A Speaker Workshop Tutorial

Speaker Building Tips

Home

This is a short tutorial on taking measurements and using the crossover portion of the program.

Before we begin, the actual Audua Speaker Workshop tutorial is at:  http://www.speakerworkshop.com/SW/Project/Main.htm
The author takes you through the steps to design, build and measure a loudspeaker.

Tutorial:

Preliminary Steps

Steps for Measuring your Driver parameters

Steps for Acoustical Measurements

Steps for designing the crossover

Acoustic Measurements on the system



Preliminary Steps:

1.  Download the program from Audua.  I have used it on WinME and WinXP with success.

2.  Install it on your machine.

3.  Check your soundcard for duplex operation.

 Before buying an expensive soundcard, open the program and look under the menu item Options.  Go down to Wizard and click Check Sound Card...  A box will open up and verify if you have duplex capability - most soundcards do.  If your soundcard is not capable of duplex operation, this wizard will tell you and you will have to buy a new soundcard.  Disable on-board sound in the BIOS if replacing it with a dedicated card.  If your soundcard doesn't have a built-in amplifier for driving speakers directly (most don't), you will need an external amp.  With an external amp, instead of connecting the soundcard output to J1, you hook it to the amplifier input (say the left aux input), and hook the amplifier's (left channel - if that's where the input is connected) speaker outputs to BP1 and BP2 and set SW1 to 20 dB attenuation.

I bought a cheap SB16 PCI from Walmart for about $40 I believe.  One reason I like this card is because it has a built-in amplifier, so I don't need a separate amp which might fry my soundcard inputs.  It had an installation issue (as many SB cards do in XP) where I had to go into the registry as follows:

Open HKEY_LOCAL_MACHINE\system\CurrentControlSet\Enum\PCI.
Open the “+” in PCI to reveal the VEN numbers.
Click the next item under each VEN number and search for Creative AudioPCI or similar in the descriptions on the right
Right click and click permissions (or Edit menu, permissions).
Make sure “Full Control” and “Read” are clicked for that VEN number.
Restart and Windows should install card.

(Update:  I now use a SoundBlaster Live 24bit soundcard, and it has a much better response, though I now use a separate amp for measuring.)

4.  Build one of the Wallin Jigs (named after Eric Wallin).

Jig #1 is at:  http://mysite.verizon.net/tammie_eric/audio/audua/audua.html
Jig #2 is at: http://mysite.verizon.net/tammie_eric/audio/jig2/jig2.html

Thanks to Dan Kerl for this suggestion.  Dan suggested tieing the vacated terminal to its neighbor so I wouldn't have to switch cables, but that didn't work for me unless I've done something wrong.  But I used his other suggestion to solder the wire from J4 to the unused pin.)
There are minor differences between the two jigs, and you may want to begin with jig #2.  I built jig #1 and it works fine.  The main difference is that if you use jig #1, you must reverse the left and right audio jacks at J2 and J3 when using a microphone to measure the acoustic output of the driver (unless you make the change below).  When you want to measure impedance, switch the left and right jacks at J2 and J3 to their original configuration.  You don't have to switch cables with jig #2; nor with jig #1 if you will just build it so that J3 and J1 are always connected, no matter which way SW4 is connected.  First, Hook the wire from J1 to the unused pin (as shown on Wallin's diagram) of switch SW4.  Then you have to solder a jumper from the VACATED pin on SW4 (where J1 used to be connected) to the other end of the switch where the 10k resistor connects from SW3.  If you make this modification to jig #1, you do not have to switch cables for acoustic measurements, nor do you have to go into Options-Preferences-Acoustic tab to reset the channels so that the mic is the left channel and the reference is the right channel.  Let me repeat, with the above modifications, don't swap cables or change left/right settings on the acoustics tab - Don't do anything.  If anyone spots an error in this, email me at the address on my homepage, and I will change it again.  Again, if you don't add the jumper to J4, then you have to swap the left and right cables for acoustic measurements, but if you do add the above modifications, then don't swap cables.  Whenever I refer to the Wallin jig, I'm referring to jig #1 so be careful of the reference.  Jig #2 may use different switch or jack designations.

