Metlako: A Small, Affordable Two-Way Unity Waveguide

Experimentally, I have found that the best performing midrange taps are taps that expand. The way that I get away with an expanding midrange tap is that it begins as a circle and ends up as a slit.

ybgIKcJ.jpg

For instance, the entrance of this midrange tap measures 1" in diameter and it's circular. That's an area of 5.07cm^2. The exit of the midrange tap measures 4" x 1/2", for an area of 12.9cm^2.

y4QXIYi.jpg

Here's the measured polar response of the midrange array in Metlako V1

tJSlKJX.png

Here's the predicted response from Hornresp, which can only simulate a conventional (straight) midrange tap. (1)

xHo8eY3.png

Here's the predicted response from Akabak for a straight (conventional) midrange tap (red) and an expanding midrange tap (black)



Here is what I see from the sims and the measurements:

  • I believe the measured and simulated results show that the expanding taps are superior to straight taps
  • I think that some additional work on the Akabak model could be useful. For instance, the Akabak sim doesn't "catch" the null that exists at 1800Hz.


(1) if there's a way to simulate the expanding taps in Hornresp, let me know.
 
I've created a GIT repo for my waveguides.

The latest version of Metlako can be downloaded here:

GitHub - crater-lake/metlako: a two way Unity waveguide

If you have a git client installed, you can get the repo like this:

johndoe@DESKTOP-FA5T5QS MINGW64 ~/Downloads/work
$ git clone https://github.com/crater-lake/metlako.git
Cloning into 'metlako'...
remote: Enumerating objects: 9, done.
remote: Counting objects: 100% (9/9), done.
remote: Compressing objects: 100% (7/7), done.
remote: Total 9 (delta 0), reused 6 (delta 0), pack-reused 0
Unpacking objects: 100% (9/9), done.


What I'd like to do is keep these online, so whenever someone is looking for the latest version, they can pull them down from git. Another nice thing about git is that I can include instructions on how to print them and build the speaker. That's the next step.
 
Last edited:
lol, thought I'd listed them!

Here's the parts:

1) 2 x waveguides for a stereo pair. Get waveguides here : GitHub - crater-lake/metlako: a two way Unity waveguide

2) 1 x MiniDSP. $95, get here : miniDSP in a BOX : miniDSP 2x4

3) 4 x MCM 55-1870 midbasses. $40.68 for four, get here : https://www.newark.com/mcm-audio-select/55-1870/woofer-5-aluminium-cone/dp/39C2182

4) 1 x Tymphany NE19VTS-04. $23.10, get here : Peerless by Tymphany NE19VTS-04 3/4" Silk Dome Tweeter

Also available from Madisound.

I think the Dayton DA135 might work as a replacement for the MCM. The Dayton is 3mm too big, so if someone went this route they'd have to grind down the mounting plate to make it fit. The MCM would definitely be easier and cheaper.
 
What would you recommend for bass reinforcement in a nearfield setup?

It really depends on how loud you want to get. This speaker should be capable of peak in excess of 110dB. So it would need a fairly beefy sub to keep up.

OTOH, if you don't need to hit 110dB you have a lot of smaller options.

At home I'm using a pair of TC Sounds fifteens in a bandpass with 350 watts along with my Cosynes. I'd like to add another sub or two if I could find room, and more power.
 
Thanks for sharing your work and 3D printing files. If you were to buy a 3D printer now, for this type of project,
what would you buy?

Funny you ask :)

Basically this project resulted from the fact that I figured out a way to cut the waveguide into three pieces, allowing for a much larger waveguide using a small printer. It looks like this:

hVdAR6x.png


But this has created a NEW problem - basically the print takes about 24 hours and if it fails, I've wasted 24 hours and an entire spool of filament.

Due to the size of the finished print, it's taking forever to produce and it takes a LOT of filament.

So far, out of five prints, I've had three fail, or 60%.

Due to all that, I'm thinking about splitting the print in half - even though it fits on a single printer - and running each half in parallel, on two printers.

I like the idea of using two printers to make ONE print. It has the following advantages:

1) if a print fails, you don't lose the entire thing, only half of it.

