How to implement my horn and my speaker

Acoustic Basics summary

Energy is not free so when we ask a compression driver to be constant in directivity in a horn, the on axis response will always have a form of bell, needing EQ with DSP.

The implementation of a speaker is not done in function of the room, The entire speaker must be flat on-axis at close distance (around 60cm).

Then the air absorption, the room will apply an effect of decay, this decay must NOT be corrected at listening position, never.

Rooms mods below 100hz, taking account of an area of listening and not at one precise position, can be EQ but nothing more.

FIR & IIR

A FIR filter can do minimum and linear phase filtering, a IIR can only do minimum phase filtering

Linear phase filtering (allowed by FIR DSP with Taps, costly in CPU resources) doesn’t touch the phase so it’s good for crossover but not for frequency correction as we want to modify the phase also when we do a frequency response correction.

When you do a cut with it the easier way to do it is to EQ it flat 1 octave after the cut with Minimum phase filtering for correct phase and response then apply a Linear Phase Filtering on it.

Minimum phase filtering (allowed by IIR DSP) modifies the phase so is very adapted to frequency response correction. In this case we will in most cases try to combine the acoustic natural fall-off with our DSP electric cutoff, see subwoofer par for example.

As the center of emission isn’t at the same depth we will need to put delay on the woofer in most cases to have in-phase response and no cancellation at X-over.

The more brutal is the cut the more a very precise delay is important between components so be very careful with this and follow the article for it, if delay is perfect the LR150db in FIR after a IIR linearization 1 octave below the cut for the compression driver, 1 octave upper for the woofer, is the best.

All steps have to be done in the correct order we will need REW and a mic :

1. The horn with my DSP

We have to choose where we will do our crossover, basically it’s when the horn starts to lose directivity control but still has SPL in his low end region.

A factor to consider is the physical distance between the acoustic centers of the woofer and the high-frequency section. Ideally, this distance should be less than or equal to roughly 66% of the wavelength at the crossover frequency. However, listening distance also plays a role. At greater listening distances, the impact of this spacing becomes less critical, As seen in vertical lobing article.

The directivity of two elements cut together must have similar directivity at the crossover frequency as see in directivity match article.

directivity horn match

If people mainly listen sit-down , the center of the horn should be around the ears position, so 94 / 96cm from the floor.

Measure the horn at 30/60cm (depending of woofer or horn size) for taking account of all his form but not the room, you can gate the measurement before the first accident see on impulse response for the most advanced of peoples:

windowed impulse

EQ it flat in minimum phase filtering, not in linear phase, from the fall bottom to the breakup. Use minimum phase filtering in IIR here will allow us to correct phase, in the good way, in the same time, as explained here: Horn & Energie

Sometimes the breakup is so present that it’s not possible to EQ flat until 20kHz, in this case do it flat until 14/15kHz, it’s not necessary to “excite” a diaphragm breakup.

minimum phase filtering (IIR)

You can add with REW a target curve, like a LR 48:

crossover

We EQ flat until natural fall (in orange) then we add a minimum phase filtering on the natural fall to match the target curve (in green), the red curve is the result.

In DSP, your cut acts like an electric adjustment, summing with the horn’s natural roll-off to shape the final acoustic response.

linear phase filtering (FIR)

To achieve an accurate acoustic linear phase filtering, a two-step EQ process is recommended.

Implementing a steep cut filter precisely is crucial. Delay alignment (point 3) requires meticulous calibration, and pre-ringing can be a valuable tool for assessment.

Additionally, the final acoustic responses of both sections after the filter must mirror each other. To achieve this mirroring effect, a minimal-phase equalizer applied one octave before the crossover corrects phase response, ensuring the electrical and acoustic cuts align perfectly. This same approach must be applied to the woofer crossover.

About very strong cut, compared to LR48 filters, LR150 offers advantages when approaching distortion limits and achieving improved thermal performance, ultimately enhancing the listening experience without introducing ear fatigue. These steep cuts are often implemented using convolution filters generated by software like RePhase and need a very precise tuning and respect of the process.

2. The woofer

Basically we will do the same as horn section with the same choice of filtering strategie, we will EQ flat at 30/60cm in front of him (so we change the height position of our mic). We cut it at the same frequency of the horn and in the same way to do, target curve can be use too but in most case the woofer will rise up in frequency so we will just have to EQ it flat one octave upper the crossover frequency.

In active filtering with delay having miror cut of the upper section is the way to do.

important : Put your woofer as close as possible to the horn to reduce vertical center to center distance, the ideal distance is below 66% of cross-over Lambda (Lambda : wavelength of a frequency) and keeping horn center at ears position height, generally at 94/96 cm sit-down. Otherwise you will have vertical lobs (seen in Xdir free software)

If your woofer response is rising up in the box (partially due to baffle step), and only in thi case, you can linearize him by adding a simple air core inductor in serie on “+” of the woofer, it will lower the H3 by reaction with the motor. It doesn’t work if the woofer have AIC (Active Impedance Control, from 18Sound).

