out of band eq filters

[Jay Barracato]

Can we discuss the theory and application?

I am really a minimalist in my approach to system eq. I really liked Jamie Anderson's suggestion of using fairly broad filters to adjust general trends rather than using a bunch for each little detail you measured.

This got me thinking about the effect of each filter on the phase response and the possibility of using out of band filters to adjust phase within the band or at a crossover.

Is this idea on track? Any hints for real world practice? Any papers to sit back and digest?

 

[Bennett Prescott]

Re: out of band eq filters

Jay,

Out of band EQ is to correct the response of the loudspeaker so it behaves better once crossover filters are added (or perhaps off axis). Many loudspeakers (especially subwoofers) have big peaks out of band, and instead of hiding them with a steep XO I EQ them out.

If I want to affect phase out of band I use a different class or slope of XO filter, or else I use an all pass. You have to use some pretty big EQs to have the same effect.

IMHO, the reason not to go after small magnitude aberrations is it wastes a lot of time, they may not actually exist, and they may not have any audible effect. Determining which ones do have an audible effect is why we listen when we measure. Determining which ones exist is why many mic positions can be your friend.

Really cutting edge DSP has very wavy EQ to make the loudspeaker truly flat, but you and I are unlikely to have the time or resources to pull those kinds of tricks. Those are things that should be done by the manufacturer. Putting a loudspeaker in a room or an array you are not going to see any of those, you are going to see large trends especially in the LF. If you need to correct that kind of thing at the loudspeaker level you probably need a different loudspeaker.

 

[John Roberts]

Re: out of band eq filters

By definition an out of band filter is out of band so not very useful for tweaking in band phenomenon. It is not unusual to use all pass filters, phase shift only with no amplitude change, to play games with crossovers and drivers.

For room EQ, indeed the best correction is exact mirror image (actually reciprocal) of the flaw being fixed, On a good (lucky) day the corrective EQ will make everything one with nature again...

It may be worth trying to understand the mechanisms causing the response error. EQ to damp feedback can be very narrow since the nature of feedback is very narrow band, the only m=limit to making it silly narrow, is temp changes in the room could change the speed of sound enough to have to retune the feedback frequency.

So I/3 GEQ are for feedback only, room EQ should be broader, and preferably parametric.

Sorry for no real info to speak of, but you're on the right track with broad gentle EQ for broad gentle problems. Sharp narrow for feedback and the like.

JR

 

[Tim McCulloch]

Re: out of band eq filters

Yes, out of band filters work because they really *aren't* out of the band... at least not completely, and they still affect both the phase and magnitude response of whatever circuit they are a part of.

As Bennett points out, it's often easier to better magnitude response and phase summation through crossover when you begin with reasonably flat magnitude response for a octave on either side of the acoustic crossover.

Harry Brill covers some of this in his SMAART classes.

Have fun, linear transforms.

Tim Mc

 

[Jay Barracato]

Re: out of band eq filters

I may need to back up a bit in my thinking, and it may be that an all pass filter is better suited to what I am thinking, and my thinking may be totally off becuase I think I am picturing this more in the electronic domain when I do most of my thinking in the acoustic domain (hows that for a sentence; I think it parses).

Here is my logic (based on some reading about eq and phase in 6O6's book):

1. Every eq filter has an effect in both the magnitude and phase.
2. The narrower the Q on the filter, the steeper the change in phase is at the center point.
3. The frequency band over which the phase is effected is wider than the band where the magnitude is affected.

So for example, if I were working on a sub/top crossover where I have managed to get the phases close in timing but they still have slightly different slopes, could a eq filter be used to modify the phase of one of the two bands to better match in slope.

Or maybe, I should back up a step and ask if anyone can think of an examples where a eq filter might be used because of the change it makes in phase as opposed to magnitude?

 

[John Roberts]

Re: out of band eq filters

You may be worrying a little too much about phase shift. Ideally EQ that corrects amplitude errors will also move the phase shift to be more correct.

All pass filters in connection with crossovers were often used in simpler times generate some effective delay to compensate for physical driver offsets, back before DSP and simple pure delay was easily available. Making multiple drivers work nicely together is far more complex than room EQ.

Another thought, unless the phase response readings are identical at every room microphone, what do you correct to?

JR

 

[Bennett Prescott]

Re: out of band eq filters

The frequency band over which the phase is effected is wider than the band where the magnitude is affected.

EQ filters, at least the kind you and I are talking about here, are minimum phase. That means that every change in magnitude will create an equal change in phase, i.e. you can take the magnitude response of a minimum phase system and draw precisely what the phase response will be, and vice versa.

So for example, if I were working on a sub/top crossover where I have managed to get the phases close in timing but they still have slightly different slopes, could a eq filter be used to modify the phase of one of the two bands to better match in slope.

The problem is that filters that are not bandpass filters (i.e. parametric and shelf EQ) only create little phase changes. I almost always find that if I need to adjust the phase slope I need to do it by tens of degrees over several octaves. The phase shift that EQ makes is perfect for counteracting errors in the loudspeaker's phase due to magnitude response ripple, but you would have to use pretty big EQ to do the kind of alignment work we're talking about. Usually you're putting a XO filter there anyway, why not use it?

For example, here is the phase and magnitude response of some roughly 1/2 octave parametrics:


And here it is for some "6dB/octave" slope shelving filters:


Here's what some Bessel filters look like (imagine that an all pass looks like the purple trace without the magnitude change):


Use the one that matches the phase shift you need!

 

[Jay Barracato]

Re: out of band eq filters

If you heard a loud pop yesterday that was the light coming on so suddenly my the bulb blew.

I wish I knew where I first was reading about the use of out of band filters to see if I misunderstood it. I thought it was saying the out of band filter was to correct the phase.

Based on what Bennett is saying, the filter corrects the amplitude so you can pick a crossoverfilter that best fits the phase.

It also appears that I was wrong about the width of the effect on the phase for the eq filters. I must have made the ultimate science, mistake of comparing graphs without making sure the axis were the same.

Based on what Bennett posted

 

[Peter Morris]

Re: out of band eq filters

Can we discuss the theory and application?

I am really a minimalist in my approach to system eq. I really liked Jamie Anderson's suggestion of using fairly broad filters to adjust general trends rather than using a bunch for each little detail you measured.

This got me thinking about the effect of each filter on the phase response and the possibility of using out of band filters to adjust phase within the band or at a crossover.

Is this idea on track? Any hints for real world practice? Any papers to sit back and digest?

Very generally, the frequency response of a transducer is directly related to its phase response – flatten the amplitude response and you will have a flat phase response (except for things caused by non- linear behavior such as cone breakup etc.)

If a speaker has a 10 dB resonant peak “out of band”, unless the crossover causes it to be at least 30dB down compared to the in band level, you will still hear it as an artifact. You may not see it in the amplitude response but you will see it the energy time curve (ETC).

If you flatten the amplitude response and get ride of the resonant peak, it will flatten the phase and there will be no out of band artifacts … and no need to modify the phase with an out of band filter.

If you have issues with phase irregularities caused by non-linear behavior within transducer, I would suggest that you really need to start looking at modifying amplitude and phase separately with FIR filters … not a trivial exercise.

Peter

Here is a bit more ...

Your focus seems to be on crossovers and phase – here are a few papers that maybe of interest.

http://www.dolby.com/uploadedFiles/zz-_Shared_Assets/English_PDFs/Professional/DLP_LinearPhaseCrossoversWhitePaper.pdf

http://www.rane.com/note147.html

http://www.excelsior-audio.com/Publications/Crossover/Crossover1.html