diyAudio (
-   Multi-Way (
-   -   Active Crossover roll-off (

ajg112 16th January 2013 01:44 PM

Active Crossover roll-off
Hi All,

I'm building a pair of active 2-way loudspeakers. I'm designing a crossover circuit and am wondering about the filter roll-off.

I'm considering using unity-gain sallen-key 2nd order filters; 1 for HF, 1 for LF. The crossover freq is 2.5kHz, this was derived using the excellent crossover guide.

Is there any advantage in using higher order filters? The roll-off will be steeper, but my gut feeling is the transition between mid-range and tweeter my become too abrupt...

Any advice gratefully received!

Pano 16th January 2013 02:01 PM

You won't get a single answer to this question, you'll get endless debate. Just be warned and be ready for it. :D

For the speakers I build and I like, either Bessel or Linkwitz-Riley acoustic slope works best. I use whatever filter function is needed to achieve the acoustic target. It's what sounds good to me. Other people have other opinions and ideas. Some love steep filters, others hate them. You should get both opinions in this thread.

For me it's the final acoustic slope that matters, not the electrical. And for me, that's 4th order acoustic.

ajg112 16th January 2013 05:37 PM

Pano, many thanks for your honest reply.

I'm an experienced electronics engineer, but unfortunately, have no 'acoustics' knowledge/experience. I suspect this will be the start of a long journey in trying different things to see what 'sounds' best for me. However, anything/everything has got to sound better than listening to CDs through the TV!!!

At this stage I guess I'm just looking out for answers like "Don't do that, it is really silly because everyone who has tried it knows it sounds rubbish" etc...


sreten 16th January 2013 08:17 PM


Active filters are no shortcut to good speaker design, and moreorless
guarantee beginners will oversimplify the correct acoustic decisions,
as well as probably not balance the amplification for an active 2-way.

Take this (good) design : Zaph|Audio - ZMV5 - MCM / Vifa 5" System

These are near the filter transfer functions you'd need for active :

(You wouldn't include the wobble at low frequencies.)

You also wouldn't include the hefty treble attenuation meaning
for this design the treble amplifier could be tiny compared to
the amplifier for the bass/mid - say <1/10 the power output.

Unity gain Sallen and Key filter topology is not the way to do it,
at least for the bassmid, you need a topology that can include
BSC, and that implies manipulation of the filter in an Emulator.

rgds, sreten.

Active filters can be built use the power amplifiers as the "op-amps".

Tordenguden 16th January 2013 10:08 PM

I'm using fully active diy speakers, though with a dsp. But my experience is that sound can change alot, when tweaking the filter - in my case clicking a few times with the mouse.
I found the sound best to my liking, when I tweaked the upper roll-off on the midrange to look very alike the lower roll-off on the tweeter - just like the picture that >sreten< posted.
I also get a very flat response when measuring my speakers with this filter-setup.
I use 24db linkwitz riley for most x-overs. But use steeper filtering for my sub-bas to control some unwanted resonances - that might could have been avoided, by building a better cabinet - but it works great with the filtering, so I dont bother.
An advice could be to look at the natural acoustic roll-off at the drivers you have choosen. Preferably measure them your self - because the frequency curves given at the data-sheet, is often measured in a way bigger baffle than the one normal speaker cabinets have - and this has big effects on the behaviour of a driver.
If a driver has a natural roll-off at say 3Khz. You have to take this into consideration, since it also changes the shape of your roll-off curve, when filters are added. And both the shape and the steep-ness has to match on both sides(mid and tweet), for best frequency repsonse flatness.
Hope it helped :) And yes be aware, there are alot of different opinions about this subject. I try to keep close to the basics - but even basics can be discussed alot :D

Tordenguden 16th January 2013 10:31 PM

Better answer your question a little better :o
A steeper filter can move unwanted "problems" away. Lets say you have a midrange like accuton or metal driver from seas. They have some breakup node that are not very nice. Normaly a combination of low x-over and steep filters can help to move the breakup down below a level where it wont be audible. I generaly hear about less steep filters when using well behaved drivers and passive filters, where you normaly try to keep the number of filter-components to a minimum. My reason for using 24db linkwitz riley, is beacuse it helps to get the phase rigth. I use the filter options in my dsp to smooth out the breakup in my midrange. But I'm still experimenting with different filter versions - no expert - still learning :)

ajg112 17th January 2013 09:01 PM

Many thanks all for your responses. I can see that this is no easy task...

sreten, please could you explain your reasons why a sallen-key based filter is not the way to go. Is this just due to baffle step correction, or are there other reasons?

As this is a learning exercise and the design is very modular, I'll probably go with basic crossover I have, get something 'working', then start on the improvements/changes. This way it'll be easier to 'hear' any improvements.

I'm using these drivers; Visaton 'W 100 S - 8 Ohm' and a 'G 20 SC - 8 Ohm' both from Visaton.

Here are their frequency responses (Tordenguden thanks for your warning with regards to their accuracy in 'real-life' use)


sreten 17th January 2013 11:27 PM


There are a number of active filter topologies and Sallen-Key
is just one of them, good for unity gain classic filter functions,
but not good for messing around with EQ and the like as well.

e.g. using a power amplifier as the "op-amp" : you can use
S-K unity gain circuit but using the inverting input as the unity
gain point, not the output, as long as that driving point has
low enough impedance. You can then manipulate the overall
response in the standard feedback network for BSC etc.

Understanding your possible choices is the key to good design.

rgds, sreten.

dumptruck 17th January 2013 11:47 PM

If you're using Visaton, you might as well play around with Boxsim a bit, as it comes with real, full measured data for their drivers ready to import and simulate. The active xover simulator is not very helpful, but you might find it useful to mess around with passive circuits, and to be able to export various data for use in other simulation software. Enclosure/baffle simulations also, of course.

Don't forget that you'll have to estimate phase shift from acoustic offset, assuming your drivers aren't going to be time-aligned, and you're not going to be determining the offset with measurements.

Tordenguden 18th January 2013 08:45 AM


Originally Posted by dumptruck (
Don't forget that you'll have to estimate phase shift from acoustic offset, assuming your drivers aren't going to be time-aligned, and you're not going to be determining the offset with measurements.

You can actually measure the deleay, with a ruler. Theres is a nice and simple explanation right here:
Loudspeaker time alignment - Wikipedia, the free encyclopedia
The problem with simply tilting the speaker backwards, is that you have to take the off-axis frequency response into account as well. Easier to just make a thicker baffle for the midrange. Normally the tilt is only needed to be around 7 degrees, så off-axis response might be well within normal specs - but again try it ;)
If there are the same dealy in the filter for both the tweeter and midrange, you can then phase the midrange 180 degrees(switch + and -) then you measure until you get the biggest possible suck-out, right in the x-over point.
Check out Troels' page, he has alot of great tricks to show off.
Scroll down, where he measures with inverted polarity on the midrange - look for the yellow line, it shows how well he made the midrange work with the tweeter:

All times are GMT. The time now is 08:03 AM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio