Linearteam WinISD Pro
WinISD pro supports behavioral simulation of various filter types, such as:
The behavioral simulation means, that WinISD simulates filter as "black box", it does not consider what black contains internally. Filter can be passive or active, as long it does what is simulated.
Any combination of cascaded filters can be simulated. Filters are connected logically in series. Individual filters can be turned on and off by checking/unchecking the checkbox in the listbox. Graphs are updated immediately, so effect can be investigated.
Filter system is logically located at electrical side. 0 dB gain at filter chain means that voltage at driver terminal is equal that is specified at "signal"-tab. If filter/EQ chain gain is 6 dB at particular frequency, then voltage at that frequency is twice that is specified, and so on.
The following lowpass filter subtypes can be simulated:
Butterworth
Butterworth is probably most common filter type. It is so called "maximally flat" filter.
This means that amplitude response stays practically flat until fc frequency is reached,
where it is -3 dB down. Response then falls rapidly according the filter order.
Linkwitz-Riley
Linkwitz-Riley is filter type, where two second order butterworth filters are cascaded as one
fourth order filter. This means that frequency response magnitude is -6 dB down at fc. It is
different from fourth order butterworth, where magnitude response is -3 dB down. Transient
response is better than butterworth, but still worse than bessel.
Bessel
Bessel filter is fundamentally different from two above. Whereas butterworth is built on
amplitude response approximation, the Bessel filter is maximally flat group delay
approximation filter. fc is frequency, where the phase is half of the maximum phase shift (90
degrees times the filter order). This filter exhibits best transient properties of analog
filters, but roll-off starts rather slowly.
Second order section
This filter type is for advanced users. Second order section can be considered as "typeless"
second order lowpass. Parameters are fc and Q. fc determines the frequency, where the filter
gain is Q or 20*lg(Q) dB. This can be used to compose new "all-pole"-filters, by adding
several sections in cascade.
Highpass type contains same subtypes as lowpass filter. See explanation above. Only difference is that cut-off is below fc, not above, as is logical.
Order determines how rapidly the response magnitude decreases after cut-off point. Greater the order, steeper the roll-off. This has it's price, however. Faster you make it roll-off, worse the group delay behaviour. 1st order filter has roll-off of 6 dB per octave, 2nd order filter 12 dB per octave, and so on. Realization of high-order filters is very difficult to achieve, due to high tolerance requirements. Generally, anything over 4th order should be considered very carefully.
fc Cut-off frequency or critical frequency. For butterworth class of filters, this is the -3 dB point.
Q is only used with Second order section-type. This makes possible for the user to experiment with filters with peaky response (however, this is seldom done with lowpass filters). Q is the quality factor of the filter. It determines the shape of frequency response near the fc frequency. One physical meaning for Q is that gain at frequency fc is Q (in linear scale, not dB).
This type of filter is commonly used as subwoofer equalizer. It suits quite well to ported or passive radiator type boxes. Because tuning frequency of the box is quite near the roll-off, response can be boosted several dB's before cut-off, extending the frequency response. After the peaking, response starts to roll off quickly. This is convenient, since it serves as an subsonic filter also. However, all this comes with a price tag: it introduces quite high amount of delay distortion depending of the amount of peaking.
Peak freq is frequency of maximum equalization.
Peak mag is amount of equalization in dB you want to apply at peak freq.
This is so called phase correction filter. It alters only phase, gain is unity at all frequencies. Only first and second order allpass filters are implemented.
Order determines how fast the delay will reduce, after cut-off point. And of course, the practical implementation is quite different.
Q or the quality factor. This is relevant only with second order all-pass. Higher the Q, higher the peak in group delay.
Delay time is desired signal delay time. Delay approaches this value asymptotically when frequency approaches zero. Higher you set this value, earlier it rolls off.
This type of filter is implemented in commonly available equalizers. They might come as
freely adjustable, where center frequency, boost/cut and quality factor are all adjustable
(they are called as parametric equalizer), or partially adjustable, where one parameter is
fixed (semi-parametric equalizer) or only boost/cut amount is adjustable (they are called
graphic equalizers).
You can boost or cut desired band.
Center freq is frequency of maximum boost or cut.
Q controls bandwidth of the boost/cut. Higher the Q, narrower the equalization is.
Gain adjusts amount of the boost/cut. Use negative values for cut, positive for boost.
This is an elegant way to equalize closed box subwoofer.The good thing about this is that it introduces no excess phase shift, which would be inevitable introduced if for example, parametric equalizer would be used. In practice, Linkwitz transform makes closed box look like the specified parameters. So called S-plane representation would be nice way to show how this thing works, but it is too difficult concept to grasp without engineering background.
f0 is fsc of original closed box system. WinISD automagically fills this field, if adding to closed box system.
Q0 is Qtc of original closed box system. This is also filled automagically, like f0.
fp is fsc of the transformed system. This can be rather freely specified, of course one should consider the physical limits.
Qp is Qtc of the transformed system. This is also up to user. General guideline is to keep this below 0.707, or otherwise peaking will occur.
this "filter" just adds described amount of dB to the eq/filter response curve. It can be used to normalize the response, to user preference. For example, there has been a debate whether the linkwitz transform should be normalized at high or low frequencies. With this, it could be either.
There are three purely EQ/Filter-related graphs in WinISD:
These are same curves than system curves, but calculated just for the filter part.