You are on page 1of 21

Loudspeaker Port Nonlinearity Testing

Or

“Just Port It1”

By

Scott Hinson

1
Alt-title credit: Jonathan Atkins

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Introduction
There’s a lot of misinformation on the internet about speakers. This may come as a shock to
anyone familiar with the various loudspeaker design forums. Or…not.

There is a lot of discussion about ports, port noise, distortion, size and shape. What isn’t
available is a lot of reproducible testing protocols, results from those tests and an idea of things
we need to do to improve knowledge on the tradeoffs of certain port design choices. I wanted to
fix that.

With this document I hope to introduce a method for testing, and show some results for common
port geometries.

One thing to note, I’m going to challenge a pretty widely held “everybody knows that” belief.
So…you might want to sit down, this could get rough.

Prior Work
Before getting into the testing I did, I thought it might be best to review some of the prior
published work. I can’t link or reproduce most of these papers since they are behind the Audio
Engineering Society paywall, but I can talk about them.

The first is a seminal piece of work by Juha Backman (now at Genelec) called “The Nonlinear
Behaviour of Reflex Ports” published in 1995 as a conference paper at the 98’th convention of
the AES. In this paper Juha demonstrates a measurement method that I used for my work in the
following sections and a extraordinary UK English spelling of behavior. He creates a test
enclosure with 8 12” woofers in an un-damped 400L enclosure tuned to a nominal 16Hz using a
single 4” internal diameter port. The idea is that the 8 12” woofers would be able to drive all of
his port geometries into non-linear behavior without running into thermal or mechanical limits
for the drivers themselves.

He tested various port geometries from a standard flush mount tube, ports with a sharp bend,
ports with small round overs, internal flanges etc. In conclusion he states: “The experiments
indicate that the commonly used reflex port with asymmetrical, sharp ends is a relatively poor
choice. The nonlinearity can be reduced to a large extent relatively easily, but the experiments
indicate that for optimal results both the inner and outer ends should be designed to reduce the
turbulence.”

I’m not sure I agree with his assessment though, I’ve summarized the results for his port studies
for compression and THD in Table 1, comparing the measurements for each type of port with the
straight flush mounted port that is standard on many speakers.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Table 1 Backman Port Performance Findings

Port Type Compression THD


Symmetric, both ends unflanged. (Port
Worse Same
tube through wall)
Symmetric, interior flanged, one end Slightly Slightly
flush with enclosure wall Better Better
One end flush with wall, one end cut at
Worse Mixed
45 degree angle.
Flush with outer wall, small radius inner
Same Same
and outer ends.
Flush with outer wall, interior flange,
Better Better
15mm radius on both ends.
Port with sharp 90 degree corner. Worse Better
Port with smooth 90 degree bend. Slightly
Same
Better
Port with 90 degree bend, outside bend Slightly
Worse
smooth inside bend sharp. Better

Of the port designs tested only the port with a 15mm radius and inner flange showed
improvement in both compression and THD showed a clear improvement. The types tagged
with slightly better performance showed truly that…only slightly better performance. The only
slam dunk improvement was the port with an inner flange and both the inside/outside port edges
radiused with a 15mm round over.

• Critical point: even though the sharp 90 degree corner port had reasonable distortion
performance…the compression performance was so bad that I cannot recommend it. If
you have to turn a port, the testing was clear a gradual bend (like using a large sweep 90
PVC bend) is preferred if you have the space to do it. You’ll be giving up a lot of bass to
make that bend a sharp right angle.

Next up I took a look at “Reduction of Bass-Reflex Port Nonlinearities by Optimizing the Port
Geometry” by a group of researchers with Philips published as a convention paper for the AES in
1998. In this paper the researchers did a bunch of research on ports with different entrance and
exit radius profiles, comparing them to a straight port. Most interesting they also compared it to
a port that was narrowest in the center, and had a slight taper as you move away. Then, at the
end of the port there is a modest radius flare.

As part of this research, they did FEM modeling to show the generation of turbulent flow. The
result of this paper is fascinating. The absolute best port at all volumes has a slight 6 degree
taper into and out of the length of the port and a 5mm radius termination. The second best
performing port is the straight. Any of the straight ports with large radius ends produced higher
levels of chuffing at high output levels..some ports substantially more. At low levels the

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


radiuses ports were extremely quiet, outperforming the straight port by quite a bit, but at high
levels the radiused ports get very noisy fast and had worse performance than the straight port.

