University of Iowa

Iowa Research Online

Theses and Dissertations

Summer 2017

Design and laboratory evaluation of an inexpensive noise sensor

Laura Ann Hallett
University of Iowa

Follow this and additional works at: https://ir.uiowa.edu/etd

Part of the Occupational Health and Industrial Hygiene Commons

Copyright © 2017 Laura Ann Hallett

This thesis is available at Iowa Research Online: https://ir.uiowa.edu/etd/5768

Recommended Citation
Hallett, Laura Ann. "Design and laboratory evaluation of an inexpensive noise sensor." MS (Master of
Science) thesis, University of Iowa, 2017.
https://doi.org/10.17077/etd.qun4tq56

Follow this and additional works at: https://ir.uiowa.edu/etd

Part of the Occupational Health and Industrial Hygiene Commons
  
 
 

 
 
 
 
DESIGN  AND  LABORATORY  EVALUATION    
OF  AN  INEXPENSIVE  NOISE  SENSOR    
 
 
 
 
 
 
 
by  
 
Laura  Ann  Hallett  
 
 
 
 
 
 
 
 
A  thesis  submitted  in  partial  fulfillment  
of  the  requirements  for  the  Master  of  Science    
degree  in  Occupational  and  Environmental  Health  in  the    
Graduate  College  of  
The  University  of  Iowa  
 
August  2017  
 
Thesis  Supervisor:        Professor  Thomas  M.  Peters  
   
 
   

 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Copyright  by  
 
Laura  Ann  Hallett  
 
2017  
 
All  Rights  Reserved  
 
   

 
 
 Graduate  College  
The  University  of  Iowa  
Iowa  City,  Iowa  
 
 
 
 
 
CERTIFICATE  OF  APPROVAL  
 
____________________________  
 
 
MASTER'S  THESIS  
 
_________________  
 
This  is  to  certify  that  the  Master's  thesis  of  
 
 
Laura  Ann  Hallett  
 
has  been  approved  by  the  Examining  Committee  for    
the  thesis  requirement  for  the  Master  of  Science  degree  
in  Occupational  and  Environmental  Health  at  the  August  2017  graduation.  
 
 
Thesis  Committee:   ____________________________________________  
  Thomas  M.  Peters,  Thesis  Supervisor  
 
 
  ____________________________________________  
  T.  Renée  Anthony  
 
 
  ____________________________________________  
  Geb  W.  Thomas  
 
 
 
   

 
 
 ACKNOWLEDGEMENTS  

I  would  first  like  to  express  a  sincerest  thank  you  to  my  advisor,  Dr.  Tom  Peters.  Thank  

you  for  helping  me  realize  my  academic  capabilities  and  encouraging  me  to  keep  going  through  

the  cloud  of  uncertainty.  I  would  have  been  unable  to  accomplish  any  of  this  without  your  

support.  I  would  also  like  to  thank  my  committee  members,  Renée  Anthony  and  Geb  Thomas.  I  

am  appreciative  of  all  the  guidance  that  you  have  both  shown  me  throughout  my  time  here  and  

I  feel  privileged  to  have  the  opportunity  to  work  with  you.    

To  my  low-­‐cost  project  partners:  Sinan,  Alyson,  Marcus,  Mitch,  Xiaoxing,  Kirsten,  Chris,  

and  Nima.  It  has  been  honor  to  work  on  this  project  with  you  the  past  2  years.  I  enjoyed  learning  

from  you  all  and  seeing  the  project  progress  to  where  it  is  today.  Marcus,  this  project  would  not  

have  been  a  success  without  you.  Thank  you  for  working  so  persistently  and  for  always  

answering  my  questions,  even  when  they  required  last  minute  action  on  your  part.  Sinan,  I  am  

so  grateful  for  all  your  help  with  experiments  and  data  analysis.  I  truly  enjoyed  working  

together,  even  on  weld  fume  generation  days.    

Thank  you  to  the  faculty  and  staff  in  the  OEH  department  that  made  my  experience  at  

Iowa  better  than  I  could  have  ever  imagined.  Thank  you  to  my  lab  mates  and  classmates  for  the  

consistent  feedback  and  friendship.  I  am  fortunate  to  have  built  lasting  relationships  with  so  

many  of  you  and  hope  that  we  can  continue  to  encourage  one  another  throughout  our  careers.      

I  would  like  to  thank  my  family,  friends,  and  derby  team  for  all  of  their  love  and  support.  

You  have  all  provided  me  much  needed  balance  through  this  process,  and  I  am  lucky  to  have  

every  one  of  you  in  my  life.  

Finally,  this  research  was  generously  supported  by  The  Heartland  Center  for  

Occupational  Safety  and  Health  Training  Grant  No.  T42OH008491  and  funded  by  the  National  

Institute  for  Safety  and  Health  grant  R01OH010533.    

ii  
 
 ABSTRACT  

Noise  is  a  pervasive  workplace  hazard  that  varies  spatially  and  temporally.  Hazard  

mapping  is  a  useful  way  to  communicate  intensity  and  distribution  of  noise  sources  in  the  

workplace.  These  maps  can  be  created  using  a  stationary  network  of  sensors,  although  the  cost  

of  noise  measurement  instruments  has  prohibited  their  use  in  such  a  network.  The  objectives  

for  this  work  were  to  (1)  develop  an  inexpensive  noise  sensor  (<$100)  that  measures  A-­‐weighted  

sound  pressure  levels  within  ±2  dBA  of  a  Type  2  sound  level  meter  (SLM,  ~$1,800);  and  (2)  

evaluate  50  noise  sensors  before  field  deployment  as  part  of  an  inexpensive  sensor  network.  

The  inexpensive  noise  sensor  consists  of  an  electret  condenser  microphone,  an  amplifier  circuit,  

and  a  microcontroller  with  a  small  form  factor  (28mm  by  47  mm  by  9  mm)  than  can  be  operated  

as  a  stand-­‐alone  unit.  Laboratory  tests  were  conducted  to  evaluate  50  of  the  new  sensors  at  5  

test  levels.  The  testing  levels  were  (1)  ambient  sound  in  a  quiet  office,  (2)  a  pink  noise  test  signal  

from  65  to  85  dBA  in  10  dBA  increments,  and,  (3)  94  dBA  using  a  SLM  calibrator.  The  difference  

between  the  output  of  the  sensor  and  SLM  were  computed  for  each  level  and  overall.    Ninety-­‐

four  percent  of  the  noise  sensors  (n=46)  were  within  ±  2  dBA  of  the  SLM  for  sound  levels  from  

65  dBA  to  94  dBA.  As  sound  level  increased,  bias  decreased,  ranging  from  18.3%  in  the  quiet  

office  to  0.48%  at  94  dBA.  Overall  bias  of  the  sensors  was  0.83%  across  the  75  dBA  to  94  dBA  

range.  These  sensors  are  available  for  a  variety  of  uses  and  can  be  customized  for  many  

applications,  including  incorporation  into  a  stationary  sensor  network  for  continuously  

monitoring  sound  in  manufacturing  environments.    

iii  
 
 PUBLIC  ABSTRACT  

Exposure  to  noise,  particularly  in  occupational  environments,  can  result  in  irreversible  

hearing  loss.  Traditionally,  occupational  noise  exposures  are  measured  on  a  worker  using  

devices  that  cost  over  $1,500.  A  stationary  network  of  these  devices  could  be  used  to  

continuously  monitor  and  visualize  how  noise  varies  over  space  and  time  in  the  workplace,  

helping  to  comprehensively  assess  the  hazard  and  prioritize  actions  to  reduce  noise  sources.  

However,  the  cost  of  these  devices  limits  the  feasibility  of  their  use.  Availability  of  an  

inexpensive,  but  accurate,  noise  sensor  would  alleviate  some  of  the  cost  concerns.  

  There  were  two  objectives  to  this  study.  The  first  was  to  develop  an  inexpensive  noise  

sensor  that  can  detect  noise  levels  within  ±2  A-­‐weighted  decibels  (dBA)  of  the  more  expensive  

reference  device.  Second,  fifty  inexpensive  noise  sensors  were  evaluated  prior  to  deployment  as  

part  of  a  stationary  monitor  network  in  a  heavy-­‐vehicle  manufacturing  facility.    

  The  inexpensive  noise  sensor  (<$100)  performed  similarly  to  the  reference  device  

(~$1,800)  across  a  75  to  94  dBA  range.  These  noise  levels  are  typical  of  what  is  expected  in  the  

heavy-­‐vehicle  manufacturing  facility  where  they  will  be  deployed.    The  versatility  of  these  noise  

sensors  coupled  with  their  accuracy  is  promising  for  future  applications  in  different  occupational  

environments.  

   

iv  
 
 TABLE  OF  CONTENTS  

List  of  Tables    ..................................................................................................................................  vi  

List  of  Figures    ................................................................................................................................  vii    
Chapter  I:  Literature  Review  ...........................................................................................................  1  
Overview  .....................................................................................................................................  1  
Noise  ...........................................................................................................................................  2  
Health  Effects  of  Noise  ...............................................................................................................  3  
Regulations  .................................................................................................................................  4  
Hearing  Conservation  Programs  .................................................................................................  5  
Evaluating  Noise  Levels  ...............................................................................................................  6  
Hazard  Maps  ...............................................................................................................................  7  
Low-­‐Cost  Noise  Measurement  .................................................................................................  10  
Gaps  in  the  Literature  ...............................................................................................................  12  
Objectives  of  Thesis  Work  ........................................................................................................  13  
Chapter  II:  Design  and  Laboratory  Evaluation  of  an  Inexpensive  Noise  Sensor  ...........................  14  
Introduction  ..............................................................................................................................  14  
Methods  ...................................................................................................................................  16  
Sensor  Design  ........................................................................................................................  16  
Laboratory  Evaluation  ..........................................................................................................  18  
Data  Analysis  ........................................................................................................................  19  
Results  ......................................................................................................................................  20  
Discussion  .................................................................................................................................  22  
Conclusions  ...............................................................................................................................  26  
Figures  and  Tables  ....................................................................................................................  27  
Chapter  III:  Conclusions  ................................................................................................................  32  
References  ....................................................................................................................................  35  
Appendix  A  ...................................................................................................................................  35  
Appendix  B  ....................................................................................................................................  45  
 

   

v  
 
 LIST  OF  TABLES  

Table  1.  Mean  sound  pressure  levels,  standard  deviation,  and  coefficient  of  variation  at  five  
target  dBA  levels  (side  by  side,  65,  75,  85,  94  dBA).  P-­‐values  are  resultant  from  one-­‐sided  
t-­‐test  with  an  alpha  of  0.05.  Acceptance  criteria  (AC)  is  defined  as  monitor  output  within  
±2  dBA  of  reference  output,  n=46  monitors.  ...............................................................................  30  
 

Table  2.  Slope  and  intercept,  correlation,  bias  and  coefficient  of  variation  results  for  lab  
checkout  of  noise  sensor  from  65  dBA  to  94  dBA  and  from  75  dBA  to  94  dBA,  n=  46  monitors.   30  
 
 
 
 
   

vi  
 
 LIST  OF  FIGURES  

Figure  1.  The  inexpensive  sound  sensor.  A  simplified  circuit  diagram  (top  panel)  shows  that  
the  electrical  signal  from  the  microphone  is  amplified  twice  before  reaching  the  sensor  
microcontroller  (Teensy).    The  actual  components  of  noise  sensor  incorporated  into  multi-­‐
hazard  monitor  are  shown  in  the  bottom  panel.  An  electret  microphone  extends  from  the  
exterior  of  the  grey  enclosure  and  the  sensor  microcontroller  is  shown  inside  the  enclosure.  
On  the  red  circuit  board,  a  monitor  microcontroller  communicates  with  the  noise  and  other  
hazard  sensors  and  with  a  database  via  WiFi.  ..............................................................................  27  
 

Figure  2.  Setup  for  laboratory  validation  of  noise  sensor.  NTI  XL2  and  noise  sensor  
microphones  are  located  within  2.5  cm  of  one  another,  centered  30  cm  from  the  center  of  
the  amplifier.  The  amplifier  is  connected  to  the  laptop  by  auxiliary  cable.  .................................  28  
 

Figure  3.  Mean  A-­‐weighted  sound  levels  from  noise  sensor  versus  reference  SLM  at  five  
target  sound  intensities.  Points  represent  mean  reading  of  all  noise  sensor  and  reference  
measurement  pairs.  Error  bars  represent  the  standard  deviation  of  sound  pressure  levels  at  
each  target  sound  intensity,  n=46  sensors.  ..................................................................................  29  
 

Figure  4.  Box  plot  of  distributions  of  differences  in  sound  sensor  output  from  the  reference  
SLM  at  five  target  sound  levels.  Error  bars  represent  the  distribution  of  sound  pressure  level  
differences  at  each  target  sound  level,  n=46  sensors.  .................................................................  31  
 
   

vii  
 
 Chapter  I:  Literature  Review  

Overview  

  Noise  is  a  prevalent  occupational  hazard,  with  the  potential  to  irreversibly  damage  

worker  hearing  if  not  adequately  controlled.  One  way  to  determine  sources  of  noise  in  the  

workplace  is  through  the  development  of  a  hazard  map.  Hazard  maps  can  be  generated  using  a  

handheld  sound  measurement  device,  a  roving  cart  of  direct  reading  instruments,  or  a  network  

of  stationary  monitors.  Data  collection  with  a  roving  cart  is  subject  to  temporal  uncertainties,  

making  a  network  of  stationary  monitors  attractive.  However,  direct-­‐read  instruments,  like  a  

sound  level  meter  (SLM),  can  be  expensive  (~$1,800),  which  severely  limits  opportunities  to  

collect  simultaneous  information  from  a  number  of  fixed  locations.    

