Basic Guide to Crossovers (Part 2)

[MiniVanMan]

Okay guys, here’s the next thread continuing on my previous “Basic Guide to Crossovers” http://www.diymobileaudio.com/forum/showthread.php?t=7160 thread. This is going to go into a bit more detail, actually quite a bit more detail.

I’m going to use a design I’m currently working on using the Dayton RS180-8 woofer, and the Seas 27TBFCG tweeter. I will also be modeling up a crossover for the Seas CA18 with the 27TBFCG tweeter as well to compare the two woofers. The reason for doing this is to show how vastly different two drivers can be. In fact the only real similarity is the 8 ohm impedance.

A few notes before I start. These are all modeled using Zaph response graphs, and measured T/S parameters. I did this for consistency of the measurement parameters. Also, understand that I’m using Zaph’s measured responses and not my own. These are modeled responses, and will not be perfect in the real world. I’ll be using Speaker Workshop for the crossover modeling. It’s a clunky program, but works very well. With some help from the FRD Consortium http://www.pvconsultants.com/audio/frdgroup.htm, I’ve been able to utilize enough modeling tools as to create a realistic simulation of how everything works together, and in turn create this tutorial.

Now, why would I go to all this trouble? Well, frankly I’m pretty tired of hearing “can I use this crossover with these drivers?”, or “this passive crossover has been optimized for a car”. Hopefully, by the end of this tutorial, you will have your answers, and many of the myths surrounding passive crossovers utilized in an automobile will be either debunked, or at least questioned.

This will in no way be a slam against passive crossovers. In fact, I plan on (with a little more tweaking) implementing this design for my own personal use. So, I’m not “anti-passive”. But you will definitely see how response changes with just a small variance in many different factors.

I’m not going to pull any punches here. This will be extremely long. I will be taking into account enclosure response, and baffle step. I will be using graphs, then some more graphs, and even more after that. I definitely recommend that you have read, and understand my previous thread “Basic Guide to Crossovers”. If you don’t, then this thread will be more or less meaningless.

So, on to the good stuff. Like I said, I’ll be using the Dayton RS180-8 and the Seas 27TBFCG. They will be going into a bookshelf speaker. So, I will be modeling this from the design of the enclosure to the design of the crossover. Then I will do the Seas CA18. Why I’m going to compare the two is because the Dayton RS180 is highly regarded, but is, as you will see, a very difficult driver to work with. Conversely, the CA18 is very easy to work with. So, on to the RS180/27TBFCG combo.

First we need basic frequency response data for the driver. Here it is. Notice the HUGE breakup starting around 3.5-4k all the up to about 10k. It’s nasty, nasty, nasty. Already, there is going to need to be very special attention paid to it in our crossover. Also, pay attention to the SPL numbers on the sides of the graphs. You’ll find that dips and peaks are relatively small (+/- 1-2db). The graphs just make them look bigger. If you can design a speaker that +/- 1 db throughout it’s bandwidth, you’ve done pretty well.

http://img.photobucket.com/albums/v641/glevii/180Trace.jpg

On to the enclosure. It’s a basic bookshelf. I’ve modeled the best response to be an 11 liter box, tuned to 46 hz. As you can see by the graphs, the driver stays under it’s Xmax limit, and thus we avoid overexcursion. We can go bigger, and tune lower, but the driver reaches it’s limits in Xmax, and it’s just something I want to avoid.

http://img.photobucket.com/albums/v641/glevii/VBResponseDaytonRS180S-8-2.gif

http://img.photobucket.com/albums/v641/glevii/VBExcursionDaytonRS180S-8-2.gif

However, let’s look at how this driver will react in a door. I’m going to use .75 cubes, which roughly translates to 21 liters to simulate a door’s volume 18” x 24” x 3”. This is only the first point where a car starts to become difficult. Anyway, here are your graphs.

http://img.photobucket.com/albums/v641/glevii/CB21literRS180.gif

http://img.photobucket.com/albums/v641/glevii/CB21liter180Xmax.gif

I’m actually fairly surprised by these results. This driver will play down to 70 hz before running out of Xmax. Now that’s at 60 watts. Throw more power at them and that number rises. Anyway, not bad overall. It is a big difference over a properly built vented box. Not bad overall though.

