTPC-based, so unfortuantely I have no experience with the nano-digi :/
theSuede,
Thanks for the response it definitely helps my understanding seeing what other people are doing. I guess I am a little lost in my understanding. If I created a FIR linear phase crossover HP and LP would not the filters be linear phase in relation to each other and over the entire audible band. Are you saying that using crossvolver you are correcting the HP and LP separately with separate corrections? Is this way you are implementing a more efficient way to do it.
I have seen some writeups on using IIR filter and then correcting phase. I think this would require the minisharc and the minidigi together or I could implement FIR filters with the minisharc alone.
Can I ask you what program you are using on a PC that does crossovers? Does it also do room eq? Also, are you saying that room eq does not correct for the non-linear phase if I used a IIR filter. Do you output through a multichannel dac or do you output digital signals for each crossover to external dacs.
Best Regards,
gwho
Thanks for the response it definitely helps my understanding seeing what other people are doing. I guess I am a little lost in my understanding. If I created a FIR linear phase crossover HP and LP would not the filters be linear phase in relation to each other and over the entire audible band. Are you saying that using crossvolver you are correcting the HP and LP separately with separate corrections? Is this way you are implementing a more efficient way to do it.
I have seen some writeups on using IIR filter and then correcting phase. I think this would require the minisharc and the minidigi together or I could implement FIR filters with the minisharc alone.
Can I ask you what program you are using on a PC that does crossovers? Does it also do room eq? Also, are you saying that room eq does not correct for the non-linear phase if I used a IIR filter. Do you output through a multichannel dac or do you output digital signals for each crossover to external dacs.
Best Regards,
gwho
I am not a real fan of room correction, but am a fan of doing crossovers before amps. The best reason for steep slopes is to restrict the bandwidth the drivers "see" which makes it easier for the drivers to perform well. Gradual crossovers allow drivers to attempt to follow well outside the "bandpass", and can cause distortion and/or overload.
If you have a spare PC, you can do your crossovers there, using free software.
If you have a spare PC, you can do your crossovers there, using free software.
jplesset,
Thank you for the response. I see your point on limiting the frequency response of a speaker with a stepper filter.
I have not heard a lot of room eq solutions myself but I did try a Dspeaker dual core in my room and it was a substantial improvement as I had substantial room modes at 50hz, 100hz etc...
Regards,
gwho
Thank you for the response. I see your point on limiting the frequency response of a speaker with a stepper filter.
I have not heard a lot of room eq solutions myself but I did try a Dspeaker dual core in my room and it was a substantial improvement as I had substantial room modes at 50hz, 100hz etc...
Regards,
gwho
Yes, the FILTERS would be perfectly in phase through the x-over region. That's not a guarantee that the summed output (including driver shifts...) will be in phase...
That works, but takes quite a bit of tinkering and fine-tuning to get the phase summation right. You have a lot less control when doing the main filtering in IIR. But you can of course build using just phase as a reference, and then equalize before sending the signal into the network.
There isn't really any significant (theoretical) difference between using:
Input >> All FIR (magnitude and phase correction, room correction) >> outputs
Input >> room correction/equalizer >> IIR x-over >> outputs
Input >> room correction >> output >> passive filters
-As long as the drivers are x-overed to get correct phase summation...
I think most of the more advanced modules give you the choice in which order (and how) you implement the filters.
What room EQ cannot do is to adjust how your drivers sum together. Room EQ (when done as one global correction) works with the sum output (or rather - full-range input) of the system. It doesn't change how the drivers add together, it tries to correct the result of that addition.
For instance - if you inverse the phase of the high-range, and get a deep cancellation null at the x-over point, room correction will try to add lots and lots of power to that null to correct it up to reference levels. You could damage your drivers long before you reach normal listening levels by doing that. You will also build large lobing peaks (frequency bands are sent out in a certain direction, but not not others...) when you amplify two out-of-phase drivers. I guess this is what made the difference for me - making sure the system adds up correctly before adding global room correction makes sure that the radiation pattern out from the speaker into the room is as uniform as your base construction allows for.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.