Recent  innovations  in  smartphone  technology  have  led  to  the  development  of  a  number  

of  SLM  applications  (apps).  Under  certain  conditions,  some  of  these  apps  have  shown  accuracy  

within  ±2  decibels  of  a  reference  SLM  (Kardous  2016).  However,  the  microphones  incorporated  

into  smartphones  have  been  determined  to  have  substantial  amounts  of  variability  in  

performance,  dependent  on  the  software  version  and  phone  hardware.  External  microphones  

can  be  connected  to  smartphones  to  increase  accuracy  and  precision  of  the  noise  readings,  but  

this  fix  does  not  work  for  all  phone  types  (Roberts  2016).  Additionally,  concerns  with  data  

privacy,  accuracy  over  time,  phone  battery  life,  and  data  export  and  storage  limit  the  versatility  

of  these  applications.  

  This  thesis  describes  the  development  of  an  inexpensive  noise  sensor  that  is  

independent  of  a  smartphone.  The  noise  sensor  was  designed  to  measure  A-­‐weighted  sound  

pressure  levels  within  ±2  dBA  of  a  reference  SLM  with  component  costs  below  $100.  Then,  50  

noise  sensors  were  evaluated  for  accuracy  and  precision  over  a  range  from  65  to  94  dBA,  sound  

1  
 
levels  consistent  with  their  intended  use  as  part  of  a  stationary  sensor  network  in  heavy-­‐vehicle  

manufacturing  facility.  

Noise  

Workplace  noise  is  unwanted  sound  produced  in  occupational  environments  (NIOSH,  1998).  

Noise  is  one  of  the  most  common  workplace  hazards,  prevalent  in  most  occupational  

environments  (NIOSH,  1998).    The  effects  of  noise  on  human  health  depend  on  the  amplitude,  

frequency,  and  duration  of  the  noise  exposure  (NIOSH,  1998).  

Sound  pressure  levels  are  used  to  describe  the  amplitude  of  pressure  change  associated  

with  a  specific  sound,  expressed  in  decibels  (dB).  Decibels  are  a  conversion  of  this  pressure  

change  relative  to  the  hearing  threshold  pressure  and  are  measured  on  a  logarithmic  scale  

(OSHA,  2013).  The  amplitude  of  a  sound  directly  relates  to  the  listeners  perception  of  loudness  

(NIOSH,  1998).  Frequency  of  a  sound  relates  to  the  pitch  perceived  by  the  listener,  expressed  in  

hertz  (Hz).  Humans  with  normal  hearing  are  able  to  discern  a  frequency  range  of  20  Hz  to  20  

kilohertz  (kHz).  Human  speech  is  typically  generated  at  frequencies  in  the  500  Hz  to  4,000  Hz  

range  (OSHA,  2013).    

Noise  rarely  occurs  as  a  pure  tone,  but  rather  consists  of  a  number  of  different  frequencies  

occurring  at  different  sound  pressure  levels.  An  octave  band  describes  a  range  of  frequencies  in  

which  the  highest  frequency  is  twice  the  lowest  frequency.  Particular  octave  bands  contain  

frequencies  that  are  more  harmful  to  human  hearing  than  others.  For  this  reason,  weighting  can  

be  applied  to  sound  pressure  levels  to  understand  the  influence  each  frequency  has  on  the  

perception  of  loudness.  There  are  three  different  types  of  weighting,  A,  C  and  Z.  A-­‐weighting  

attenuates  low  frequencies,  so  that  the  sound  pressure  level  output  is  indicative  of  that  

perceived  by  a  human  (OSHA,  2013).  When  a  sound  pressure  level  has  been  modified  using  a  

2  
 
particular  weighting  network,  the  units  are  written  as  dBA  or  dBC,  the  last  letter  indicating  the  

type  of  weighting  in  use.  

Duration  of  noise  exposure  is  described  as  either  continuous  or  impulse.  Continuous  noise  is  

characterized  by  small  fluctuations  in  the  overall  noise  level  over  the  observation  period.  

Impulse,  sometimes  called  impulsive,  noise  occurs  for  less  than  1  second  and  results  from  a  

sharp  rise  and  rapid  decay  in  sound  levels  (NIOSH,  1998).  The  effects  of  impulse  noise  on  

hearing  are  dependent  on  the  repetition  rate,  duration  of  exposure,  and  personal  characteristics  

that  may  enhance  susceptibility.  Impulse  noise  levels  greater  than  119  to  125  dB  are  considered  

critical,  and  hearing  loss  is  more  prevalent  with  exposure  to  these  sound  levels  (Dunn  1991).  In  

many  industrial  environments,  there  is  a  combination  of  both  continuous  and  impulsive  noise  

occurring  at  the  same  time  (NIOSH  1998).  

Health  Effects  of  Noise  

Hearing  loss  can  develop  when  there  is  a  disruption  in  the  ability  of  the  ear  to  convert  

physical  vibration  into  a  nerve  impulse  (Alberti  1992;  WHO  2001).  In  the  United  States,  an  

estimated  one  in  four  adults  has  evidence  of  noise-­‐induced  hearing  loss,  as  indicated  by  a  notch  

on  audiometric  test  results.  Additionally,  one  third  of  workers  who  identified  as  working  in  a  

noisy  environment  had  evidence  of  noise-­‐induced  hearing  loss  (Carrol  2017).  The  three  main  

types  of  hearing  loss  resulting  from  occupational  noise  exposures  are  conductive,  sensorineural,  

and  mixed  hearing  loss  (OSHA,  2013).  Conductive  hearing  loss  occurs  when  sound  waves  are  

unable  to  pass  through  the  outer  or  middle  ear.  Sensorineural  hearing  loss  occurs  with  damage  

to  the  physiological  processes  in  the  inner  ear.  Mixed  hearing  loss  occurs  when  there  is  a  

combination  of  both  conductive  and  sensorineural  hearing  loss  (CDC,  2015).    

3  
 
 Over  22  million  workers  in  the  United  States  experience  occupational  exposures  to  

hazardous  noise  each  year  (Tak,  Davis,  &  Calvert,  2009).  In  2014,  the  incidence  of  hearing  loss  in  

all  industries  was  1.9  events  per  10,000  full-­‐time  workers,  with  20,900  reported  cases  of  hearing  

loss  (BLS,  2016).  Industries  that  have  higher  proportions  of  worker  exposures  to  hazardous  noise  

levels  include  mining,  lumber  and  wood  product  manufacturing;  rubber,  plastics,  and  leather  

production;  utilities,  and  repair  and  maintenance.  While  a  lower  proportion  of  workers  in  the  

manufacturing  industry  are  exposed  to  hazardous  noise  levels,  the  greatest  number  of  workers  

exposed  to  hazardous  noise  levels  work  in  manufacturing,  estimated  at  5.7  million  workers  at  

risk  in  2009  (Tak  et  al.,  2009).  

Regulations  

A  number  of  occupational  exposure  limits  (OELs)  have  been  developed  to  protect  worker  

hearing.  The  Occupational  Safety  and  Health  Administration  (OSHA)  sets  legally  enforceable  

noise  exposure  limits.  OSHA  requires  that  employers  implement  a  hearing  conservation  program  

when  employees  are  exposed  to  sound  levels  greater  than  85  decibels  (dB),  A-­‐weighted  (A),  as  

an  8-­‐hour  time  weighted  average  (TWA).  An  exchange  rate  of  5-­‐dB  is  used  for  OSHA  dose  

calculations.  OSHA  also  specifies  a  Permissible  Exposure  Limit  (PEL)  not  to  be  exceeded  of  90  

dBA  TWA  (OSHA,  1974).  If  worker  noise  exposures  exceed  90  dBA  TWA,  employers  must  provide  

hearing  protection  devices  to  attenuate  noise  levels  to  at  least  the  PEL.    

The  National  Institute  of  Occupational  Safety  and  Health’s  (NIOSH)  Recommended  Exposure  

Limit  (REL)  for  noise  is  85  dBA  TWA  over  an  eight-­‐hour  work  shift.  Additionally,  NIOSH  

recommends  that  no  noise  exposure,  continuous  or  impulse,  exceed  140  dBA.  The  NIOSH  REL  is  

based  upon  a  3-­‐dB  exchange  rate,  which  represents  how  sound  intensity  doubles,  compared  to  

the  less  protective  5-­‐dB  used  by  OSHA  (NIOSH,  1998;  OSHA,  1974).  

4  
 
 Hearing  Conservation  Programs  

The  most  effective  way  to  prevent  hearing  loss  is  to  directly  eliminate  the  noise  hazard  (Tak  

et  al.,  2009).  Elimination  is  not  always  the  most  feasible  solution,  so  other  programs  can  be  

implemented  to  protect  workers  from  noise  hazards.  Both  OSHA  and  NIOSH  recommend  

hearing  conservation  programs  when  workers  are  exposed  to  sound  levels  louder  than  85  dBA  

TWA  (NIOSH,  1998;  OSHA,  1974).  Exposure  monitoring  in  these  environments  must  include  all  

continuous,  intermittent,  and  impulsive  noise  within  an  80  to  130  dB  range  and  be  collected  

during  a  typical  working  day  (OSHA,  1974).  OSHA  1910.95(c)(1)  and  1910.95  Appendix  A  specify  

the  requirements  for  collection  of  noise  dose  information.  There  is  no  required  instrument  type,  

but  audio  dosimeters  are  most  commonly  used  to  collect  information  on  worker  dose.    

Another  important  component  of  a  hearing  conservation  program  is  the  establishment  of  an  

audiometric  testing  program.  Clinically,  hearing  loss  presents  itself  as  a  reduction  in  the  ability  of  

a  person  to  recognize  high  frequency  tones  between  4000  to  6000Hz  on  pure  tone  audiometry  

testing  (Kurmis  &  Apps,  2007).  In  order  to  discern  a  loss  in  hearing  ability,  employers  must  

conduct  baseline  audiometric  exams  within  6  months  of  exposure  to  sound  levels  at  or  above  

the  action  limit.  Then,  annual  audiometric  testing  is  performed  to  monitor  hearing  health  and  

ensure  that  the  program  is  effective  in  preventing  hearing  loss  (OSHA,  2013).  Employees  that  

record  an  average  shift  in  either  ear  of  10  decibels  or  more  at  2000,  3000,  and  4000Hz  have  

experienced  an  OSHA-­‐defined  standard  threshold  shift  (STS).  Per  OSHA  CFR  1904.10(a),  if  an  

employee’s  audiogram  indicates  that  they  have  an  STS  in  one  or  both  ears,  and  the  total  hearing  

level  is  25  dB  or  more  above  audiometric  zero  in  the  same  ear  as  the  STS,  the  hearing  loss  is  a  

reportable  illness.  State  worker’s  compensation  for  hearing  loss  varies  in  frequency  and  

magnitude  of  the  audiometric  shift  measured.  

5  
 
 Evaluating  Noise  Levels  

There  are  a  number  of  instruments  suited  for  noise  exposure  measurements.  Most  

commonly,  a  sound  level  meter  or  a  dosimeter  is  used  to  determine  worker  noise  exposures.  

