Linkwitz-Riley filters may sum well electrically, but not necessarily acoustically when they are added to a speaker. Better to recommend them as targets to achieve a unity crossover and choose appropriate filters for each individual application.
So why do people always recommend Linkwitz-Riley filters? When we look at a measurement like this and make sense, right? And we look at the sum between them and assuming that we have the necessary polarity inversion, it’s beautiful, right? It seems like a perfect result, but we should really be careful about recommending these filters when they’re used on speakers, because the acoustic result
Will not be the electrical result. 9 times out of 10, it’ll be something slightly different. So what I want to talk about with you today is how you can get started playing with filters investigating the results
And while you should probably never recommend one specific kind of filter, unless you are fully familiar with all of the circumstances around its implementation. So today I’m going to be doing this demo and Crosslite, which is definitely a piece of software that I recommend, but you can also do this in free software.
So I’m going to do this demo and Crosslite, but then at the end, I’m also going to show you how to do it in REW. So if you want to download REW or you already have it, you can open that up and you can do all of these steps along with me, just on a slightly different platform. Okay. So learning how to use filters. It’s definitely a good idea to start playing around with them in the electrical domain. What is the result? If we have these 2 36 dB per octave Linkwitz Riley filters, what does that look like? What does it look like?If we change the topology, the slope? What if we change the frequency between them investigating all those things in the electrical domain is really good, because then you’re just focusing on the combination of these two shapes.
The next step then is to look at what happens when we start applying these to speakers. And I’ll tell you now that what I teach my students is to start thinking about these electrical filters. Not as specifically what you want to apply, but as the target, this is the result that you might like to achieve.
If you like the shape, if you like the sound of this, if you would like to create a unity class crossover, then this is a great target to pursue. So here in Crosslite, I’ve just created these two separate channels to serve as our targets.
And so what I’m going to do now is I’m going to insert a main and sub speaker that I downloaded from Tracebook, because I wanted you to be able to use the exact same data. So let me show you how I did that. So if you go to chase book.org and you go to the search page, then you can just search for subwoofers. And that’s what I did here. It just grabbed one of the first ones. And then I downloaded the CSV file. That’s just a text file that has magnitude phase and coherence values, all of this stuff that you see in this graph here.
And then I downloaded this speaker because I measured it and I’m familiar. And you’ll notice that these are two speakers from two different brands. They are not designed to work together. And I thought that would make an interesting demo since they are quite different. Okay. So I downloaded the CSV files and then I imported them here into Crosslite. So here’s my main and sub and here they are on top of my targets. So let’s start just working on the main speaker. Okay. So I know this is what, where I want to end up. So what do I need to do? I know this is a 36 dB per octave slope, but if I add a 36 DB per octave filter, is that going to give me the result that I want? Let’s try it. So I’ll change this to Linkwitz Riley 36 dB per octave. We know we’re heading towards 80 Hertz, so I’ll insert that and.
This is not what we want. We’ve gone a little bit too far, so I could play around with the frequency here, but, we can get this part to match, but not this part to match. So maybe I like this and I could insert a filter here, but I say, you know what, let me keep playing around with this. What if I go to a less steep slope so I’ll go to 24 DB per octave. And now I play around with this and I see, wow, this gets me. Really close. This gets me part of the way there. And now I’m looking at this and I’m thinking, you know what? This is not something I want to EQ because this is just some kind of reflection. There’s a little bit of ripple here, but I am curious to see what this would look like.
Without the ripple. So I’m going to do two things. Number one, I’m going to run the noise reduction function here in Crosslite. I’ve done that already. And here’s what that measurement looks like. So a little bit less ripple here. And now this is totally improper EQ, but I am going to just apply some little filters here, just cause I want to see what this would look like without this ripple, as I mentioned.
So I’m going to put in some little filters here just to make that ripple all go away. And now I have a nice fit here. So ideally it looks like I would have only had to use one filter that Linkwitz Riley 24 DB per octave filter, maybe one other parametric filter here to achieve my target.
So as you can see, we wanted a 36 DB per octave result. But we didn’t use a 36 dB per octave filter to get there because this speaker naturally, already has some amount of roll off. You’re never going to measure any speakers that are just flat from DC to daylight, that they’re just flat all the way across and then you can add filters.
That’s not the way filters work. You’re always adding on top of something that’s already there. And so we can look at the math of that and that’s easy to figure out and we can do the numbers, but it’s also just fun to play with and just add things and see what gets you to the target as efficiently as possible.
And sounds good. All right. Let’s take a look at the sub. Let’s zoom in a little bit here. What do we need to do here? This is sometimes some of the most challenging work because here at the crossover frequency, we are already locked in. By the way, it might help to have some kind of a target here. So over here, I’m going to turn on the target today. I’m just going to use a flat target.
That’s not what I always use, and you don’t have to use the same thing, but that’s what I’m using today. So I have this target. Here’s the top of it. And then I turn on this target Delta so that I can see where the -6 dB down point is. Okay. And I see that my sub is already hitting that point. And so what do I want to do here?
What if I use something super sharp?