So, now we’ve calculated enclosure response. Now for those that don’t know, when mounting a speaker onto a baffle of finite width, you get a nasty rise in response known as baffle step. This can be accurately calculated in many circumstances, but in a car is impossible to know. But let’s first see how the driver reacts to the enclosure we’re going to build for it. I’ve decided to go with a baffle 13.5” x 9”. Now look at my combined response. This is a combined response taking into consideration the baffle, and the enclosure.

Here's the predicted baffle step response.

http://img.photobucket.com/albums/v641/glevii/180Baffle.jpg

Now here’s the same response for a very large baffle. Like I said before, it’s impossible to accurately predict baffle step in a car, so I’m going to go as large as I can. 40” x 40”. This is not accurate, as a car can react like a true infinite baffle. The following graphs do show how a change in baffle size affects overall response. Also, listening angle has an affect, so I put a 45 degree listening angle graph in as well. There’s a huge difference in response. You can also see the huge difference in response between the small baffle and the large baffle. So, when mounting a speaker, it's fairly critical that you know the listening angle and the size of the actual mounting location.

Here's 0 degrees.

http://img.photobucket.com/albums/v641/glevii/180Car0Degrees.jpg

Here's 45 degrees. (It's the highest the program will let me model for)

http://img.photobucket.com/albums/v641/glevii/180Car45Degrees.jpg

Next, here’s my summed response. This includes enclosure response and baffle step. This is ugly for any alignment. Some serious crossover work is going to need to be done. I'm not going to do a combined response for a car door. There is just no way to accurately predict. Suffice to say again though, that response can change dramatically from car to car. The following response is predicted for a well built enclosure. Every parameter entered is accurate, so the prediction is fairly accurate as well. It simply just can't be done for a car without extensive measuring equipment.

http://img.photobucket.com/albums/v641/glevii/Summed180Resp.jpg

So, now that I have a completed response for my mid, I'll move onto the tweeter. So, those of you still with me, head on to the next post, and read on.

[MiniVanMan]

I picked the Seas 27TBFCG tweeter due to it's low cost, high performance. It has tested well in several places, and is a favorite of Zaph as well http://www.zaphaudio.com/. He uses this tweeter in his L18 bookshelf design, of which we will be doing a direct comparison later. When playing with this tweeter, I can see why it's so well liked. It's very easy to work with. Response remains flat up to about 15k, where it starts to dip off. That helps take away a lot of the shrillness that a metal dome can exhibit on the high end. Also, the breakup node is out of the hearing spectrum, so it won't have to be dealt with.

Anyway, on to the tweeter response graphs. There is no need to model the tweeter in an enclosure, so we can skip that part. So, here's the measured response of the tweeter.

http://img.photobucket.com/albums/v641/glevii/27TTrace.jpg

Here's the predicted baffle step response.

http://img.photobucket.com/albums/v641/glevii/27TBaffle.jpg

And finally here is the summed response of the tweeter when mounted on our baffle.

http://img.photobucket.com/albums/v641/glevii/Summed27TResp.jpg

Really not a whole lot to complain about here.

Next is the CA18RNX. I'll keep this brief as most of the notes have already been stated with the RS180.

Here's the measured response.

http://img.photobucket.com/albums/v641/glevii/CA18Trace.jpg

Here is our enclosure. This is quick, and I didn't play with it for very long. This works though.

http://img.photobucket.com/albums/v641/glevii/VBResponseSeasCA18RNX.gif

Here is the response when placed in our leaky, sealed door. Remember, I guessed 21 liters. It could be more, or it could be less.

http://img.photobucket.com/albums/v641/glevii/CB21literRespCA18.gif

http://img.photobucket.com/albums/v641/glevii/CB21literXmaxCA18.gif

Notice there is about a 2-3db gain in low end response (80 hz) over the RS180. In my experience this is about correct.

Okay, here's the predicted baffle step measurement.

http://img.photobucket.com/albums/v641/glevii/CA18Baffle.jpg

And, here is our summed response.

http://img.photobucket.com/albums/v641/glevii/CA18Combined.jpg

As you can very easily see, this driver will be much easier to work with. There is no large breakup node, so the crossover should be much simpler.