Sound  level  meters  (SLM)  measure  sound  pressure  levels  in  units  of  dB.  A  SLM  consists,  at  a  

minimum,  of  three  components:  microphone,  a  frequency  selective  amplifier,  and  an  indicator  

(NIOSH,  1998).  The  American  National  Standards  Institute  specifies  standards  for  performance  

of  SLMs  in  ANSI  S1.4  and  indicates  three  different  grades  of  instrument:  Type  0,  1,  and  2.  Type  0  

SLMs  are  used  as  laboratory  reference  standards.  Type  1  instruments  are  designed  for  field  or  

laboratory  use  and  are  allotted  an  overall  error  of  +/-­‐  1.5  dB  from  the  Type  0  instrument.  Type  2  

instruments,  or  general-­‐purpose,  are  designed  for  field  use  and  allowed  an  overall  error  of  +/-­‐  

2.3  dB.  General  field  measurements  are  assumed  to  be  collected  in  environments  in  which  high  

frequency  sounds  do  not  dominate  (ANSI,  1983).  OSHA  regulations  require  the  use  of  at  least  a  

Type  2  SLM  for  noise  exposure  measurements.  If  a  SLM  is  used  to  determine  personal  noise  

exposure,  time-­‐weighted  averages  must  be  computed  using  time  and  motion  information  on  

workers  to  ensure  that  the  exposure  estimate  accurately  reflects  worker  experience  of  noise.    

Dosimeters  are  the  primary  instrument  used  for  noise  exposure  compliance  measurements  

(OSHA,  2013).  Noise  dosimeters  measure  sound  pressure  levels  similarly  to  the  SLM,  then  record  

and  process  this  information  over  the  sampling  period  to  compute  the  noise  dose  and  TWA  

experienced  by  the  wearer  (NIOSH,  1998).  Different  threshold  and  exchange  rates  can  be  

programmed  into  a  dosimeter  to  measure  noise  dose  based  upon  either  OSHA  or  NIOSH  criteria.  

By  attaching  the  microphone  directly  to  the  worker,  the  calculated  dose  is  representative  of  the  

noise  experienced  by  the  worker  as  they  move  throughout  their  work  environment.  

6  
 
 The  work  environment  can  inform  the  type  of  sampling  needed.  In  a  workplace  with  

homogenous  noise  levels,  area  samples  can  be  collected  using  a  sound  level  meter,  and  worker  

exposures  estimated  by  the  percentage  of  their  workday  they  spend  in  each  area.  For  workers  

that  have  more  variety  in  tasks  or  are  exposed  to  a  more  complex  environment,  dosimeters  are  

a  better  tool  to  measure  their  exposure  dose,  as  they  are  worn  by  the  worker  at  all  times  (OSHA,  

2013).  

Hazard  Maps  

  A  method  growing  in  popularity  is  the  use  of  hazard  maps  to  display  hazard  information  

as  it  varies  over  space  and  time.  Hazard  maps  provide  an  easily  understandable,  graphic  

representation  of  hazard  location  and  intensity  in  the  workplace,  aiding  in  communication  of  

workplace  risks  to  management  and  workers.  Noise  mapping  can  aid  in  the  understanding  of  

noise  levels  in  context  of  sources.  Within  the  OSHA  technical  manual,  there  are  guidelines  for  

the  creation  of  noise  maps  and  the  designation  of  the  “hazard  radius”  (OSHA  2013).  

Previous  studies  have  generated  hazard  maps  using  a  mobile  cart  equipped  with  a  

number  of  direct-­‐read  instruments  (Liu  &  Hammond,  2010;  Peters  et  al.,  2012).  These  

instruments  have  the  ability  to  quickly  provide  information  on  contaminant  levels.  However,  

portable  direct-­‐read  instruments  have  the  potential  to  suffer  from  a  number  of  measurement  

errors,  including  poor  accuracy,  lack  of  precision,  and  bias  from  interferences  (Kirsten  A  Koehler  

&  Volckens,  2011).  Additionally,  generating  hazard  maps  with  this  process  requires  considerable  

time  and  personnel  support.  Collecting  measurements  point-­‐by-­‐point  also  reduces  the  ability  to  

make  temporal  inferences  on  hazard  levels  in  the  mapped  area  (Hakala,  Kivela,  Ihalainen,  

Luomala,  &  Gao,  2010).  

7  
 
 Noise  hazard  maps  have  been  created  to  display  noise  hazards  in  both  environmental  

and  occupational  applications.  Environmental  noise  maps  are  generated  to  identify  and  monitor  

noise  sources  present  in  the  non-­‐occupational  environment.  While  there  is  no  standard  to  

create  environmental  noise  maps  in  The  United  States,  the  European  Noise  Directive  

2002/49/EC  requires  the  development  of  environmental  noise  maps  every  five  years  in  each  of  

the  EU  member  states.  The  goal  of  this  mapping  directive  is  to  provide  communities  with  the  

information  they  need  to  preserve  noise  levels  where  they  are  not  harmful  and  prioritize  

controls  for  areas  that  exceed  applicable  exposure  levels  through  the  creation  of  strategic  noise  

maps  (Directive,  2002/49/EC).  Environmental  noise  maps  are  commonly  created  for  areas  

surrounding  busy  roads  or  airports.    

  Occupational  noise  maps  have  been  generated  for  a  variety  of  different  workplace  

types,  from  rice  mills  in  India  to  engine  testing  facilities  in  Colorado  (Lake,  Zhu,  Wang,  Volckens,  

&  Koehler,  2015;  Prasanna  Kumar,  Dewangan,  Sarkar,  Kumari,  &  Kar,  2008).    A  mobile  individual  

measures  sound  pressure  levels  at  each  point  on  a  grid,  and  then  the  measurements  are  

integrated  in  mapping  software  to  generate  hazard  maps.  This  method  provides  a  good  visual  

interpretation  of  the  distribution  of  hazards  in  the  workplace.  However,  generating  a  hazard  

map  with  this  method  requires  personnel  support  and  time  and  can  be  cost  prohibitive  for  a  

large  workplace.  This  method  also  provides  a  brief  snapshot  of  sound  levels,  without  capturing  

information  on  sound  level  variability  at  a  particular  location.    

The  representativeness  of  a  hazard  map  can  differ  dramatically,  dependent  on  the  path  

that  the  researcher  takes  through  the  facility  and  the  timing  of  the  measurements  (K.  A.  Koehler  

&  Peters,  2013;  Lake  et  al.,  2015).  Hazard  maps  make  communication  of  risks  easier,  but  this  

benefit  is  lost  if  the  data  were  collected  in  a  way  that  fails  to  capture  the  whole  story.  A  study  

conducted  by  Lake  et  al.  (2015)  found  that  noise  hazard  maps  produced  for  workplaces  with  

8  
 
steady  sound  levels  looked  similar,  whether  the  data  were  collected  using  static  or  roving  

monitors.  On  the  contrary,  hazard  maps  produced  for  workplaces  that  had  more  variety  in  

sound  levels  and  intermittent  peak  noises  were  very  different,  dependent  on  the  method  by  

which  the  data  were  collected.    If  hazard  maps  are  to  be  used  to  inform  decisions  on  controls  or  

interventions,  it  is  crucial  that  they  are  representative  of  the  risks.  An  underestimation  of  risk  

may  lead  to  increased  risk  for  the  worker,  while  and  overestimation  may  result  in  additional  

costs  for  the  employer.  

Historically,  noise  hazard  maps  have  been  used  to  identify  areas  of  a  workplace  that  

have  sound  levels  exceeding  85-­‐dB.  Within  these  identified  hazard  zones,  all  workers  are  

required  to  wear  hearing  protection  devices.  While  a  valuable  tool  to  characterize  noise  in  a  

work  environment  with  stable  sound  levels,  collecting  information  based  on  a  short  duration  

average  does  not  provide  adequate  information  on  the  variability  of  sound  levels  over  a  longer  

time  period.  This  method  limits  the  true  understanding  of  worker  noise  exposures.  By  mapping  

sound  levels  over  time,  not  just  as  snapshots,  the  temporal  variability  in  sound  levels  can  be  

identified  as  they  change  over  the  work  shift,  day,  or  month.      

Stationary  networks  of  noise  sensors  have  been  used  to  collect  information  on  

environmental  noise  pollution,  particularly  in  urban  areas.  A  study  conducted  by  Santini,  

Ostermaier,  and  Vitaletti  (2008)  investigated  the  feasibility  of  a  stationary  sensor  network  for  

measurement  of  environmental  sound  levels,  specifically  along  roadways.  The  investigators  

found  that  the  evaluated  noise  sensors  could  capture  information  on  noise  pollution  levels,  but  

they  are  subject  to  errors  based  on  directionality.  A  study  by  Filipponi,  Santini,  and  Vitaletti  

(2008)  investigated  the  suitability  of  wireless  sensor  networks  to  capture  information  on  noise  

pollution  and  established  requirements  for  a  sensor  network  to  comply  with  noise  pollution  

monitoring  applications.  The  investigators  recommend  a  dense  network  of  sensors  deployed  

9  
 
over  a  small  area,  like  the  façade  of  a  building,  to  collect  the  most  accurate  information  on  noise  

pollution  levels.    

Low-­‐Cost  Noise  Measurement  

Many  low-­‐cost  noise  measurement  devices  have  been  developed  and  deployed  for  

environmental  noise  monitoring  purposes.  A  study  conducted  by  Van  Renterghem  et  al.  (2011)  

investigated  the  suitability  of  a  number  of  different  microphones  for  environmental  noise  

monitoring,  ranging  in  price  from  less  than  $1  to  over  $2,000  USD.  The  researchers  found  that  a  

number  of  inexpensive  microphones  placed  in  an  outdoor  setting  were  able  to  detect  sound  

pressure  levels  with  only  small  additional  error  compared  to  the  error  experienced  by  the  much  

more  expensive  reference  microphones.  

Electret  microphones,  like  those  evaluated  in  the  Renterghem  study,  are  frequently  

incorporated  into  phones  because  of  their  low  cost.  The  increased  use  of  smart  phones  in  recent  

years  has  improved  access  to  low-­‐cost  noise  measurement  applications  (apps).  An  estimated  

77%  of  Americans  owned  a  smartphone  in  2016,  an  increase  from  just  35%  in  2011  (Pew  

Research  Center,  2017).  Smart  phone  users  can  download  noise  measurement  apps  on  their  

phones  and  infer  information  on  sound  pressure  levels.  User-­‐friendly  interfaces  make  it  possible  

to  collect  noise  survey  information  in  many  different  locations,  from  the  outdoor  environment  

to  the  workplace  (Maisonneuve,  Stevens,  Niessen,  &  Steels,  2009).    

A  popular  use  for  SLM  apps  is  the  generation  of  crowd-­‐sourced  environmental  noise  maps.  

Nast,  Speer,  and  Le  Prell  (2014),  evaluated  how  accurate  smart  phone  sound  level  meter  apps  

were  for  environmental  noise  monitoring.  The  investigators  tested  the  accuracy  of  SLM  apps  

using  six  octave  bands,  generated  at  three  intensities.  The  criterion  for  a  good  response  from  

the  app  was  ±5  dB  from  the  Type  1  reference  SLM.  The  results  showed  differences  in  accuracy  of  

10  
 
each  app  across  the  different  octave  bands  evaluated.  Murphy  and  King  (2016)  used  broadband  

white  noise  to  measure  SLM  app  response  at  background,  50,  70  and  90  dBA  sound  levels.  Mean  

differences  in  app-­‐reported  sound  levels  from  the  reference  SLM  differed  across  the  range,  with  

the  poorest  performance  at  the  extremes  of  the  range.  A  significant  difference  between  the  

types  of  phone  platform  used,  iOS  or  Android,  and  the  mean  output  at  each  sound  level  was  

found,  indicating  possible  inconsistencies  in  the  microphone  quality  of  each  phone.  The  

investigators  determined  that  these  noise  apps  are  potentially  a  good  tool  for  use  in  crowd  

sourced  environmental  noise  monitoring,  but  should  not  be  used  to  replace  reference  SLMs.    

The  potential  for  use  of  smartphone  apps  in  occupational  settings  has  been  evaluated  a  

number  of  times  by  Kardous  and  Shaw  (2014;  2016).  The  first  study  with  this  team  evaluated  the  

accuracy  of  62  smartphone  apps  for  occupational  noise  measurements.  SLM  apps  were  

evaluated  at  sound  levels  from  65  to  94  dBA,  indicative  of  the  typical  occupational  noise  

exposures  encountered  in  the  workplace.  Apps  were  not  calibrated  prior  to  use.  The  

investigators  identified  two  apps  that  had  unweighted  mean  differences  within  1  dB  of  the  

reference  SLM  across  all  testing  conditions.  The  investigators  also  found  that  smartphones  

performed  differently  based  on  model,  similar  to  those  results  shown  by  Murphy  and  King.  

