Context and Market

The project began with the simple brief of “intuitive and engaging interactions for a portable speaker”. Three main categories of interactions were considered:

Extension/Compression Gestures
Deforming Capacitive touchpad could provides
Soft bodied speaker where Vertical motion to control volume endless control options
warping it modifies the sound + Symbolic: speaker size and volume + Could be reinforced with
+ Extremely novel become intrinsically linked sensorial feedback eg: haptics
+ Playful and Expressive + Both intuitive and engaging + Simple to implement
− Acoustical challenges could − To provide reasonable travel you − A well explored control
render it too gimmicky are limited to vertical, tall forms mechanism so novelty
− Engaging but not intuitive − This motion alone cannot control is challenging
all the functions of a speaker − Possible to be neither
Selected for potential and feasibility! engaging nor intuitive if not done well
Refined Brief
A premium portable speaker with 360° sound, beautifully controlled with vertical motions that reveals the speaker and acts as power and volume control

Target User: Technical Specification

Medium-High End Buyer: Individuals who have disposable income. • Even acoustic distribution: Avoid variation in volume around the speaker
Budget speakers have limited flexibility for innovation. • Full range frequency response: Must replicate deep base, smooth
Design Conscious: Product aesthetics should be an key purchasing midrange and crisp treble frequencies
motivator, this will be a primary focus of the speaker concept • Connectivity: Offers a flexible range of methods that provide an equal
Experience Driven: Those that see the value in control interfaces that experience; bluetooth, wifi and wired
go beyond traditional button controls • Protective: Resilient to drops, scratches, and dust. Waterproofing is limited
Casual to Semi-Audiophiles: The speaker should sound excellent but due to the inclusion of moving parts, thus unsuitable for wet conditions eg:
convincing serious audiophiles to move away from traditional wired box watersports
speakers may be challenging. • Thermal management: Heat from amplifier and battery must dissipate

The Market Where should my speaker aim to fit within existing market offerings? Brands and products were analysed Bang and Olufsen aim
for visual uniqueness,
USP is a flexible this and the high price
ecosystem seperate them from the
Cheap The best option
combining rest of the market
speakers from available for a
The 360° speaker 360° speaker. both portable
countless Affordable, shown here is and static
Excellent
random unglamorous considered an speakers.
benchmark
brands on speakers “design” speaker: Monotone,
despite being
websites like designed for evidence of how practical
outdated- it still
Amazon the outdoors stale the market is uses USB-micro! design

£100 £200 £350 £500 £1500

Indicative price. Generally, cost =
audio quality but this relationship is My speaker will position itself somewhere within this range,
not always proportional aiming for aesthetics to rival B+O without the exorbitant price
 Interactions
The experiential USP of the speaker. These interactions offer a unique way to control your sound.

Play/Pause
A short press plays or
pauses the current
track. The speaker gently Volume Down/Power Off
springs back, unless Softly compress to turn the volume down, and compress
more force is applied all the way past a final “click” to power off. This also reduces
in which case it will air space within the speaker, slightly boosting middle
compress, triggering frequencies and limiting the bass (at lower volumes bass is
the volume control much less noticeable anyway)
interaction

Volume Up/Power On
Extend upwards to power on and further to
Track Navigation increase volume. This also increases internal
Clockwise to skip forwards, air space, significantly boosting bass which
anticlockwise to skip backwards. Quick becomes more noticeable at higher volumes.
motions change track whilst slower The speaker is bottom heavy so this gesture
ones scrub through current track shouldn’t lift the entire speaker up
Forms, Colours Honeycomb Pattern
• Hexagonal mesh provides
structure whilst gaps
provide airflow
Motivation Behind Forms • Fabric covering
protects from debris
• Adds subtle textural
Vertical Bars Curved Notch interest to the otherwise
• Provides the illusion of
depth despite the bars
• Informed by ergonomics to
provide a comfortable way to
unassuming base Cloud
White - Grey
only being 5mm deep extend the speaker Timeless, simple elegance
• Slightly scatters higher • Adds a sculptural quality;
frequency sound the speakers stand out Ember
waves [1] for subtle aesthetic choice Orange - Copper
softness and tonal Bold, warm and contemporary
warmth
Shimmer
Pale blue - Violet
Hero colour with hidden depth

