Xsim Crossover

The crossoverdesign can make or break any multi-way loudspeaker. Having the world’s best drivers with a poorcrossover design will yield poor results. It is important to approach crossoverdesign with humility because it truly is an art that requires experience toperfect. Yes, it is possible to get decentsound by sending driver data sheets off for a generic crossover design. However, without proper analysis of thedrivers in the intended cabinet it is a game of roulette. This design overview neglects many designconsiderations a professional designer may consider such as time alignment,distortion and polar response but provides a starting point.

It depends because there are several tests that you should do and some people prefer one method or another there are a few ways to get there. For example, many people use the least components possible in theory every component in the signal path w. With the changed component values in the premium crossover, two speakers with different crossover tunings are now compared. (Just to be clear: The audibility of the shown low-Q resonances is of course not directly comparable to a modified Encore crossover, it should only illustrate that tonal changes can be perceived well below the 3dB you.

Before designing acrossover, the cabinet must be fully assembled including ports anddrivers. I typically run a separatespeaker wire from each driver out of the port and label each wire. For sealed loudspeakers, connect the wooferto the binding post as this is the first driver measured in the cabinet. Using SoundEasy, each driver is measured inthe cabinet to obtain frequency response and electrical impedance. These parameters are obtained following thetest methodology outlined in our LoudspeakerMeasurement Standard. Pleasenote that any measurements taken near field must factor in the effects ofbaffle diffraction. SoundEasy allowsmodeling and application of baffle diffraction estimates to any frequencyresponse data. Getting good measurementdata is one of the hardest parts of loudspeaker design. It is crucial to validate your results bytaking measurements using both near field and quasi-anechoic gated techniquesand compare the measurements before proceeding. The measurement data is subsequently used in SoundEasy’s crossoverdesign tool.

XSim Crossover and Graphs (I have the amplifier set at 60W @ 4ohms to see the power distribution using a certain amp) Woofer. A little bit about my choices: the crossover is at about 1700hz. The tweeter Fs is 830hz, so i believe i will be alright, and the tweeter with the 60W amp is getting only 12W. Click 'Target number of base meshes' input-box and set the value to '500000'. Click 'Computational domain' input-box and set the coordinate to (50, 55, 0). Next, we will set layer mesh settings. Select 'Wall' as region from a combobox in 'Laye mesh settings' and set 'Number of layer' to '3'. Then click Set button.

DIY Crossover Installed in Cabinet

Although CAD hascome a long way, designing a loudspeaker crossover benefits from somebackground in electronics. SoundEasy has2-way to 5-way crossover templates covering 1^st order to 4^thorder crossover topologies. Additionally, there are templates for a wide array of typical filtersand compensation circuits. The CADframework allows implementation of any circuit or filter topology imaginable solong as it uses inductors, resistors, capacitors, operational amplifiers,potentiometers or logic gates. If youhave not ever studied filter design or do not know what the previously-mentionedelements are, you may want to consider studying up to improve your chances ofpulling off a working and affordable solution.

The first step Itake when designing a crossover in SoundEasy is to determine workable crossoverfrequencies and filter orders. Thisrequires careful inspection of the driver data sheets, frequency responsemeasurement and driver impedance. It isimperative to make sure that response anomalies, such as the breakup of a metalcone driver, are sufficiently attenuated by the crossover so that they are notaudible. It is generally a good idea tokeep the crossover frequencies low enough to prevent driver beaming but highenough to prevent driver failure. Forexample, crossing over most tweeters below 1.5kHz is a bad idea if it is meantto play program material at 110dB or has a resonant frequency above 750Hz. Crossing an 8 inch bass driver at 5kHz willresult in poor dispersion and transient response at higher frequencies. As a rule of thumb, the max frequency for agiven driver diameter can be equated by taking 13560 and dividing it by thedrivers effective diameter. For an 8”driver with an effective cone diameter of 7”, this means that the maximumcrossover frequency should not exceed 1937Hz.