2) it prints twice as fast

Right now I have three 3D printers. Two of them are Monoprice "Maker Select" :

Monoprice Maker Select 3D Printer v2 - Monoprice.com

I upgraded it with an all metal print head from Swiss Micro. I think that step is a necessity if you print with ABS+ or PETG. I'm also printing on a glass bed.

I've heard good things about the Creality printers.

In summary: I have a couple of Monoprice printers that I'm fairly happy with after I've upgraded them. My biggest challenge lately has been prints failing after 12-20 hours. Basically the large size and complexity is leading to failures. There isn't much I can do to improve that except break the print into pieces that are more manageable. At the moment, I don't thing a printer upgrade would make a whole lot of difference.
 
Funny you ask :)

Basically this project resulted from the fact that I figured out a way to cut the waveguide into three pieces, allowing for a much larger waveguide using a small printer. It looks like this:

hVdAR6x.png


But this has created a NEW problem - basically the print takes about 24 hours and if it fails, I've wasted 24 hours and an entire spool of filament.

Due to the size of the finished print, it's taking forever to produce and it takes a LOT of filament.

So far, out of five prints, I've had three fail, or 60%.

Due to all that, I'm thinking about splitting the print in half - even though it fits on a single printer - and running each half in parallel, on two printers.

I like the idea of using two printers to make ONE print. It has the following advantages:

1) if a print fails, you don't lose the entire thing, only half of it.

2) it prints twice as fast

Right now I have three 3D printers. Two of them are Monoprice "Maker Select" :

Monoprice Maker Select 3D Printer v2 - Monoprice.com

I upgraded it with an all metal print head from Swiss Micro. I think that step is a necessity if you print with ABS+ or PETG. I'm also printing on a glass bed.

I've heard good things about the Creality printers.

In summary: I have a couple of Monoprice printers that I'm fairly happy with after I've upgraded them. My biggest challenge lately has been prints failing after 12-20 hours. Basically the large size and complexity is leading to failures. There isn't much I can do to improve that except break the print into pieces that are more manageable. At the moment, I don't thing a printer upgrade would make a whole lot of difference.
You could get a known-reliable printer like the MakerGear M2. Would be a worthwhile investment if you plan to keep making 3D printed audio parts.

The M3 also has a dual-head option if you want some niftier aesthetics for your car audio.
 
Over on Facebook, there was a debate about the audibility of phase. Here's a synopsis of why I invest so much effort to get a flat-ish phase curve:

TkPI0ug.png

This is a Genelec 8351A. It's a triaxial loudspeaker. There's a tweeter nested in a midrange which is nested in a woofer.

FE96Ee5.png

Here's the polar response of the 8351A. It's about as good as it gets. What I notice here, is that there's no apparent transition from midrange to tweeter, or woofer to midrange. In a conventional loudspeaker you'll typically see a change in directivity at the xover point. This is particularly noticeable on the vertical axis of most speakers.

NsyXOui.png

This is a Polk 5JR. It's a fairly typical two way loudspeaker.

xHjsNZM.png

Here's the horizontal polars of the Polk 5JR. What I see here, is that the on-axis frequency response is fairly good. But the polars have issues. First, there's an obvious crossover point around 2500hz. I can tell it's there because the polars are a mess in the octave between 2500hz and 5000hz.

The polars are a mess due to three things:

1) there's a directivity mismatch between the tweeter and the woofer. Basically the woofer has a narrow beamwidth because it has a large diaphragm, and the tweeter has a wide beamwidth because it has a small diaphragm.

2) At the crossover point, the loudspeaker is significantly LOUDER at 45 degrees off axis than on. (See how the green line dominates?)

3) The speaker is beaming at high frequency. For instance, at 12khz, if you listen 30 degrees off-axis the speaker is attenuated by 6dB.

In a nutshell, the Genelec response curve OFF axis is very similar to how it sounds ON axis. This has two advantages. First, when you walk around the room, the speaker still sounds good. Second, the reflected energy behaves the same as the primary wavefront.

6HpfcG8.png


Here's the vertical polars of the Polk. The xover point is apparent and that dip is going to be audible. The dip is a fundamental problem which is caused by the phenomenon I describe here : What Causes Off-Axis Nulls?