Three  of  the  62  tested  apps  were  found  to  report  mean  differences  within  ±2  dBA  of  the  

reference  SLM,  leading  to  the  conclusion  that  these  apps  are  adequate  for  noise  assessments  in  

some  occupational  settings.    

A  more  recent  study  by  Roberts,  Kardous,  and  Neitzel  (2016),  further  investigated  the  

accuracy  of  smart  phone  apps,  this  time  incorporating  an  external  microphone  and  calibration  

procedure.  The  apps  were  downloaded  onto  5th  generation  Apple  iPods,  devices  virtually  

identical  to  the  iPhone  save  the  ability  to  communicate  with  a  cellular  network.  The  internal  

smartphone  microphones  reported  values  exceeding  their  acceptance  criterion  of  ±2dB  

11  
 
deviation  from  the  reference  microphone.  They  also  found  an  upper  noise  limit  around  90  dB  on  

a  number  of  the  apps,  limiting  the  potential  to  use  the  app  in  an  occupational  or  loud  

recreational  environment.  While  the  external  microphones  were  found  to  improve  

measurement  accuracy,  additional  instrumentation  comes  with  additional  expense.  

Gaps  in  the  Literature  

The  hazards  associated  with  noise  exposure  are  well  documented,  and  regulations  and  

best-­‐practice  recommendations  exist  to  protect  workers  from  occupational  noise.  A  popular  

way  for  employers  to  communicate  noise  hazards  to  workers  is  through  the  generation  of  

hazard  maps.  The  research  conducted  on  noise  hazard  mapping  has  indicated  that  the  

representative  of  noise  hazard  maps  is  greatly  dependent  on  the  method  of  collecting  sound  

levels,  the  variability  of  sound  levels  in  the  workplace,  and  the  specific  path  taken  through  the  

area  of  interest.  A  method  for  mapping  that  can  account  for  temporal  variations  would  improve  

understanding  of  how  sound  levels  fluctuate  over  time  in  a  workplace.    

The  advantages  that  increased  static  sound  measurements  can  have  on  the  

representativeness  of  noise  maps  for  both  occupational  and  environmental  applications  is  clear.  

In  the  past,  mapping  required  the  movement  of  a  single,  expensive  SLM  through  a  workplace.  

This  method  fails  to  capture  adequate  information  on  the  variability  of  noise  sources  over  time.  

Inexpensive  microphones,  like  those  incorporated  into  smartphones,  are  stable  and  accurate.  

The  advent  of  the  smartphone  and  the  abundance  of  sound  level  meter  apps  available  to  

consumers  has  provided  hope  that  there  are  accurate  and  inexpensive  ways  to  collect  dynamic,  

time-­‐dependent  information  on  sound  levels  over  a  large  work  area  without  purchasing  an  

expensive  reference  SLM.  However,  there  are  limitations  that  arise  from  the  model  of  

smartphone  used,  version  of  app,  and  the  condition  of  the  smartphone.  With  the  constant  

12  
 
development  of  these  applications  and  their  updates,  there  needs  to  be  a  standardized  way  to  

evaluate  app  accuracy.  Additionally,  technology  is  rapidly  changing,  so  it  can  be  challenging  for  

app  developers  to  keep  up  with  phone  operating  system  updates.    

It  would  be  valuable  to  separate  the  noise  sensor  from  the  smartphone  and  optimize  

the  circuitry  to  deploy  in  an  occupational  environment.  By  developing  the  noise  sensor  as  a  

stand-­‐alone  component,  opportunities  open  to  use  the  sensor  technology  for  a  variety  of  

different  purposes,  like  a  stationary  network  of  noise  sensors.  

Objectives  of  Thesis  Work  

  The  objectives  of  this  thesis  were  twofold:  (1)  develop  an  inexpensive  noise  sensor  that  

measures  A-­‐weighted  sound  pressure  levels  within  ±2  dBA  of  a  reference  SLM  with  component  

costs  below  $100;  and  (2)  evaluate  50  noise  sensors  for  precision  and  accuracy  prior  to  their  

deployment  as  part  of  a  stationary  sensor  network.  The  sensor  developed  and  evaluated  in  this  

study  was  optimized  for  a  heavy-­‐vehicle  manufacturing  environment,  where  sound  levels  were  

anticipated  to  be  between  65  to  95  dBA.  

   

13  
 
 Chapter  II:  Design  and  Laboratory  Evaluation  of  an  Inexpensive  Noise  Sensor  

Introduction  

Each  year,  over  22  million  workers  in  the  United  States  experience  occupational  

exposures  to  potentially  hazardous  sound  levels  (Tak  et  al.,  2009).  Occupational  exposure  to  

hazardous  sound  levels  can  result  in  disabling  hearing  loss  (Hong,  Kerr,  Poling,  &  Dhar,  2013).  

Noise-­‐induced  hearing  loss  occurs  when  there  is  disruption  in  the  ability  of  the  ear  to  convert  

physical  vibration  in  the  air  into  a  nerve  impulse  (Alberti,  1992).  Between  2003  and  2012,  the  

prevalence  of  hearing  loss  across  all  industries  approached  13  percent  ({Masterson,  2016  #116}.  

In  order  to  protect  worker  hearing,  the  Occupational  Safety  and  Health  Administration  (OSHA)  

sets  a  legally  enforceable  permissible  exposure  limit  (PEL)  of  90  decibels,  A-­‐weighted  (dBA),  as  

an  8-­‐hour  time  weighted  average  (TWA)  (OSHA,  1974).  Additionally,  OSHA  requires  employers  

to  implement  a  hearing  conservation  program  when  sound  levels  exceed  an  action  level  of  85  

dBA  TWA.    

Traditionally,  occupational  noise  exposure  is  quantified  using  a  dosimeter  or  a  sound  

level  meter  (SLM).  These  instruments  can  cost  upwards  of  $2,000  USD.    Dosimeters  are  affixed  

to  the  worker’s  collar  and  provide  a  percentage  of  the  full  noise  dose  experienced  by  the  

worker.  Sound  level  meters  provide  information  on  area  noise  levels  and  are  primarily  used  for  

screening  purposes  to  determine  where  dosimetry  should  be  performed  or  to  designate  “high  

noise”  areas.  OSHA  requires  that  sound  level  meters  meet  the  American  National  Standards  

Institute  (ANSI)  Standard  S1.4,  “Specifications  for  Sound  Level  Meters,”  in  which  three  different  

types  of  SLM  performances  are  identified.  Type  2  meters  are  most  commonly  used  in  

occupational  environments  and  have  an  accuracy  of  ±2  dBA,  the  minimum  level  of  precision  to  

comply  with  the  OSHA  noise  standard  (OSHA,  2013).    

14  
 
 An  increasingly  common  way  to  visualize  hazards  in  the  workplace  and  determine  

sources  of  noise  is  through  the  development  of  hazard  maps.  Hazard  maps  have  be  generated  

with  a  handheld  direct  reading  instrument,  a  roving  cart  of  direct  reading  instruments,  ,  or  a  

network  of  stationary  sensors  (Lake  et  al.,  2015;  Liu  &  Hammond,  2010;  Peters  et  al.,  2012).  

Uncertainties  can  arise  when  measuring  sound  levels  by  mobile  mapping,  especially  concerning  

the  temporal  distribution  of  hazards.  Systems  that  can  map  sound  both  spatially  and  over  time,  

such  as  a  stationary  sensor  network,  are  more  attractive  (Lake  et  al.,  2015).  However,  the  price  

of  direct  reading  sound  instruments,  like  a  dosimeter  or  SLM,  can  be  cost-­‐prohibitive,  severely  

limiting  the  opportunity  to  collect  simultaneous  information  from  many  locations  necessary  to  

characterize  a  workplace.    

Developments  in  low-­‐cost  technology  present  opportunities  to  use  inexpensive  

microphones,  like  the  electret  condenser  microphones  incorporated  into  cellular  phones,  to  

measure  sound  pressure  levels.  A  number  of  SLM  applications  (apps)  are  available  for  download  

onto  personal  smartphones.  However,  concerns  have  arisen  to  the  applicability  of  these  apps  for  

occupational  environments.  Studies  assessing  the  validity  of  these  apps  have  found  that  the  

accuracy  and  precision  is  dependent  upon  the  phone  model  and  the  software  version  (C.  A.  

Kardous  &  Shaw,  2014;  Murphy  &  King,  2016;  Nast  et  al.,  2014).  The  variability  in  performance  

of  some  apps  may  be  improved  with  the  installation  of  an  external  microphone,  but  this  is  not  a  

viable  solution  for  all  phone  types  (Roberts  et  al.,  2016).  Additionally,  the  battery  life  of  a  

smartphone  is  not  sufficient  to  keep  a  computationally  demanding  app  continuously  running  

over  an  extended  period.  

While  the  advancements  in  smartphone  SLM  apps  are  promising,  there  are  a  number  of  

limitations  to  their  use  in  a  stationary  sensor  network.  Roberts  et  al.  (2016)  found  an  upper  limit  

noise  limit  around  90  dBA  on  a  number  of  the  apps  they  evaluated.  With  the  OSHA  PEL  for  noise  

15  
 
set  at  90  dBA  TWA,  it  is  essential  that  a  sound  sensor  designed  for  use  in  an  occupational  

environment  provide  accurate  sound  level  readings  at  levels  near  and  above  the  PEL.  

Additionally,  the  reliance  of  these  SLM  apps  on  the  smartphone  increases  the  cost.  Isolating  the  

microphone  and  computational  components  of  the  SLM  from  the  phone  would  allow  for  the  use  

of  this  low-­‐cost  technology  in  a  variety  of  different  environments  and  for  a  variety  of  uses,  like  a  

stationary  sensor  network.    

Thus,  the  primary  objective  of  this  work  was  to  design  a  noise  sensor  that  measures  A-­‐

weighted  sound  levels  within  2  decibels  of  a  Type  2  reference  SLM  with  components  costing  less  

than  $100.  The  second  objective  was  to  evaluate  50  of  these  sensors  prior  to  field  deployment  

to  assess  sound  levels  in  a  heavy-­‐vehicle  manufacturing  facility.      

Methods  

Sensor  Design  

We  developed  the  noise  sensor  to  integrate  into  a  multiple  hazard  monitor  that  includes  

sensors  for  sound,  gasses,  aerosols,  temperature,  and  relative  humidity  (Figure  1).  These  

monitors  will  be  incorporated  into  a  sensor  network  to  map  multiple  hazards  in  an  occupational  

setting,  specifically  a  heavy-­‐vehicle  manufacturing  facility.  Preliminary  sound  pressure  levels  in  

this  facility  were  all  greater  than  65  dBA.  Thus,  we  optimized  the  sensor  circuitry  to  measure  60  

dBA  and  greater.  Additional  design  criteria  for  the  sensor  include  compact  design,  independence  

from  a  smartphone,  and  component  cost  less  than  $100.      

The  noise  sensor  consists  of  a  microphone,  amplifier  circuitry,  and  a  sensor  

microcontroller.  A  20  Hz  to  20  kHz  omnidirectional  analog  electret  condenser  microphone  

(CMA-­‐4544PF-­‐W,  CUI  Inc,  Tualatin,  Oregon,  USA)  projects  from  the  exterior  of  the  monitor  

enclosure  (Figure  1).  Analog  voltage  from  the  microphone  passes  through  two  amplifiers  in  

16  
 
series  (4.93x)  and  is  then  acquired  by  a  sensor  microcontroller  (Teensy  3.2,  PJRC,  Sherwood,  

Oregon,  USA).  The  sensor  microcontroller  samples  1024  amplified  voltages  (N)  over  a  0.023-­‐sec  

period  (∆t).  The  acquired  sample  voltages  are  then  processed  with  Fast  Fourier  Transform  (FFT)  

to  a  power  spectrum  with  512  frequency  bins,  ranging  from  0  to  22.05  kHz  to  encompass  the  

range  of  human  hearing.  The  FFT  code  was  developed  using  the  Teensy  Audio  Library.  An  A-­‐

weighting  coefficient  is  then  applied  to  each  bin  of  the  power  spectrum  to  determine  the  A-­‐

weighted  power  spectrum  (XA).  The  A-­‐weighted  decibels  for  the  sampling  period  is  obtained  by  

integrating  the  frequency  spectrum  as  follows  (Lanman  2005):  

!
! !
Signal  Level  in  dBA = 10𝑙𝑜𝑔!" !
!!! 𝑋! 𝑘 + 𝐶                      Equation  1  
!∆!

where  C  is  a  calibration  constant.    