Now, on to the crossovers themselves. The next post will be pretty scattered, and may be hard to follow. There are so many variables that it's impossible to cover them all. I'll try and keep it as simple as possible.

[MiniVanMan]

Now, for the really, really hard part. Crossovers themselves. The whole point of this tutorial is to show that crossovers are not as simple as throwing an inductor and capacitor together and you have a crossover. Technically you do, however, as you've seen with the RS180, and CA18, the requirements are going to be vastly different.

Please don't comment on my crossovers. I'm not going for perfection here. I'm looking to prove some points. I've titled this thread as "Basic" for a reason. I'm not going to go into phasing, and I won't be posting phase plots. I'm not overly concerned with impedance at this point. I'm mostly keeping it to "inductor/capacitor" networks.

So, let's start simple here. A simple 6 db slope for our RS180.

http://img.photobucket.com/albums/v641/glevii/6dbnetwork180.jpg

Here's the response.

http://img.photobucket.com/albums/v641/glevii/6db180.jpg

There are a couple of things to note. If you plug in an inductor value of 3 mh into a crossover calculator, you'll get a crossover point of 425 hz. However, we are compensating for a rise due to baffle step. As you can see, this simple inductor vastly improves our response, and we achieve a relatively flat response out to 2k. It can be said that we've just compensated for out 6-7 db of baffle step. We still have that nasty breakup node to deal with though.

On to a 12 db per octave slope. Maybe this will help our breakup node.

http://img.photobucket.com/albums/v641/glevii/12db180network.jpg

http://img.photobucket.com/albums/v641/glevii/12db180resp.jpg

Okay, it helps a little bit, but not much. It's still there, and at this point will still be audible.

However, it's been tamed quite a bit. So there could be a couple of ways to approach this. We could notch the breakup, which is what Zaph does in his L18 design, or we could move to a higher order. First, let's work with a proven Zaph design, and notch this thing. http://www.zaphaudio.com/audio-speaker17.html

http://img.photobucket.com/albums/v641/glevii/ZaphL18Woofer.jpg

http://img.photobucket.com/albums/v641/glevii/ZaphL18WooferResp.jpg

As you can see, this design is not bad. However, one of the key differences between the L18 woofer and the RS180 is the breakup. The L18 is also a bit more efficient, but, that's not a concern right now. Zaph states that his design implements about 5 db of baffle step. As we can see, and by comparing with my first graphs with the larger inductor, this is pretty accurate. Response is still rising into the 1-2k range. However, the real problem is still the breakup node. The L18's breakup has a much smaller bandwidth than the RS180s. This poses a problem as you can see, that we still have breakup in the 5-6k range, and again at the 8k range. The 5-6k is quite troublesome as when you add 20 db of volume (100 db listening level), that breakup will present itself at 80 db. That's audible. The 8k breakup will be around 70 db. Still moderately audible. This crossover IS NOT a viable solution for the RS180. However, it is important to note that the notch did in fact do it's job. (The notch filter is actually just L3 in the circuit).

So, to go on, while Zaph's crossover works well for the L18, it won't be suitable for our RS180. So, more designing is going to be needed. We can try to tweak that design a bit and see what we come up with.

http://img.photobucket.com/albums/v641/glevii/Notched12db180network.jpg

http://img.photobucket.com/albums/v641/glevii/Notched12db180Resp.jpg

Here we recentered the notch. It gives the appearance of almost eliminating the breakup. It's a tough call though. Zaph concluded with his design. On to reality. Measured response matches modeled response fairly close. As mentioned above, the notch does not take the extra high frequency garbage out as much as expected, but it's still 30 dB down and essentially off the radar. So that being said, knowing that the RS180s breakup is worse, I'm not going to chance it. Let's move on to the next order.

