0 ratings 0% found this document useful (0 votes) 2K views 15 pages I Have Things in My Ears - How The Heck Do They Work? (Draft)
A graphic essay on hearing aid technologies. This is the final draft; there is a section missing from p. 10 and I am waiting on the bibliography. Comments welcome!
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content,
claim it here .
Available Formats
Download as PDF or read online on Scribd
Carousel Previous Carousel Next
Save I have things in my ears -- how the heck do they w... For Later “Tf HAVE THINGS
IN MY EARS =
HOW THE HECK do THEY worK>”™
7
A COMIC ABOUT HEARING AIDS FOR MY SIX YEAR-OLD SELF
WITH LOVE FROM 29-YEAR-OLD ENGINEERING PHD STUDENT ME —>
by Mel Chua - cc-s7 2015
please don't let
mom and dad catch
you with Hak screwdriver,
When | was b, \ got in touble
for disassembling my hearing aids
with an eyeglass screwdriver,
Mind you, |
Never wanted te
STi rye
. 2 ¥
waa Se ee
\ grew up to be an engineer
electrical £ compute?) with a
builk- in auditory lows pass filter.
CTeanslation: t'm deaf. )
The picture on the right is
called an audiogram, and it
(sort of) shone what | can hear. Son Tere
SAY MAKES NOISE,
AND 1 SEUEVE THEM
This axis shows how much louder
Soveds have 4o be fr me 40
hhear then, compared *2 4 “normal”
hearing pevon.
12088 ty OW OW JET Engine OW
ave ve bere)
Go See piste tec, Sis elon ae
frequency (He)
mp
This ante
Shows the
Frequencies
| ean heer
= Jow avtes igh notes >
Cpiane Keyboard
pane MIDDLE C fobviously not
ven 5
cas (201.6 He) to scale)
You can think of the audiogram | Lely of 1 OCTAVE Agove
Uke @ giant eqvaliter, ‘bass: MIDDLE C (623.3)Drums don't have Sometimes, babies will
Fie alacms make
@ Faint, barely
delectable, and nok
hi-hats or cymbals
open Weir raouths for
(as far as 1 can dell),
+ long Hime, and all
| Bass deums make a Me Neaving people ill fervibly unpleasant
| noise, though. Stat Plincking ee
| (Bass/ kick) W's amusing.
It also means that things like water (and the “t" sound) often feel
like figments of the rest of the world's imagination.
Ws not a terrible hardship or anything.
| dov'+ feel broken.
I'm just curious.
jest curiow
So lets stat with a
dissection of my
EARHook
MICROPHONEFirst of all, heaving aids look very different from the outside.
Ce (es:
‘COMPLETELY mw WE HAE Ste
IN CANAL cana BEHIND THE EAR
Cote) cate) Hs) Gre)
They come in all serts of form factors depending
on the type of amplitication you manly hort
jon want +o be able + conte! it, whether you
Want +40 heok it vu te your cell phone, ete.
Generally speaking, louder and mere featvre- full
hearing aids are bigger.
This is mine. H's what
really loud hearing aids
look like.
Secondly, aad more important fo my exginecring well,
the picture on the eudside doesn't say much abord
ihat happens % the inside,
This is a better picture.
“mee. {sce
Ree
Virmree
AC
This is called a beck diagram, It explains tre funchonal subcomponents of |
the hearing aid iia terms of hew jt converts sounds cant hear info sounds
| Csemetimes) can.
Youll notice seme obvious Vatags ave missing.
wo button, i: That's because this diagram
a battens re
© : oe ines 2 is all about intormapen
K pai S ie
Farrery! inTeReAces! iiuare aeSound start3 out as ad aaaleg signal, S kinda like how
Gnd needs te be Converted tea digital one digital picture...
Se that Computers can understand it.
| 2 ote oi
Unsueg singly, the component at dacs ti
is called the..
sare
made of
pixels.
eeccoee-
pe) Crus |e"
Apc |e
(ky | have input? ad
Ka ge get it digiticee! Chek ist hing) saa
| gr ee) rae [wchrand it has body gvards| and it has [wchrand it has body gvards|
| Grecognize these guys from the block diagram on p-3?)
|
A
The fist problem is That the
mic is so tiny, its signal is
5° soft and hard +e hear,
$0 we Send them Thrmigh a preanp boost them op.
[Then the ton bodyguard
blocks check for..