The  calibration  constant  was  determined  using  a  1  kHz  tone  and  side-­‐by-­‐side  

comparison  of  the  sensor  output  with  a  Type  2  SLM  (XL2  Audio  and  Acoustic  Analyzer,  NTi  

Audio,  Tigard,  Oregon,  USA)  and  microphone  (M4260,  NTi  Audio,  Tigard,  Oregon,  USA).  This  SLM  

was  selected  because  it  was  compatible  with  a  previously-­‐developed  data  logging  program.  A  

sample  of  three  noise  sensors  were  evaluated  to  determine  if  the  same  calibration  constant  

could  be  applied  to  all  sensors.  A  calibration  constant  of  127.5  was  applied  to  all  sensors  to  

adjust  the  A-­‐weighted  frequency  spectrum  to  the  signal  level  in  dBA.    

The  noise  sensor  (microphone  and  sensor  microcontroller)  can  be  operated  as  a  stand-­‐

alone  device  with  a  form  factor  of  approximately  28  mm  by  47  mm  by  9  mm.  In  this  work  we  

evaluated  the  noise  sensor  integrated  into  the  multi-­‐hazard  monitor.  Within  the  monitor,  a  

program  on  the  monitor  microcontroller  (Seeeduino  Cloud,  Seeed  Development  Limited,  San  

Leandro,  CA,  USA)  requests  data  from  the  sensor  microcontroller  once  every  2.5  s,  triggering  the  

sensor  microcontroller  to  send  the  most  recently  calculated  A-­‐weighted  decibel  sound  level.  

17  
 
This  2.5-­‐s  data  can  be  accessed  directly  from  the  monitor  by  USB  connection.  For  purposes  of  

the  sensor  network,  the  monitor  microcontroller  collects  the  2.5-­‐s  data  from  the  sensor  

microcontroller,  averages,  and  then  sends  15-­‐minute  averages  of  the  sound  levels  to  a  monitor  

database.    

Laboratory  Evaluation  

The  sound  pressure  levels  measured  with  50  noise  sensors  embedded  within  the  larger  

monitors  were  compared  to  those  measured  with  a  SLM  using  a  pink  noise  test  signal.  Each  

noise  sensor  was  tested  individually  inside  an  isolated  office.  The  microphones  of  the  SLM  and  

noise  sensor  were  placed  within  2.5  cm  of  each  other  and  30  cm  from  the  center  of  an  amplifier  

(2311000000,  Fender  Musical  Instruments  Corporation,  Scottsdale,  AZ,  USA).  The  amplifier  was  

connected  to  a  laptop  computer  with  an  auxiliary  cord.  The  standard  operating  procedure  for  

the  laboratory  evaluation  is  included  in  Appendix  A.  

The  SLM  used  for  laboratory  comparison  was  newly  purchased  at  the  start  of  

experiments  and  calibrated  prior  to  each  testing  day  using  a  114  dB  test  tone  generated  by  a  

sound  level  meter  calibrator  (SCAL1356,  General  Tools  &  Instruments,  Secaucus,  New  Jersey,  

USA).  The  manufacturer  stated  performance  range  for  the  SLM  and  microphone  is  29  dB  to  144  

dB  with  a  resolution  of  0.1  dB.  For  all  experiments,  the  SLM  was  programmed  to  report  the  A-­‐

weighted  equivalent  sound  level  (LAeq).  No  threshold  sound  level  was  designated  in  the  SLM  

settings.    

Five  target  sound  levels,  ambient,  65  dBA,  75  dBA,  85  dBA,  and  94  dBA,  were  generated  

to  test  each  noise  sensor.  These  sound  levels  were  selected  to  match  the  range  used  by  Kardous  

and  Shaw  (2014)  in  their  evaluation  of  smartphone  applications  and  to  meet  the  needs  of  the  

heavy-­‐vehicle  manufacturing  project.  For  levels  from  65  dBA  to  75  dBA,  pink  noise  was  

18  
 
produced  by  playing  a  computer  sound  file  (NTi  Audio  Test  Signals  for  Audio  and  Acoustic  

Analyzers  V1.0).  The  sound  level  was  adjusted  using  the  volume  settings  on  the  laptop  and  then  

verified  on  the  reference  SLM  before  each  testing  period  began.  The  94-­‐dBA  tone  was  

generated  using  a  sound  level  calibrator  (SCAL1356,  General  Tools  &  Instruments,  Secaucus,  

New  Jersey,  USA).  Each  test  tone  was  generated  for  30  s  before  moving  on  to  the  next.  Both  the  

noise  sensor  and  SLM  reported  one  sound  pressure  level  measurement  every  2  s  over  each  30-­‐s  

testing  period,  n=50  sensors  x  5  test  levels  x  1  measurement  every  2  s  x  30  s=  3,750  paired  

measurements.    

Data  Analysis  

For  each  sensor  at  each  test  level,  we  calculated  the  mean  and  standard  deviation  of  the  

sound  level  reported  by  the  sensor  (dBA !"#!$! ; SD!"#!$% )  and  the  SLM  (dBA !"# ; SD!"# ).  The  

difference  between  the  means  of  the  noise  sensor  and  the  SLM  was  compared  to  our  

acceptance  criterion  of  ±2  dBA.  This  criterion  was  adapted  from  ANSI  standard  S1.4  -­‐  1983  

(R2006)  for  Type  2  SLMs  as  an  indication  of  accuracy  (ANSI,  1983).    

Further  analyses  were  conducted  on  sensors  passing  this  acceptance  criterion.  For  test  

level  over  all  sensors,  we  calculated  the  mean  and  standard  deviation  of  individual  sensor  

means.  As  an  indication  of  sensor  precision,  we  calculated  the  coefficient  of  variation  as:  

!"!"#!$%  
%  CV!"#!$% = ∗ 100%                                            Equation  2  
!"#!"#!$%

As  an  indicator  of  accuracy,  we  calculated  bias  for  each  sensor  and  test  level  as:  

!"#!"#!$% !!"#!"#  
%  Bias = ∗ 100%                                                Equation  3  
!"#!"#

19  
 
 For  each  sensor,  simple  linear  regression  analysis  was  conducted  to  determine  the  slope  

and  intercept  between  the  sensor  and  SLM  outputs.    As  an  indicator  of  the  correlation  between  

the  sensor  and  SLM,  the  Pearson  correlation  coefficient  was  calculated.  

To  determine  the  overall  performance  of  the  noise  sensors,  data  were  also  aggregated  

over  two  ranges  relevant  to  occupational  exposure  assessments  (65  dBA  to  94  dBA  and  75  dBA  

to  94  dBA).  Overall  statistical  measures  were  calculated  for  coefficient  of  variation,  bias,  slope  

and  intercept  as  the  mean  of  individual  values.  Calculations  were  conducted  done  using  

Microsoft  Excel.  

A  paired  sample  t-­‐test  was  conducted  to  test  the  hypothesis  that  the  noise  sensor  sound  

level  was  equivalent  to  the  SLM  sound  level  at  each  testing  level  with  an  alpha  of  0.05.  

Significant  p-­‐values  indicate  a  difference  in  the  mean  of  the  noise  sensor  from  the  SLM.  All  t-­‐

tests  were  conducted  on  SAS  (Cary,  NC).  Codes  and  outputs  from  the  SAS  procedures  are  

included  in  Appendix  B.    

Results  

Ninety-­‐two  percent  of  the  noise  sensors  (n=46)  passed  our  acceptance  criterion  of  ±2  

dBA  deviation  from  the  SLM  from  75  dBA  to  94  dBA.  The  noise  sensors  that  did  not  meet  criteria  

were  removed  from  the  monitor  pool,  repaired,  and  retested  prior  to  field  deployment.  In  this  

manuscript,  we  report  only  on  the  results  from  the  initially  accepted  sensors.      

The  mean  A-­‐weighted  sound  level  measurements  of  the  noise  sensor  versus  the  SLM  are  

presented  in  Figure  3  and  Table  1.  At  the  ambient  test  level,  the  difference  between  the  noise  

sensor  mean  and  SLM  mean  was  nearly  12  dBA,  the  noise  sensor  over  reporting.  As  the  sound  

level  increased,  the  difference  between  mean  sensor  and  SLM  sound  pressure  levels  decreased.  

However,  the  mean  measurement  of  the  noise  sensors  was  statistically  different  from  that  of  

20  
 
the  SLM  at  all  testing  levels  (p  <  0.05).  Variability  in  the  measurements  decreased  as  sound  level  

increased,  as  indicated  by  the  reduction  in  the  standard  deviations  as  the  testing  level  increased.    

Mean  percent  bias  at  the  target  sound  levels  ranged  from  18.3%  at  ambient  to  0.48%  at  

94  dBA,  with  monotonic  decrease  in  bias  as  sound  level  increased  (Figure  1).  Percent  bias  close  

to  zero  indicates  only  slight  differences  between  the  values  reported  by  the  inexpensive  noise  

sensor  and  the  SLM.  The  percentage  of  monitors  that  met  acceptance  criteria  also  increased  as  

sound  level  increased.  Very  few  of  the  noise  sensors  met  acceptance  criteria  at  ambient  sound  

levels,  less  than  60  dBA,  where  100  percent  of  the  monitors  met  acceptance  criteria  for  85  dBA  

and  94  dBA.    

Overall  performance  statistics  of  the  noise  sensors  across  two  different  ranges  of  sound  

pressure  levels  are  presented  in  Table  2.While  the  95%  confidence  intervals  for  both  slopes  

included  unity,  slope  measures  were  improved  in  the  more  restricted  75  dBA  to  94  dBA  range  

(0.98  ±  0.03  dBA)  as  compared  to  the  65  dBA  to  94  dBA  range  (0.94  ±  0.05  dBA).  This  result  

indicates  that  the  noise  sensor  was  in  better  agreement  with  the  SLM  at  sound  pressure  levels  

greater  than  75  dBA.  The  mean  correlation  between  the  sensor  and  SLM  was  slightly  stronger  

from  75  dBA  to  94  dBA  (1.00  ±  0.001)  as  compared  to  the  65  dBA  to  94  dBA  (1.00  ±  0.004)  test  

range.  The  mean  percent  bias  of  the  noise  sensors  improved  from  1.47%  to  0.83%  when  the  65  

dBA  testing  level  was  excluded  from  the  range.  The  mean  coefficient  of  variation,  CV,  improved  

from  1.18%  to  0.76%  when  the  65  dBA  testing  level  was  excluded  from  the  range.    

A  box  plot  of  the  distributions  of  differences  between  the  reference  SLM  and  the  noise  

sensor  at  each  target  testing  level  is  presented  in  Figure  4.  Differences  between  the  output  of  

the  noise  sensor  and  the  reference  SLM  ranged  from  nearly  25  dBA  at  ambient  sound  levels,  to  

less  than  2  dBA  at  the  higher  target  noise  levels.  The  range  of  differences  decreased  as  noise  

21  
 
level  increased,  similar  to  the  trend  observed  in  Figure  3.  A  difference  of  zero  between  the  noise  

sensor  sound  level  and  the  SLM  sound  level  is  indicative  of  agreement  between  the  two  devices.    

Discussion  

The  inexpensive  noise  sensor  (~$30  for  components)  provided  similar  sound  pressure  

level  measurements  in  dBA  to  a  substantially  more  expensive,  Type  2  reference  SLM  (~$1,800)  

from  75  to  94  dBA.  The  stand-­‐alone  nature  of  the  noise  sensor  developed  in  this  work  with  

embedded  measurement  and  processing  on-­‐board,  coupled  with  their  small  size  and  low  cost,  

offers  great  potential  for  use  in  a  variety  of  applications.  Ninety-­‐two  percent  of  the  noise  

sensors  were  within  ±2  dBA  of  the  reference  SLM  from  65  to  94  dBA.  This  range  encompasses  

the  sound  levels  expected  in  manufacturing,  for  which  the  noise  sensors  were  designed.    

The  accuracy  of  the  noise  sensors,  represented  as  mean  bias  compared  to  the  reference  

SLM  was  1.47%  from  65  to  94  dBA.  Bias  further  reduced  to  0.83%  across  75  to  94  dBA.  The  bias  

determined  in  this  study  is  equivalent  to  less  than  ±1  dBA  deviation  from  the  SLM  across  the  75  

to  94  dBA  range.  Kardous  and  Shaw  (2016),  found  similar  accuracy  for  two  SLM  apps  with  the  

addition  of  an  external  microphone,  SoundMeter  and  SPL  pro.  Differences  for  these  apps  from  

the  reference  were  between  ±1  dBA  across  the  65  dBA  to  95  dBA  test  range.  SPLnFF  and  Noisee  

apps  had  wider  variations  in  sound  pressure  levels,  especially  at  65  dBA  and  75  dBA  (Chucri  A.  

