Best way to design any crossover is build the box, install drivers and THEN measure with a calibrated mic.

Then SIMULATION of crossover designs using these curves will get you 95% there seeing as off axis plots are very important. A flat on axis response only can still sound like utter garbage that can't be eqd out.

This of course needs to be listened to as a flat curve may still have nasty sound thanks to drivers etc.

Using this method I turned basic plastic passive speakers into very high performance passive boxes that have no brand name etc and are very low key(and cheap by comparison)

Edited by LunchieTey - 26 March 2017 at 10:51am]]>

Best way is to look them up in a reference design table as a modification to the standard formulae you mentioned in the first post.

If you want to know where they come from then draw out the transfer function of a series LC lowpass filter, you will get a result with a polynomial on the denominator. This structure is what all more complicated filters are broken down into, because it is the basic filter building block. Ok a 1st order is really the simplest, but with one pole only you can't manipulate it as much.

Once you have the polynomial (coefficient of the s^2 term should be one to be in standard form). The coefficient of the linear term can then give you the Q factor of the filter. So if you want a certain Q and Wo, you reverse engineer from this point to get your L and C. Or you can just look them up as you did!

More reading here, Matlab (or Scilab=free) is very handy for evaluating filter designs and generating nice visual plots. Python is probably quite good too, not used it myself for this purpose.

Edit - Forgot link

Edited by odc04r - 13 March 2017 at 10:32am]]>

thanks Mik, will show you what I come up with. Hopefully model once, build once ;)]]>

Phil

bench with many components here still if you want to try things empirically!

Mik]]>

won't pretend I understand why but the Bessel values are a factor of root3 different to the LR values - just verified this by tapping numbers into the online calculator. so for normalised 2nd order bessel:

L=root3.Z/2.Pi.f

C=1/2.root3.Pi.Z.f

okay the 2nd order LR components are the same value as the first order butterworth components. obvious when you think about it as LR is two cascaded Butterworth filters. Still stumped on the Bessel values.]]>

guys, I don't want to start a debate about crossover design. I'm actually quite into it and am trying to set up some calculations for filter components loaded with different zobel networks. I need the formulas not online calculators. does anyone know them or should I start trying to work them out from the transfer functions?

btw all the crossover i've designed in akabak have come out so near to what I expected that it wasn't worth tweaking them.

]]>
Well they will get you into the back row of seats if not the actual ball park...

When I worked for a major speaker manufacturer (long time ago) we would design crossovers using an iterative process. After examining the frequency response and impedance curves of the drivers (and maybe building zobel networks), the starting point would be a "calculated" set of filters built with component substitution boxes.

Then of course, lots of measurements - using either B&K chart recorders or the original TEF analyser (I said it was a long time ago...)

The substitution boxes made it easy to change components just by setting switches, so although it involved a lot of adjustments it was not difficult, just very time-consuming. But I used to get through an awful lot of chart paper...

Edited by madboffin - 09 March 2017 at 12:13am]]>

No they wont get you even close to the ball park.

Crossovers are made with acoustical measurements, not by calculators.

]]>
I'm sure I have seen some online calculators that will give the theoretical filter component values. But the only use for a calculated crossover is as a starting point to get you in the ballpark before doing a proper design exercise - for the reason DJK gives in the previous message.

The component values will need a lot of tweaking before they work properly with real drive units.

Edited by madboffin - 08 March 2017 at 11:34pm]]>