5.  You will also need a microphone and a separate microphone pre-amp.

Don't even consider using the soundcard as its response is horrible.  Most people buy one of the Panasonic electret microphone capsules (WM-60 or WM-61) usually for about $25 for 10 from various electronic supply companies.  If you are soldering the terminals yourself, you may want 10 mics as you will damage a couple just trying to solder those itty-bitty tabs.  The WM-60 has a flatter response that rolls off a little on top.  The WM-61 actually has a peak of about 3 dB or so in the treble, then rolls off past 20 kHz.  Various generic calibration files can be found on the net.  Here are some graphs but you will have to create .frd files or find the .frd files elsewhere - search Audua's Forum.  Or you could order a calibrated mic from someone like Kim Girardin.  Do a search on the Madisound board or the PE board for his email address.  It will cost around $40 to $50.  Besides the calibration, one big benefit is that you don't have to worry about soldering on those little asprin-sized elements, because the leads will already be soldered for you.  Here is a DIY microphone tutorial.  Here is a link for powering your electret condenser microphone.  I used the circuit titled Balanced electret microphone circuit (about $10 in parts) which gets power from phantom power and I use a Behringer MX-602A mixer for a pre-amp.  The MX-802 mixer could also be used, but it's more expensive.  Run one of the tape out jacks to J4 on the Wallin jig.  If you would rather buy a microphone, buy a Behringer ECM8000 from any professional music store.  Musician's Friend sells the Behringer mixers and the microphone for about $100 total.  Check back from time to time as they always have sales.  I bought my MX-602A from them for $50.  Parts Express now sells Behringer equibment.  They sell the UB802 mixer and the ECM8000 mic.  Someone has said that the MX-502 does not have phantom power so beware of this if purchasing it.  PE sells the UB-502 which does not appear to have it either (contrary to the implication in the description).  They will need a separate power source for the microphone.

You can also build one of the Wallin Microphone Preamps instead of buying a microphone mixer:
Wallin Preamp #1
Wallin Preamp #2

Suggestion:  Buy a Behringer ECM8000 and a MX-602A or MX-802A for ~$100 total (maybe less).  You can find some generic calibration curves on the net or have someone like Kim G. calibrate it for you for ~$40 or so.  This will avoid having to build a powering circuit, microphone and mic preamp (~$70 uncalibrated or ~$100 calibrated).  Oh, don't forget some microphone cable!

Test your jig according to Wallin's instructions.  And I will reiterate what he states - make sure to set your volume levels appropriately!  Somewhere there is a VU meter to help in setting levels.  Usually it's to the lower left.  If it's not open go to the menu and click View-VU meter.  When you record your test signal, the VU meter will show left/right maximum, minimum and average levels  It just shows numbers, it is not a "dancing bar."  Make absolutely sure that these numbers never go above 32,000!  As a matter of fact, with sine waves, don't let the levels go above about 16,000.  And don't clip the tops!!!  Expand the graph width if necessary.  Clipping shows up as flat tops on the sine waves.  Lower your levels until clipping disappears.  MLS type signals will have higher levels than sine waves (usually around 22,000).  You could actually make your test wave an MLS or pulse signal, and watch that you don't go over about 22,000.  To save your mixer level settings so you don't have to re-set them every time you start SpeakerWorkshop, download Quickmix from Martin Saxon's web site.  Once you have the correct levels, run quickmix and save them.  Whenever you run SW (Speaker Workshop), run Quickmix and load your saved file to set the levels.  I make another file called normal, so I can return my mixer settings to their previous state.  One last thing, sometimes it seems as though Speaker Workshop doesn't respond properly.  For instance, it may quit optimizing your network.  If it starts acting weird, I usually save everything and close it, then re-open it.  I think it runs into memory problems after operating for extended periods.  Remember, this is Beta software, and it still has some bugs, but it is still extremely useful.  As a matter of fact, I wouldn't even dream of trying to design a speaker without it now.  It is a really humbling experience to see speaker design done properly.