Let's move on to an 18db per octave slope. This will add another inductor in series.

http://img.photobucket.com/albums/v641/glevii/18db180network.jpg

http://img.photobucket.com/albums/v641/glevii/18db180resp.jpg

Now we're getting somewhere. The breakup looks like it's starting to get controlled. However, the way the breakup is behaving, this network looks more like a 12 db per octave slope. Still not enough, so we may be relegated to the granddaddy of them all, the mighty 24 db/octave slope.

http://img.photobucket.com/albums/v641/glevii/24dbwoofernetworkcrossover.jpg

http://img.photobucket.com/albums/v641/glevii/24dbwooferNetwork.jpg

Here we have a decent response it looks like. Breakup looks controlled. However, even though the filter is technically a 24 db/octave network, it's behaving more like a 12 db up to about 8k. This is definitely a product of the breakup. It may be reasonably acceptable though at this point. If we wanted to move further we could add a notch and control the breakup even further. However, we're starting to get expensive here. Inductors are pretty expensive, and we're utilizing 2 of them for this network. We're talking about $25.00 in inductors. For consistency though, let's look at a notched network.

http://img.photobucket.com/albums/v641/glevii/Notched24db180Network.jpg

http://img.photobucket.com/albums/v641/glevii/Notched24db180Resp.jpg

There we go. Now this is a network I can live with. There's not a lot of guessing here anymore. This will definitely do the trick, and the breakup is eliminated. Like I said before, though, this is an expensive crossover. The RS180 just lost a lot of value. Do demontrate this, I'll move on to the CA18. You'll be surprised.

[MiniVanMan]

Refer to the modeled CA18 responses above. I'm going to move straight on to the network. Pretty simple here.

http://img.photobucket.com/albums/v641/glevii/CA186dbNetwork.jpg

http://img.photobucket.com/albums/v641/glevii/CA186dbresp.jpg

I rushed this a bit, but as you can see, one simple inductor and we have a pretty decent response. To smooth it out a bit, let's see what a 12db network would look like.

http://img.photobucket.com/albums/v641/glevii/CA1812dbNetwork.jpg

http://img.photobucket.com/albums/v641/glevii/CA1812dbResp.jpg

Completely different animal than the RS180. This is a much cheaper crossover. The RS180 with the extensive network will still come out cheaper than the CA18 and it's network, but the CA18 is definitely easier to work with.

You can definitely see that the RS180 IS NOT compatible with the CA18s network. The CA18 may work with the RS180 network. Let's try that out.

http://img.photobucket.com/albums/v641/glevii/RS180CA18network.jpg

http://img.photobucket.com/albums/v641/glevii/RS180CA18Resp.jpg

Could definitely be better. This is not an optimal network, even in a controlled environment. It would work, but as you can see there is a dip in the midrange area. So, again, just another example on non-cross compatibility.

[MiniVanMan]

Now the tweeter at this point is just a formality. I think I've proven what I've wanted to prove. I'll save you the extra graphs and pictures, and just go straight into my conclusion and summary.

Myth - "This passive crossover has been optimized to work in a car."

Reality - No it hasn't. As we've seen, there is no way to build a blanket crossover that is optimized for something as dramatically different as a vehicle's interior space. With reflections, curves, angles, cancellations, etc, the only way to build an optimized crossover is to do extensive measuring of each driver in it's mounting location as it will be played. After doing the measurements, then you build the crossover to suit. I don't care how expensive the component set is, the crossover IS NOT optimized.

Myth - "As long as the impedance and crossover point are consistent, you can interchange crossovers."

Reality - Ummmmm, no. I think this has been abundantly demonstrated, and I don't need to say anything more about it.

Please understand that this is still very basic when it comes to crossover design. I can hope that Werewolf or npdang chime in here and reinforce that statement. Crossover design can be an art. To do it right, you really need to have extensive measuring equipment. You also need to know exactly where the speaker will be placed. Placing a speaker against a wall can really affect your baffle step. In fact the above RS180 crossover would require the speaker to be far away from any walls. Put it next to a TV, about a foot from the wall, and the response changes.

All in all, my goal here is to prove why some of the common myths are false. Hopefully, I've done that. Also, understand that an active crossover faces the same challenges. An active crossover can help you achieve an optimal crossover point, but it can't give you an optimal response. You still need a good amount of EQ.

If you see something blatantly wrong with anything I've posted, please feel free to correct me. If you feel like being nitpicky, then please save me the aggravation. I've spent a lot of time on this, and will correct mistakes, but I know this is not all inclusive. It wasn't meant to be. It was meant to be a simple explanation of a complex topic.