Next up is “Maximizing Performance from Loudspeaker Ports” by Doug Button, Alan Devantier,
and Alex Salvatti from JBL and Infinity. This is the most substantial paper of the group for
practical information based on experimentation. They tested a ton of ports a ton of different
ways including measuring air velocity, distortion, ability to cool the loudspeaker box. First
published as a conference pre-print in 1998 it was finally released as a full Journal of the Audio
Engineering Society paper in 2002.

They have 10 general conclusions for the paper, I won’t reproduce them all here, but some
critical ones are:

“Vast historical data and the results presented in this paper suggest that the largest port area
allowable by a given design should be employed to keep the air velocity down if low port
compression and low distortion are desired.”

“For best distortion performance at higher levels, make sure both sides of the port are
symmetrical. Adding a small flange to the inside of a flush-mounted port makes a dramatic
improvement in distortion.”

“Maximally radiused ports have the best low-level performance, but they have poor high-level
performance due to excessive turbulence within the port, near the ends. This will lead to
compression and tuning shift due to the shortening of the apparent length of the port”

“Large ports with a taper designed to minimize turbulence will act poorly to exchange the air
(and heat) in the box. Ports that are in fact overdriven under maximum use and located at the
top and bottom of the box would be preferred thermally.”

“There are many approaches to finding a port profile that will provide excellent performance.”

These three confirmed the Roozen paper saying that highly radiused ports have better
performance at lower levels and add a pretty big potential negative. The fact that the turbulent
air changes the effective port area and therefore tuning frequency means that any EQ you apply
might need to be changed dynamically with program level. This is especially important for PA
applications where you’re likely to be using them over a broad range.

Next up is incredibly important for PA applications…the large taper ports in the paper were very
poor at exchanging air in the box and only a sealed box was worse for driver heating under
heavy, high power usage. Interestingly undersized ports have a tendency to exchange large
amounts of highly turbulent air cooling the speaker quite well. (This explains a lot of the
commercial 12”/1” PA two ways with tiny ports….)

So what have we learned? That sticking a port in a box has a lot more science and potential for
complex interactions than we expected? YES! That port geometry can impact performance at
overload? YES! That highly radiused ports might not give you the best high output solution?

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


YES! (Wha?)

Really. And it’s highly variable on the size of the radius. One important thing to note is that
you have to separate compression and distortion for this discussion. The
Button/Salvatti/Devantier paper shows this pretty clearly. Most of the radiused ports they tested
outperformed a straight port for compression. However, when you look at the distortion
performance of ports with radius of different size performance changes pretty dramatically. That
means there’s optimum(ish) choice. In their distortion testing they found that small radius ports
actually hurt performance, in some cases worse than the straight port.

What makes it difficult to compare the results for the JBL paper (or even my testing) to the one
published by the engineers at Philips is that Button and his team didn’t really test equal port
designs. The Philips teams applied radius to the end of a straight pipe. The JBL team made
their radius part of the port length calling it a Normalized Flare Rate or “NFR”

NFR = port length/2(flare radius) , 0 < NFR < 1 .

For the tested designs they found a port that had an inlet/outlet NFR of .5 to be the best
compromise for distortion and compression. It’s not clear from the three published papers
(without substantially longer review) that anyone tested identical port geometries except for a
straight port.

Test Setup
So, after all this theory what did I do? I wanted to test the performance of common port shapes
more than end radius treatments. I did test some radius ends on the 2” ID PVC port, but I was
more interested in the performance of rectangular and triangular ports. To do the testing I
needed the ability to completely overdrive the ports into their non-linear range. I constructed a
12” x 12” x 18.5” test enclosure out of .75” Medium Density Fiberboard (MDF). I used two
Dayton DCS305-4 8” subwoofers..these drivers have a linear xmax specification of nearly 9mm,
150W power handling and distortion reducing copper shorting rings in the motor. All ports
were cut to 7.5” in length yielding an effective tuning frequency (subtracting driver and port
volumes) of 29-30Hz for all designs. All ports were designed to have the same area.