Other useful settings:

In Options-Calibrate, you will find three calibration boxes.  One is for microphone calibration, one is for channel balance and one is for amplifier calibration.  Place the jig in loop mode and run channel calibration and follow the instructions.  Under the system folder, you can see the channel calibration file (measurement.calib).  If the phase goes way off by like 900 in the treble, you may have to set the interchannel time delay (on the general tab).  Change it in the positive direction if the phase is going negative, in the negative direction if the phase was going positive.  Otherwise, this file is just the difference between the channels.  SW has to know this because it uses one channel for a reference so it must know the normal difference between the channels.  Also, try to get signals to start in under 5 msec (preferably 2 msec) by recording a sine wave, and seeing where it starts (same signals as the ones used for level settings above).  Use the latency setting in the preferences window (options-preferences-debug).  Keep changing it until the sine wave starts within 2 to 5msec.  The microphone calibration is just a text file (.frd usually) that has the mic's frequency and phase difference from flat.  SW will apply that correction file to all measurements.  Just choose Resource-Import from the menu and pick the calibration file.  This puts the calibration file in your project.  Go to Options-Preferences-Calibrate, and in the mic calibration area, click browse and get the calibration file.  Click okay.  Likewise, you can enter an amplifier calibration.  Previously, I had stated this isn't usually critical as amps are generally pretty flat; however, I would suggest that unless you know your amp is flat, and the tone controls centered or off, go ahead and do the amp calibration.  Besides, it only takes a moment.

Note:  To see the amplifier calibration curve, look in the system folder for a file named, Measurement.Pulse.  Open it and on the menu go to Calculate-Frequency Response...  Hmmm, I notice something called "waterfall" in that menu that you can use on pulse responses... :-)  I'll let your imagination take you where it will.  Oh, where was I... Oh yes, Under Windowing, click uniform, and under Calibration, use channel calibration.  This will create a Measurement.pulse.frequency graph in the system folder which is your amplifier response.

If you are using a separate amp, always use the attenuation switch (SW1) or you will fry your soundcard inputs!!!  To set levels on an external amp, set all your computer mixer levels full up (record, wave, line in, main volume).  Turn the amp volume all the way down.  Depending on the output of your soundcard, you may have to bring the soundcard mixer volume level down to keep from clipping the tops of the waves.  Watch the "name".in.r graph (where "name" is the signal name) to see this.  Create a signal and record it as you did in the Wallin Jig Tutorial.  Open the left "name".in.l and right "name".in.r where "name" is the name of your signal you created.  The "in.l" signal will be low-level  noise.  Slowly turn up the amp volume a little, then re-record the signal.  Keep repeating until the signal reads approximately 16k.  Remember to use the attenuation switch to protect your soundcard.

On the Preferences window (Options-Preferences), click the Measurements tab.  You can adjust the sample rate and sample size.  For the rate, I usually set it to 48,000.  Note that some soundcards can't sample that high, so you may need to back it down to 44,100.  I usually set the sample size to 16k or 32k.  More samples will be slower, but give better resolution in the bass.

The Impedance tab is where you calibrate your jig.  See the Wallin jig tutorials for this.

To print anything in SW:

Open the curve, driver, network or whatever you want printed.  Then click the print button.  SW prints the active window whether it's a graph, driver data or whatever. 

One last thing, you can still use SW even if you don't buy a mic/preamp, but you must manually create driver file data.  And without phase data, crossovers are still just a best guess, but it's better than "standard formulas."  Check out the FRD Consortium, especially their SPL Trace program.  You can scan frequency response graphs and import them into SW.  Use SPL Trace to trace the graph and it will create the .frd file for you.  To use it, just set the reference lines (for SPL and frequency) using your mouse.  When the program begins tracing, you just move the mouse up or down.  The program will slowly move you right in tiny increments.  Move the mouse up or down, and when the horizontal line (your mouse pointer) intersects the driver's graph, click the mouse.  It's a tedious process, but is better than "eyeballing" points and writing them down!

