Original subtitles

If you have ever tried to integrate a

subwoofer with speakers, you already

know the problem. There are hundreds of

tutorials online, dozens of recommended

crossover methods, and yet most systems

still don't sound flat, don't align

properly, and don't sound right. In this

video, I'm going to show you a correct,

repeatable, and optimal way to combine a

subwoofer and a speaker. We will use

room EQ Wizard. And what I will show you

here is fundamentally different from the

random methods you'll find all over the

web. So, let's start.

W beta 1112. And here is an actual

speaker and subwoofer measurement with

correct timing. When I say correct

timing, it means

the impulse responses

are exactly as measured. So speaker

starts and the subwoofer speak is

quite a bit delayed

around 60 milliseconds in my case. So

it's important that you should measure

speaker and subwoofer with acoustic

timing reference or loop back timing

reference. Impulse responses should be

timed correctly. If you don't use

acoustic reference, every single

measurement will show up at time zero

and you will lose the relative timing

information which makes to align a

subwoofer and a speaker to each other

impossible.

So now the first step is target curve

suitable for your speaker and sub.

You select speaker. This blue tag here

shows when you can double click any

measurement in RW the blue tag moves all

the operations you're doing are based on

the one with the blue tag. So if I click

info window right now for example this

is showing subwoofer

right now info window is showing the

speaker

similarly IR windows

that's its functionality

so let's start going to EQ windows with

the speaker

change to 1 over 48

octave And so we see the proper response

of the speaker in the highest detail

possible.

Now let me

pick up my favorite curve. Calculate

target level from response. NRW already

gave us a target level. By the way, if

you use no smoothing, the target level

will be slightly lower. Look calculate

target level for most cases but I prefer

one or in any other smoothing it will be

the same only no smoothing will make it

a bit lower

regardless. Now what is the rolloff

frequency correct crossover frequency

where the bass starts rolling off with

the speaker? First you have to know that

the speaker is designed to roll with 12

dB per octave butter on its own and you

add that another butterwork 12 dB per

octave highp pass filter. I will explain

why soon

and let's say let's start from 80 Hz.

This is a pretty small speaker. It's one

of my focal satellites.

As you see with this 12 dB per octave

butterwork highp pass filter, the

speaker's rolloff moved from this one to

this one. Okay.

And the target type obviously

has to have 24 dB per octave.

So what you're doing here is to

understand where

this line correctly

tracks this line.

So it's something like 120. Let's say we

can remove this one.

What will it be if 120? So this should

also be 120.

And as you see, it's even higher than

120. This is a very small speaker.

Let's try 150

and 150 here.

Now, this is too much. I think something

like that. So, this is the place where

you're going to spend the most time.

This one. So something like let's say

140 htz

and calculate target level. The bass

response will not change the target

level. It's calculated from high

frequency response.

But this is your speaker's correct

crossover frequency. Only after as you

see

the butter word 12 dB per octave highp

pass filter is applied.

That should coincide with this 24 dB per

octave LF cutff here with full range

speaker. Now you can generate

measurement from target shape. We know

the speaker's target level. It's 71.8 dB

for this particular target curve

and just select none here and generate

another one. So we have the full target

curve on screen. Now we close EQ windows

and as you see we have a target with the

rolloff and also a part proper target

like so.

Now the second thing is to determine the

subwoofer correct volume for this given

target care.

For that you need to

determine subwoofer's target level.

First

select subwoofer and you should know

give or take the roll of frequencies of

your subwoofer.

You can find it in its text specs. this

subwoofer.

So this is the other way around. 200 Hz

is it maximum and 40 Hz is it minimum

according to spec. But every subwoofer

has different rolloffs and rolloff

slopes. So I think this one mine is

here.

This is also some part you have to spend

some time on.

Now when I'm changing this, I'm looking

at the slope of this with this one

and trying to find a suitable area. And

also on this side, it's important. For

example, it's definitely not 24.

It's definitely not 12. It's probably

24

And I think it's more like 182.

And then calculate target level. Every

time you calculate target level and also

don't forget to make this 148 and then

now if you calculate target level it's

going to go a little bit up which you

want. Yes. So normally because you don't

even know the 40 and 200 that easy. You

should also do these calibrations and

ideally I should do that because this

subwoofer's measurement shows that it's

not really 40. It's even

something like that probably. But when

you calculate target level, it will move

depending on these numbers. So it's a

continuous calculate target level.

Readjust calculate target level until

you see it perfectly represents your

subwoofer's output, measured output.

And this looks like quite okay, but

let's stay with the spec.

Maybe like this

and calculate.

So this is it and now generate

measurement from target shape and also

select none and generate measurement

from target shape. Now let's go back to

our main screen and as you see the

speakers the subwoofer target curve

is higher than the speaker target curve.

