Crossover design - TOTALLY lost...

[WMcD]

Well, I dunno. My spelling is off sometimes. I have not found anyone here obnoxious; and least of all the named parties.. Maybe it shows I'm in the same boat of obnoxious mis-spellers. If so, it is a good crew and I'm pleased to be part of it.

= = = =

In my view the nay sayers are a bit too negative. You can learn the basics from reading the cook book stuff.

EE classis certainly help. They teach circuit analysis without giving much info on the real world. Unfortunately it takes a lot of reading before the geshtalt hits and you can see that many descriptions describe the same things in different ways.

One good insght is the "power factor". When voltage and phase are not matched, there is a failure to transfer power. Crossovers are filters and they create this. That is how they work.

Another is that when acoustic power output of drivers (or anything) are out of phase, for any reason, they don't add up properly.

= = = =

There are known problems in real world systems. I believe they can be broken down with some sanity.

1) Much of what you read about crossover design assumes that the speaker load at the terminals is a perfect resistance. I.e. the filter is "seeing" a perfect load. (Big Problem #1) Even if it is a perfect load, what is its value. If, per what others say there is an inductance at high frequencies, how do we gauge it.

2) It also assumes that any signal put into the electrical terminals come out of the speaker the same way in sound pressure. (Big Problem #2). The speaker is already rolling off. If we have a response target caused by the filter, and the speaker is already doing that without the filter, the target shifts.

3) It also assumes the acousitic centers of the drivers are at the same distance from the listener (Big Problem #3). This is the relation between time and phase of just the acoustics of driver offset.

= = = =

All the above show that classic designs do not account for known problems. So people throw up their hands and say it is all none sense.

None the less, very sophisticated desgn systems do exist, such as the LMS/LEAP system. It is not cheap or easy. It integrates the driver input impedeance and acoustic output, plus the driver distance, and interreacitons of the bass, mid, and tweeter, and filters. Once there is a prototype, info has to be rechecked, and tweeked.

= = =

Does this mean you can't do things at home and do it well? I'm not sure. PWK used just a first order with caps and inductors. .

Does this mean once we have a decent design, does it have to be perfect? I hope not.

I set up a bass horn and horn treble unit. Theory says that the blend at the crossover point is better if you reverse the polarity to one of them. Otherwise, there is a dip. Making measurements that was indeed the case. There is a dip when the the things are fed in phase, and none when they are fed in phase.

None the less, I get the mental heeeby geebies about getting eveything else out of phase between them just to worship a little glitch.

That is what I think.

Best,

Gil

.

[kenratboy]

Learning experience.

I am not trying to make new speakers for myself. I have Chorus

II's and a nice Synergy system in my room - I would not be able to MAKE

anything in my lifetime to top this stuff!

[Al Klappenberger]

kenratboy,

John Albright started another thread that suggested that the place to start the learning process is with a simple 1st order netwrok. I think he's right. It's pretty had to screw up the design on a 1st order filter!

Dean,

Those links you posted look like an explanation of how PWK did the AA tweeter filter (Constant K). Look at it as a history lesson and not a good way to actually design higher order filters. I know becasue I tried that method early on. It doesn't work worth a crap. This includes the "m-derived" methods. The right way is using the tables of lowpass prototypes, or "normalized lowpass filters". These are actual filters tabulated for every order for a cutoff frequency of 1 radian / second (1 / (2 X Pi) = .159 Hertz) and 1 Ohm impedance. The values are in Farads and Henrys. These can be used to "scale" virtually any kind of filter to high orders. This includes lowpass, highpass, notch and all sorts of bandpass filters. You just need to know how. If your interested, I can scan a typical table of filters and post it. I got lots of them here. I don't use them now becasue my filter program calculates all of them, on demand, as I need them.

BTW: I also have a small "exe" program that calculates these normalized lowpass filters and displays them for any type, Butterworth or Chebyshev. I think the attachment feature of the board would allow me to make it downloadable in exe or zip form. I'll upload it if anybody wants it.

Al K.