Top of Page

Steps for Measuring your Driver parameters:

SpeakerWorkshop opens the last project entered.  A project contains multiple drivers, networks, graphs, etc...  If you want a new one, close the one that opens.  Information is stored in the "file tree" in the left pane.  As you measure or create drivers, signals, networks, etc..., those elements will be listed on the left.  You can create folders (say a woofer file, a tweeter file and a crossover file), then drag elements to the folders to keep things neat.  The menu changes depending on what element (driver, network or graph) is open.  It is a little weird to get used to, but quickly becomes easy due to its intuitiveness.  Items you create:  the individual drivers, individual crossover networks, maybe a signal file for measuring components and various graphs from acoustic measurements.

1.  First measure the driver's DC resistance using a good ohmmeter.

2.  In Speaker Workshop, create a driver and give it a name.

3.  Hook the driver to BP3 and BP4.  Put jig in impedance mode and go to the menu and click Measure-impedance.  Now you have an impedance curve which also shows phase.  You can right click on the chart and choose chart properties and adjust the chart to your liking.

4.  Right click on the driver and choose Properties.  Click the Data tab.  Click the option to "Use this DC resistance."  Input the DC resistance you measured with the ohmmeter.

5.  Now you can estimate parameters (other than Vas at the moment) by going to the menu and click Driver-Estimate Parameters.

I haven't yet measured Vas, but when I do, I'll write up a procedure if it is more than the obvious - measure the driver again with a known weight applied (delta mass method).  I am more concerned with acoustical measurements and crossover work than with box design at the moment, and we won't need Vas for that.

Top of Page

Steps for Acoustical Measurements:

1.  IF you haven't attached a jumper to SW4 and moved the wire from J4, switch the left and right soundcard inputs at jacks J2 and J3.  Plug the microphone mixer "tape out" jack (either one) into J4.  If you are using an external amp, turn the volume down and connect it at BP1 and BP2 and adjust settings as stated above.

2.  Set up your mic and driver for measuring.  I set the mic about 2 to 4 feet or so from the driver and on center with it.

3.  Measure a pulse.  With the driver window open, on the menu click Measure-Pulse response.  You can also click the "P" button (first button) on the button bar.  BTW, if you don't see the button bar, look under View-Resource Bar and click it (or the forth icon from the left).  This gives us a pulse to set our markers.  Open the "drivername.pulse" graph.  Drivername is the name you gave the driver.  You should have two red vertical lines.  If you don't, on the menu go to Options-Preferences and click the Markers tab.  Under Time, click the visible option.  You can move these lines by clicking them and dragging them where you want.  We want to drag the left one to the point just prior to the start of the spike waveform.  If you follow the curve to the right, you may see smaller (sharp) ripples fairly evenly spaced.  (They're easier to see if you right click the graph and change the y-axis min. and max. to +200/-200 or something similar.)  These are reflections.  We will move the right line to just before the first ripple (reflection).  This will give us a time window in order to ignore reflections.  Reflection time can be estimated by counting the number of feet from the microphone to a reflecting surface (say the floor) and then the number of feet back to the speaker.  The total number of feet is roughly the milliseconds, so if you don't see clear reflections, and the reflection distance is 3 feet to floor, 3 feet back up, you should set the marker to around the 6 millisecond mark.  Note that you can click on a graph and use your mouse roller (or right-click-zoom) to enlarge or shrink the viewing area.  When setting markers, zoom in on the area and make sure you are not cropping the start of the pulse.  This is important - SW cannot give good measurements when part of the signal is missing!

One other thing, you should try to get at least a 3 or 4 millisecond wide window to get decent curves.  This is why big open spaces far from walls and off the floor is desired.  For a while, I had a 6 foot step ladder in my vaulted ceiling living room!  Out in the middle of a room, 4 feet off the floor would allow for about an 8 ms (minus mic distance) window to be achieved.

Note that you should watch the VU meters and set the volume of the microphone mixer or amplifier so that the VU meter reads about 22,000 when measuring the pulse.  If the levels are correct, you can now measure the response. 