I used a Mackie M1400i amplifier bridged into the woofers wired in series and the source signal
was provided by REW. Within the power limits of the driver simulations said I could theortically
achieve port velocities of 50+ m/s. My normal design “rules of thumb” dictate that I use a 2”
port for a single 6-7” class woofer. I’ve been known to use a 3” port for a high performance 7”
driver…so this port was severely undersized for actual use. For studying the performance at
high levels though it was perfect.

The port designs I tested were:

• 2” PVC pipe with no radius.


• 2” PVC pipe with no radius and a 6x6” flange on the inside end.
• 2” PVC pipe with ¾” roundover on external end, no flange inside

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


• 2” PVC pipe with ¾” roundover on external end, 6x6” flange with ¾” roundover.
• 3:1 ratio rectangular port.
• 10:1 ratio rectangular port.
• 30:1 ratio rectangular port.
• Triangular port.

One of the things you have to be careful about for testing ports in non-linear operation is that
they do fun things like “vortex shedding” and “jet formation”. This means that you can easily
overload the microphone with wind noise rushing past it. To minimize the impact in testing I
placed the microphone ~50 degrees off axis and about 6” away from the port. I used a miniDSP
UMIK-1 calibrated to a Cross Spectrum labs reference mic.

This ended up being one of the more challenging data collection efforts to date. At low
frequencies it’s easy for external noise to creep into your measurements. I would be making a
measurement and one of my kids would open the garage door, or the entrance to the garage and
the sudden pressure change/noise would cause the measurement to be corrupted. Early on I
discovered how critical microphone placement was, I ended up building a jig to improve
placement to make it more repeatable. That meant I had to re-start the whole process again so
that the measurements were directly comparable. At the end of each set of measurements for
each port, I would re-check the low level measurement to make sure nothing had happened
during the test and that I hadn’t started pushing the drivers into compression. I probably took
data for about 10 hours over the course three days.

I would have like to included a THD + Noise figure as well, but this is one case where you really
do need the background noise of an anechoic chamber or an otherwise quiet space.
Unfortunately, that limited my ability to do testing on audible chuffing to subjective means vs.
objective testing. I have the data, but the results are a bit unreliable because the ports at low
level are quiet enough where the background noise floor was higher than the measurements.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 1 Example Test Ports

Results
For each port I tested over a 20dB range (sometimes a bit more to ensure I was starting to see
clear compression/distortion degradation). Testing started at 2.33V RMS and continued upward
until I hit ~10% THD from the port. All tests are referenced back to the plain, easy base case of
a 2” port with no roundover. One thing to note, I expect I’ll get a bunch of questions along the
lines of:

“What did port bllaahahagrgh do?”

Well…I didn’t test that port...it’s impossible to test all possible combinations let alone every
combination that folks can think up. Not only that, port position will matter as well…a port in
the corner of an enclosure will see two walls extending the length of the port. I tried to do the
ports that seem to be most common, and I limited the testing to configurations where the port
was centered in the wall of the enclosure.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 2 Straight Port

The internal volume differences based on the port displaced volume was less than 3% across all
ports.

Port 1: 2” Port vs 2” port with ¾” roundover

The first test case looks at a 2” port with a ~3/4” roundover. As with all tests the starting
voltage is 2.33V RMS (-30dB REW signal generator level, max gain on the Mackie amp) which
I picked rather arbitrarily, but it was low enough to keep all the ports from being in
compression/high distortion regions. I moved up in 3dB steps until I got to -15dB then I moved
in 1dB steps.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 3 2" Port With Roundover

As you can see this smallish, one side only roundover actually made distortion performance
slightly worse. Even though this port doesn’t follow the roundover calculation in the Button
paper you can see that it follows their conclusion that an undersized port radius can make things
worse. It does seem to also follow their assertion that a roundover helps with compression, the
squared entry port taking a sharp dip in output at -15dB input level, while the roundover port has
the capability for reasonable output for a bit longer.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 4 Onset of Audible Chuffing Straight Flat Port

As a quick sanity check for my general design guideline that port chuffing, a broadband
mechanical noise is audible at around ~17m/s I paid close attention to the 2” port with no
roundover. One thing to note is that the chuffing started to become audible just before the port
nose-dived in terms of compression and harmonic distortion took off at input levels of -15dB.
The chuffing sounds are a broadband noise, and can occur from formation turbulent air both in
the port and at the port exit when the resonant air mass meets the outside air. From this testing it
appears that chuffing is an audible method that signifies the onset of non-linear port behavior.
There’s one giant caveat to this statement…this is non-linear behavior meaning the output
doesn’t follow a simple equation rule based on input stimulus. I suspect that for other ports your
mileage may vary using port noise as a predictor of the start of high levels of port distortion.

