Patman

Here's a couple of photos of my DIY center channel speaker project, which I've dubbed: The SunCC.

This CC speaker uses the same tweeter (Morel MDT30) and midwoofers (Peerless HDS 850439's) as my other DIY speaker projects, a pair of 2.5-ways (SunOnes) for the front speakers, and a pair of 2-ways (SunTwos) for the rear speakers. I've incorporated the Morel MDM-55 midrange in an effort to help its dispersion pattern for a speaker that lies horizontally.

No crossover is done yet (you really need to measure the drivers installed in the front baffles to get something to work with in crossover design). My intention is to shoot for crossover points of 500Hz and 4000Hz, but that's subject to change, but stay tuned for updates.

Photo 1

Photo 2

The tweeter and midrange drivers are recessed, but not the midwoofers (supposedly it helps in getting their acoustic centers closer in line, plus recessing tweeters help in combatting diffraction effects.

When I get some more free time, I'll post all of the construction photos in a webpage (but it's pretty much the same as my other enclosure designs, with a few differences to regain some more volume I lost by going with a shorter and narrower front baffle. I do change some things around given the constraints of the project, like locating the port (not shown yet) on the backside of the enclosure.

The dimensions of the enclosure are 24"x9"x14.75" (WxHxD).

palebluedot

This looks good as usual...well except for you know what


Where do you get your parts from?

bfrank

SunCC
SunOnes
SunTwos
SunSub ?

Patman

I'm telling ya, it's what's on the inside that counts!

I got the MDF from Home Depot, the tweeter from Zalytron (a year ago in the hopes of finally getting some time to create this monster), the midrange and midwoofers from Parts Express.

Patman

I see a pattern emerging as well... (Everyone wants a line of speakers with their name on them).

bfrank

Just wait until I can show you what I am working on right now!

Lets just say - "extreme highend"

(around the end of the year)

Patman

Oh nice thread hijack!

Patman

Okay, here's a shot of the work-in-progress known as the crossover (I got smarter, and made something that allowed me to swap out the crossover elements without the need for soldering them in place - I just need to find a local place that sells small alligator clips...):

Photo 3

And here's a rough look at the frequency response (this was all done with textbook filters, plus taking into account the acoustic roll-offs of the drivers themselves), and I won't certify how valid the graph is, but it got me into the ball park, and my ear tell me that it's relatively useful given that I am only using 1/3 octave test tones and a Radio Shack SPL meter, a notepad, and Excel to graph the results):

Photo 4

I need a little help in the 2000-5000Hz region with the low pass on the midrange (perhaps I need to go to a higher order filter to get a sharper knee in the corner frequency. But for a first pass at a crossover design with rudimentary tools, I don't think I did all that badly for now. Can't wait to get on a real measurement system one of these days...

palebluedot

Pat,

Can you give me a primer in crossover design?

Also you can get your clips at McMaster Carr...a most awesome place for everything. http://www.mcmaster.com/catalog/108/html/0640.html

Patman

That is a loaded question! You can do it the hard way (like me), or do it the more scientific way.

The scientific way requires a PC, with a good soundcard (for input of the mic measurements and output of the test tones), a small amp to drive the drivers/speaker, a wand mic (calibrated would be preferable), and measurement software. There are some "free" or cheap software out there (like the FRD Consortium which is sort of advanced but really good stuff, orSpeaker Workshop).

The measurements give crossover programs the data files that is used to come up with filters for each driver, with the goal of integrating the drivers with the filters so that the speaker sounds accurate (or you can flavor it to your listening taste by making the speakers sound warm or bright).

Beside the SPL measurements of the drivers, you also need the phase measurement (phase can be thought of as the delay in frequencies, but much more complicated than my feeble mind can handle), and impedance measurement. The impedance measurement is important because a crossover filter depends on the impedance at a target frequency.

Some woofers require what's called a Zobel, which is a R-C circuit that's connected to the woofer's positive and negative input terminals after the filter and before the driver. The Zobel will flatten out the driver's impedance profile so that it act like a resistive load to the amp. This makes designing a filter for the woofer easier, but I seldom use them because in 2-way speakers, the crossover frequency is usually around 2KHz or higher, and I don't find myself needing the Zobel. But in a 3-way, with a low crossover point for the woofer, a Zobel might be the way to go.

Some drivers require lots of little circuit elements to resolve peaky output from cone breakup (running the driver in a too-high frequency area) called a notch filter to tame the peaks. There's a series version of a L-C-R circuit that is connected in parallel with the driver which is used to remove a peak by inducing a valley at the target frequency, click here for a small discussion on it. There's also a parallel version of a L|C|R circuit that's connected in series with a driver - I forget which does what. Doh!

There's also the Baffle Step Compensation (BSC), usually just a R-L circuit in series at the front of the low pass filter for a woofer. The Baffle Step is the loss in output due to the wavelengths of low frequencies (usually 500Hz and under) radiating all around the speaker in 4 pi space, while the rest of the speaker's output will radiate in the front of the speaker in 2 pi space. The BSC shelves down the woofer output above 500Hz to make the speaker sound more fuller in the bass region, and thus, less bright, but it will rob the speaker of overall speaker sensitivity (you'll have to pad down the tweeter in little more to conpensate for the smoothing out of the bass output).

There are quite a few way to go about designing a crossover. Some camps subscribe to less components, the better. Other camps subscribe to alway use the drivers in their intended frequency region for proper pistonic operation of the driver, and if that means using higher-order filters to "break off" the unwanted frequencies, so be it. Some camps fall in between those two camps.