[WMcD]

I don't disagree with Al, of course. However, we should look for simple things behind inponderable equations. The issue of scaling normalized values gets too difficult.

One thing is that we have to know the reactance of inductors and caps.

Xl = 2 pi F H

Xc = 1/(2 pi F C).

Do not worry about the j operator you see so often..

The load is just R. Granted the driver doesn't have a pure R and it may be difficult to appoximate the real one from spec sheet.

None the less. A first order filter, high pass and low pass, works to a cossover pont where

R = Xc

R= Xl

You have to back calculate things. You pick an F and an R. Then the H and C can be calculated with high school algebra.

A second order circuit is not too much more difficult. The values Xl and Xc are 1.414 (square root of two) time R.

I'd like Al to check me on this..

Gil

[Al Klappenberger]

Guys,

Gil has it correct as usual. For a first order network, if used the normalized table method, the normalized tables would show the value is simply 1! Divide it by 2 X Pi X frequency. Then for an inductor you multiply that number by the filter impedance. For a cap, you divide it by the filter impedance.

Example: A 6000 Hz tweeter filter for an 8 Ohms tweeter:

1 /(2 X Pi X F) = 1 (6.28 X 6000) = 2.2653X10^-5

2.653X10^-5 / 8 = 3.316X10^-6 = 3.316 uF.

One other detail: The normalized numbers from the tables need to be inverted (1/G) for a highpass filter. In the case of a first order filter, the value is 1, so 1 / 1 is still 1.

In the case of A 2nd order filter, the normalized values are 1.414 Henrys and .7071 Farads,

Inverting them you get .707 and 1.414. Then you scale them out to get your filter! It's actually very easy. It's like instant powered filters. You just add water!

Al K.

[hwatkins]

You guys are too much fun!!!

This is another thread to add to my favorites. Kerant - there is good information throughout this and thanks to all for posting the links.

Look past the arguments and you will find good stuff on both sides (if you read closely I think there are four sides, but that is another thread). I, for one, do not think you will find a 'Holy Grail' of crossover that serves a universal purpose. In all but my first crossover modification attempt (which by pure luck worked very well - it was a bit like catching the River in Holdem) I have chosen the drivers I wanted based on performance, sound and frequency response and built the boxes first. I then have them professionally measured one driver at a time before I start thinking of what I want my crossovers to accomplish. Most often the decision is based on how dramatic a slope I want when blending two drivers.

That said, I always vette my calculations through knowledgable folks and I have not, as yet, had any of my designs go into service without someone that knows much more than me managing the final tweaks. I find that whilst these folks like to argue they tend to take their passion to helping folks like you and I - the uneducated masses.

Good luck.

[sfogg]

Henry,

" I, for one, do not think you will find a 'Holy Grail' of crossover that serves a universal purpose."

IMO the closest to that would likely be an active line level unit.... probably DSP based. That way its crossover points and slope have a huge amount of options where it can really be tailored to the requirements of the system.

Shawn

[Deang]

Al, that's how all the online calculators work.

Fine for two-ways. However, when you go to three-way designs the reactive component plays a greater role and most of the calculators don't factor it in.

I'm mathematically declined. Here are some of my favorite calculators:

http://www.kbapps.com/audio/speakerdesign/calculators/Default.htm

http://www.pvconsultants.com/audio/targetgen/pcdc.htm

http://ourworld.compuserve.com/homepages/Bill_Bowden/ohmslaw.htm

http://www.1728.com/resistrs.htm

http://www.kusashi.com/calculators.php

Sites or articles that helped me more than others:

http://www.silcom.com/~aludwig/contents.htm

http://www.bcae1.com/

http://www.peavey.com/support/technotes/concepts/impedance.cfm

http://www.hal-pc.org/~bwhitejr/

http://www.thielesmall.com/database.asp

[filmboydoug]

How does one design a crossover for a horn tweet and mid? winisd, xover pro, etc dont list horns in their database, and the manufacturers dont list much for specs on their websites.