Canopy
Emerald - Light Green
Vibrant and verdant style

Eclipse
Rich Blue - Dark Blue
An opulent Art Deco homage

Colourways
For a premium product, individual expression
is crucial. 5 distinct colours cater to a range of tastes.
Pragmatic users could be drawn to the monochrome “Cloud”,
whilst the bold “Ember” could be for someone looking to draw attention
to their speaker. “Shimmer” is the powerhouse colour used for most visuals: the
original with the most colour variance caused by anodised film. The base is always
grey-black, providing both contrast and visual consistency across the range.
Components Custom “D” Subwoofer
Cross Sectional and Exploded Views
The curved notch poses a space
restriction on the top of the speaker.
Rather than limit surface area with
a smaller circular sub, a custom D
shaped subwoofer maximises the
available space, to provide bass that Link Piece
can live up to the speakers name! The bars slot into place,
fabric can then line the
inside, forming the structure
of the extending section

This is all
Linear Sensor/Encoder
extra airspace
Hybrid to provide position data (informing volume) and to
to amplify detect rotations for skipping tracks. A pivotal component,
subwoofer
vibrations!
made of a titanium alloy for excellent strength

Custom Wideband Driver

Handles midrange and treble. These
frequencies are much more directional so
3 drivers are spaced around the speaker to
ensure an even 360° coverage. The custom Base Mesh
tall “O” shape widens vertical directionality Stamped and roll formed.
Provides the structure of
the base in addition to the
honeycomb pattern
Friction Ring
Embedded around the lower inside surface of
the bars. Assists the linear sensor to provide
Coil Springs
Bi-directional resistance
the exact right amount of resistance
to the skip-track twisting
motions
PCB+Shield
Amplifier, crossover circuit,
bluetooth, wifi, USB-C and aux; Battery Unit
everything the speaker needs. 3 x 21700 cells provide
A central button interfaces 12600mAh of capacity,
with the linear sensor to detect estimated to provide 1 1 hours
short presses (to pause/play). of audio. The casing has a hole
The protective shield has cut- through it to allow the linear
outs to aid heat dissipation sensor to pass
Validation +100%

FEA Simulation Relative

Theoretical “Perfect” Response
A physics simulation of a slightly simplified Loudness
(Converted from
speaker model was set-up in Comsol dB for readability)

Multiphysics. The key findings are summarised.

More depth on later pages. Blue:
-100% Frequency
response when
Frequency Response: fully extended

Extended vs Compressed
Loudness at 1m distance was taken at 1/6 octave -200%
intervals. The extended space boosts the base Green:
performance by around 100% for the subwoofer, with Frequency
response when
a minor trade-off being a slight decrease in midrange fully compressed
loudness. This supports the core concept, validating an -300%
expendable enclosure as beneficial.

Acoustic Radiation Pattern Bass frequencies. Boosted by

around 100% when the
200Hz Middle frequencies. Much
less dependant on the extra
2000Hz Treble

speaker is extended extension space

Loudness was computed in a spherical area around Frequency
the speaker. The minimal difference in loudness above
and below the subwoofer confirms the validity of its
upward-facing position. As expected, directional variance
massively increase at higher frequencies, reinforcing the
need for an array of three wideband drivers.
User Validation
I wanted to confirm I had met
the aesthetic brief of a premium
speaker. Armed with a printed
render, I decided the best
79.9 dB approach to reaching my target user was the dubious
method of asking people walking out of the Bang and
Maximum Olufsen on Bond St: “Can I get a quick reaction to the
difference in design of this speaker?” Here are some answers:
directional
sound level “It looks like modern art but the colour isn’t really for me”
less than
2dB, (20% “If I knew it sounded good I might buy it”
loudness) “It reminds me of the stuff in there, you know, flashy
and not in my budget”
“Oh this is really nice. I’ll take 10”
78.2dB
“My quick reaction would be that I like it”
Most people who didn’t ignore me had a positive
Radiation at 200Hz reaction, but more thorough research is needed to
confidently validate that people would buy the speaker
Sketch Exploration
Alongside much rough sketching, these concept sketches show how the design evolved (in chronological order)

Line based texturing and Triangular shape with small Rounded square shape: The first design that really
rounded rectangular slot lip to extend the speaker. Both utilitarian and plain. Circular inspired me. Heavily iterated on
both revisited later aspects felt cumbersome shapes became preferred

Moodboard
Textures, forms and visual styles that
caught my attention, saved for inspiration
in my exploration and visualisations

Circular bars with more The bars are orientated in Cladding of vertical, uniform An off-piece experiment, opens
pronounced curve, a parallel with a curve, this curve bars in a circular pattern. with more of a twist. Needs two
combination of previous 2 felt too subtle Inspired future designs hands , ergonomically poor