Are your Frequencies
ceo low or +20 high?
att
AUSSo now we have this.
Mic aS
PREAMP
er 4
ww eh || DIGITAL
ea, || SIGNAL
PROCESSOR'
The DSP, of Di
ignal Processor, is the brains of the aid
H's basically a really, really small computer.
going se the
whele place
rocks out
together.
Not all hear
In fact, they all used 4o be
@ ot
G )
The SleCk Nv
keeps a beat af
Z
es)
aids are digital’.
Analog- ne computer involved, just
of small electrical components.
And the
memory
stores
mings
Wee Pre
progeam,
settings.
lookup
tables...
any thi
the system
Needs to
remember.
The gore is where
all the mixi
Processing happens.
and
There are some
real arguments
for analog aids.
And arguably,
digital aids
benefit fom Hae
~ placebo effect.
|But there are still a lot of advantages
My First hearing aids
te digital hearing aids over analeg ones. ae were analog.
9 over analog ones. Keay)
DIGITAL Roe ye 8 of hy
— Fs
~ SMALLER!
~ LOWER power!
= Connect To Your
PHONE, commurer, Ete!
Sea
You didn't just
disassemble then,
ae ee
+ SIGNAL SHAPING! a Gar) OL
~ ADAPTABLE serTIngs | ao Get) x Ase
~ REPROPUCIBLE AND * “ &
COMPARABLE! fe
+ you “programmed”
Because of this, most modern- day hearing aids
them with one, by
are digital, so that's what we'le going “o feus on
twisting small
potentiometers
on the back.
Back in Club DSP, the place where eveythi
Our DS, the Core (processor), has a lot oF
ing happens...
Bs
We got te keep up) <— No matter what,
5 Pup
she's Ahinkiag
about this.
< FREQUENCY RESPONSE >
CEELeee
how low/high a sound cat yeu
near, and hew smooth is it
4 in between?
D Signal level D Tw
INGE mieasear et ia nomunene, S313
Y Géests at least gous) B DEBUG HY — 0 accvsstely ae
eie> of
can you hear loud and the frequencies
a aae and +l the the “reat” sound map
difference between them? to the frequencies
of the “precesnea” enc?
aeeaat PROCESSING DELAY
INTERNAL NOISE > how long does it take fr sound
Tena’ aga
Comini mize, please) Ao get through the syst
mosHy +nis comes Cs it enough > 19-3em3 ~ that
from he microphone; on can notice it slipping? )
adding signal that PRT
wasn “really there” OO NERSevensthing’s a trad
microphones
Similarly, the core might be trying te do
Nvise reduction... but reducing noite
is a (nonlinear) signal distortion.
Lets tale about noise reduction fer a bit, First, definitions.
NOISE+ meaas backgrovad noise, things SIGNAL! is the part we do want
we dont want te hear. Fer the te hear, Im this case, ve
Purposes of this section, it vsvally anume is speech,
means nef speech.
“There are some obviovs Ii mitatens
do Hhis approach.
Py
Q:
~Q [\ & a
For these sitvatiens,
neise reduction
doesnt work 50 well,
For instance, vnusic isn't Also, we might not want
speech, but we de want to near al) the speech
te hear ite all the me,
We care about +ne relative levels of these ~
the signal-te-nsise rate, or CNR,a
To de this, the core needs fo gure out wha’ oe and whats Pile,
cele She sorts this eat
o by channel - findia
perv Be Sam fr a book
Secs
:
ee of frequencies at
Sy BRENT.
DCD fa
Ep ‘agonan) | HIGH
Lew wig] | Sours
a“
Channel channel chawnel chanel channel
1 34 5
Geehnically, the channels can overlaps
ML any given p
in Hime, each ch2nnel
has a certain cate
* SIGNAL
ot that makes iF really herd te draw.)
alla aloo [pom vn nl
Now, we Knew a lob about the sounds of speech. How fast they are. How leud
they should be, and at what frequencies (and thus what channels they should
show up in.)
So if Hae SNR for q specie channel crosses a thetaheld ak some point,
the con's alyerithan dells it us
Ew, noise’
1
Dump it or muting the Velume on that chanel for
a a I @ little while,
— BW Wet There are Still many design decisions fe make,
Such ase
Aap .. How steeply and cagidly S
May + Should it fur the
™~ 1 volume down?
