Phase Plug for Planars

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Recently, I've had very good results using phase plugs on midrange drivers. (Car Audio | DiyMobileAudio.com | Car Stereo Forum - View Single Post - Imitation)

Phase plugs on tweeters has been (mostly) solved, so I thought I'd examine what happens on a planar.

First off, what's a phase plug for?

I would argue that a phase plug is generally designed to minimize phase differences between different points on a radiator. For instance, if you are listening to speaker with an 8" woofer, and the speaker is rotated 45 degrees, the FAR side of the cone is 5.66" further away than the near side of the cone. That pathlength difference creates a dip at 1193Hz due to the phase difference. This is because 1193Hz is 11.33" long, and one side of the cone is radiating sound that's one half wavelength out-of-phase.

That dip at 1193Hz will basically limit how high you can listen to your speakers.

A phase plug can improve things, by equalizing those pathlengths.

This article covers this well : Phase Plugs

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

I think the BG NEO 8 is a particularly good candidate for a phase plug. This is because the radiator is capable of playing up to 10Khz and beyond, but the geometry of the radiator limits it's upper response. (Because it's so big.)
 
w9EP1S5.png


NVFWIhf.png


NQVeXjG.png


kGLTml1.png


Cixlrd5.png


Here's the frequency response and distortion of the NEO8 with six different configurations.

I'll explain what the different options were shortly.
To my eyes, there are three good solutions here:

1) the bare driver performs fairly good as it is

2) The Costas array phase plug reduces the high frequency peak of the NEO8, but without the cost/complexity of a capacitor. The frequency response is also a little smoother.

3) "Phase Plug 2" extends the highs, but at the expense of frequency response which is lumpier overall

I'm not real excited about the results of phase plug three and four. Basically the distortion goes up when too much of the radiator's surface is covered, just as it does in a horn.
 
Last edited:
07foCRY.jpg

This is the Costas phase plug. A Costas array is an array where the elements are literally random. I considered flipping a coin 48 times to come up with this pattern, but wound up using a random number generator instead.

The appeal of the Costas array can be seen in the patents of Voishvillo at JBL. Basically, dips and peaks are frequently caused by geometry. If you have two points on the radiator that are radiating the exact same frequency, the pathlength difference can sometimes lead to a peak or a dip. (Depending on whether the pathlength difference is one wavelength, or one half wavelength.)

So the Costas array works very simply, it randomly alters the radiator by removing the radiation from some points on the surface.
 
b4lGly0.png

Phase plug #2 was designed to mimic the effect of a log spaced array.

Basically 50% of the top and the bottom of the NEO8 is masked off.
25% of the center is masked off.
The idea is to get something like power tapering, where the edges of the array have the power reduced to widen the beamwidth and extend the highs of the array overall.
 
GnUX9F7.png

Here's phase plug number three.

This is simply an evolution of phase plug two.

I don't recommend phase plugs 3-4. If you look at the distortion performance, you'll see that the distortion is rising due to compression that's too high. Phase plug #3 has distortion that's about 5dB higher than the Costas Phase Plug.

Oddly enough, the distortion on the Costas Phase plug appears to be a little bit better than the NEO8 with no phase plug at all.
 
Last edited:
VjDfM7t.png


pha9coc.png


PIr11uf.png


jsCgBqe.png


Here's the polar response of the four phase plugs.

To me, the Costas phase plug is the winner here. In the midrange and the lower treble, the Costas phase plug response is virtually the same as the 'bare' NEO8. But in the top octave, the Costas phase plug is doing a couple of things that's positive:

1) it is raising the output level off axis. This is because of the geometrical cancellations of the large NEO8 surface

2) it is flattening the peak at 11khz. For the same reasons.
 
SVEoeZG.png


I was curious to see how the NEO8 would react to EQ. I've noticed that it is frequently impossible to use EQ to raise the output of a tweeter above 10khz or so. This is because the rolloff isn't electrical, the rolloff is caused by cancellation. So you can pour all the power you want into the driver, it just won't get louder. (And this is a great way to blow a ribbon to bits, literally make it explode.) I've only blown up two drivers during testing in my life, one was a planar magnetic, the other was a ribbon.

But there's good news here! The NEO8 *does* appear to be "eq-able". In the pic above, I've used 10dB of boost to bring up the top end of the NEO8. BTW, be careful with this in the real world, 10dB of boost is pressing your luck, from a durability standpoint.
 
well, the referenced article seems to refer specifically to "horn loaded speakers* - (NOT the silly polished bullets sometimes seen on hi-fi speakers)" :D -
*which a dipole planar tweeter used as shown here definitely is not.

I'd be concerned about resonances induced in the moving diaphragm by randomly sealing the apertures - up to 50%, if I read correctly

but thanks for yet again another creative use of 3M Blue painter's masking tape.
 
In the midrange and the lower treble, the Costas phase plug response is virtually the same as the 'bare' NEO8. But in the top octave, the Costas phase plug is doing a couple of things that [are] positive

Since you want the tweak to target top octave problems, foam might be better - something that's fairly transparent in the lower frequencies. Basically clone these:

https://www.madisoundspeakerstore.c...-ribbon-tweeter-with-optional-amorphous-core/

Foam might allow you to block a lot of area (to widen dispersion), without having an increase in distortion.
 
I agree.

P1010775.jpg


One neat thing about masking off the NEO8 is that we're not actually eliminating the radiation from that point on the radiator. We're just making the pathlength longer.

For instance, if you mask off one of the "cels", the sound has to travel about one centimeter longer than normal. That doesn't sound like a lot, but it's 180 degrees of phase shift at 17000Hz, and ninety degrees of phase shift at 8500Hz. I believe these random shifts in *some* of the phase is what smooths out the polar response between 7000Hz and 14,000Hz, while leaving the rest of the response curve unmolested.

Obviously, it would be nice if BG would just curve this dumb thing, that would work a lot better :) But that would easily double the cost.


Long story short: foam may help, but it reduces the efficiency. Phase plug helps, but doesn't affect the efficiency below 10Khz.
 
Interesting points: things that hadn't occurred to me, but make sense now that you've explained.

I believe these random shifts in *some* of the phase is what smooths out the polar response between 7000Hz and 14,000Hz, while leaving the rest of the response curve unmolested.

TL;DR version: I concur.

Long version:

I had one experiment that agrees nicely: as shown in in post 4 of this thread,

http://www.diyaudio.com/forums/multi-way/290092-tweeter-coaxial-mounting-midrange-horn.html

slapping a 4" driver on a 2" throat horn was OK but gave a ~4.5kHz notch.

I assumed it would be hard (too much work) to eliminate, but found that I could mostly fix it with a few scraps of dense foam stuck just inside the horn throat. The foam broke the symmetry / introduced path length differences, softening the -15dB notch to a much more benign -3dB dip. This allowed the cross point to be lifted from 4kHz to 7kHz ...and was much less effort than the fix I thought this would require: rebuilding the horn with a bigger throat + a proper (lathed / 3D printed) phase plug :)
 
Why not combine them? And have the bars cover both the vertical and horizontal axis (make the entire transition of the oval openings a smooth transition)

Also I wonder how much more quickly distortion ramps up as power levels go up with the costas array vs standard.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.