Design Opportunity/Constraint
As I learned more about sealed loudspeaker design,
an interesting opportunity emerged. Compact
speakers with small internal enclosures limit bass
The next
vibrations. If a subwoofer was integrated in the design
extending section and the enclosure could expand turned out
dynamically, bass performance could be significantly to be the
improved [2]. Current designs didn’t allow for this so final...
future designs had to allow space for a driver at the
top and an expandable internal volume Designed cross sectionally, practical A poor attempt at a “bar” design
but lost the allure of previous designs that fit with the new constraint
 Modelling and Prototyping
Modelling Experiments Prototyping Pt 1: Prototyping Pt 2:
In Fusion 360, the name of my CAD model that ended Concept/Interaction Testing Ergonomics of Curved Forms
up being final is “experiment 7.5”. Below are some of Method: 3D printing and foam modelling Method: Air dry clay modelling
the 7 models that came prior, informed by sketches,
rendered in Keyshot Testing the validity of some designs and the Before the final design, I wanted to experiment with smooth
feeling of the extension interaction, plus curves to find more comfortable ergonomic geometry for both
“Experiment 2” exploring additional interaction methods extension and rotation. I modelled only the top sections, at
2/3rds scale to save material
Gold bars felt too Art Deco.
The inward facing bars lack Extension Rotation
design flair. I was glad to Small Contour
see the bars obscuring the
fabric them so well Space efficient but
awkward for both
interactions, it feels
“Experiment 3” too cramped
The cloudy plastic
was chosen to subtly
reveal the bars when Fits three fingers comfortably
enough and extension is easy Continuous Contour
compressed, but it
looks cheap but the untapered edge looks Good for rotation but
and feels jarring to maintain space for
the subwoofer the
“Experiment 5” Hard to extend due to the small contour must still be
[Link] concave top surface small, thus it is still
This design was killed when I
lends slight visual interest but cramped
realised it looked like a fusion of
an Apple Homepod and an aero- exemplifies the edge problem
press. The anodised metal and a Large Contour
dark base for contrast both showed The gap was too small. Rotation is
potential Someone testing it twisted comfortable and it
the top; showing the obvious feels easy to lift, with
solution for track skipping. The plenty of space even
foam had slight give when you at this scale. Needs
“Experiment 7” pushed down, and this became
A proof of concept for my final idea. Showed further refinement
the idea for play/pause
great potential although the curved notch
didn’t feel correctly implemented. It None of these concepts felt perfect. But A developed version of the larger contour was the clear choice.
needed careful consideration, prompting interacting with them revealed how the The final concept can use parallel bars, each shaped to the
prototyping part 2 speaker could be controlled beyond volume desired contour to achieve both ergonomics and aesthetics
Final Design Process Bars and Enclosure

Sketch
Rough drawing
based off previous
designs and
prototyping helped
plan the modelling
process
Base
T-Spline modelling to Parellel lines on a circle, Interior cavity and Form was elongated Each bar was Base modelled with
create a smooth form extruded up, interested fabric lining modelled and joinery pieces smoothly filleted curvature that allows
with previous surface to were modelled some rotation
make the smooth form
Components

USB-C and
Aux Port added
Sourced a sensor, batteries and coil PCB-Shield made Rough PCB modelled with Custom drivers modelled using a and bars modified
springs, modelled a battery case with t-splines standard components variety of techniqiues to accomidate

Assembly of Base material

components Base Material Graph
The fabric is a combination
Final CAD of pbr texturing plus a All the
procedural geometry components
node that physically model were textured
indvidual fabric hairs, with and then it
denser regions showing the was ready for
hexagonal pattern rendering!

Bar Material Branding

A subtle bump and
roughness map, with an
anodised micro-film for color
varience. Enabling rounded
corners improved reflections Warped text symbolise the unique
Importing into Keyshot and overall allure. curves of the speaker
Default With rounded corners
Advert
Showing off the speaker design
as well as colour options
PART TWO