It’s one of the best tricks in the active filtering system and is explained here: Speaker Break-Up and associated distortion. The self is around 1.2/1.7 mH, then does a precise EQ at 30/60cm with gated measurement.

In order to simulate the inductor network in VituixCAD with impedance curve and frequency response and to do the precise EQ at the end with EQ we want to mesure the woofer with the baffle step but not the room with it. So we have to measure relatively close for have the baffle step effect and gate the measurement for don’t have the room:

windowed impulse

Be careful to not have reversed the phase (+ and -), you can check it by reverting on purpose, measuring and see that there is a cancellation at cross-over.

For brace and stuffing you can follow our guide : braces and stuffing

3. Find the delay between horn and woofer

There is several way to do it, we will talk about the simplest, advanced user can look how to do it with the sound card Loopback method but the result is in fact the same.

In your DSP configuration, voluntarily add 10ms on the compression driver to “move” his impulse pick further.

Put your mic at 1m50 at an average height between horn and woofer and basically align SPL level (we will do it precisely in the next step for this) with elements crossed.

Remove any crossover but keep the frequency corrections, do a large sweep around desired crossover with both composants playing together.

For example from 300hz to 5kHz (if you let the sweep goes up it will be easier to see) if you want to cross around 700hz, not need to have a huge volume as there is no crossover, and we don’t want to break something as there is no crossover on the compression driver.

Look the impulse response on REW, you will see two impulse peaks, one at 0ms and another one around 10ms plus or minus something, just remove the 10ms, for example :

windowed impulse

9.54ms is the delay between the two impulse pick so the delay is -0.46ms, at the speed of sound in air it’s around 15 cm.

Put the right delay on the right component (the one that is the closer to the front baffle) in the DSP and don’t forget to put back the crossover on each elements.

4. Align level

With only one speaker, put your mic at 1m50 of the speaker at the average height between horn and woofer (so he won’t move from step 3), we will align dB level between the two components thanks to measurements iterations.

The first attenuation on compression driver or even a tweeter in active filtering should be done in passive to bring protection to the high frequency driver and imporve S/N ratio, we use for this a L-pad network:

L-pad
Thanks to frequency response and impedance curve we can simulate it in VituixCAD.

Then we affine it in the DSP by step of 0.2 dB in stereo, with the two speakers, without use of your mic by listening at listening position, if the sound looks muffled add 0.2 dB and try again, if it looks light too bright (or “agressive”) then remove 0.2 dB.

Why in stereo: when the two speakers will play at the same time some frequencies, the lower ones, will sum easier than the highest frequency, so it should be taken into account.

Do it after putting acoustic treatment, because it will lower energy in high frequencies, we generally can add 0.2dB more after adding a good acoustic treatment by absorption (see 7. Acoustic treatment).

Without acoustic treatment it can be better to add a slight high frequency decrease when we measure a constant directivity horn at 30/60cm in point 1.

It allows the tuning of energy balance due to room acoustics, this aspect is facilitated by a perfect polar plot pattern exempt of off-axis accidents, midrange narrowing or beaming.

Of course it makes sense only if you have done all the previous steps.

5. Speaker position

The acoustic center of the horn must be at the same height of your ears at your listening position.

So generally the acoustic center of the horn must be close to 94 or 96 cm height from the ground.

You must toe-in your speaker at your main listening position, like this :

speaker position

You also should listen your speakers inside the Critical Distance area and don’t put your sitting position to close from the rear wall, let a gap at to at least 1m.

But, this triangle is an ideal, if you can have 3m/3.50m with 1/1.5M gap behind your listening position it will be better than just follow the theoretical triangle.

Following the ideal triangle must not be done on the depend of listening at critical distance.

Rear gap is important in any cases.

6. Subwoofer (optional)

Even if we have a FIR DSP, so linear phase filtering, we don’t use it here as it will be very costly in TAPS as the frequency is lower and useless about phase at this frequency. We will search to cut between 50 and 60 hz optimally.

Measure the delay in cm and convert it to ms if the subwoofer is far, be sure to not have a reversed phase as we said for the woofer, you can check it with the same techniques.

You can correct the woofer response at the listening area by doing several measures by moving the mic in the listening area, not at just one position.

A regular EQ in minimal phase (“IIR”) placed on a modal effect with a quality coefficient (Q) equal to the modal effect will not only improve the frequency response but also the temporal behavior, reducing trailing bass and improving global bass experience.

Be carreful when you correct your subwoofer to not introduce masking effect.

You can add an electric (in DSP) Butt 12 on your natural woofer natural fall to have a real (acoustic) Butt 24 in minimum phase filtering, linear phase filtering is not necessary at these frequencies. Then add Butt 24 on the subwoofer.

Is Groupe Delay (GD) audible in a bass reflex system?

It depends on the cycle length. Generally, it’s recommended that the GD should not exceed 1 or 1.5 cycles of the frequency, as this is related to its wavelength.