The port noise audibility corresponds well with a calculated port air velocity of between 15 and
20 m/s. At high levels the software just keeps predicting higher and higher velocities….I don’t
believe the software is accurate at these levels. However, jet formation was quite strong at high
input level, I could actually blow out a candle at a distance of 6 feet using the 2” round port with
roundovers on both ends. It wasn’t making a lot of noise, that jet meant that although the air
velocity was high, the effective coupling of the port air mass to the surrounding environment was
very low. The single roundover port started audible chuffing at about the same input level as
the straight port.

2” Port, Internal Flange, No Roundovers

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Next up I tested a round port with an internal 6” x 6” flange. No roundovers were added to
either the inside or outside port ends. This port showed excellent low level distortion
performance, but interestingly the overall efficiency showed a slight drop. The port with the
interior flange was a bit lossy compared to one without. I tested this three times to confirm the
performance and every time the results were the same.

The transitional behavior from linear behavior to non linear behavior was pretty interesting as
well. There seemed to be a slight increase in distortion before leveling off and then increasing
quickly. This port allowed ~2dB more input before chuffing became audible compared to the
reference straight port.

Figure 5 Internal Flange Performance

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 6 Internal Flange

3:1 Ratio Rectangular Slot

Next up I tested a 3:1 ratio WxH slot port. These are common in pro sound subwoofers, and
even home speakers. This port is shows even higher losses than the previous two ports. This
time however distortion performance was decreased and is worse than the round port reference.
Port chuffing started to occur about 1-2dB lower than the round port, but didn’t follow quite the
same pattern of low noise, some noise and very quickly high volume chuffing sounds. The onset
of the noise for this port was more gradual so it’s tough to say exactly when it was intolerable
compared to the other ports. If it weren’t for the relatively high losses the performance of this
port was pretty good. I can see why multiple manufacturers use this design or designs similar to
it with regularity.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 7 3:1 Rectangular Port

Figure 8 3:1 Slot Port

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


10:1 Ratio Slot Port

The 10:1 Slot port was one I wanted to test for a while. Years ago I did a subwoofer with a high
excursion Dayton Titanic 10” woofer and a slot port with roughly these proportions. It was
horrible…I ended up rebuilding the box. I didn’t take measurements at that time with the rigor I
do now, but it sounded bad, and measured bad. I had a feeling this time would be the same, but
I wanted to make sure I had numbers to back it up.

The pattern measured before continues…even higher losses and distortion at lower levels. This
port began audible chuffing a bit lower in input even than the 3:1 port. I cannot recommend this
design when a simple 2” port so clearly outperforms it.

Figure 9 10:1 Slot Port Performance

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 10 10:1 Slot Port

30:1 Slot Port

This port was tested because of embarrassing reasons. Mostly, because I did the math for the
original port ratios one morning at 5:15-5:30 AM before the kids got up and I had to make
breakfast for school. I thought I had created a 10:1 port. I built it that night and was looking at
it, thinking that’s…awfully thin. So I checked the area…yep…same. Checked the tuning
frequency…yep…same. Divided height by width? Nope not 10:1. Darn pre-coffee math!!! I
would never actually use this port, but I had the port, and I tested it. There are several
commercial designs with ultra high aspect ratio slot openings like this..so at least we’ll be able to
see how they compare.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 11 30:1 Slot Port Performance

Figure 12 30:1 Slot Port

First thing to note is that this port has even higher levels of loss and compresses even earlier.
Port noise again started earlier in input level, way before the round port. I can’t recommend this
design in any way shape or form based on these measurements.

Unless…

Well..