So, what does this have to do with phase?

The answer is that it's nearly impossible to get polars like the Genelec if your phase response sucks.

For instance, a LR4 filter has 360 degrees of phase rotation at the crossover point. This rapid shift in the phase response makes it nearly impossible to get the seamless transition that we see with the Genelec.

22 years ago, John Dunlavy said "We participate regularly in the recording of our symphony orchestra here in Colorado Springs. We've also recorded instruments like violins and cellos and timpani in our big anechoic chamber using instrumentation-quality microphones and equipment. And we find that in order to reproduce those sounds with a level of accuracy such that you can not literally hear any difference between the live and the recorded sound, you have to have a speaker that exhibits almost perfect impulse and step responses. The only way to do that is to time-align the drivers very, very accurately, usually within a matter of a few microseconds, then use a minimum-phase, first-order crossover network and get everything right. And you have to have an on-axis response of better—well better—than ±2dB."

I mostly agree with Dunlavy, but there's one big opportunity we have in 2019: cheap DSP. Because DSP is trivially cheap, we have filter options which offer phase response very close to what Dunlavy achieved with 3rd order filters. What I do is described here.

In summary:

No, I don't have magic ears. I can't detect a time difference of 0.01 milliseconds. But I *can* detect problems with directivity, and problems with the crossover between a midrange and tweeter. In particular, I noticed that Unity horns are exceptionally good at intelligibility and percussion. The reason that this is the case is because the midrange and tweeter basically behave as if they're a single full-range unit. If you look at the phase and the polar response of a good Unity horn, it doesn't look like two or three drivers, it looks like one. A conventional loudspeaker like the Polk, it can't do that. Not with DSP, not with EQ, it's geometrically impossible. The tweeter and the midrange are too far apart to do what the Genelec does.

uBAlGAk.png


D19Fjin.png


EsD0RT7.png


Here's the polars, the frequency response, the phase and the distortion of my current project. Is this as good as the Genelec?

NO

It's not even close. But it's in the ballpark. I still have a lot of work to do. The performance of the Genelec can only be achieved by a coaxial or triaxial speaker with well behaved phase. There's no other way. It cannot be done with a conventional two-way or three way.

BTW, this is also why many of the loudspeaker manufacturers out there are converging on a similar format:

M1RWdH4.jpg
 
Over on Facebook, there was a debate about the audibility of phase. Here's a synopsis of why I invest so much effort to get a flat-ish phase curve:

This is a Genelec 8351A. It's a triaxial loudspeaker. There's a tweeter nested in a midrange which is nested in a woofer.

Here's the polar response of the 8351A. It's about as good as it gets. What I notice here, is that there's no apparent transition from midrange to tweeter, or woofer to midrange. In a conventional loudspeaker you'll typically see a change in directivity at the xover point. This is particularly noticeable on the vertical axis of most speakers.

This is a Polk 5JR. It's a fairly typical two way loudspeaker.

Here's the horizontal polars of the Polk 5JR. What I see here, is that the on-axis frequency response is fairly good. But the polars have issues. First, there's an obvious crossover point around 2500hz. I can tell it's there because the polars are a mess in the octave between 2500hz and 5000hz.

The polars are a mess due to three things:

1) there's a directivity mismatch between the tweeter and the woofer. Basically the woofer has a narrow beamwidth because it has a large diaphragm, and the tweeter has a wide beamwidth because it has a small diaphragm.

2) At the crossover point, the loudspeaker is significantly LOUDER at 45 degrees off axis than on. (See how the green line dominates?)

3) The speaker is beaming at high frequency. For instance, at 12khz, if you listen 30 degrees off-axis the speaker is attenuated by 6dB.

In a nutshell, the Genelec response curve OFF axis is very similar to how it sounds ON axis. This has two advantages. First, when you walk around the room, the speaker still sounds good. Second, the reflected energy behaves the same as the primary wavefront.

Here's the vertical polars of the Polk. The xover point is apparent and that dip is going to be audible. The dip is a fundamental problem which is caused by the phenomenon I describe here : What Causes Off-Axis Nulls?

So, what does this have to do with phase?