A 4^thorder crossover rolls off frequencies at 24dB per octave where as a 1^storder crossover rolls off frequencies at 6dB per octave. Lower order crossovers have less phase shiftbut tradeoff power handling and masking out of band response anomalies. The designer has to determine the best set oftradeoffs for an intended design goal. SoundEasy has built-in filter templates for delay networks if a designerdesires to design a time aligned loudspeaker for higher order filters. It is also important to note that the filterorder does not have to match for each driver in a system but it does simplifythe design.

Once approximatecrossover frequencies and filter orders are determined, a filter can be developedin CAD. For the purposes of this designanalysis, a 2^nd order crossover was designed which approximatelymatches a Butterworth filter target. Additional response shaping elements were added where needed to meet thedesign requirements. This circuit wasdetermined based on using a 2^nd order low-pass for the woofer,band-pass for the midrange and high-pass for the tweeter. Due to the response anomalies and differencesin sensitivity, additional elements were added. The final circuit design is shown below.

Crossover Circuit inCAD

Xsim Crossover Download

The componentvalues for the circuit design are determined using SoundEasy’s ingeniousoptimization technique. It is as simpleas setting a target response for a single driver and selecting the componentsSoundEasy should manipulate to attempt to meet the target response. In this example, a 2^nd orderButterworth band-pass filter target was setup in the optimization dialog asshown in black in the figure below. SoundEasy will modify the selected crossover components and employ trialand error to arrive at the optimum component values required to meet thefrequency response target. Theoptimization technique will attempt thousands of combinations and may be runmultiple times to determine the best component values. While running optimization, it is important topay attention to the component values set by SoundEasy because it may setvalues that are very high inductances or capacitances and therefore expensive. This may take some playing around to getright. If you are unable to get close tothe target curve you might consider changing the filter topology. The midrange for the sample project tracksthe target curve pretty well but is not perfect at the high frequency knee dueto response problems with the driver itself. It is possible to shape the response further but more crossovercomponents equate to a more expensive crossover. It is easy to spend hundreds or eventhousands of dollars on crossover components. It is often better to solve extensive frequency response problems withbetter drivers or a more in-depth diffraction analysis.

Midrange CrossoverOptimization

The optimizationprocess is applied to the low-pass, band-pass and high-pass filters and acombined response is generated. Due toresponse issues, it may be necessary to attenuate a target or move thecrossover frequency slightly to obtain a flatter response. Ultimately, it may take several rounds ofmoving crossover frequencies and even filter orders to get to an acceptablesystem response. The combined systemresponse after optimization for this design is shown below.

3-Way CrossoverSummed Response

Xsim Crossover

During the designprocess, it is important to pay close attention to system impedance graphs. It is no fun perfecting frequency when yourdesign has a minimum impedance of 1.5 Ohms. SoundEasy has an option to set the minimum acceptable impedance duringoptimization to help reduce the risk of developing a network that dips too lowfor the amplifier you are using.

As a final step,it is absolutely necessary to set the values of capacitors, resistors andinductors in the CAD editor to values you can actually purchase. This usually involves rounding up or down tovalues you can purchase. If thecrossover parts are expensive, you can attempt to reduce component values andsubsequently plot their effect on the final response. This is time consuming but can save you a tonof money. A 3-way crossover of thiscomplexity is not cheap. Using cheapelectrolytic capacitors this crossover costs close to $150 for componentsalone.

3-Way Crossover Billof Materials

Conclusion

Unfortunately, the tweeter used in this design did not measure close tothe data sheet at all and required pushing the crossover frequency up. Ultimately the dip at 4kHz is fixable but thecost required outweighs the benefit.

Personally, Iprefer loudspeakers that have a frequency response that tilts slightly downwardso I find the response of this speaker pleasing. It did not ultimately meet the design goal of+/-3dB frequency response deviation but it plays low bass with authority andsounds very good overall. Even though itis a ton of work, there is something very rewarding about designing aloudspeaker from the ground up and the owner of the speakers discussed in articleis extremely happy with his one of a kind creation.

Xsim Crossover Manual

DIY speaker Finished Product

Ultimately,designing your own loudspeaker is something you do because you like thecreative side of designing things. Ifyou are after bang for your buck sound, it is much cheaper and less time consumingto use someone else’s well-established and well-documented DIY design withpre-fabricated cabinets.