2 main crossover design packages are: LspCAD, and LEAP. LspCAD can be bought for $129 for the normal version, while the pro version is around $500. LEAP is also pricey (and has a hardware lock to prevent piracy). Madisound offers a LEAP crossover design service if your speaker project uses drivers that they carry and sell. I think they charge $60/design (you can spec out the front baffle dimensions, and driver "center-to-center/CTC" distances. There are also some others, like CALSOD (DOS interface), and others which I can't remember the names of, but a google search might help.

The CTC distance can determine how high or low you can cross 2 drivers over. The larger the CTC distance, the lower you'll need to cross. I try to stay within one wavelength of the crossover frequency. For example, (units in inches), if the CTC was 5", then you'd want to cross over no higher than 13585/5 = 2717Hz. If you ever see someone using 8" wide midwoofer, they are usually limited to crossover over low, like 2000Hz or below.

A filter just slopes off a driver's output at a target frequency that is determined by the designer. You have low-pass filters that will choke off the higher frequencies (woofer/midwoofer/midbass), and you have high-pass filters that will filter out the low end frequencies (tweeters, or midranges). When you mate the output of the driver output, you are hoping you get smooth integration in the crossover region(s) without much of a dip or bump.

Most of the time, the tweeter output will be louder than a woofer/midwoofer's output, so you'll need to pad down the tweeter with either a L-pad, or a series resistor. A L-pad is current divider resistor network that simply moves some of the current from the tweeter into a resistor that's connected in parallel to it, thus padding the output of the tweeter. To keep the same "nominal" impedance load, the L-pad consists of 2 resistors, a Rs (resistor in series) and a Rp (resistor in parallel) that not only reduces the current being fed to the tweeter, but maintains the nominal impedance of the tweeter load. A series resistor can be used to also pad down a tweeter, but it's harder to integrate without slightly retuning the crossover filter for a difference impedance value as different series resistors are used to pad down the tweeter. L-Pad make for less guesswork on the original filter being used.

The filters are varied, and depending on target designs, can be limited to how you want to use them.

Every inductor or capacitor used in a filter adds an "order" to the filter. So if you used a L|C lowpass filter, it's an electrical 2nd order lowpass filter. L|C denotes an inductor in series with the amp input with a capacitor hooked up in a parallel fashion.

At the "corner" frequency of a filter (where the filter starts to roll off the output), usually named Fc, the output should be -3dB for Butterworth filters (BW filters), or -6dB for Linkwitz-Riley filters (LR filters). But there's an "dampness" associated with the corner, and that's the "Q" of the filter. The higher the Q, the less damped, or controlled the driver is at that Fc. The lower the Q, the more damped or controlled the driver is at that Fc. BW are considered critically damped with a Q = 0. 7071 (sharp roll-off), and LR are considered over damped with a Q = 0.5 (more gradual roll-off). There's also a Bessel filter that imparts a Q = 0.577 to the Fc.

Each order should impart -6db/octave of slope to the filter. So a 4th order LR filter should have a -24dB/octave of slope, and would be down 6dB at the Fc of the filter. An octave is the doubling of the frequency going up the frequency scale, or halving the frequency going down the frequency scale. So if the Fc was at 2000Hz, the 4th order LR filter would make the output at 4000Hz be -24dB from 2000Hz, and -48dB at 8000Hz.

Some designers have crossed woofers really steeply to make sure the woofer isn't working too hard at the higher frequencies, so you may see designs that have -48dB/slopes to protect the woofer. The downside is that it makes the component part count go up, which can cost money (major speaker makers try to keep part counts low while still making speakers sound good enough to the consumer). For DIYers, component part counts shouldn't be a deterrent to get the crossover done right.

When you sum filters (low-pass and high-pass), a rule of thumb is that summing -6dB slopes will produce a flat output response at the crossover frequency (i.e. LR filters), summing -3dB slopes will produce a 3dB bump at the crossover frequency (i.e. BW filters). But it all depends on the drivers you use, which will dictate what's the better filter/slope to use for a particular design.

The name of the game is to design for the final acoustic output of the drivers. Sometimes you'll hear that a speaker has 3rd order acoustic crossover slopes, but used an electrical 2nd order filter to achieve it. What happened was the designer incorporated that natural acoustic roll-off of the driver when they designed the filter for use in the network.

Go here for a site that will enable you to do quick-n-dirty L-Pad and filter components calculations.

There's a whole bunch more to speaker crossover design, like phase is one topic that I have no firm grasp on, other than making sure that I remember to either reverse the polarity of drivers when I need to due to the order of the acoustic roll-off of the filter, each order gives you 90 degrees of "phase-shift" and you have to imagine rotating around a circle (clockwise for highpass, counter-clockwise for lowpass) to see where the phase shift is for each driver, and if they are 180 degrees out of phase, you need to reverse the polarity of one of the drivers (in a 2-way)

A 2nd order design will most likely require the flip in polarity on one of the driver, but a 4th order will not. A good test in the integration of the filters in the crossover region is to flip the polarity and see if there's a sharp deep dip in the response of the speaker output.

If you go mucking around in the Speaker Building section of Parts Express, there's some good articles on this stuff. Plus the Parts Express Tech board is good for questions (but more of a novice nature, though there are some advanced gurus there too). The Madisound Audio Discussion board tends to be more advanced in discussion and DIY project undertakings.

I'm afraid I've probably confused more than illuminate, though...

Patman

Or you can be jealous of BFrank, with access to all sorts of wonderful speaker design tools....

palebluedot

Het Patman...can you go over that one more time?

Thanks buddy...it's a start. I have been wanting to build some speakers for a while now...but I am limited on tools. I will basically be doinf everything with hand tools. Cabinet design and speaker choice I think I am set on but the cross-over design is still a mystery to me.

Eh...BFrank takes all the fun out of with all his hoity toity gear...he thinks he is better than all of us.

Patman

Well, when you can injection mold your front baffles, yeah, color my jealous.