4.  On the menu click Measure-Frequency response-On axis.  You can also click the "f" button on the button bar.  Likewise, click on Farfield and also Gated (ff and fG on the button bar).  You can get accurate measurements of the woofer and port and combine them to give an accurate frequency response curve all the way down to the speaker's F3 point.  I'll let you figure those out.  I'm more concerned with measuring for crossover design for now.  The bass response will always be affected by the room.  You can move the microphone to within a 1/4" or so of the driver and measure the nearfield response.  You can measure tweeters or woofers this same way.

Note that you can click on a hard to read graph (say the farfield response) and go to the menu and click Transform-Smooth and choose the smoothing octaves (usually 1/3 octave smoothing) to see what the curve looks like.  However, undo this after you have looked at or printed the curve because SW needs un-smoothed responses when modeling crossovers.  I may be over-reacting to this because SW actually uses the on-axis curves, but I don't want to take chances.

5.  If you already have a test baffle with the woofer and tweeter mounted - Great!!!  You can measure both the woofer and the tweeter at the same time using the same microphone position.  Place the microphone centered between the woofer and the tweeter about 2 or 3 feet away, or on axis with the tweeter.  Do everything above except for the nearfield response - We can't move the microphone or this won't work!  We can measure the nearfield response later.  Hook up and measure the tweeter, then the woofer.  Observe the correct polarity and do not move that microphone!  When measuring both drivers this way, the offset is taken into account in the measuring process.  This means that as we play with different crossovers, the drivers will be modeled with any driver offsets taken into account - the Holy Grail of crossover design!!!  This is the preferred method in my opinion.  Actually, instead of making a test baffle, you could just put the drivers in the their actual box instead and measure them there.

Top of Page

Steps for designing the crossover:

Note:  This assumes you've measured the tweeter and woofer at the same time on the same baffle.  If not, you have to find the offset distance between the woofer and the tweeter and enter it on their driver properties page.  If you don't, there is no way for SW to combine the woofer and tweeter curves to get meaningful results.

1.  Create a network.  Go to the menu, click Resource-New-Network.  Give it a name (say high-pass or low-pass if you like).  Click okay.

2.  Insert driver and connect it to the source as follows:

With the network window open, on the menu click Network-Insert-driver and choose a driver to insert.  You can also right click in the network window and choose this.  Click on the driver icon.  Click (and hold) in the circles and drag to the circles on the source.  This will connect the driver to the source.

Don't immediately put both drivers and their respective crossover components in the same network. Design a high-pass network, then create a new low-pass network for the woofer (in the same project, of course). The reason is to be able to use the crossover optimizer. SW won't optimize a network containing both (or more than one) drivers. At least I had trouble, but maybe it's just me.  :-)  Later, while the frequency response graph for the network is open, we can go to the menu and choose Calculate-Combine and add the curve of the other network's response (in phase + or out of phase -) to the one that's open thereby generating the total system response.

3.  Add components.

For the woofer go to the menu and click Network-Insert-Impedance compensation (or right-click the network window to choose these options).  For the tweeter you might add an L-pad.  Then add stock crossovers.  SW appears to want to input driver side components first, then move toward the crossover.  In other words, do impedance compensation first (or attenuation networks), then response shaping circuitry, then finally, the crossover.  (Otherwise, you will have to disconnect and reconnect components to get them in the correct order.)  If you want to add some kind of response shaping circuitry, you will have to individually add inductors, resistors and capacitors of the appropriate values.  You can change values by right-clicking the network and choosing properties.  Then click the components tab.  The boxes with yellow and blue rectangles shows possible orientations of the component and its value and name.  Click on the appropriate orientation to change the component's alignment from horizontal to vertical, etc...  You can also change a components value and name by clicking in the appropriate boxes and changing them.  Make as many changes as you want but don't click apply!  The apply button doesn't seem to work.  Click okay when you are finished.  If you need to disconnect components, just click the component, then click in it's circle and unclick without dragging a line anywhere.  All connections to this point will be disconnected.

