Linear distortion and edge coating
Copyright 2019-23 © Troels Gravesen

The listening test

20-01-2021: CNO-GRANDE and CENTER-641 NOW updated with new SEAS W18NX003 and new edge-coated W12CY006 drivers. 

There's a story to this W12CY003 file and it starts with the image shown below. A 4" midrange driver (not the W12CY003 and don't ask which, please) with its intrinsic frequency response (red) and the same driver after having had an edge coating to the rubber surround eliminating the all too familiar resonance phenomenon at ~700-1500 Hz depending on driver size, cone mass and rigidity, spider and rubber surround properties, mass/elasticity/resilience. All of these things will impact the place and size of this deviation from linear response.
Linear distortion is the non-linear frequency response as seen from the red graph. Non-linear distortion adds new frequencies that were not present in the signal fed the driver. Non-linear distortion is not the subject here, but has been measured and display no significant differences between the two drivers.

The thing is this: When the driver cone is moving in and out of the basket, the surround is supposed to move exactly the same, but at a certain point the movement of the cone becomes too fast and the mass of rubber surround becomes too heavy to follow the cone, thus when the cone moves forward the rubber will slash backwards - and vice versa. This means the rubber surround will produce sound out of phase with what comes from the cone. As the surround is not nearly as big as the cone, we'll see a minor dip at some frequency depending on the driver properties as stated above. The cone itself is part of the phenomenon from its inherent break-up nodes. We see quite a few hard-cones not having the problem, but then they usually have horrible break-up nodes at higher frequencies. It can be mentioned that foam surrounds are less prone to this problem.

Also read here:

By adding a proper amount of glue with proper elasticity and resilience* and in the right place on the inner side of the surround, the energy coming from the cone can better be absorbed and the result is a more linear frequency response - and hopefully, more fidelity. The deviation from linear frequency response is really not the key issue here, rather the narrow phase shift.
*The source of material and application technique shall not be revealed, so please do not ask.
The driver below used for the investigation is particularly well suited for this due to having almost identical frequency response below and above the trouble zone. This is not always the case, where the SEAS 4" has an elevated response in the ~700-1200 Hz before the dip, thus not so easy to align for comparison.

Listening test

Frequency response of 4" driver shown from 300-5000 Hz, red = as-is, green = edge coated.

Adding mass to the moving parts of a driver reduce driver sensitivity and to compare the two drivers sonically the non-edge-coated driver was slightly attenuated to produce the same sound pressure level except for the dip at 1500 Hz. The two drivers were set up in identical 15 litre cabinets side by side and run full-range for sonic evaluation. The cabinets were covered with a speaker cloth to enable blind tests and by a switch and a relay, test persons could choose between the two drivers. Initially they were only told they had to listen to two 4" drivers run full-range, no more.
A wide range of musical genres were presented and my visitors were asked to make a sonic evaluation of the two drivers.
After they had done their evaluation they were shown the frequency response of the two drivers, basically showing only the dip @ 1500 Hz, but still not knowing which driver they were listening to.
First of all all visitors were chocked to learn that two driver with virtual identical frequency response, except for the narrow 1.5 kHz dip, could sound that different. In fact, they said it was almost unbelievable.
Next, there was no clear winner when it came to evaluating the sound as such from the drivers, although after some considerations a minor preference for the edge-coated driver. Generally my test persons were utterly confused when the got to know the full story and that so little could do so much. What the heck was going on?

Here's my own take on what is happening:
First of all, these two drivers sound like night and day. Knowing all the details I cannot believe two drivers, so identical, can sound so different.
Having a slight rise in amplitude and a rapid phase shift in this frequency range produce a quite forward sound. A sound presumably full of energy and with lots of presence - which can be quite seducing. It makes the sound stand out in front of the driver, where it should really start at the baffle plane and exceed backwards as deep as the rest of the system and recording allows. The downside of this is that this forwardness can turn sharp and edgy and make some records unlistenable.
I remember sitting in a church listening to a choir and the church acoustics were horrific and I was sitting in a unfortunate place. At certain notes by ears would click as direct and reflected sound would be 180° out of phase, same thing as I experienced when testing a Lowther EX3 driver.
We can be seduced by an energetic loudspeaker with linear distortion, which may "bring the music into the room". It may make us hear things we didn't hear before and we take it for energy, transparency and detail, where we may rather be seduced by linear distortion. But taste cannot be argued.
The edge-coated driver is much more neutral sounding, which particularly on vocals bear fruit and to my ears it presents a deeper soundstage with a better portrayal of the recording room - and the evaluation was done on material not recorded in a studio. The edge-coated driver is no less dynamic, but it doesn't appear to make things sound better than it really should. It's no less transparent, no less capable of delivering micro-detail - to my ears. Micro-detail is a dubious word so often read in speaker reviews - and we really can't define what it is. It can be what it is, a true low-distortion speaker's ability to reveal details never heard before or it may be simple linear distortion making us hear things we didn't hear before.

If my visitors were chocked I was no less the first time I set up the experiment. It will forever change my attitude when I enter a room at a hifi show to hear an unknown speaker for the first time.