Some of mypersonal favorites include:

http://www.linkwitzlab.com - Siegfried Linkwitz is a renowned loudspeaker designer that developsDIY loudspeaker projects for fun in his retirement. The new LX521 and older Orion loudspeakerprojects are considered by some to be among the best speakers available, evencommercially.
http://www.troelsgravesen.dk - A Danish designer responsible for many well regarded mid to very highend DIY designs. Most require building acabinet but the designs represent very high value.
http://zaphaudio.com - Although this site is notupdated much anymore, John Krutke has many good designs available as fullkits. He has also tested a slew ofdrivers if you need unbiased raw driver data.
http://parts-express.com - Many full DIYkits and parts available directly from Parts Express
http://madisound.com - Many full DIY kits andparts available directly from Madisound. They also offer a crossover design service using very good crossoverdesign software.

Comments

Xsim Crossover

You've got a line driven through that inductor shorting it out.
I'm not deaf, I'm just not listening.
Well , hell . Thank you for the quick response , that did it. Much appreciated dB
Sorry to bother again guys , but where can you view the phase of the drivers as you go about the design process?
Go to the FR window, pull up the 'Curves' menu, and open the woofer curve.
There is a checkbox there for 'show phase'. There is a similar checkbox for the 'System' curve.
Don, Donno, or 'Hey you' all work for me, But never 'Mr Johnson'
Thanks for all the info gents , Ill poke around some more with this when i get home tonight .
Thanks again guys , would anyone want to interpret the phase on this two way? Call me out on anything that looks off.
S2 looks like the tweeter and the phase doesn't look right. I would expect it to wrap at the knee. Did you extract min phase?
John H, thanks to JP I did get that email
I'd say they were about 105-110* off at the crossover point, what's it look like with reversed phase?
deadhorse - leviathan - harbinger - shockwave (wip)
I guess you could get this phase shape with a large offset added to the tweeter
John H, thanks to JP I did get that email
How was phase obtained? It appears far too smooth considering the driver response, almost like the driver files contain no phase information and you are only looking at the phase of the electrical filter. Also, where the phase 'wraps' is unimportant, what is important is that your driver phase doesn't line up at the crossover point, you have ~75 degrees between drivers. How I would look at this is that the speaker forward axis is not the listening axis (straight ahead), if you go off-axis where the phase does line up (the speaker forward axis), you will now have a peak at the crossover frequency. Unfortunately, limits of Xsim prevent this sort of visual analysis.
I'm not deaf, I'm just not listening.
Thanks for the reply guys , I thought the phase on the tweeter looked off as well which is what promted the post. Since my upgraded version of SE does not want to play nice on my PC ,Ive decided to Let Bohden keep the money and move on , hence XSIM ans REW. Anyways .... sorry kind of went off the road there. All FRD and ZMA measurements were pull from the manufacturer specs and traced in . None of these measurements are measured in/on a baffle. Im just, at this point , trying to get my feet wet with XSIM and see what it can do . I really like it so far!
Ok, so if you've traced the frequency response, the standard process for this would be to take the manufacturer infinite baffle response, apply your baffle diffraction, apply HBT (tails), then extract minimum phase to get data accurate enough for crossover simulation use.
If it's just a traced frequency response, then you have missed a few steps, the results obtained in Xsim won't be anywhere close to reality since the data input doesn't include the baffle step or any phase data. That being said, Xsim is a super simple program, especially compared to SE, so it looks like you've got a handle on it.
As far as going through the steps above, for free solutions there's a few options out there, if you like spreadsheets Jeff B has put together a FRD response blender spreadsheet which has a simple diffraction simulator built in, so you can load your traced FRD, apply baffle step, tail the response and extract minimum phase all in one sheet.
I'm not deaf, I'm just not listening.
No, Charlie Laub made the Blender. Jeff's program is called Response Modeler.
My spreadsheet says they both made it, but that's beside the point.
I'm not deaf, I'm just not listening.
The Blender is more difficult to add the baffle to FRDs as you can only add the baffle diffraction into the data loaded in the LF data tab.
It's easier to add the baffle in Response Modeler to the factory FRD. Adjust the tails in the Blender by only loading the FRD saved from RM into the HF tab.
Unfortunately it's still a two spreadsheet process.
John H, thanks to JP I did get that email
edited January 2018
John, you can use the FRD blender, and skip the HF tab entirely since we are not actually using the sheet for blending.
One spreadsheet process using FRD Blender, step-by-step
1. Goto low frequency tab and load traced FRD
2. Goto diffraction tab and mess with the diffraction thingy, then press 'use diffraction model'
3. Goto control tab, you should see your response here now with the diffraction included. Go ahead and adjust the tails as needed to make a clean low and high end response.
4. Extract minimum phase.
5. Save the file
Easy peasy.
I'm not deaf, I'm just not listening.
The main problem I have with the Blender, is that you can't have other Excel sheets open when you extract phase with it. Since I have everything open when I'm doing a sim, it's a pain to have to close all but the Blender, do that one step, and then open them all back up again. This is why I never moved from using RM for a majority of my baffle/diff/phase routines.
Did you get a high frequency tail to adjust? dcibel said:
John, you can use the FRD blender, and skip the HF tab entirely since we are not actually using the sheet for blending.
One spreadsheet process using FRD Blender, step-by-step
1. Goto low frequency tab and load traced FRD
2. Goto diffraction tab and mess with the diffraction thingy, then press 'use diffraction model'
3. Goto control tab, you should see your response here now with the diffraction included. Go ahead and adjust the tails as needed to make a clean low and high end response.