How much higher? Let's check.

This one is four. This one is six. So

four. Double click and bring the blue

tag and info. And here you can read it.

Data offset 71.81

dB. And number six, the subwoofer

info data offset 73.21.

So 73.21

minus 71. What was it?

81

1.4 dB is the exact difference. So

subwoofer volume is 1.4 dB higher than

the speaker volume or this target curve.

So what you do is right click on

subwoofer and here minus 1.4

add to data.

Now speaker and subwoofer are volume

aligned. Okay, this is the first step

over the chosen target here. This is how

you volume align a subwoofer. It's

important for all the good reasons

because otherwise you will be eqing

forcing distortion to the subwoofer. If

it's not loud enough or if its volume is

higher than it should be then you will

be applying too much filter cutffs. All

of them harmful for the bass sound. So

it's important to find out the correct

rolloff frequencies and rolloff slopes

of the subwoofer you're using to be able

to volume align the subwoofer with the

speaker. Second step is time aligning

subwoofer with the speaker and the most

important step for the best bass sound.

Now the speaker

is applied a butter 12 dB per octave at

140 hertz. This is our crossover

frequency. We decided that

and generate measurement from predicted

under filter tasks.

This applies this filter to the speaker

and generates the response on the screen

here as you see from this

it went to this.

So apply the 12 dB per octave meaning

octave per octave is half the frequency.

So from 100 in the info you can see

140 htz

shape butter slope 12 dB per octave.

This means from 140 to half of it which

is 70 htz it drops by 12 dB. This is

what it means by 12 dB per octave. So

you can check this 140 is somewhere

here.

Let's say here and from here to 70 htz

which is here.

It's this says it drops like 12 dB.

Well, it shows like minus 255 dB here,

but it's because of the peaks and dips.

So, it's not like an exact science. But

if you do a one one thing here, you will

see a more

Let's check 140.

Let's say here is 140 and here is 70.

here. So comes up here

and it's more like

minus3 dB. So it's a straight slope is

minus 12 dB and also from 140 to

no that's this is for a highp pass

filter. For low pass filter it goes the

other way. So you double the frequency

and you will see the drop from here 200

280 this time uh 12 dB drop. This

becomes a low pass filter. This is a

highp pass filter.

Only high frequencies pass. Everything

else is then down. And a low pass filter

only low frequencies pass and high

frequencies are cancelled.

A band pass filter is both sides are

cancelled and only the band stays alive.

Um so this is our speakers after the why

are we applying this weird slopes and

let me show you now for the subwoofer.

Before that we go to EQ and this time

apply the low pass filter at 140 Hz our

crossover and this time link widths

Riley

filter type not butter word and this

time 24 dB per okay so a sharper one and

with a different so and generate

measurement from predicted

let's see

as you see this was the original

subwoofer response and this is with the

LR24

low pass filter applied to the

subwoofer.

Now why do we do these apply these

weird slopes?

Now on in the speaker EQ instead of

generate measurement from predicted if

you select generate measurement from

filters it will only create the filter

itself which is

You can't see it but here

as you see this is the actual filter and

here you will see perfect from 140 for

example

to 70 htz

perfect 12 dB drop

well it's not my day but I 70 is here

sorry

Sorry. From 140 to at 70 it is perfect

12 dB. See cuz this is flat

from zero here. So 12 dB at exactly 70

hertz. That is what 140 butterward 12 dB

per octave highp pass filter does. And

for the subwoofer again instead of

generate measurement from predicted

generate measurement from filters. I

click it and this is the subwoofer

filter again 24 dB per octave 140 which

means from at 280 double the 140 htz

let's find 280

I'm looking at here when I'm using the

cursors

81 okay and here it drops 24 dB already.

That's what it means. Go to zero as you

see 24 dB drop at double the frequency

and this is how it works on the

subwoofer

this filter. So now why these two

filters? Because the speakers are

designed most of them okay there are

exceptions with 12 dB per octave normal

roll off in the bass and you added

another 12 dB. So it becomes 24 dB. When

you add these two let's start the do

vector sum

as you see this is not flat. We have

some loss here.

But the speaker as we said has already a

12dB per octave its own rolloff which

means we can simulate it by adding

another highp pass here at 140 Hz

another but 12 dB per octave and now

generate measurement from filters.

So this is actually

how the speaker filter will be given the

speaker's own neutral rolloff. And when

you sum these two vector sum here as you

see it's perfect flat.