The formula for calculating the maximum acceptable GD is:

Δt max = n / f

Where:

Example:

You can use VituixCAD Box Simulation to visualize the GD according to the box volume and port, and consider this information in your design and choices.

In practical cases, we don’t exceed the recommended GD limits if we follow standardized alignment so it will not be audible in standard usage.

Positioning Subwoofer

Subwoofer position can be hard, we want room gain but without room modes, but the first comme rarely without the second.

We should place subwoofer at 20% of each side wall on the speaker wall as here, if we have 4 subs we regroup it by pair:

speaker position
From : getting the bass right - Harman

The alternative of this is to put a line on sub horizontally on the ground, if possible integrated to the wall, completely from one side wall to the other, each subwoofer unit as close as possible from each other, it can require a lot of 18", it will allow to create a plane wave radiation subwoofer. It’s more technical and very expensive. It can be coupled (same chanel) with Divatech style wall of sub.

We have to choose, depending on our room and SPL max goal, between one or two 12",18",21":

If we prioritize max SPL we go for pro audio subwoofer, if we need flat responding subwoofer so non pro speaker as car audio.
If we place the subwoofer on the center of the room we will have less mods but also less room gain so a flat responding subwoofer will be better.
Generally, except very big room we search for flat responding subwoofer.

Here is a list of soms, for vented ones we will search a Qts between 0.33 and 0.36 and a low Fs :

Vented:

Sealed:

7. Acoustic treatment

For this section we will start from the principle that we are in a regular room, a living room, not a dedicated room or studio record.

We will avoid coupled volume, close doors are alway better to avoid to create a coupled volume.

A basic living room is already very reverberant, so the first main focus is to absorb, not to diffuse.

The best place for absorption is the wall behind the speakers, it will bring 80% of what we need: lowering the room RT.

We will use melamine (BASF Basotect G+) or Woodrock (Rockfeu REI, it’s bi-density so put low density to the external side) on a big part of the wall, prioritizing exposed surfaces near angles. The treatment must be massive, at least 8/10cm deep on at least 50% of the wall, only absorption.

Then if we want to extend the lower bottom of the absorption we can add the same things on the wall rear the listening position.
Don’t put your sitting position right against the wall, let 1m ditance at least.

Then we can think on the side walls by an assembly of diffusing parts side by side with absorbing one. Don’t use a product that does “the two”: It gives a “V” shape absorbing response that we don’t want.

Deal with first or early reflection on side walls

The goal on the side is to play with ITDG (Initial Time Délai Gap) it’s, the time the early reflection comes to you.

The early reflection is the “first” reflection on a surface that will directly go from the speaker to your ears with only one or two rebound so they are close in time from the direct sound and can cause phantom image, dysmetria soundstage when the room is not symetrical and speakers close to side walls…

early reflection

This reflection on walls should not be specular (aka not like a mirror) but diffuse, it’s why, in room and not in recording studio, we put side by side absorption and diffusion on sides. In a recording studio we will absorb everything as we want almost only direct field.

We need early reflection for the sound stage but not too much and not too close in time comparing to the direct field (so drive away speakers from side walls) or the tone quality will be degraded and a finger response will appear at medium/high frequencies.

Pinch the speakers as we said upper and use horns adapted to your listening distance will reduce the intensity of this early reflection.

The brain has a TPI (Time Period of Integration) when sound is fusionned, you will have more information about TPI here.

Listening position measurement

As discussed in point 4, there are adjustments that can be made at the listening position to achieve precise SPL (Sound Pressure Level) balance between the compression driver and woofer. The compression driver’s response curve might also exhibit a slight downward slope at the listening position, the degree of which can vary depending on factors such as listening distance and room acoustics. You may even prefer a more pronounced downward slope to balance energy according to listening position and room acoustics.

This interplay of factors effectively creates a personalized target curve, tailored to your specific listening environment, speaker placement, speaker directivity, and personal preferences.

It’s crucial to understand that a one-size-fits-all ideal target curve does not exist. The final sound you experience in your listening area will be distinctly shaped by the unique combination of your speakers, room acoustics, and personal preferences.

Except for frequency adjustments in the subwoofer region (for mods), we don’t correcting or filtering our speakers at the listening position. Even then, caution is advised, as room cancellations can vary significantly with even slight changes in microphone position.

Therefore, it’s essential to consider the listening area as a whole rather than relying solely on measurements taken at a single point. Multi-point measurements or MMM within the listening area can be helpful for identifying mods below 100 Hz or for checking the response at specific locations.

However, the design and development of a speaker must not be based on measurements taken at the listening position.

As seen in upper sections, a crossover slope target curve is used to match the crossover slope of each individual element at 30/60 cm, measured alone. This is the only time we use a target to match a frequency response.

Point of attention

Be careful of the reversed phase, check each element alone with the mic and look at the impulse response in REW, it must go up firstly, if it goes down in first place it’s reverted.

copyright