John H, thanks to JP I did get that email
Tails are applied to the 'combined response', in my screenshot you can see the high frequency tail starts at 9kHz.
I'm not deaf, I'm just not listening.
The high frequency tail does not function for me unless I load data into the HF tab
John H, thanks to JP I did get that email
You just have to make sure the start freq for the tail is less than the last data point in your data.I verified it works in Office 2003 and 2016. If you can't make it work that way, take 3 columns, copy and paste from LF sheet to HF sheet and off you go again.
I'm not deaf, I'm just not listening.
Ha, thanks! My data was only going out to 19,900 Hz in most every FRD I tested!
Time to put RM back on the shelf...
John H, thanks to JP I did get that email
For most woofers and some mids, the tail should be before the end of the measurement so you remove the noise floor from where the driver rolls off naturally. For drivers that have response to 20kHz you are just providing the data trend beyond the measurement.
I'm not deaf, I'm just not listening.
I trim the LF tail on most tweeters and mids to clean up the response but for tweeters I'm mostly concerned about the phase implications. I'm looking to get closer to a true minimum phase response.
John H, thanks to JP I did get that email
Great info guys , thank you . Im sure it will make more sense once I have blender downloaded and start messing with it. This is getting bookmarked
Not trying to hijack your thread, but I have a question that I believe applies to the thread in general.
Is there a way to adjust x,y, and (particularly) z axis in xsim, since that will affect delay and ultimately phase alignment?
Not trying to hijack your thread, but I have a question that I believe applies to the thread in general.
Is there a way to adjust x,y, and (particularly) z axis in xsim, since that will affect delay and ultimately phase alignment?
You can't adjust X,Y in the current version, but you can adjust Z offset. Go to 'tune' the driver popup and there is a field there. Billl is currenty working on geometry and off axis response, per his thread a diyaudio.
Don, Donno, or 'Hey you' all work for me, But never 'Mr Johnson'
edited January 2018
To convert from PCD to XSim, I get good correlation using the right triangle theorem to calculate the (hypotenuse (measurement distance to the woofer) + Z offset) minus the tweeter measurement distance equals the XSim delay.
Just watch for the change in signs and the conversion to inches from meters.
John H, thanks to JP I did get that email
Not trying to hijack your thread, but I have a question that I believe applies to the thread in general.
Is there a way to adjust x,y, and (particularly) z axis in xsim, since that will affect delay and ultimately phase alignment?
You can't adjust X,Y in the current version, but you can adjust Z offset. Go to 'tune' the driver popup and there is a field there. Billl is currenty working on geometry and off axis response, per his thread a diyaudio.
XSim 3D with polars. Wow!! Thanks for the reference.