Thank you for your time, and for those that have made it this far, I salute you.

[tyroneshoes]

Very interesting read. Ive been building crossovers but have never designed them. If you would like, I can post a crossover Roman Bednarek designed for me using the rs180 and seas 29TAF/W metal dome. It sounds great in my setup and Im curious how it might differ.

[MiniVanMan]

Very interesting read. Ive been building crossovers but have never designed them. If you would like, I can post a crossover Roman Bednarek designed for me using the rs180 and seas 29TAF/W metal dome. It sounds great in my setup and Im curious how it might differ.

Sure, go ahead. This isn't necessarily a crossover design tutorial, but it would be good to compare working crossovers for the same drivers.

[tyroneshoes]

http://img476.imageshack.us/img476/6396/rs18029taf2wayid9.jpg

they seem somewhat different. Excellent sounding bookshelves though.

[MiniVanMan]

http://img.photobucket.com/albums/v641/glevii/RS180AltNotch.jpg

Very similar actually. A bit less baffle step compensation, which is what I'm going to go with in the end, as I will have these speakers acting as fronts for my HT. Other than that, he just notched it differently. In fact, his design is a bit cheaper, though the breakup isn't as neat and tidy and the inductor notch I used. Still inaudible on his end though.

Honestly, in a design like this, the only components that need to be pretty accurate are the initial inductor and the notch filter. The other components have quite a bit of flexibility. In fact playing with his design, I can illustrate this. Every value except the first inductor is different. However, the results are fairly similar, but I got the notch another 10 db down, and I eliminated the shelf at 4k which is where the breakup starts. Honestly though, it's all semantics at this point. Using the capacitor to notch is a great idea for this design.

http://img.photobucket.com/albums/v641/glevii/RS180AltNotch2.jpg

I quite like this design. Thanks for posting.

[tyroneshoes]

Still have a lot to learn. Thanks for this tutorial.

These are my fronts in my home theatre as well and it is an excellent sounding speaker. Very content with them.

[a$$hole]

MiniVanMan,

Excellent !! Thanks, for taking all the time to explain this.

Wowser, caps, inductors, resistors,Zoebel networks.

Energy equals mass times the speed of light squared...

umm, yeah, I'll take fries with that!!:)

[seagrasser]

Excellent write up

Any way you could send me the file(s)? So I could learn poke at it to gain some valuable information.

I just dowloaded speaker workshop and it looks overwhelming.

Cheers,

Jason

[MiniVanMan]

Speaker Workshop is VERY overwhelming. It's a very clunky program, and not very intuitive, especially when doing your own measuring of drivers.

Here's a pretty easy way to get started though. Read through this link.

http://www.rjbaudio.com/Audiofiles/FRDtools.html

It's pretty much step by step instructions on how to create a loudspeaker using just measurement data. Of course, that's not going to be optimal, but it will be close.

It will get you to a point where you can see how various components of a crossover will affect a drivers response. As for, a tutorial on how to use Speaker Workshop, here it is.

http://www.audiodiycentral.com/ntutorials.shtml

It's enormous, so plan on being overwhelmed again.

[seagrasser]

Great this will get the interactive experience I need to "poke and prod" at my next project using p18's

Thanks,

Jason

[hh47]

Awesome job! Great read, and I learned some shit.

So I guess the passive crossovers that come in the box with a component set are "optimized for vehicles" or whatever by using some generic simulated response of an "average" vehicle with speakers in most likely mounting locations and so on...

You also need to know exactly where the speaker will be placed. Placing a speaker against a wall can really affect your baffle step. In fact the above RS180 crossover would require the speaker to be far away from any walls. Put it next to a TV, about a foot from the wall, and the response changes.

I'm curious as to how you would adjust things if you were planning on having the speakers close to the wall as you mentioned. Is there a general rule of thumb for designing around that?

[mk1982]

bloody awesome. thanks !

[azngotskills]

Just wanted to BUMP this up....great info here and appreciate the time it took you to put it together :)

[iyamwutiam]

Hi-what would be the phase relationship (for example at 24db slope it is 18- degrees) with a 30 db slope- also many dbs up/down would it be at the xover point?