To connect the added components, you must hold down the shift key while you connect them!  Click a component (say a resistor).  Circles will appear at the ends.  Click (and hold down) the circle and drag a line to where you want to connect.  Hold the shift key down and let up on your click inside the desired connection point.  If you don't hold the shift key down while connecting, all other connections will be disconnected!

4.  Create a Network Goal. This is important!

We need a target response to which we may compare the network's response, and for the network optimizer to model.  Click in a network window.  On the menu, choose Network-Create goal...  Here you can choose the type and order.  Note - this is not the actual crossover type.  This is a goal we are aiming for - the final response we desire.  I believe most people would agree that a 4th order Linkwitz-Riley is generally a nice goal so I usually just choose that.  Luckily, this goal is usually attainable with lower order networks.  The acoustic rolloff combines with the electrical network to give us the final crossover point and slope.  Choose high-pass or low-pass as appropriate.  Don't forget that step!  In the open window, you will have to click high-pass or low-pass.  SW will try to make a low-pass network model a high-pass network if you tell it to!  If you just made a goal for a woofer, and now you will make a goal for the tweeter, open the tweeter network window (or click in it).  Create goal, but choose high-pass (it will still say low-pass).  All other options should stay the same (same x-over point, type, order).  The only confusing thing here is the section called "Set level via."  Usually I create a goal for the woofer, choose "dataset" and click the box with "?" on it.  Then I pick the "woofername.onaxis" where woofername is the name I gave the woofer.  I choose a start frequency of around 200, 300 or 500 Hz depending on where the curve starts.  Then I choose an upper limit below any woofer peaks or ringing.  Important!  Also leave this as the tweeter's "Set level via."  This does not choose the data for the goal.  It is merely setting the level where the target response will be flat.  And we want both the woofer and the tweeter targets at the same relative level (the same SPL).  I choose the woofer to set levels because the tweeter will have to be padded down to its level.  I guess the frequency range doesn't matter too much, but I avoid the cone breakup region because I don't want that region to raise the level of the woofer goal.  There would be no way to bring up the woofer's level. 

5.  Graph the network's response using the driver's measured acoustic response.

To get a frequency response graph for a network, from the menu (with the network open) choose Network-Calculate Response (or right click the network and choose Calculate response).  A couple of graphs will be created.  I usually open up the graph named for the network (not the "total" one - I'm not sure what the difference is except that I use the "total" one to combine curves since it says "total").  This shows the network's response.  Now let's compare it to our goal.  Right-click the graph and choose add.  This feature just places other graphs on the one you have open.  It doesn't add (combine) the curves together.  It just allows you to see separate graphs on the same graph.  For the curve to be added, choose that network's goal (if it's a high-pass network, choose the high-pass network goal).  Now we can see the network's response against the goal.  As we make changes to components, we can right-click the network and choose Calculate response and see if we are getting close to our goal.  Fascinating!!!  Note that if you save everything and quit SW, the added graphs disappear.  You will have to re-add them when you open SW again.  To avoid this, SW lets you create graphs.  Click on Resource-New-Chart and add your graphs to it.

6.  Optimize the network. 

With the network open, go to the menu and choose Network-Optimize network.  A window opens that gives us some choices.  First, choose the frequency range.  For all intents and purposes, this could be 20 Hz to 20 kHz, but I usually enter the range the driver covers (say 200 to 12 kHz for a woofer, 600 Hz to 20 kHz for a tweeter).  For the Number of Points, choose something in the low 1,000's if you have a moderately fast computer.  Choose something less than 1,000 if you have a slow computer.  The more points you pick, the better the optimizing, but the slower it will go.  Here is also where we pick a target to model.  Click the box with "?" on it and choose the "network.goal" file, where network is the name of the network you are optimizing.  This is why we created that goal.  The optimizer will pick components to match that goal.  Remember to change the goal when you do the next network.  If you just optimized a low-pass network, and now you try to optimize a high-pass network, the goal will still say low-pass, so change it to high-pass-network.goal, where high-pass-network is the name of the network!  Finally, we can uncheck components we don't want optimized.  If you don't have a lot of components, you can probably leave them all checked.  However, I've found that it will give odd component values at times, especially when you have many components.  In this case, optimize small areas of the network at a time.  Perhaps model an attenuation network first.  Just leave the two attenuation resistors checked and de-select everything else - impedance compensation, parallel traps, etc...  Click okay on the optimizer.  Then repeat this with the response shaping networks, then the actual crossover network, etc...