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Okay…so here’s the deal. You may be able to use this performance in a fairly clever way. Let’s
take for instance a very high excursion driver and you realize that in a cabinet it’ll have
reasonable response if it’s sealed or vented in the same sized(ish) box. Then let’s say you have a
desire to give it some marketing pizzazz and tell folks it’s got an excellent low frequency F3 or
3dB down point, knowing full well than very few folks have the equipment or know how to
actually check.

Let’s also say…you realize that a very narrow slot port at high enough drive levels effectively
closes off, making your vented box a lossy sealed box…now with less worry about over-
excursion at low frequencies damaging the woofer.

Huh…you might want to make a narrow slot port if you don’t think your customers will notice
the relatively poor performance in the transition region. But that might just be my snarky side, I
cannot confirm that sequence of design choices for commercial product. Though…you can tell
I’ve thought about it.

Triangular Port

You see triangular ports a lot in the corners of PA subs. I chose purposefully to make this port in
the center of the panel as I mentioned because I wanted to change only one thing at a time. I
must credit Marjan Milosevic with MM-Acoustics for being the first one to really give me the
heads up on how bad these were. He warned me….and I didn’t listen. I should have. Again,
you’ll see companies make them, but I suspect it’s for cost and ease of assembly reasons. It
can’t be for performance.

One thing that’s beyond the scope of this article is a study comparing a smaller circular port
tuned to the same frequency fitting inside the area of the triangle. I wonder if the loss of area
means that the overall performance would be the same. I kind of think the smaller circular ports
have a shot at out-performing the triangular ports based on how bad they are.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Figure 13 Triangular Port Performance

Figure 14 Triangular Port

Compression wasn’t as bad as I thought it would be, but distortion starts off poor and turns
horrible quickly. The port chuffing also sets in way earlier ~6dB or so than a round port and
unlike the narrower slot ports doesn’t increase slowly. Once it sets in it gets loud fast.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Round Port with flange and ¾” roundover on both ends.

Last I decided to put a roundover on both ends of a flanged 2” round port.

Figure 15 ¾” Roundover on Flange and Port Entrance, 2” Round Port

In some ways this was the best performing port of the bunch…in other ways…on closer look not
so much.

This port had the most output for equal input voltages to the woofer. Distortion performance at
lower levels improves on the straight port, and it shows less compression. You can clearly see
that it exhibits some gain in output before the compression sets in, Button et. al. showed that I
their paper. To me this might explain some perception of boominess at high drive levels. This
port also starts to compress earlier than the straight port, again the papers talked about this for
several flared port geometries. I suspect the roundover was too small to be truly effective for
adding performance.

Conclusion
If you can adequately size your port for a reasonable area, a straight round port, cut flush to the
baffle with a moderate interior flange is a very safe bet for performance. For simplicity
eliminating the interior flange appears to decrease output before chuffing sets in, but otherwise
the performance of the two ports are very similar.

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


None of the other ports presented here represent particularly clear improvements that support the
potential complexity increase in construction, and most are clearly worse. The literature I looked
at supports the findings in most cases. The Button/Salvatti/Devantier paper is pretty specific that
placing too small of a radius on a port termination can cause more problems than it fixes. They
are also clear that the radius needs to be applied in a pretty specific ratio with the length for
optimum performance. None of the roundover port designs presented here follow that guidance.
Additionally, I’d like to know how this scales with larger port areas and with placement along
the sidewalls or corner of the enclosure. I might have to test 4” diameter equivalent ports with a
double 18” test box and 6” diameter equivalent ports with two 21” woofers...but all of that will
have to wait for another paper.

Anyone want to build those boxes for me?

Bibliography
Salvatti, A; Devantier, A; Button, D (2002). Maximizing Performance from Loudspeaker Ports.
Journal of the Audio Engineering Society, 50(1/2), 19-45

Backman, J, 1995, The Nonlinear Behavior of Reflex Ports, paper presented to the 98’th AES
convention, Paris, February 1998.

Roozen, N.; Vael, J; Nieuwendijk, J, 1998, Reduction of Bass-Reflex Port Nonlinearities by


Optimization the Port Geometry, paper presented to the 104th AES convention, Amsterdam, May
1998

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson


Appendix: All Ports Plotted

Loudspeaker Port Nonlinearity Copyright 2019 Scott Hinson

You might also like