So this is why almost all audio video

receivers at the crossover apply a

butterwork 12 dB per octave highp pass

filter to the speaker and a link with

Riley 24 dB per octave low pass filter

to the subwoofer. And if you're using a

mini DSP and if you have the freedom to

choose your crossover slopes and types,

you should go with these ones unless you

have a very good reason not to like a

special driver that you're using or a

DIY speaker or subwoofer and whatnot. I

will also show you for example, it's not

they don't always sum up flat. If you

try other things, let's say you use 24

here, generate filters and with

subwoofer because these are now 36.

Imagine now you're using instead of

links 24, you're using 36 generic

filters.

Let's see what happens when you sum

these two up. This one and this one. But

they also seem very similar. and you sum

them,

look what happened.

So this is the reason that these two

slopes are used. This is how you time a

line. You first apply the correct slopes

to speaker and the sub at the crossover

frequency. So you're simulating what's

going to happen in your system when you

apply a crossover. So these three are

useless. Remove selections.

And we also showed all this

and this. Yeah, these are also useless.

We can make use of them later, but I

don't want to confuse you. So, you have

the Butterword applied speaker and link

with Riley applied subwoofer now.

And this is where

I don't like these colors. Let's set

colors to default and go to alignment

tool. These are not the original

responses. Once again, speaker

originally is like that and subwoofer

originally is like that. These are

rolloff slopes applied to them. And now

we go right click alignment tool. Here I

select the speaker EQ that is the

speaker with the butterwork 12 dB per

octave highp pass filter. And here I

select the subwoofer with the low pass

filter. and I go to 140

hertz which is our crossover frequency.

It's not stopping at 140. Then what I do

is make it larger. And now I can find

perfect 140. I guess

it's still not. So

now

yeah 140.

So wherever this cursor is impulse

alignment not phase alignment and align

IRS impulse responses at cursor which

means align subwoofer and speaker at 140

Hz to each other. I click this and look

already did its magic and

it didn't invert polarity. It would

invert polarity of the subwoofer if it

needed to. And you have to listen to

that if it says so. The polarities are

match and the delay required is 38

milliseconds.

Okay. If your DSP system is capable of

applying this much delay, you're good to

go. If not, you have to enter a range

here, minimum and a maximum, and RW will

do its best to align the sub to the

speaker within the given delay range.

And what is really happening here is

these dashed lines. I don't know if you

can see

the dash line here

is the theoretical maximum that this

speaker and the subwoofer can combine

and produce the maximum output.

Theoretical maximum is impossible. It

means at every single frequency

subwoofer and speaker are in phase with

each other which is not possible in any

room. But what RW did for example from

all these alignments that's possible.

Look how much how you get a dip suddenly

at around

120 htz with just a little change of

subwoofer delay.

And theoretical maximum always stays

fixed. And as you see when your cursor

is 140 and you just click align IRS

instead of playing with these sliders

already gives you the best possible

delay that will give the closest to the

theoretical max. And as you see it's

almost everywhere identical to

theoretical max. So this is a this is a

good alignment. So you create here an

aligned copy which will create the

subwoofer with the new time alignment.

And if you want, you also create an

aligned sum which is speaker and

subwoofer combined at this crossover

frequency with the subwoofer's new time

alignment. So also create an align. It's

the same thing as

vector summing. Let's I will show you

again here. These are the subwoofer and

the speaker with their like vector sum

this.

And now vector sum the aligned subwoofer

speaker.

Okay, it was this and it became from

this to that just by s of red line. We

completely removed this dip here and

these dips here and this is the target.

But align sum already gives us that

almost that. Now let's see what this 38

millisecond

minus 38 millisecond you can read in the

notes

alignment actually does so you can

properly apply subwoofer delays in your

system. This was the subwoofer with the

low pass filter and this was after it's

aligned as you see frequency response is

identical where we can see the time

alignment in impulse response. So we go

to overlays

impulse

and in the bigger picture it's easier to

see this was where subwoofer was the

peak was here and now after alignment it

shifted here. So

you will see that this is 38

milliseconds here 38 milliseconds. So

minus 38 milliseconds means

doing something so that subwoofer fires

38 mconds earlier. So in time now it's

peak comes earlier than original. How

you do that? You apply a distance like

13 m distance if you have a distance

setting in your DSP to your sub so that

the sub fires first and then your

speaker fires 38 milliseconds after the

sub. So it may sound confusing but just

know that almost always the subwoofer

should have a larger distance than the

speaker. So it fires up earlier than the

speaker because bus waves are long and

slow. So subwoofers take their time.

Also they have internal filters a lot of

them which causes extra delay. So

subwoofers should be fired first. And

when you apply a higher distance to

subwoofer in your for example AV

receiver you're actually not delaying

it. you're delaying the speaker.

Okay?

If you enter 0 m as distance, then you

are delaying the subwoofer. If you enter

13 m as your subwoofer distance, then

you're delaying the speaker by 13 m. We

could have done, let me just create a

response copy of this unaligned

subwoofer. Okay, we could do offset t=0.

