|
|
| Re: Been learnin' about crossovers [message #69348 is a reply to message #69347] |
Mon, 05 September 2011 22:13   |
 |
Wayne Parham
Messages: 18793
Registered: January 2001
|
Illuminati (33rd Degree) |
|
|
One thing I've learned about crossovers - Textbook garden-variety filters are never optimal. The best crossover will always have some fudged values, not textbook ones. Getting the phase angles right for the complex load and specific path lengths always means juggling the values to dial-in the lobe. So there's little meaning for me of the basic slopes described as Butterworth, Linkwitz-Riley, Bessel, Chebyshev, Elliptic, etc. All describe a specific transfer function, but I never implement any of them in practice. Mine are always modified slopes, hybrids, dialed in to set the transfer function and forward lobe precisely where I want it to be.
|
|
|
|
| Crossover design is science and art! [message #69349 is a reply to message #69347] |
Tue, 06 September 2011 08:40 |
spkrman57
Messages: 522
Registered: May 2009
|
Illuminati (1st Degree) |
|
|
I agree with Wayne that you can't just throw a textbook crossover together without taking the drivers/amps/environment into consideration.
Right now I'm trying to decide if I want to build Wayne's new 4 Pi crossover or reduce the LF section to be a 2nd order w/heavy zoebel. Either design is modified for 16 ohm use!
JBL 2226J and 2446J/Be/500hz round tractix Edgarhorn are the drivers and it has been mentioned before that the round horn has a different dispersion pattern and that affects the lobing.
In the past I have had success with the old 4 Pi crossover w/single coil on the 2226. I also built the new design years ago and I felt the whole did not equal the sum of the parts. Bear in mind the different compression drivers/horns I used.
I will probably build 3 versions of the LF crossover section and listen to each one for a week or so and see which I prefer in the long run.
I'm currently using a modified version of the 3 Pi crossover as it worked very well in my previous system using a EV SP-12. However the 2226 and SP-12 are 2 completely different drivers.
I just had to break-in some 2226's I just swapped for recently. The part that was really hard to handle is using a pair of JBL 2426's on a 1" to 2" JBL adaptor for my 2nd pair of 500hz Edgarhorns.
"Titanium vs Berilium" (sp?) is a no brainer, I just did not want my good horns/drivers out while swapping things around and testing.
The Be diaphrams are everything a diaphram can be. I'm spoiled big-time with them and listening to the Ti phrams now is depressing!
Sorry to go so far OT Bill/Wayne, I just have not posted much lately and trying to catch up!
Regards, Ron
|
|
|
|
|
|
| Re: Been learnin' about crossovers [message #69365 is a reply to message #69363] |
Tue, 06 September 2011 16:14 |
|
Wayne Parham
Messages: 18793
Registered: January 2001
|
Illuminati (33rd Degree) |
|
|
I always start with textbook values. It's definitely the place to start. Then I manipulate them slightly, sort of fine-tuning the forward lobe.
If you scan through my Spice models, you'll see both the chosen values and a list of textbook figures, not even rounded to the nearest available component value. As an example, look at the model for the three π loudspeaker in the link below:
The lines that begin with an asterisk are comments. The ones that begin with a component value are actual circuit elements. You can see, for example, that the tweeter circuit uses 6.8uF, 1.0mH and 20uF components in the core high-pass splitter filter.
Now scroll down a little bit and you'll see a whole lot of commented out values. Those are textbook values for crossover between 1.0kHz and 2.0kHz, first-order through fourth-order. They're not even rounded to the nearest available component value. I always start off with one of those, and work through a process to find the best readily available component values, i.e. 1.0uF, 4.7uF, 6.8uF, etc.
The first thing I do is to measure a speaker with a filter having exact textbook values in the crossover, as a start, and see where the lobes and nulls fall. Then I manipulate the values to get the lobe and nulls precisely where I want them to be. After that, I round each textbook value to a readily available component value and measure again. If the lobe/nulls shift too much, I'll manipulate other values in the circuit until I find the best fit.
|
|
|
|
|
|
|
|
|
|
|
|
Posting Rules
Members
Search
Help
Register
Login
Get a T-Shirt!