Simulate Experience
Simulation Setup Model Simplification
The complex geometrical CAD model, shown bottom left, poses a problem for the
Software Selection simulation. It makes replicating the physics more technically challenging as there
are more physics interfaces to consider. It also significantly increases the complexity
Different software packages were evaluated for my simulation needs of the CAD mesh representation, which could exceed the capabilities of my PC and
lead to a failed simulation. For these reasons, the speaker model was simplified to
• Pros: I have prior experience. Easily
represent the most important parts. This is a summary of the changes:
accessible general FEA simulation tool
• Cons: Acoustic capabilities are limited and • Midrange/tweeter drivers were omitted, and the model was halved to reduce
loudspeaker simulation resources scarce mesh size and ease computation.
• The curved notch was removed for mesh simplicity. Minimal impact expected
as most of the enclosure is below the notch anyway
• Pros: Non-FEA tool specifically designed for a • A standard driver shape was used since resources for simulating custom
range of loudspeaker design applications. Free drivers in Comsol were unavailable.
• Cons: Only works with traditional box enclosure •
shapes, so my custom cylindrical shape poses • An air cavity was added, intersecting
a unavoidable problem with the base so air space expands as
the sliding section is moved upwards
• Pros: Extremely powerful and versatile FEA • A single driver was used to simplify the
tool, audio industry standard with plenty of model, idealized to cover 20-5000Hz, for
loudspeaker simulation resources a full range frequency response. In reality,
• Cons: Very complicated, hard to access (no one driver could not effectively do this
student access, free trial must be used) without distortion

Conclusion: Comsol Multiphysics is the only tool I was confident could accurately
simulate and produce useful outputs, so it was chosen despite its challenges

Custom drivers
simplified and
combined
as one
standardised
model [4]

Simplification

Air

Solid Extension
significantly
increases
enclosure
air space
Simulation Setup Pt.2 Physics Setup
The simulation uses three physics interfaces which are then linked with
multiphysics couplings. All 3 interfaces have a symmetry condition to
Domain Setup accurately capture the behaviour of the speaker.
Comsol simulates physical systems
using domains which can be defined Shell Interface
directly from CAD geometry (such Efficiently models thin, flexible components using 2D shell elements
as the enclosure), using operators since meshing thin elements as solids becomes very expensive
like intersection (the air domain is an • Features: Cone, surround, spider (see image)
example of this) or with faces (such • Thickness’s ranging between 0.4-1.5mm are applied
as the symmetry boundaries). Some • Damping: energy dissipation is defined by isotropic loss fac-
examples are shown. Materials can then tor but now they are frequency dependant, e.g. for the foam
be assigned to each domain. A lot of the material that is applied to the cone surround, it is:
domains are for parts of the driver. These
are labelled with a

Solid-Thin Coupling
Links displacement between thin and solid ele-
ments
Mesh Setup Solid Mechanics Interface
3 meshes were set up, for 3 possible Models 3D deformations and the forces in bulky structural
extensions of the speaker (minimum, middle components.
and maximum). Significant mesh refinement • Features: Enclosure, voice coil, magnets, basket (see image)
was needed to get an mesh that accurately • The electromagnetic force equation is applied as a body load
to the voice coil using the following equation:
captured fine details while still being able to run
• Structural
on my PC. This is why the triangles appear finer • solutions
around the intricate geometry of the driver and Maximum • Damping: energy dissipation is still defined by material iso- (less
Important)
larger on the solid enclosure where the physical tropic loss factor, now frequency independent (constants)
properties are uniform
Middle Acoustic-Structure Coupling
Transfers structural acceleration to driving acoustic pressure
Minimum
Acoustics Interface
Simulates the acoustic pressure field within the bounded interior
cavity of the enclosure across two sub=interfaces .
• The first sub-interface (FEM) simulates the acoustic pres-
sure field within the enclosure air space (including the shell
domains). The interface models the narrow slits around the
voice coil that can significantly affect acoustic impedance
(and thus would be inaccurate to ignore)
• The second (BEM) simulates the external radiation into an
infinite domain without the need to mesh the external space.
Essential for determining the frequency response

Outputs acoustic solutions (frequency re-

sponse, radiation, acoustic pressure)
Simulation Results Acoustic Radiation Radiation Pattern at 200Hz
Acoustic Pressure At 200Hz, the decibel reading
The below plots show how SPL distribution varies at different frequencies. As varies only by 2dB. This differ-
theory dictates, low frequencies have less nodes and anti-nodes causing much less ence becomes even less signif-
variation in SPL than for high frequencies. icant at lower frequencies. This
result validates an upwards
Acoustic Pressure at 200Hz Acoustic Pressure at 2000Hz facing subwoofer as perfectly
feasible.
For higher frequencies the dif-
ference becomes an issue; at
2000Hz it is around 30dB. This
validates the need for an array
of midrange/tweeter drivers.
The full unsimplified design has
3 around the base which should
provide a even distribution of
radiation based off the directivi-
ty shown here.
This data could be used in the
MaxMSP experience to vary the
volume accordingly as the cam-
era orbits the speaker, but this Radiation Pattern at 2000Hz
felt too ambitious for now.