This is time alignment in ref. And here

what you would enter was 38

milliseconds, not minus 38. Apply and

close

as you see

you just brought it on top of the

aligned copy. Minus 38 in the alignment

tool is a 38 millisecond delay that you

use with offset t=0 here. Positive

delays move the subs or speakers to the

left on this impulse window. negative

delays more than to the right. To the

left means you want the subwoofer to

fire up earlier and to the right because

this is the time horizon. To the right

means it can arrive later but we want it

to arrive earlier so it goes to the

left. Just don't get confused with all

this sign of delay values and how it's

converted to distances. For many Denon

and Marance receivers out there, the

older models for example, this

38 milliseconds is not 13 m because they

use a different speed of sound which you

can for example in REW preferences you

can go come here and decide that speed

of sound is not 343 but it's 300.

And now when you measure here this

distance it is still 38 milliseconds.

Sorry,

it's not easy to work on a small shrink

screen just to

follow up YouTube resolution. So, it's

still 38 mconds, but it's now 11.4 m.

So, in this old model of amplifier, you

would enter 11.4 4 m distance to

subwoofer to generate 38 millconds of

delay to the speaker for them to

maximize their output with lowest group

delay possible. This is the 140 Hz

crossover combination of this speaker

and this sub.

Let me just fix this back because I

forget that and then all my distance

calculations show up wrong.

And that's it. You now time aligned the

subwoofer to the speaker.

Next thing you do is the EQ filters.

Now for the EQ filters,

this is your aligned sub. This is your

speaker. You don't have to align it. It

fires first. It fires last. No matter

what. This is your target.

Remember target subwoofer is something

else actually we should delete that or

we should just drop these two 1.4 was it

minus 1.4 dB. So these are the new

correct targets for speaker and

subwoofer.

This is how it looks like.

Now again for the EQ subwoofer aligned

you generate a minimum phase version of

that and replicate data outside range

and

this is the target curve. At the on the

target curve, you generate the

subwoofers crossover filter which was at

140 link with Riley

and 12 dB to 24 dB per octave generate

from predicted

and you will shortly understand why

we're doing this.

So now from the target I made the same

slope as the subwoofer. And now on this

target with this correctly volume

aligned and time aligned and minimum

phase version generated subwoofer

response I can comfortably invert.

This is the target. This is my subwoofer

and divide.

Maximum gain 5 dB from 40 Hz to 200 htz.

Okay, something like that all the way

generate

and apply no smoothing. So this is the

filter for your subwoofer. It's not

minimum phase. So you have to generate

minimum phase version of that with

replicate data outside range. And now

this is your subwoofer filter.

And to test your filter, this was the

subwoofer response with the crossover

applied

but not minimal phase version yet. So

here and this is the filter that you're

going to apply to it which is

multiplication

and this is your subwoofer response.

Similarly,

you already have this

for the speaker. So, generate speakers

minimum face copy

and not this one. We already created

this wrong target curve. So

you divide

number three with number 19.

Now this time you don't really have to

start it at 40 Hz but somewhere

here maybe

94

and up to 200 I think this. So don't go

too much over shredders frequency. It's

not going to sound good.

And in fact, you can even apply where I

was smotting here. So it doesn't do very

sharp cuts in higher frequencies

and apply no smoking now for the filter

to to see the correct shape of the

filter as you see.

Oh, sorry.

This is still applying the filter

through unsmooted

speaker. So filter is too choppy. What

you do is EQ speaker

first variable smoothing

and then generate minimum phase with

replicate data outside range. So you

freeze the smoothing of the speaker and

then with this frozen

speaker minimum phase response this is

going to stay where I will smooth it.

You can do that. And now when you use no

smoothing

apply as you see filter is still the

same. This was not the correct filter.

This is the correct filter. Now you have

to generate minimal phrase copy of this

filter when it's unsuited and this is

frozen smooth actual filter that you

will apply to your speaker. Now let's

apply it to the actual speaker response

which was

EQ speaker here and this is the final

filter that will apply multiplication.

And this is the speaker final response.

This was the subwoofer.

And when you add them with vector sum,

this is your equalized final speaker and

subwoofer properly combined. You will

usually find yourself having to align a

subwoofer to left and right speaker

pair. And if when this is the case, all

you have to do with alignment tool

because you can only align two responses

to each other in the alignment tool.

take the vector average of the left and

right speakers and align the subwoofer

to this vector average. You should still

apply the highp pass filter to this

vector average speaker response and you

should time align the speakers before

you take vector average, but the rest is

exactly the same. So that's it for this

tutorial. Don't forget to comment,

subscribe, and like. See you soon.