Kardous  &  Shaw,  2016).  Nast,  Speer,  and  Le  Prell  (2016)  found  a  number  of  SLM  apps  that  

reported  differences  ranging  from  3  to  10  dBA  higher  than  the  Type  1  SLM  across  their  selected  

testing  frequencies.    

The  reference  SLM  equipped  with  a  Type  2  microphone  has  an  allowable  error  of  ±2  dBA  

deviation  from  a  Type  0  laboratory  standard.  The  noise  sensor  designed  in  this  study  performed  

well  compared  to  the  Type  2  reference,  but  this  reference  may  be  off  by  2  dBA  from  the  actual  

22  
 
sound.  Our  sensor  introduces  additional  uncertainty  and  bias  (1.15%  for  sound  levels  greater  

than  65  dBA)  that  must  be  considered.  While  the  designed  noise  sensor  is  not  intended  for  use  

in  determining  compliance,  if  decisions  were  to  be  made  regarding  worker  health  a  +4  dBA  

safeguard  should  be  applied  to  the  reported  sound  pressure  level  output.    

The  sensor  response  at  sound  levels  lower  than  75  dBA  was  not  consistently  within  our  

acceptance  criterion.  There  was  a  large  distribution  in  the  differences  between  the  sensor  and  

SLM  at  ambient  and  65  dBA  target  levels,  as  observed  in  Figure  4.  As  the  sound  level  increased,  

the  spread  of  the  distribution  of  differences  between  the  sensor  and  SLM  output  decreased.  

Preliminary  measurements  collected  at  the  heavy-­‐vehicle  manufacturing  facility  that  these  

sensors  were  designed  for  resulted  in  sound  level  measurements  consistently  greater  than  75  

dBA.  Additionally,  sound  levels  less  than  80  dBA  are  not  of  primary  concern  for  controlling  noise  

hazards  that  may  cause  noise-­‐induced  hearing  loss.  The  sensors  are  within  our  acceptance  

criterion  for  the  sound  range  they  were  customized  for  and  perform  similarly  to  the  Type  2  SLM.    

Modifications  can  be  made  to  the  sensor  circuitry  to  allow  for  application  in  a  variety  of  

workplaces  with  differing  sound  levels.  The  circuitry  of  these  sensors  could  be  customized  using  

high  or  low-­‐pass  filters  to  improve  accuracy  at  the  tail  ends  of  the  range.  While  sound  levels  less  

than  80  dBA  are  not  a  primary  concern  for  noise  exposure  in  the  workplace,  they  may  be  

relevant  for  other  environments,  such  as  patient  care  in  a  hospital.  Additional  refinement  could  

be  made  to  improve  the  accuracy  over  the  entire  65  to  94  dBA  range,  or  a  custom  range  for  a  

desired  environment.  Adjustments  can  be  made  to  the  sample  period  of  the  FFT  to  collect  

voltages  according  to  the  specifications  for  impulse,  slow  and  fast  weighting.    

Inexpensive  microphones,  like  the  electret  condenser  microphone  used  in  this  noise  

sensor,  are  incorporated  into  a  variety  of  consumer  products,  namely  smartphones.  Some  SLM  

23  
 
applications  on  phones  have  shown  good  agreement  with  reference  SLMs,  especially  with  the  

introduction  of  an  external  microphone  (Roberts  et  al.,  2016).  However,  the  accuracy  of  these  

applications  varies  significantly  with  smartphone  and  app  software  as  well  as  smartphone  

hardware  (Roberts  et  al.,  2016).  Additionally,  a  number  of  studies  have  shown  that  SLM  apps  

may  not  be  adequate  for  noise  assessments  in  occupational  environments  exceeding  90  dBA  

(Murphy  &  King,  2016;  Roberts  et  al.,  2016).  The  noise  sensor  evaluated  in  this  study  showed  

excellent  agreement  with  the  reference  SLM  at  sound  levels  higher  than  75  dBA,  indicating  that  

it  is  more  acceptable  for  use  in  a  noisy  work  environment  than  the  previously  evaluated  SLM  

apps.    

While  a  number  of  accurate  SLM  apps  have  been  identified,  there  are  limitations  to  

their  deployment  into  the  workplace  for  noise  monitoring.  The  smartphone  required  to  utilize  

the  app  can  be  expensive  and  bulky,  posing  cost  and  functionality  challenges.  Moreover,  the  

SLM  apps  are  computationally  demanding,  and  require  a  great  deal  of  battery  power.  

Additionally,  concerns  have  been  raised  regarding  data  privacy,  accuracy  over  time,  and  data  

export  and  sharing  (Chucri  A.  Kardous  &  Shaw,  2016;  Nast  et  al.,  2014).  Removing  the  noise  

sensor  from  the  smartphone  alleviates  a  number  of  the  concerns  associated  with  SLM  apps  and  

opens  the  door  for  application  to  other  projects  beyond  a  stationary  monitor  network.  For  

instance,  the  small  form  factor  offers  potential  for  repackaging  the  device  into  a  wearable,  

personal  model.  Additionally,  the  availability  of  FFT  output  enables  access  to  data  required  to  

provide  the  end  user  with  octave  band  information  that  may  help  identifying  determinants  of  

exposure  rather  than  just  sound  level.  For  this  study,  the  noise  sensor  was  incorporated  into  a  

multi-­‐hazard  monitor.  The  noise  sensor  components  could  stand  alone  with  the  addition  of  a  

power  source  and  data  storage  capabilities.  The  microcontroller  selected  cannot  accommodate  

24  
 
a  USB  flash  drive,  but  an  alternative  microcontroller  with  USB  storage  capacities  could  be  

selected  for  use  in  future  versions.  

The  purpose  of  this  inexpensive  noise  sensor  is  to  provide  reasonably  accurate  sound  

pressure  level  measurements  in  real-­‐time  so  that  an  occupational  health  professional  can  collect  

more  spatial  and  temporal  information  on  sound  levels  in  their  facility.  The  importance  of  

temporal  and  spatial  resolution  for  the  representativeness  of  hazard  maps  has  been  established,  

especially  in  non-­‐homogenous  workplaces  (Lake  et  al.,  2015).  A  clearer  understanding  of  the  

variations  in  sound  levels  throughout  the  facility  and  across  shifts  with  a  stationary  monitor  

network  will  allow  for  better  allocation  of  resources  when  conducting  personal  noise  exposure  

measurements.  However,  the  cost  of  SLMs  has  been  prohibitive  to  the  widespread  applicability  

of  a  stationary  monitor  network.  The  noise  sensor  we  evaluated  provides  a  relatively  accurate  

and  inexpensive  alternative  to  the  reference  SLM,  and  is  versatile  for  use  in  a  variety  of  

applications,  like  a  stationary  monitor  network.  

There  were  several  limitations  in  this  study.  Testing  of  the  noise  sensors  took  place  in  a  

quiet  office,  not  a  reverberant  chamber  as  other  noise  sensor  testing  studies  have  used  (C.  A.  

Kardous  &  Shaw,  2014;  Murphy  &  King,  2016;  Roberts  et  al.,  2016).  Reverberant  chambers  

create  a  diffuse  sound  field,  removing  the  influence  that  positioning  or  reverberation  off  

surfaces  can  have  on  the  sound  pressure  level  readings.  In  our  study,  the  microphones  of  the  

noise  sensor  and  reference  SLM  were  setup  close  to  each  other  to  limit  the  influence  of  position  

and  reverberation.  The  pink  noise  test  signal  selected  for  this  study  allowed  insight  into  the  

sensor  performance  over  a  range  of  frequencies,  but  is  not  a  perfect  surrogate  for  real  world  

noise  (Nast  et  al.,  2014).  Additional  evaluation  of  the  sensor  in  an  occupational  environment  

should  be  conducted  to  account  for  differences  in  temperature,  humidity,  increased  

intermittent  noise,  and  object  interference-­‐  variables  not  present  during  the  laboratory  testing.  

25  
 
These  noise  sensors  are  also  not  individually  calibrated  prior  to  use.  The  calibration  offset,  

added  on  the  sensor  microcontroller  during  development  of  the  sensor,  improved  accuracy  of  

the  sensors  and  allowed  for  consistent  measurements  of  noise  levels  within  2  dBA  of  the  

reference  SLM.  Future  work  may  look  to  incorporate  a  calibration  procedure  for  individual  noise  

sensors,  further  improve  the  accuracy  of  the  sound  measurements  across  the  target  range  and  

assess  sensor  performance  over  time.  

Conclusions  

A  new,  inexpensive  noise  sensor  (~$30  for  components)  developed  in  this  work  

responded  similarly  to  a  substantially  more  expensive  reference  SLM  from  75  dBA  to  94  dBA.  

The  independence  of  this  noise  sensor  from  a  smartphone,  coupled  with  the  small  form  factor  

and  low  price,  allows  for  use  in  a  variety  of  applications,  including  incorporation  into  a  sensor  

network.  Increasing  the  number  of  measurement  points  in  a  stationary  network  will  improve  the  

representativeness  of  hazard  maps  and  better  inform  decisions  on  where  further  sampling  

should  occur.  Future  work  may  look  to  optimize  the  noise  sensor  for  different  work  

environments  or  further  evaluate  the  sensor  performance  in  field  conditions.      

   

26  
 
 1@I78<?!69B!-64;<?!

! !
!

!
4<=>?8!9:!-><!@9<VC<9?@J<!?E79B!?<9?E8R!(!?@AC;@D@<B!H@8H7@=!B@6I86A!c=EC!C69<;d!?>EL?!=>6=!=><!
<;<H=8@H6;!?@I96;!D8EA!=><!A@H8EC>E9<!@?!6AC;@D@<B!=L@H<!4<DE8<!8<6H>@9I!=><!?<9?E8!
A@H8EHE9=8E;;<8!c-<<9?5dR! -><!6H=76;!HEACE9<9=?!ED!9E@?<!?<9?E8!@9HE8CE86=<B!@9=E!A7;=@[
>6Z68B!AE9@=E8!68<!?>EL9!@9!=><!4E==EA!C69<;R!(9!<;<H=8<=!A@H8EC>E9<!<V=<9B?!D8EA!=><!
<V=<8@E8!ED!=><!I8<5!<9H;E?78<!69B!=><!?<9?E8!A@H8EHE9=8E;;<8!@?!?>EL9!@9?@B<!=><!<9H;E?78<R!+9!
=><!8<B!H@8H7@=!4E68BS!6!AE9@=E8!A@H8EHE9=8E;;<8!HEAA79@H6=<?!L@=>!=><!9E@?<!69B!E=><8!>6Z68B!
?<9?E8?!69B!L@=>!6!B6=646?<!J@6!X@1@R!!

! !

MP !
!
!

!
4<=>?8!;:!$<=7C!DE8!;64E86=E85!J6;@B6=@E9!ED!9E@?<!?<9?E8R!'-%!2)M!69B!9E@?<!?<9?E8!A@H8EC>E9<?!
68<!;EH6=<B!L@=>@9!MRa!HA!ED!E9<!69E=><8S!H<9=<8<B!bN!HA!D8EA!=><!H<9=<8!ED!=><!6AC;@D@<8R!-><!
6AC;@D@<8!@?!HE99<H=<B!=E!=><!;6C=EC!45!67V@;@685!H64;<R!

! !

M_ !
!
!

9PP!

TO!

TP!
.8IHK?!.K>IJ!(8F87U!J/"!

SO!

SP!

RO!

RP!

QO!

QP!

OO!

OP!
OP! OO! QP! QO! RP! RO! SP! SO! TP! TO! 9PP!
08M8?8IV8!.K>IJ!(8F87U!J/"!
!
4<=>?8!@:!G<69!([L<@I>=<B!?E79B!;<J<;?!D8EA!9E@?<!?<9?E8!J<8?7?!8<D<8<9H<!$)G!6=!D@J<!=68I<=!
?E79B!@9=<9?@=@<?R!3E@9=?!8<C8<?<9=!A<69!8<6B@9I!ED!6;;!9E@?<!?<9?E8!69B!8<D<8<9H<!
A<6?78<A<9=!C6@8?R!#88E8!468?!8<C8<?<9=!=><!?=69B68B!B<J@6=@E9!ED!?E79B!C8<??78<!;<J<;?!6=!
<6H>!=68I<=!?E79B!@9=<9?@=5S!9k^j!?<9?E8?R!