I asked this question on another forum and was told this was due to erratic impedance curves. Was further told to use a zobel, then try textbook crossovers and go from there.[8-)]

[hwatkins]

I asked this question on another forum and was told this was due to erratic impedance curves. Was further told to use a zobel, then try textbook crossovers and go from there.[8-)]

I have built a zobel filter before and it can have some value - especially as a cheap stand alone (mostly standalone - impedence matching may require some sort of padding) way to make a reasonable Xover on a two way where the woofer has a pretty high extension in the higher frequencies. In my experience, to use a zobel you need a couple of speakers that are pretty darn close on where they start and stop. Also, a zobel is most effective on the lower freq driver - it manages to give you a flatter impedence as you get to the higher frequencies thus allowing those higher freqs to be reproduced without such a dramatic fall off. In many respects my dirty little filter solution is a 'crossoverless' design and can be done quite pleasingly if you spend time on getting the right drivers.

However, a zobel solution by itself is a very limited thing. In a real crossover solution you may use it as part of the network, but most folks get drivers to match in a way that they already work strong in a 1/2 to whole octave above/below their chosen crossover point. This limits the overall value of a zobel focused solution.

As to the horn stuff, I suggest you go here:

http://www.pvconsultants.com/audio/frdgroup.htm

And take a look at some of the calculation programs and see if something can help. I thought they had hornres, but my first glance doesn't see it. To design your crossover you will need to find out how the horn shape influences the response. That is probably why folks get into bunches of mumbo jumbo when talking about the networks for two horns.

good luck

[CONVERGENCE]

Hi,

The engineer is right ,it's more of an art than a science.

The easy way out is to go active instead of passive.

Theres a group of technicians that have designed upgradre passive XO for klipsch.

Here is their web site.

http://alkeng.com/es_xo.html

........................................................

[kenratboy]

Now could one buy a passive X-over, use it to get a desired sound, then turn the data over to the proper person to then make a passive X-over?

[Wrench722]

Try this. Howstuffworks "How Speakers Work"

[CONVERGENCE]

Hi,

The anwser to your question is yes. Anny good sound technician can draw a schematic on a sheet of paper using an existing passive XO.

The problem with this forum and I don't want to sound condescending. There are not enough professionals implicated in these discussions.

[Elso Kwak]

Hi, I designed a new crossover for the K-horn which is totally different from Klipsch or ALK.

I like the sound of it; it' s optimised by ear. No auto-transformer, no audiophile parts! A lot of parts though.

You can get the schematic by sending me an email.

Edit: I also use EV 1824M and T350 instead of original drivers.

[ironwoods]

Hi, The problem with this forum and I don't want to sound condescending. There are not enough professionals implicated in these discussions.

And where might one go to seek this professional help? [:|]

Al; aka-(http://alkeng.com/es_xo.html) has been participating in this disscussion.

I think Ken may have miss-worded the question, as I'm thinking he wants to know if a passive filter could be derived from the settings of an active one. In this case, I doubt the answer to be a blunt "yes".

[Al Klappenberger]

"Hi, The problem with this forum and I don't want to sound condescending. There are not enough professionals implicated in these discussions."

This is very true. The technical people at Klipsch do not participate here AT ALL! I believe they should.

AL K.

[seti]

"Hi, The problem with this forum and I don't want to sound condescending. There are not enough professionals implicated in these discussions."

This is very true. The technical people at Klipsch do not participate here AT ALL! I believe they should.

AL K.

I've often wondered why we don't see more Klipsch employees on here. These discussions are over my head but I can see where it would be mutually benefitial for the tech people and forum members not to mention good PR.

[kenratboy]

If anything, Klipsch could step in and say:

"Designing a crossover is a vastly difficult, time consuming, and

technical exercise - dozens of people spend huge amounts of time

designing, testing, and listening to various designs. It is not

uncommon to spend many weeks or months on the process."

See, not only do they discourage potential competition (LOL!!!!!!! []

), but they make a shameless plug as to the amount of time, effort, and

money that goes into even the least expensive Klipsch product!