[bmaupin]

Hi-what would be the phase relationship (for example at 24db slope it is 18- degrees) with a 30 db slope- also many dbs up/down would it be at the xover point?

I am not sure what you're asking, but I will try to respond.

At the crossover point and assuming that the slopes (high & low pass) are the same, the phase will be 45 degrees times the order of the slope (1st order = 6db, etc). So for a 24db (4th order) slopes, the phase will be 180 degrees at the crossover point (and then eventually 360 degrees). I believe that Linkwitz-Riley filters can only be even order.

I also believe that part of the definition/characteristic of a Linkwitz-Riley filter is that no matter what the filter slopes are, the response is always 6db down at the crossover point.

Also note that since loudspeakers do not have flat frequency/phase response, what really matters is the effect of the filter to produce a resultant target acoustic response. Say you are shooting for 4th order slopes, I find that the norm when creating a passive crossover is that a 2nd order filter on the woofer and a 3rd order filter on the tweeter will result in the target acoustic response.

If you are dealing with active crossovers, basically take the target response, subtract the driver's response and that is what the active filter needs to produce. This is where processor units come in and do this for you.

Someone please correct me if I am wrong on any of this.

Hope this helps :) Google "Linkwitz-Riley" and you will probably find an understandable article somewhere.

Brad

[solacedagony]

Also, consider that that's only the ELECTRICAL response. Acoustic response could be completely different.

[iyamwutiam]

Also, consider that that's only the ELECTRICAL response. Acoustic response could be completely different.

Yes- thats the poitnt - I am sure the acoustical response may vary- but I am thinkiing about the phase changes (which I guess would be acoustical?).

Another question is - over-lapping xvers or perhaps the word cascading is better. For example if I use a 24 dbs slope at 2000HZ thru a processor and then use an outboard xover at the same settings (24 dbs @ 2000) -- would I get a 48 db slope at 2000Hz?

[bmaupin]

Another question is - over-lapping xvers or perhaps the word cascading is better. For example if I use a 24 dbs slope at 2000HZ thru a processor and then use an outboard xover at the same settings (24 dbs @ 2000) -- would I get a 48 db slope at 2000Hz?

Simple answer: No. Why? Simply look at the response at 2kHz as the result of the cascaded filters - the response is 12db down instead of 6db. The crossover point is more like 1.8kHz or so.

I do not know of any software that allows simple creation and simulation of cascaded filters besides SPICE (circuit simulator), but maybe someone knows of one and will chime in - I'd be interested as well.

[solacedagony]

Simple answer: No. Why? Simply look at the response at 2kHz as the result of the cascaded filters - the response is 12db down instead of 6db. The crossover point is more like 1.8kHz or so.

I do not know of any software that allows simple creation and simulation of cascaded filters besides SPICE (circuit simulator), but maybe someone knows of one and will chime in - I'd be interested as well.

Like you said, the simple answer is no. If you strictly consider the -3db point to be the crossover point, then that's true. Otherwise, he is correct because your -3db point is now a -6db/-12db point from the cascading filters.

Also, when you're talking about phase and FR WITHOUT taking into consideration environmental factors, you're talking about electrical response (or your theoretical/calculated response). Do a measurement in your car/house and you've got acoustic/actual response.

Edit: Sorry for the kind of off-topic babbling, MiniVanMan. If you want this in another thread, we can surely bring it there.

[iyamwutiam]

Speaking of slightly off topic babbling - If one takes a comp set and throws away the passives - the question is are these passives on say QSD/Rainbows/Lotus/Dynaudios - already packaged with Zoebels- to decrease a rise in impedance -especially when it comes to tweeters and woofers as you approach the resonance frequencies?

So when you use active - how do you calcualte for the increasing impedance as you aproach the Fs of your speaker- more power (a lot more power) - how does this affect what we 'think' we should be hearing/

[azngotskills]

from what werewolf was saying in an older post: zobel networks dont matter when you going active

[Sephiroth619]

This was an excellent read. Minivanman, you are great with words. Everything just kinda flowed and kept me interested throughout the entire read. Very informative.

I wonder why I rarely visit the Tutorial forums.

[MiniVanMan]

Bump!!

This has fallen way down the list, and I've seen some questions popping up that this can help with.