Obviously I have been discussing edge-coating with people very skilled in the art of speaker driver manufacturing and there are objections. The primary issues brought forward are these: (1) loss of band-width, (2) loss of micro detail, (3) glue not sticking properly to rubber over decades - and (4) cosmetics.
So far none of the drivers I have done edge-coating to has lost any band-width, they stay as broad-banded as they were. They loose some 0.5-1 dB sensitivity, but this is no problem on at least these midrange drivers as they're too sensitivity anyway and need attenuation in their final application.
Loss of micro-detail is discussed elsewhere on this page and comparing the sound of the two brands of drivers presented here with/without edge-coating is comparing apples and pears. To my thinking, they're simply too different to claim one superior in micro-detail over the other. I'm tempted to say the linearised driver helps us in hearing micro-detail as out brain is not disturbed by the phase irregularities.
Edge-coating of drivers have been done by the skilled ladies at ScanSpeak for decades and I struggle hard to be as good as they are. With modern coating materials (glues) - some made to adhere impossible materials like plastic and metal foils (coffee bags) - it is not a problem finding glues that will stick to almost whatever (except PTFE) for decades and at the same time retain elasticity and resilience. It can be done. 
Appearance? Well, the customer is always right and if he doesn't want a driver "smeared" with glue - he's right - and choose to live with what he can get.

There is a radical solution to the problem: Choosing the right rubber surround! There are rubber materials that may solve the issue and forever eliminate the problem by having the right properties of mass and elasticity and my only comment to this is: Then please go do it!
I have seen rubber surrounds with a special profile, where part of the inner rounding was made thicker to do the same thing as added coating. It can be done - but may take a lot of moldings to get it right - and cost a lot of money.

There are probably very good reasons manufacturers skip this final step in speaker design: Cost and customers ignorance. Customers' ignorance may be well-considered as his customers won't have the opportunity to hear the difference anyway. "Good enough" rules!
Production costs cannot be ignored, absolutely! Finishing speaker manufacturing can be costly and having skilled personnel doing edge-coating costs. And having 10 different rubber surrounds made so see which works and which doesn't may cost a fortune.
Another invisible and not so costly way of solving the problem may be adding the coating to the rear side of the surround, but this obviously must be done before assembling the driver.

I want to stress I don't want to point fingers here. There may be many reasons manufacturers do what they do. I just want to tell a story of my surprise to how different two drivers can sound, just from a very small, but narrow dip in frequency response and subsequent shift in phase.

There's an interesting comment here by Nelson Pass on negative phase:
"So why is the phase important? Well, it's a subtle thing. I don't suppose everyone can hear it, and fewer particularly care, but from listening tests we learn that there is a tendency to interpret negative phase 2nd [harmonic] as giving a deeper soundstage and improved localization [of images] than otherwise. Positive phase seems to put the instruments and vocals closer and a little more in-your-face with enhanced detail."

Above the frequency response of the two drivers as presented to visitors.
It has to be said the frequency response graphs of the two drivers outside the range shown here were identical.
I just do not show it all as you'll be able to guess the driver - and this is not interesting in this content.

Left: Frequency response 500-5000 Hz (red) and minimum phase (green) of the driver as-is.
Right: Frequency response 500-5000 Hz (red) and minimum phase (green) of edge-coated driver.
(use google to find out about minimum phase)
As can be seen on the graph to the left, there is a sudden phase shift (green) over a very short frequency range from ~1300-1700 Hz.

Based on literature studies I think we can conclude that the human ear is relatively tolerant towards variations in amplitude where rapid phase shifts is another matter (e.g. Richard Vandersteen). Rapid shifts in phase response does something to our perception of sound and it can be seductive - and it can turn us off. All depends on where and what and how much, but it is something that a driver should not do as it was not present in the signal it is fed.

And now to the SEAS W12CY003 midrange driver:

Above the SEAS W12CY003 midrange driver, as-is.
Click images to view large.

I've done the W12CY003 before in a tiny loudspeaker, Eekels Mini, where I had some major trouble getting the midrange reasonably flat. On the small baffle it had a humongous peak at 900 Hz and I interpreted this as a result of minimum baffle and to some extent intrinsic driver response. Here, many years later, I started placing it flush mounted on my test bed, a 34 litre test cabinet with a 70 x 80 cm baffle being short of a true infinite baffle, but 70 x 80 cm will produce similar results down to around 500 Hz - enough for evaluation. The result was this from four test samples:

Above the four W12CY003 drivers at hand. A slightly elevated response at 1 kHz and a significant dip @ 1.5 kHz. Actually the orange driver display a dip of 7 dB over just 500 Hz. And from 1300-1700 Hz the minimum phase shifts some 40 deg. This is audible!

Orange = sample 3, frequency response. Green = minimum phase.


Above the edge-coated drivers.
Click images to view large.

Above two edge-coated drivers on my 70 x 80 cm baffles.
Actually these two 4" midrange drivers are now the best 4" drivers i have ever tested. Period!
+/- 1 dB from 200-5000 Hz. And no serious peaks at all above 5 kHz.
The minor dip at ~500Hz comes from my limited baffle width.


Above the same driver w/wo edge-coating.

Above the same driver from 200-10000 Hz and minimum phase (green).

Having finished the CNO-GRANDE, I can only say it pays off having an extremely linear small midrange driver. Apart from its addition to dispersion in upper-mid and lower-treble, it adds to the fidelity and correct timbre of all instruments - and it makes it possible to use very shallow crossovers, where 1st order filters are the only filters not having phase distortion. Narrow phase shifts in upper-mid to the highest treble add their marks to the sound perceived.