Stress Analysis
Basket Stress Plot at 50Hz
Analysing the basket yielded some interesting results; at
high frequencies the stress was distributed evenly and
too small to be an issue. Low frequencies are causing
some significant stress concentration on some corners
near where the coil lies (see arrows on image).

Stress concentration at low frequencies is concerning

for a subwoofer. It will cause fatigue which could lead to
unpleasant resonance or even breakage.

The shell interface defined the thickness of the basket

as 0.8mm. Clearly this is insufficient. This can inform
further designs; the subwoofer basket must be
significantly thicker.
Note: for all these results, the extension of the speaker had minor
impacts but nothing interesting enough to be explicitly mentioned
 Simulation Results Pt.2
Frequency Response (Sensitivity)
This was the most important output from the simulation. To determine this, a dB SPL reading was taken by Comsol 1 meter away from the driver at a range of frequencies starting
at 20Hz, then repeating 1/6 Octave intervals up until 4000Hz. I repeated this for the minimum, middle and maximum meshes. The plots were overlaid on each other for readability.
Interesting features that need to be replicated in MaxMSP are annotated.

An unexpected result was the low and

middle extensions having a louder
low-mid response (200-500) than A narrow peak in the response
the maximum extension. This may occurs around 3000Hz. This
be caused by the resonance of the varies by around 3dB between the
enclosure affecting the standing waves extensions
inside. This suggests that the optimal
response occurs somewhere between
the middle and maximum extension,
rather than the at the maximum
extension as predicted A dip occurs around 2000Hz which
is constant for all extensions; an
interesting feature to replicate. It may
prove to be beneficial and reduce
sound harshness

There is a wide peak around 1000Hz

which has constant boost magnitude
but gets wider with the larger extensions.
Minimum extension shows a This should boost vocal clarity and
steeper low-frequency cut-off with certain instruments like guitars
the least boost, while maximum Maximum Extension
extension offers the most extended
response and highest boost
The middle extension falls in KEY: Middle Extension
between, as expected. The
difference between maximum and
minimum is significant; around 8dB, Minimum Extension
almost twice as loud

Note: the frequency response for

higher frequencies is less accurate
to the real speaker as the mid-highs
would have their own drivers, but it is
still interesting to discuss and replicate
Filter Representation My setup for tuning the filter parameters to match the possible
frequency responses of the speaker

Filters in MaxMSP are usually applied with a biquad object, but these are limited to one filter
e.g: one notch filter. The filtergraph~ object can represent a series of notch filters which
cascade~ then can combine and output.

To control each filter the parameters of filtergraph can be edited with lists in this format.

I designed a system with six distinct filters, each targeting specific features observed at
different points in the speaker frequency response curves. I used slider objects to control
gain, and—in one case—the Quality factor (effectively the frequency range the filter affects).

By manually adjusting the gain range and Q factor for each filter, I was able to closely match
the behaviour of the maximum, middle, and minimum extension responses. See image top
right for the setup. The goal was to create a flexible system that could smoothly transition
between these different extension responses using a single control slider.
The final sub-patch for controlling the frequency response for any
Maximum Middle Minimum possible extension with one slider
Extension Extension Extension

The above images show the results of my tuning to match the frequency response. I then
determined the range each slider needed to output. The master slider outputs a float
from 0 to 1, which is then modified for each filter using basic math operators to output the
required range. This principle also controls the volume and the animation of the speaker, all
using the one slider.

For the volume control, a desired range was determined. So the subtle changes in spectral
envelope can be observed without volume interfering, I implemented a simple override
feature which sets the gain at a fixed 0dB regardless of the slider level. I added a similar
override to hear the unfiltered flat audio response to allow for comparison, and the sub-
patch was complete.
Jitter Scene Implementation

Jitter objects to generate a scene with an environment, a table and

both parts of the speaker. Some simple animations are triggered when
playing/pausing or when skipping tracks to demonstrate all the gestures
used to control the speaker. The extension animation is controlled using
the same master slider from before. I found working with materials in jitter
to be highly unintuitive; so this is an area that could be expanded on to
make the final outcome more photo-realistic.
 The patches were combined with a

Final Experience
simple audio player. UI was designed to
emulate a phone app

Watch a video of the experience: [Link]