Noob question : what exactly is the "Q" of an active filter ?

[youyoung21147]

As I am building my high-end opamp preamplifier for my gainclone, I thought I could easily include an active filter in it.

My idea was to have the main output left unfiltered, and to have a second output just for the subwoofer.

I have found on Analog devices website a very nice applet to design active filters, but there is a parameter I don't understand : the "desired Q"

What's this ?? If I change its value, then all the components in the circuit see their value changed ! A little bit problematic when ordering components

I have found other applets on the web, but none of them explains the role of Q.

Help plz !!!

[janneman]

http://www.passivecrossovers.com/index.htm

[sagarverma]

ability of filter to store energy effectively.

[theChris]

Q determines how selective a bandpass filter is. the definition makes the most sense here. Q = Fc/Bw = Carrier Frequency / Bandwidth. thus a 1Mhz BW @ 2.4ghz would need a filter with a Q of 2.4E9 / 1E6 = 2.4E3 = 2400.

of course Q can be used in lowpass and highpass filters as well. High Q stages are more resonant.

The demoniator of lowpass filter will have a "second order term". s^2 + s (wn/Q) + wn^2

wn need not be the -3db point of a given filter. for a butterworth filter it will be -- all poles of a butterwoth filter will be on the circle of radius wn.

i'm not sure how much you know about things like the laplace transform. or how technical you want this to be.

Basically, you can invision Q,F as the coordinates of the paramters of a filter. butterworth fitlers have all these paramters at the same value of F, with different values for Q for every 2 orders of the filter. this a 4th order filter has Q at 1.616 and 0.616 (i think these are the Qs). The Q values are determined by what kind of filter you want.

http://www.ee.siue.edu/~cstahl/e592b.pdf is an example of a project i did for class. a 5th order butterworth filter.

[youyoung21147]

Thanks everyone for the answer !

I have found good tools from Analog and TI.

The 3d order (12dB if I understand well) Bessel filter looks like having the flatest phase overall and the lowest overshoot.

The early roll-off could help compensating baffle losses and decrease in frequency response I think.

Is it a wise choice ? (knowing that it reverses phase, I will just have to use an opamp in inverting mode)


I also have another noob question : if I use the Sallen-key topology, it will be likely that some feedback gets into the input (well that's logical)

Will it affect the other unfiltered opamp input, or will I need to use a buffer opamp before the filter ?

[jcx]

buffering is probably necessary, any source impedance is in series with the input network and changes the filter response

preamps and other sources may vary in their output impedance, 50 - 100 Ohms is common to prevent cable C interaction with the output stage causing oscillation, my SACD player has 600 Ohms series R on the outputs, some tube preamps are rumored to go as high as 5K

[theChris]

Originally posted by youyoung21147
Thanks everyone for the answer !

I have found good tools from Analog and TI.

The 3d order (12dB if I understand well) Bessel filter looks like having the flatest phase overall and the lowest overshoot.

The early roll-off could help compensating baffle losses and decrease in frequency response I think.

Is it a wise choice ? (knowing that it reverses phase, I will just have to use an opamp in inverting mode)


I also have another noob question : if I use the Sallen-key topology, it will be likely that some feedback gets into the input (well that's logical)

Will it affect the other unfiltered opamp input, or will I need to use a buffer opamp before the filter ?

butterworth designs for a "maximally flat" response. the derivitive of the amplitude response as a function of freqeuncy is minimized.

bessel designs for "linear phase" or constant group delay. this ends up meaning a less flat response in the pass band.

chebchevy maximizes the slope from passband to stopband. this means some ripple in the freqeuncy response and a fairly poor phase response.

2nd order filters are -12dB/oct at freqeuncies signfigantly below the cutoff. 3rd are -18db/oct.

look into Linkwitz riley filters which attempt to force the two signals to be in phase and sum to 0 acousitcally.

[youyoung21147]

OK, I see clearer through all this opamp stuff. I've searched around the net for existing designs, and all seem to use many many opamps in series (from 4 to 10,12...)

So maybe minimalist is not possible in active filters !

I think I've found a good compromise for my fullrange+sub :


volume control :->
-> AD797 buffer for fullrange output
-> volume control -> AD743 buffer -> Sallen key NE5534 -> Sallen-key NE5534 -> subwoofer output

the sub would be filtered at 24dB/oct (4th order)

Do you think it's a good idea ?