Eno soon te 4 Glarmening on foFE
ages tells BS Sounds aa, hew
wand it does + which means turning dewa
Hoot long does iF
wait? Hod shacply
Neen gradval is the fade?) dacs it bring Yee
if nokices noite, volume back on
how long should agaat
it wait Zhe a's
foo fast, it may This happens constantly
od ett How far dowa does it for every frequent)
sou, iF missed umn the volume? channel,gm)
Yoo
nnn the noise from gait
here in the Rok place.
Thai's why (DIRECTIONAL. MicRoPHoNES! =i
cas
Basicalh, every micephone
has a ditterent cardieid
pattern it listens “more
in seme directions than othes,
People do this too. Vue ae tern? j
oT 50 Peery ,
“mee DS. mee | a A]
[WE we add the cardiotd patterns of muttigle microphones, we can make therm
| Supplement each other or cancel each other out in certain places, and end up
with a virtual SUFERMIC! through He power of Raa
N L/S svt ILS
Soe
Spale "sensi 92
a UA ibe\\y Nabe S.
TVVANS
‘The core can play the role of conductor /director,
using delays and modulation to ‘steer the bea
formed by multiple mics, This is called adaphve
directienatitg,
She can alse make decisions about ditect onalit
:
ee
without user intervention; this is automate Airtctonality,
(A hearing aid might have directionality that is neither, both, or one but not the Her.)
15 THE Pou PATTERN ADTUSTABLE?
en -
y= i's fined Glurn yowr head)| ws adjuttable This sounds grest, but
BET you can turn a YOu tay ell the mic | personally alse wont
¥§ 38 Wrecttonal mie which direction, +o snanval overrides for
Be Ha Na eee Test in evenp thing.
23 El a directional aie the. micophones
[FE Figs Teel an Senet rn
BE gk fig Heel ont ‘onich direc A, jl
ae ae Mh ten hs Any jlto
It can get ever more complex- you can oe multiband directional i where
the cone computes a ditterent directionality pattern tor each Prequencg chamel-
Lets cee thereh a lobof
high-frequency wind noise,
so I sch Haat channel +8
be very narrendy frtuseds
on whatt in font of 05.
hundeeds of inion
oF computations /dedsions
Om) sueeud te fue 4hiy
and divechanality
is ely one smail
Piee of what she
deals with.
(eo
BOE (Just another
ya wirt |
@ Pewee
The core is abe ting 2 lvanited 7S to
radera computers
20K Phonak.
heading cid chinse yee
No wid- Freqvenoy
noise? Aweseme-
Perfect...
The core iy makin
at
ew
Evy what's Ahad Loo tvging
on the left? Cut the bass
From there and patch it
over from the right.
fer another
Few milli seconds.
urt0 Bc Bott Intel we
TAT chigae b euNow that we have a signal with (we hope) some noise removed, the next thing
we should do is make it louder... and that should be easy, right ?
‘whoa whoa whoa!
Tom up
ALL THE
Mel, remember your
NOLUME!!
audiogram on p.1?
Reed to make
fheze louder =
Cdoesn everyone talk to Heir hearing ald's
core processor as a cartoon character?)
Then that’s easy- if it's about the gap
between wy hearing and “normal” heaving,
all we have to de is make that Frequency
louder by that decibel (48) amount.
People with hearing lors sil lave the same threshold for lovdness pain tolerance as
Perple with “normal” hearing. IF we simply ven up the volume, any sound that's
more thon “soft becemes unbearably, loud. high notes
© x aay
IN That's why i'm looking at
\ cochlear ienplonts, by the wat»let's assume we've done
lail the signal processing
that would be use fal
MEMog Y_
ig . [=I where does the
signal go fom
=a here?
III
: hey, #5
wall i¥'s ghW digifatn andieg ‘again!
Pot it needs te beceme 6 FRRO
Fo the changing room!
but why is it well, when a
all... strange? al the shi
©
(hee)
SR corners and edges, especially
ee for lower- fesludion samplings
turn into jitter.
(low-pass) f
| signal gets converted to an an ones
edges and eseaess of the Sleps 2h Bin
FILTER
and we let the Waiter
(not exactly the same ii
3
its @ cousin 40 the fest)
Once The livailer decides its not too bud for the recchver]
Cor makes it se), the Signal hits the reeriverssRECEIVER AR |
psp: Eed—-@
NZI RECONSTRUCTIVE
SSNS Filter,
e
wnect to other
i