EQ filters

WxMan1 · Mar 20, 2009, 01:54 AM

How does one properly set Q for "X"-pass filters ("X" being either high or low)?

I understand that for band-pass filters, F_center = SQR [ F_min * F_max ], i.e., the square root of the product of the -3dB cutoff thresholds for the specified bandwidth. Q, then, is F_center / [ F_max - F_min ], where the denominator is the bandwidth of the filter.

It is my understanding that the "X"-pass filters are of second order type, i.e., implying -12dB/octave attenuation. So what does Q do with respect to "X"-pass filters?

What would be the most effective means to eliminate any signal <50Hz? On the one hand, the 4th order cross-over suggests and inherent -24dB/octave attenuation at the crossover point, but uses significantly more resources than a notch filter. While the notch filter uses as much resource as a high-pass filter, and I'm clear on how to configure Q for a notch filter, I'm unsure whether implementing a high-pass filter instead wouldn't be more appropriate. Perhaps its 6 of one, and 1/2 dozen of either the other two?

Right now I'm running a notch filter with F_center at 20 Hz, and Q = 0.5 so that should get the job done.

FWIW, the whole cause of my consterntion was noticing distortion in the sub-woofer when the EQ 10-band 32Hz slider had a bit too much gain. Digging deep into the rules I discovered the freq. response for my Monsoon MM-700 speaker system runs 50Hz - 20kHz. Drilling deep into the rules for the CT4780 soundcard suggests that the 10k1 DSP has a freq. response between 20Hz - 20kHz. There's an equation that describes power with respect to frequency, not putting the energy the woofer can't handle into the active 3rd-order cross-over would be conducive to minimizing unecessary stress on said components and retain fidelity when taching the system to the red-line (if you know what I mean).

WxMan1 · Jul 23, 2010, 01:30 AM

Fingered out what the deal is is concerning Q for the x-pass filters.

Q defines the resonance of the transfer function at the cutoff frequency.

At Q = 0.75 defines maximally flat response, i.e., -3 dB @ cut-off freq. Q < 0.75 defines maximal flat time delay (i.e., -3 dB progressively beyond cut-off freq), and Q > 0.75 creates resonant peak at cut-off freq (w/associated progressive gain increase beyond cut-off).

The frequency domains defined by all values of Q coincide with roll-off rate attributable to 2nd order filters (12 dB / Oct) at ~0.3 of cuf-off freq.

http://i.cmpnet.com/planetanalog/200...t1-Figure4.gif

The graphic iluminates what I'm describing: Q>0.75 and Q=0.75. The curve for Q<0.75 is not shown, however, the freq response is below the lower curve - the -3 dB level farther to the right - and all three curves coincide at about the same point and roll-off w/the same slope.

Roger that?

Mar 20, 2009, 01:54 AM	#1
WxMan1 HardwareHeaven Lover Join Date: Dec 2007 Posts: 221 Rep Power: 40	EQ filters How does one properly set Q for "X"-pass filters ("X" being either *high* or *low)? I understand that for band-pass filters, F_center* = SQR [ *F_min* * *F_max* ], i.e., the square root of the product of the -3dB cutoff thresholds for the specified bandwidth. Q, then, is F_center / [ F_max - F_min ], where the denominator is the bandwidth of the filter. It is my understanding that the "X"-pass filters are of second order type, i.e., implying -12dB/octave attenuation. So what does Q do with respect to "X"-pass filters? What would be the most effective means to eliminate any signal <50Hz? On the one hand, the 4th order cross-over suggests and inherent -24dB/octave attenuation at the crossover point, but uses significantly more resources than a notch filter. While the notch filter uses as much resource as a high-pass filter, and I'm clear on how to configure Q for a notch filter, I'm unsure whether implementing a high-pass filter instead wouldn't be more appropriate. Perhaps its 6 of one, and 1/2 dozen of either the other two? Right now I'm running a notch filter with F_center at 20 Hz, and Q = 0.5 so that should get the job done. FWIW, the whole cause of my consterntion was noticing distortion in the sub-woofer when the EQ 10-band 32Hz slider had a bit too much gain. Digging deep into the rules I discovered the freq. response for my Monsoon MM-700 speaker system runs 50Hz - 20kHz. Drilling deep into the rules for the CT4780 soundcard suggests that the 10k1 DSP has a freq. response between 20Hz - 20kHz. There's an equation that describes power with respect to frequency, not putting the energy the woofer can't handle into the active 3rd-order cross-over would be conducive to minimizing unecessary stress on said components and retain fidelity when taching the system to the red-line (if you know what I mean). Last edited by WxMan1; Mar 20, 2009 at 02:28 AM.

Jul 23, 2010, 01:30 AM	Thread Starter #2
WxMan1 HardwareHeaven Lover Join Date: Dec 2007 Posts: 221 Rep Power: 40	Re: EQ filters Fingered out what the deal is is concerning Q for the x-pass filters. Q defines the resonance of the transfer function at the cutoff frequency. At Q = 0.75 defines maximally flat response, i.e., -3 dB @ cut-off freq. Q < 0.75 defines maximal flat time delay (i.e., -3 dB progressively beyond cut-off freq), and Q > 0.75 creates resonant peak at cut-off freq (w/associated progressive gain increase beyond cut-off). The frequency domains defined by all values of Q coincide with roll-off rate attributable to 2nd order filters (12 dB / Oct) at ~0.3 of cuf-off freq. http://i.cmpnet.com/planetanalog/200...t1-Figure4.gif The graphic iluminates what I'm describing: Q>0.75 and Q=0.75. The curve for Q<0.75 is not shown, however, the freq response is below the lower curve - the -3 dB level farther to the right - and all three curves coincide at about the same point and roll-off w/the same slope. Roger that? Last edited by WxMan1; Jul 24, 2010 at 02:38 AM.