The answer is that it's nearly impossible to get polars like the Genelec if your phase response sucks.

For instance, a LR4 filter has 360 degrees of phase rotation at the crossover point. This rapid shift in the phase response makes it nearly impossible to get the seamless transition that we see with the Genelec.

22 years ago, John Dunlavy said "We participate regularly in the recording of our symphony orchestra here in Colorado Springs. We've also recorded instruments like violins and cellos and timpani in our big anechoic chamber using instrumentation-quality microphones and equipment. And we find that in order to reproduce those sounds with a level of accuracy such that you can not literally hear any difference between the live and the recorded sound, you have to have a speaker that exhibits almost perfect impulse and step responses. The only way to do that is to time-align the drivers very, very accurately, usually within a matter of a few microseconds, then use a minimum-phase, first-order crossover network and get everything right. And you have to have an on-axis response of better—well better—than ±2dB."

I mostly agree with Dunlavy, but there's one big opportunity we have in 2019: cheap DSP. Because DSP is trivially cheap, we have filter options which offer phase response very close to what Dunlavy achieved with 3rd order filters. What I do is described here.

In summary:

No, I don't have magic ears. I can't detect a time difference of 0.01 milliseconds. But I *can* detect problems with directivity, and problems with the crossover between a midrange and tweeter. In particular, I noticed that Unity horns are exceptionally good at intelligibility and percussion. The reason that this is the case is because the midrange and tweeter basically behave as if they're a single full-range unit. If you look at the phase and the polar response of a good Unity horn, it doesn't look like two or three drivers, it looks like one. A conventional loudspeaker like the Polk, it can't do that. Not with DSP, not with EQ, it's geometrically impossible. The tweeter and the midrange are too far apart to do what the Genelec does.

Here's the polars, the frequency response, the phase and the distortion of my current project. Is this as good as the Genelec?

NO

It's not even close. But it's in the ballpark. I still have a lot of work to do. The performance of the Genelec can only be achieved by a coaxial or triaxial speaker with well behaved phase. There's no other way. It cannot be done with a conventional two-way or three way.

BTW, this is also why many of the loudspeaker manufacturers out there are converging on a similar format:

Genelec just released a revision to the 8351 that measures even better. They are leading the way with direct-radiating coaxials IMHO with their attempts to eliminate external surrounds.

Another coaxial that falls just behind the KEF and Genelec stuff is the Technics flat coax:

technics-7.jpg


Technics Premium Class SB-C700 loudspeaker Measurements | Stereophile.com

It has better directivity and linearity than the KEF LS50 up to 10kHz, but top-octave control falls a tad short of Genelec and KEF - I suspect because the throat profile and phase shield is nowhere close to either. The way I see it is that Synergies/Unities and their variants like the Metlako have the advantage of higher SPL, headroom, lower compression and less SPL-dependant directivity (as the dispersion pattern is dominated by the horn profile due to the boost provided). The direct radiators might have lower diffraction and possibly better FR and directivity at low SPL, but should have more SPL-dependent performance with excursion, as the midbass/midrange driver changes the throat profile the tweeter sees. Uni-Q mitigates it with a sorta fixed throat + lens that reminds me of a funky whizzer. Technics uses a flat radiator to minimise the change in profile with excursion. Genelec loads their midrange in a waveguide that takes up the entire front baffle.
 
Last edited:
I think it will be difficult for Kef to "catch up" to Genelec without DSP; to me, it looks like Genelec is using FIR filters on their speaker to get those crazy-good polars.

Having said that, it might not be audible.

For instance, the LS50 images really well. It's only obvious shortcoming is that the bass is boomy. There's only so much bass you can coax out of one small woofer.
 
GitHub - crater-lake/metlako: a two way Unity waveguide

I've release an updated version of Metlako here.

Here are the changes:

1) There are no changes in the size. This is by design; the old crossover will still work fine. (Speaking of, I need to add the crossover to the project...)