! !

M` !
!
  

Table  1.  Mean  sound  pressure  levels,  standard  deviation,  and  coefficient  of  variation  at  five  
target  dBA  levels  (side  by  side,  65,  75,  85,  94  dBA).  P-­‐values  are  resultant  from  one-­‐sided  t-­‐test  
with  an  alpha  of  0.05.  Acceptance  criteria  (AC)  is  defined  as  monitor  output  within  ±2  dBA  of  
reference  output,  n=46  monitors.  

  Mean  ±  SD  (dBA),  CV  (%)        

Average   %  within    
Reference  SLM   Sensor   p-­‐value  
    Bias,  %   ±2  dBA  
Ambient   50.6  ±  1.9,  3.8   62.0  ±  3.2,  5.2   18.3   <0.0001   2.1  
65  dBA   65.4  ±  0.45,  0.69   67.7  ±  1.7,  2.4   3.22   <0.0001   62  
75  dBA   75.4  ±  0.13,  0.18   76.3  ±  0.55,  0.72   1.15   <0.0001   98  
85  dBA   85.2  ±0.12,  0.13   85.9  ±  0.37,  0.43   0.83   <0.0001   100  
94  dBA   94  ±  0,  0   94.4  ±  0.49,  0.52   0.48   <0.0001   100  
 

Table  2.  Slope  and  intercept,  correlation,  bias  and  coefficient  of  variation  results  for  lab  
checkout  of  noise  sensor  from  65  dBA  to  94  dBA  and  from  75  dBA  to  94  dBA,  n=  46  monitors.  

      Mean   Std.  Dev   5th  Percentile   95th  Percentile  

Slope   0.94   0.05   0.84   0.98  
Intercept,  dBA   5.69   4.84   2.49   15.6  
65-­‐94  dBA   Correlation,  r   1.00   0.004   0.99   1.00  
Overall  Bias,  %   1.47   0.77   -­‐   -­‐  
Overall  CV,  %   1.18   0.87   -­‐   -­‐  
           
Slope   0.98   0.03   0.94   1.00  
Intercept,  dBA   2.65   2.45   -­‐0.41   6.67  
75-­‐94  dBA   Correlation,  r   1.00   0.001   1.00   1.00  
Overall  Bias,  %   0.83   0.46   -­‐   -­‐  
Overall  CV,  %   0.76   0.27   -­‐   -­‐  
 

   

30  
 
!

! !
4<=>?8!A:!*EV!C;E=!ED!B@?=8@47=@E9?!ED!B@DD<8<9H<?!@9!?E79B!?<9?E8!E7=C7=!D8EA!=><!8<D<8<9H<!$)G!
6=!D@J<!=68I<=!?E79B!;<J<;?R!#88E8!468?!8<C8<?<9=!=><!B@?=8@47=@E9!ED!?E79B!C8<??78<!;<J<;!
B@DD<8<9H<?!6=!<6H>!=68I<=!?E79B!;<J<;S!9k^j!?<9?E8?R!

! !

bO !
!
 Chapter  III:  Conclusions    
 

The  inexpensive  noise  sensor  (~$30)  designed  in  this  study  performed  similar  to  a  Type  2  

sound  level  meter  (SLM,  ~$1,800)  over  a  test  range  from  65  dBA  to  94  dBA.  Ninety-­‐two  percent  

of  the  sensors  passed  our  acceptance  criterion,  reporting  sound  pressure  level  readings  within  

±2  dBA  of  the  SLM  across  the  test  range.  The  passing  sensors  will  be  deployed  as  part  of  a  

stationary  sensor  network  in  a  heavy-­‐vehicle  manufacturing  facility.  These  sensors  will  collect  

information  on  the  temporal  and  spatial  variability  of  noise  levels  in  the  workplace  and  provide  

measurements  to  help  inform  the  allocation  of  control  and  exposure  assessment  resources.  

This  noise  sensor  was  developed  as  a  potential  solution  for  the  lack  of  spatial  and  

temporal  information  in  workplace  noise  hazard  maps.  The  high  cost  of  traditional  direct-­‐

reading  instruments  places  budgetary  limits  on  the  installation  of  a  stationary  network  of  

monitors.  Inexpensive  noise  measurement  applications  (apps)  have  been  developed  for  

smartphones  and  provide  reasonably  accurate  noise  level  measurements,  especially  with  the  

installation  of  an  external,  calibrated  microphone.  However,  there  are  limitations  to  the  use  of  

smartphones  in  a  stationary  network,  such  as  data  privacy,  accuracy  over  time,  and  data  export  

and  sharing  (Chucri  A.  Kardous  &  Shaw,  2016;  Nast  et  al.,  2014).  The  inexpensive  noise  sensor  

developed  in  this  work  is  independent  of  a  smartphone  with  a  small  form  factor,  offering  

potential  for  use  in  a  variety  of  applications.        

  There  were  a  number  of  limitations  in  this  study.  Other  noise  sensor  testing  studies  have  

conducted  their  evaluation  inside  of  a  reverberant  chamber,  while  our  testing  took  place  in  a  

quiet  office  environment.  Reverberant  chambers  create  a  diffuse  sound  field,  removing  

potential  for  position  and  reverberation  to  affect  results.  To  address  this  concern,  we  placed  the  

32  
 
microphones  of  the  sensor  and  SLM  within  2  cm  of  each  other,  and  positioned  the  microphones  

to  point  directly  at  the  amplifier  producing  the  test  signal.    

The  selected  pink  noise  test  signal,  while  providing  an  insight  into  sensor  performance,  

is  not  a  perfect  surrogate  for  real  world  noise.  Workplace  noise  comes  from  various  sources  

ranging  in  frequency,  intensity,  and  duration  (NIOSH,  1998).  Additionally,  environmental  

conditions  in  a  factory  are  not  comparable  to  those  in  the  testing  office.  Differences  in  

temperature  and  humidity  in  the  heavy-­‐vehicle  manufacturing  facility  could  have  an  impact  on  

the  sensitivity  of  the  inexpensive  microphone  incorporated  into  the  sensor.  Field  performance  

of  these  sensors  should  be  evaluated  to  ensure  that  the  measurements  accurately  measure  real  

world  noise  levels.    

The  importance  of  pre-­‐  and  post-­‐use  calibration  for  noise  measurement  equipment  is  

well  defined  in  CFR  1910.95(d)(2)(ii).    The  calibration  offset,  added  on  the  sensor  

microcontroller  during  development  of  the  sensor,  improved  accuracy  of  the  sensors  and  

allowed  for  consistent  measurements  of  noise  levels  within  2  dBA  of  the  reference  SLM.  A  

calibration  constant  unique  to  each  individual  sensor  may  further  improve  accuracy.  However,  

current  sensor  performance  is  adequate  for  use  in  a  stationary  sensor  network  and  to  provide  

insight  as  a  surveying  tool.  Data  collected  by  these  sensors  can  be  aggregated  to  produce  hazard  

maps  and  bring  attention  to  areas  of  the  facility  where  intervention  is  most  needed.    

The  versatility  of  these  noise  sensors,  coupled  with  their  low  price  and  small  form  

factor,  offers  potential  for  their  use  in  projects  beyond  a  stationary  monitoring  network.  For  

example,  the  device  could  be  repackaged  from  the  stationary  monitor  into  a  wearable  personal  

model.  Additionally,  access  to  the  FFT  output  allows  enables  access  to  octave  band  information  

that  may  be  useful  in  identifying  determinants  of  exposure,  rather  than  just  noise  level.  The  

33  
 
circuitry  can  be  optimized  using  high  or  low-­‐pass  filters  to  capture  information  on  noise  ranges  

of  interest.  The  FFT  sample  period  can  also  be  modified  according  to  the  specifications  for  slow,  

fast,  or  impulse  weighting.    

The  performance  of  these  noise  sensors  from  75  dBA  to  94  dBA  is  promising  for  the  

future  of  occupational  noise  assessment.  Collecting  more  information  on  the  noise  hazards  

present  in  the  workplace  allows  us  to  make  the  best-­‐informed  decision  to  protect  worker  

health.  Incorporating  inexpensive  technology,  like  the  sensor  developed  in  this  study,  into  a  

stationary  network  of  monitors  will  allow  for  the  collection  of  real-­‐time  data  on  hazard  levels.  

With  this  information,  employers  can  feel  well  informed  in  their  decision-­‐making  and  cater  their  

exposure  assessment  and  hazard  abatement  plans  to  what  is  most  needed  for  their  facility.      

  Throughout  this  project,  I  have  learned  a  great  deal  about  the  importance  of  efficiency.  

The  small  inefficiencies  that  are  not  readily  noticeable  during  the  evaluation  of  one  sensor  

become  exceptionally  clear  when  you  scale  up  to  50  sensors.  Designing  data  collection  and  

storage  processes  with  simplicity  in  mind  will  make  the  acquisition  of  data  from  many  low-­‐cost  

sensors  more  straightforward  and  much  faster.  

 Another  lesson  learned  from  this  project  was  bridging  communication  gaps  between  

departments.  Working  closely  with  the  engineering  department  on  the  software  and  hardware  

of  the  sensor  allowed  us  to  ensure  the  sensor  was  designed  with  scientific  principles  in  mind.    

However,  it  was  often  challenging  to  agree  on  what  the  next  step  should  be,  given  the  diverse  

backgrounds  of  the  project  team.  I  learned  to  clearly  state  my  opinions  and  back  them  with  

evidence.  This  skill,  along  with  the  lessons  learned  in  simplifying,  will  help  me  to  become  a  more  

effective  and  efficient  communicator  and  researcher  in  my  future  career.      

   

34  
 
 References  
 

Alberti,  P.  W.  (1992).  "Noise  induced  hearing  loss."  British  Medical  Journal  304(6826):  522-­‐522.  

ANSI,  American  National  Standards  Institute.  (1983).  ANSI  S1.4:  Specifications  for  Sound  Level  
Meters.  7  CFR  1755.22(s)(3)(v).  

BLS,  United  States  Bureau  of  Labor  Statistics.  (2016).  Table  6.  Incidence  rates  and  numbers  of  
nonfatal  occupational  illnesses  by  major  industry  sector,  category  of  illness,  and  ownership,  
2014.  

Carrol,  Y.I.,  Eichwald,  J.,  Scinicariello,  F.,  Hoffman,  H.J.,  Deitchman,  S.,  Radke,  M.S.,  Themann,  
C.L.,  Breysse,  P.  (2017)  "Vital  Signs:  Noise-­‐Induced  Hearing  Loss  Among  Adults  -­‐  United  States  
2011-­‐2012."  MMWR  Morbidity  and  Mortality  Weekly  Report  2017;66:139-­‐144.        

CDC,  Centers  for  Disease  Control  and  Prevention.  (2015).  "Types  of  Hearing  Loss."  from  
https://www.cdc.gov/ncbddd/hearingloss/types.html.  

Directive,  C.  (2002/49/EC).  Declaration  by  the  Commission  in  the  Conciliation  Committee  on  the  
Directive  relating  to  the  assessment  and  management  of  environmental  noise.  

Dunn,  D.  E.,  R.  R.  Davis,  C.  J.  Merry  and  J.  R.  Franks  (1991).  "Hearing  loss  in  the  chinchilla  from  
impact  and  continuous  noise  exposure."  The  Journal  of  the  Acoustical  Society  of  America  90(4):  
1979-­‐1985.  

Filipponi,  L.,  S.  Santini  and  A.  Vitaletti  (2008).  Data  Collection  in  Wireless  Sensor  Networks  for  
Noise  Pollution  Monitoring.  Distributed  Computing  in  Sensor  Systems:  4th  IEEE  International  
Conference,  DCOSS  2008  Santorini  Island,  Greece,  June  11-­‐14,  2008  Proceedings.  S.  E.  
Nikoletseas,  B.  S.  Chlebus,  D.  B.  Johnson  and  B.  Krishnamachari.  Berlin,  Heidelberg,  Springer  
Berlin  Heidelberg:  492-­‐497.  

Hakala,  I.,  I.  Kivela,  J.  Ihalainen,  J.  Luomala  and  C.  Gao  (2010).  Design  of  Low-­‐Cost  Noise  
Measurement  Sensor  Network:  Sensor  Function  Design.  2010  First  International  Conference  on  
Sensor  Device  Technologies  and  Applications.  