Sometimes you just want the response curve changed in a specific area.  Say you added a response shaping network (a parallel resistor, cap and inductor) to remove a tweeter peak.  This is where the "Frequency range" boxes in the optimizer window come in handy.  If the bothersome peak is from 6 kHz to 10 kHz, insert those values into the optimizer and it will calculate the component values to reduce the peak in this range.  It may still affect other frequencies so you should include some flat response areas around the peak if possible.  If it's difficult to see the dip, use a high value for your resistor (say 2,000 ohms), and then optimize manually by changing the value of the capacitor or the inductor until the dip is exactly where you need it.  Then change the value of the resistor to something like 3 or 8 ohms.

If you keep getting odd values for certain components, the optimizer may be telling you that you can do away with that component.  For example, if you have a third order network and it keeps making the second capacitor 1000F, you will probably reach the target response with a second order network.  Just right-click the 1000F cap and delete it.  Then re-run the optimizer.  If you have a second order network, and it's trying to delete one of the components, but you don't want to (say because of tweeter power handling), just make that component some moderate or moderately large value that you can tolerate.  Then run the optimizer, but unclick everything (including this component), except for the other component in the second order network.  The optimizer will optimize it to work with the component you left in.  When the response overlaps the goal, we are ready to combine (add) the networks together to get the total combined frequency response for the loudspeaker design.

7.  Combine the curves to create the system response.

Click on one of the network's frequency response graphs.  I usually choose woofer.totalfreq and combine the tweeter to it.  Click the graph, and on the menu click Calculate-Combine.  This will bring up a window from which you can choose a curve to combine.  I click the network.driver.freq file for the other network.  You can combine these in many ways.  (I'm not sure why you would use "*" or "/" to combine.)  Choose "+" for in-phase (drivers connected in-phase) or  "-" for out of phase.  If you optimized your crossover networks well, one way will have fairly flat response, and the other will have a dip or null of -10 or -20 dB or perhaps more.  If you want to compare individual network sections here, right-click the graph and click add.  Then add the network responses or even the goals if you want.  If you measured the actual system, you could add it here and compare the theoretical with the actual!

Remember, if you combined with "+", the tweeter should be wired in-phase.  If the flattest graph was combined with "-", wire the tweeter out of phase.

Once you get the hang of it, be creative. Do you have a slight bump at the crossover?  Separate the high-pass and low-pass resonant frequency (F3) goals slightly, then recalculate. Basically you create the network, calculate the frequency response using the selected driver or curve, create a Goal, optimize to the Goal, then combine the responses trying both polarities.

Top of Page

Acoustic Measurements on the system:

Basically, this is similar to measuring the individual drivers.

1.  Create a "driver" (I usually label it "Complete Speaker").

2.  Set the microphone on a line centered between the woofer and tweeter, or on the tweeter's axis if preferred.

3.  Measure a pulse.

4.  Adjust the gates (red lines).

5.  Measure the farfield, gated, on axis, 300 and 600.  Of course you need to rotate the speaker or move the microphone appropriately for the 300 and 600.

Cool, huh?

I hope this hasn't been too boring.  And I hope I haven't mislead you in anyway.  I've learned most of this from trial and error.  If something is in error, please email me at the address on my Home page.  As with anything use this information at your own risk.  Don't blame me if you mess up your registry trying to edit it, blow up your computer, electrocute yourself, etc...  I'm also not responsible for your divorce, mugging, robbery, car-jacking, lost soul (well, I can point you to the Bible on this), etc...

Well, with lawyers now-a-days, you never know what you might get blamed for.  ;-)

Top of Page

Speaker Building Tips

Home
Last updated on 07/19/06