2) There's a series of minor changes designed to make it print better. These include:
a) I added a cylindrical "shield" of plastic around the edges of the waveguide. I got this idea from Reddit, someone was doing this to keep ABS prints from cracking. Basically it's a one millimeter thick piece of plastic that's designed to "shield" the edges from any air motion inside of the print enclosure.
b) I added a 1.414 millimeter chamfer to the edge of the waveguide. Nothing that will be audible. The idea is that curved edges seem to split less than flat edges do.
c) I added an additional skirt to the print. Skirts seems to be the best way to keep the first layer from peeling up.

On the downside, this will require additional finishing work at the end of the 3D print. But it seems to be required to get this thing to print reliably. This object is WAY more error prone than my last project. Basically it's using every last millimeter of the print bed, and prints tend to fail when you do that.
 
Metlako has been reliably printing in PETG but I really want to print most of them in ABS+, if possible. ABS is a nightmare to print, but the finished print just looks so much better.

jYjGfNY.jpg

ABS+

zbnRUIV.jpg

ABS+

w3GEwLI.png.jpg

PETG

For comparison's sake, here's the last project in PETG and ABS+


To get my printer to semi-reliably print ABS+, I had to do the following:

An externally hosted image should be here but it was not working when we last tested it.


The enclosure for my printer is an Ikea Lack. This isn't mine, but mine is similar.

I'd stuck my printer in a corner, and only two of the four walls were enclosed. The idea was that the walls of the corner would take care of the other two walls.

To improve on that, I added half a wall to one side, and the remaining wall I added a full inch of foamular insulation. I included a HDF barrier, so that the foamular isn't exposed.

So it's basically a sandwich of HDF, then an inch of foamular, then more HDF.

The idea here, is that I don't want ANY heat radiating through the walls of the print enclosure.

So front window is glass. It *does* feel a little warm to the touch, so maybe at some point I need to get a double-pane window.

The "trick" with printing ABS and ABS+ seems to be that the ambient temperature needs to be ridiculously hot. Like an oven. The reason for this is that ABS isn't remotely as dense as PETG. Due to the low density of ABS, what happens is that it SHRINKS.

151s_1_MainForWeb-1.jpg


Does anyone remember "Shrinky Dinks?" Yeah, that's what ABS does. It SHRINKS.

As it shrinks, it splits your print. That's why many of my ABS prints look like crap:

bl8cqLF.jpg


(plain ol' ABS on the left.)

A lot of people online were recommending that I use a nozzle temperature of 250c, and a bed temperature of 100c to print ABS reliably.

That sounds like great advice, but then I ran into two new issues.

First off, the closer that I got to 250c, the more the print would jam. I have an all-metal print head but as I approached 250 it would just stop printing reliably. So I back that down to 245 and so far, so good. (crosses fingers.)

The next issue was the print bed. What I found was that the printer would generally crash if I went that high. I think the problem is that the power supply can't generate enough voltage to get the print bed to 100C.

The solution that I am using for this seems to be somewhat unique, I haven't seen anyone else doing this:

I made some heat lamps.

I worked fast food in high school, and we used to keep food under warmers. So I just did the same thing with my 3D print.

I wanted to start small, so right now I am only using a single heat lamp. It's a 40W incandescent bulb.

Living in California, it's difficult to find incandescent bulbs. My heat lamp seems to be working nicely, so I'll probably add one or two more.

The idea here, is fairly obvious: my printer can't get the bed to 100C without some help, so I'm adding some heat to the enclosure to raise the ambient temperature. The exact same thing that restaurants do with a heat lamp.

By adding a single 40W heat lamp, I've been able to raise the temperature in the enclosure by about 15c. The bed used to "max out" at 80C and now I'm printing at 95C.

The enclosure is getting really cramped, there's literally five millimeters of clearance between the lamp and the printer. So getting another two in there might be difficult. One obvious solution would be to use halogen bulbs, but I'm a little freaked out about leaving a halogen bulb running near flammable foams and plastics. Back in the 90s, halogen bulbs would routinely burn houses to the ground.

Here's a list of various bulb types and their temperatures:

1) 300 watt halogen bulb : 340C / 1004F

2) 75 watt incandescent bulb : 130C / 266F

3) 60 watt incandescent bulb : 178F

4) 100 watt ceramic heat emitter : 140F



data from : Halogen lamp - Wikipedia and Watts, heat and Light: Measuring the Heat Output of Different Lamps