Imagining the shape here and I’m thinking what if I come in with a filter that’s super sharp and maybe just push this part down this way a little bit. So let’s try that first. So I put in this 48 DB per octave link what’s Riley filter, and that gets me a fit here, but now I’ll still have to work on this region and this region, and maybe that’s fine, but I’ve played with this a little bit earlier before I started this video and I came up with what I think is. A more efficient solution, which is first applying a six DB per octave first order low-pass filter.
So answer this and I play around with this a little bit and oops, I’m probably moving the wrong filter. Let’s see. I need the red guy. There it is. There we go. So with this first order filter, now I’ve got this top getting closer here. I’m getting closer to the bottom here and now I just need one filter here to make a slightly better match. And now I could keep playing around with these until I find a closer match. Okay. So let’s say that I’m happy with that.
And now let’s look at everything together. so here’s both of my measurements now with the targets. Now I’m going to mute the targets. And now I can look at the sum between these guys and I already want to have kind of a picture in my mind of what I expect. So at this -6 dB point, I expect 6 dB of summation, which should hit up here, and then I’m hoping for a nice flatline transition across here. That is the goal of the unity crossover.
So I turned on the sum, and I see that I have a big problem here. And I remember that’s right. 36 DB per octave target. I’m going to need a polarity and version. And then I take a look at the max sum here, and I know it’s a little bit difficult to see because both of these lines are black, but I’m turning. Max on and off max signifies the total potential summation, if everything was perfectly phased aligned.
So I see that I still have some work to do here in the high end. So what can I do about this? I’ll turn off the sum for a second and let’s place some markers here, just so I can see this is the crossover region that I’m going to focus on for the moment. And I take a look at the phase here and I see, oh, this is slightly problematic here.
By the way, before I go down this path, I should just point out that this some that we had, this could be good enough and I could just move on, but if I want to keep playing with this. I could try different filters. I could try a different combination of filters to see if there’s something that would achieve a better phase compatibility here.
But for the moment, I’m going to try a solution with just a second order, all pass filter here. And when I move this around, maybe I can find a solution here. A different compromise. So it seems like we have a little bit better response through the entire crossover region there. And now when we take a look at the sum and we turn the max off and on now we have a much closer result.
Okay. So again, I’m not going to say that this is the best solution. This is just what I came up with in the few minutes that I’ve been working on it with you today. If I spent another couple hours on this and tried out a bunch of different variations, I might find something that I liked better.
And then ideally I could deploy all of those as presets and then do listening tests and find the one that I actually liked the best. And doesn’t just look the best on the screen. So you saw here that both with the main and the sub the filters that I applied. Didn’t necessarily have anything to do with the target. So I wanted to end up with this 36 dB per octave result that gave me this nice flat result here through the crossover region, but I just did whatever was necessary. I was just playing around and experimenting with filters until I got to that result.
So this is why it’s probably a bad idea to ever recommend. Just one kind of filter. Don’t tell people, just use Linkwitz-Riley filters, but you could tell them head towards this kind of result, especially if people are just starting out. They’ve never heard of this kind of idea, but. Yeah, totally fine.
To just start with some kind of a template, use Linkwitz-Riley filters to experiment with them in the electrical domain. Look at the result and then play around with that and use that as a target. Okay. So I promised I would also show you how to do this in REW, in case you don’t have Crosslite at home and you want to do this work along with me.
So let’s try that. So I’ve got an REW open and before I actually import those measurements, I want to show you how you can also play around with electrical filters in REW. So you open the EQ tab and then you can just insert filters. Down here. It says crossover filters and I can use the exact same thing I was using in Crosslite here. We see the filter prediction, and now if I want to actually play with that, then I go over here and I click generate measurement from filters, and then I’ll do the same thing for my low pass filter. Generate measurement from filters. There we go. Now I can play around with these and look at the result and compare phase and all that stuff. Just like I was doing in Crosslite. All right. Let’s import those measurements. So I’ll just grab them from the finder and I’ll just drop them here and here they are. So you can see this is this, the raw data. And so now I would like to take the next steps to change the gain between them so that they line up and they look more like this, what I was doing in Crosslite and then start experimenting with filters to achieve this target.
And let me just show you how to take the first steps with that. So here over on the left, do you want to select whichever one you want to apply the filter? And then go to the EEQ window. And then from here, you’re going to choose a target. So I typically choose the speaker driver, and then, we’re going to choose our high pass filter, which is Linkwitz-Riley six. And then set this at zero or three, line it up with your measurement here. And so now we have a target here and now we can start applying filters in pursuit of that target. So I think over in Crosslite we added a fourth order filter. Let’s try that. Yeah, there we go. So that’s getting pretty close.
All of these audio analyzers and modeling and prediction software, they all look slightly different, but the data is all very similar. We see a magnitude response here. We have a target, we have a prediction, so we need some way to practice this stuff. REW is great. Crosslite is great.
There’s plenty of tools out there. So pick one of them and figure out a way that you can play with this stuff at home. And so now that I’ve shown you how to get started here in REW, maybe go back to the beginning of this video and go through step-by-step as I was doing it in Crosslite and you can do the same thing in REW.
Okay. Let me know what questions come up for you. Let me know if you have any suggestions for me and thanks for watching.