Hong,  O.,  M.  J.  Kerr,  G.  L.  Poling  and  S.  Dhar  (2013).  "Understanding  and  preventing  noise-­‐
induced  hearing  loss."  Disease-­‐a-­‐Month  59(4):  110-­‐118.  

Kardous,  C.  A.  and  P.  B.  Shaw  (2014).  "Evaluation  of  smartphone  sound  measurement  
applications."  J  Acoust  Soc  Am  135(4):  EL186-­‐192.  

Kardous,  C.  A.  and  P.  B.  Shaw  (2016).  "Evaluation  of  smartphone  sound  measurement  
applications  (apps)  using  external  microphones  –  A  follow-­‐up  Study."  The  Journal  of  the  
Acoustical  Society  of  America  140(4):  EL327-­‐EL327.  

Koehler,  K.  A.  and  T.  M.  Peters  (2013).  "Influence  of  analysis  methods  on  interpretation  of  
hazard  maps."  Ann  Occup  Hyg  57(5):  558-­‐570.  

35  
 
Koehler,  K.  A.  and  J.  Volckens  (2011).  "Prospects  and  pitfalls  of  occupational  hazard  
mapping:‘between  these  lines  there  be  dragons’."  Annals  of  occupational  hygiene  55(8):  829-­‐
840.  

Kurmis,  A.  and  S.  Apps  (2007).  Occupationally-­‐Acquired  Noise-­‐Induced  Hearing  Loss:  A  Senseless  
Workplace  Hazard.  International  Journal  of  Occupational  Medicine  and  Environmental  Health.  
20:  127.  

Lake,  K.,  J.  Zhu,  H.  Wang,  J.  Volckens  and  K.  A.  Koehler  (2015).  "Effects  of  Data  Sparsity  and  
Spatiotemporal  Variability  on  Hazard  Maps  of  Workplace  Noise."  Journal  of  Occupational  and  
Environmental  Hygiene  12(4):  256-­‐265.  

Liu,  S.  and  S.  K.  Hammond  (2010).  "Mapping  particulate  matter  at  the  body  weld  department  in  
an  automobile  assembly  plant."  J  Occup  Environ  Hyg  7(10):  593-­‐604.  

Maisonneuve,  N.,  M.  Stevens,  M.  E.  Niessen  and  L.  Steels  (2009).  NoiseTube:  Measuring  and  
mapping  noise  pollution  with  mobile  phones.  Information  Technologies  in  Environmental  
Engineering:  Proceedings  of  the  4th  International  ICSC  Symposium  Thessaloniki,  Greece,  May  
28-­‐29,  2009.  I.  N.  Athanasiadis,  A.  E.  Rizzoli,  P.  A.  Mitkas  and  J.  M.  Gómez.  Berlin,  Heidelberg,  
Springer  Berlin  Heidelberg:  215-­‐228.  

Masterson,  E.A.,  Bushnell,  P.T.,  Themann,  C.L.,  Morata,  T.C.  (2016)    Hearing  Impairment  Among  
Noise-­‐Exposed  Workers  —  United  States,  2003–2012.  MMWR  Morb  Mortal  Wkly  Rep  
2016;65:389–394.  DOI:  http://dx.doi.org/10.15585/mmwr.mm6515a2  

Murphy,  E.  and  E.  A.  King  (2016).  "Testing  the  accuracy  of  smartphones  and  sound  level  meter  
applications  for  measuring  environmental  noise."  Applied  Acoustics  106:  16-­‐22.  

Nast,  D.,  W.  Speer  and  C.  Le  Prell  (2014).  "Sound  level  measurements  using  smartphone  "apps":  
Useful  or  inaccurate?"  Noise  and  Health  16(72):  251-­‐256.  

NIOSH,  National  Institute  for  Occupational  Safety  and  Health.  (1998).  Criteria  for  a  
Recommended  Standard-­‐  Occupational  Noise  Exposure.  U.  S.  D.  o.  H.  a.  H.  Services.  Cincinnati,  
Ohio.  

OSHA,  Occupational  Safety  and  Health  Administration.  (1974).  Occupational  Noise  Exposure.  29.  
1910.95.  

OSHA,  Occupational  Safety  and  Health  Administration.  (2013).  OSHA  Technical  Manual.  

Peters,  T.  M.,  T.  R.  Anthony,  C.  Taylor,  R.  Altmaier,  K.  Anderson  and  P.  T.  O'Shaughnessy  (2012).  
"Distribution  of  Particle  and  Gas  Concentrations  in  Swine  Gestation  Confined  Animal  Feeding  
Operations."  Annals  of  occupational  hygiene  56(9):  1080-­‐1090.  

Pew  Research  Center,  P.  (2017).  "Mobile  Fact  Sheet."  from  http://www.pewinternet.org/fact-­‐
sheet/mobile/#.  

Prasanna  Kumar,  G.,  K.  Dewangan,  A.  Sarkar,  A.  Kumari  and  B.  Kar  (2008).  "Occupational  noise  in  
rice  mills."  Noise  and  Health  10(39):  55-­‐67.  

36  
 
Roberts,  B.,  C.  Kardous  and  R.  Neitzel  (2016).  "Improving  the  accuracy  of  smart  devices  to  
measure  noise  exposure."  Journal  of  Occupational  and  Environmental  Hygiene  13(11):  840-­‐846.  

Santini,  S.,  B.  Ostermaier  and  A.  Vitaletti  (2008).  First  experiences  using  wireless  sensor  
networks  for  noise  pollution  monitoring.  Proceedings  of  the  workshop  on  Real-­‐world  wireless  
sensor  networks.  Glasgow,  Scotland,  ACM:  61-­‐65.  

Tak,  S.,  R.  R.  Davis  and  G.  M.  Calvert  (2009).  "Exposure  to  hazardous  workplace  noise  and  use  of  
hearing  protection  devices  among  US  workers-­‐-­‐NHANES,  1999–2004."  Am  J  Ind  Med  52.  

Van  Renterghem,  T.,  P.  Thomas,  F.  Dominguez,  S.  Dauwe,  A.  Touhafi,  B.  Dhoedt  and  D.  
Botteldooren  (2011).  "On  the  ability  of  consumer  electronics  microphones  for  environmental  
noise  monitoring."  Journal  of  Environmental  Monitoring  13(3):  544-­‐552.  

WHO,  World  Health  Organization.  (2001).  "Occupational  exposure  to  noise:  evaluation,  
prevention  and  control."  

   

37  
 
 Appendix  A  
 

SOP2040.00:  Laboratory  Checkout  of  Noise  Sensors  

1.   Purpose  and  Applicability    

This  document  outlines  a  procedure  for  the  lab  checkout  of  the  inexpensive  noise  
sensors  prior  to  deployment  in  the  low-­‐cost  sensor  network  

2.   Safety  and  Operating  Precautions  

N/A  

3.   Equipment  and  Materials  

• NTI  XL2  Sound  Level  Meter  and  M4160  microphone  (NTi  Audio,  Tigard,  Oregon,  
USA)  
o AC  power  supply  cord  
o mini-­‐USB  cable    
o NTi  Test  Signal  CD  
• Sound  Level  Meter  Calibrator    
o SCAL1356,  General  Tools  &  Instruments,  Secaucus,  New  Jersey,  USA  
• Fender  amplifier    
o 2311000000,  Fender  Musical  Instruments  Corporation,  Scottsdale,  AZ,  USA  
• Laptop  computer  
o Auxillary  cord  
• NTi  Audio  Test  Signals  for  Audio  and  Acoustic  Analyzers  V1.0  
• Cardboard  footprint  for  setup  
• Wooden  block  to  elevate  SLM  
4.   References  

  N/A  

5.   Definitions  

  SLM:  Sound  Level  Meter  

38  
 
Q:! 1?KV8J>?8!

9: 1?8C5?8!DB8!8_><C]8ID!MK?!D8HD<I=
OROR K6;@486=<!=><!$)G!
OROROR!-789!=><!$)G!E9!45!B<C8<??@9I!=><!+9q+DD!47==E9S!6?!@9B@H6=<B!@9!1@I78<!O!

!
1@I78<!OQ!*7==E9!D79H=@E9?!ED!$)G!
!
ORORMR!0?<!=><!8E=685!L><<;!=E!>@I>;@I>=!=><!A6@9!A<97!6=!=><!=EC!ED!=><!?H8<<9S!
69B!=><9!C8<??!#'-#,R!$H8E;;!=E!>@I>;@I>=!K6;@486=@E9S!=><9!C8<??!#'-#,!
6I6@9!=E!EC<9!=><!H6;@486=@E9!?H8<<9!!

!
1@I78<!MQ!+C<9@9I!=><!H6;@486=@E9!?H8<<9!
!
ORORbR!G6Y<!?78<!=>6=!=><!?H8<<9!?65?S!r0?<8!K6;@486=@E9!6=!OO^RN!B*Rs!0?<!=><!
8E=685!L><<;!=E!>@I>;@I>=!,0'S!47=!BE!9E=!C8<??!#'-#,!5<=R!

b` !
!
1@I78<!bQ!K><HY!H6;@486=@E9!?<==@9I?!69B!>@I>;@I>=!,0'!
!
OROR^R!-789!=><!$)G!K6;@486=E8!E9!=E!=><!OO^RN!B*!?<==@9I!69B!D@=!=><!EC<9@9I!ED!
=><!H6;@486=E8!E9=E!=><!$)G!A@H8EC>E9<R!."4)-`!%&-)\!*K!IKD!C>D!DB8!
V57<6?5DK?!>C!DK!LK>?!85?!

!
1@I78<!aQ!3EL<8!E9!$)G!K6;@486=E8!69B!C;6H<!E9=E!$)G!A@H8EC>E9<!
ORORaR!38<??!#9=<8!=E!4<I@9!8799@9I!H6;@486=@E9R!-><!?H8<<9!L@;;!@9B@H6=<!
rK6;@486=@E9!8799@9Is!69B!=><9!r$7HH<??D7;;5!D@9@?><B€s!L><9!HEAC;<=<R!

!
1@I78<!jQ!K6;@486=@E9!8799@9I!69B!HEAC;<=<!?H8<<9?!

^N !
!
 ORORjR!0?<!=><!8E=685!L><<;!=E!>@I>;@I>=!=><!A6@9!A<97!6I6@9S!=><9!>@I>;@I>=!=><!
$)G<=<8q,-(!EC=@E9!69B!C8<??!#9=<8!=E!8<=789!=E!=><!>EA<!?H8<<9!
ORMR $<=7C!DE8!=><!)64!K><HYE7=!
ORMROR!)65!=><!H68B4E68B!4;7<C8@9=!E9!=><!=64;<!69B!D@;;!@9!=><!?CE=?!L@=>!=><!
HE88<?CE9B@9I!<F7@CA<9=!

!
1@I78<!PQ!$<=7C!ED!<F7@CA<9=!DE8!;64!H><HYE7=!C8EH<??!
!
ORMRMR!KE99<H=!=><!6AC;@D@<8!=E!=><!HEAC7=<8!L@=>!=><!67V@;@685!H64;<!

!
1@I78<!PQ!)EH6=@E9?!=E!HE99<H=!(02!H64;<!E9!6AC;@D@<8!69B!;6C=EC!
ORbR KE99<H=!=><!$)G!=E!=><!3,&!C8EI86A!!

^O !
!
 ORbROR!+C<9!=><!AE?=!8<H<9=!J<8?@E9!ED!$<8@6;E4\<H=R<V<!E9!=><!;6C=EC!
ORbRMR!K>69I<!=><!HE9D@I786=@E9?!E9!=><!3,&!C8EI86A!
ORbRMROR #J<9=!'6A<Q!'E@?<K><HY[5555AABB!
ORbRMRMR ,<CE8=!(J<86I@9I!-@A<Q!M!?<H!
ORbRMRbR "6=6!36=>Q!KQÅ0?<8?Å)64[6BA@9ÅB<?Y=ECÅ)6786Å3,&!+7=C7=?!
ORbRMR^R $=6=@E9685!GEB<!

!
1@I78<!_Q!)65E7=!69B!HE9D@I786=@E9!ED!3,&!C8EI86A!DE8!;64!H><HYE7=!
!
ORbRbR!/<8@D5!=><!K+G!CE8=!6??@I9<B!=E!=><!$)GR!)<D=!H;@HY!69B!?<;<H=!r-<?=s!=E!
<9?78<!=><!B<J@H<!@?!HEAA79@H6=@9I!L@=>!=><!C8EI86AR!

!
1@I78<!`Q!K><HY!=><!K+G!CE8=!6??@I9<B!=E!=><!$)G!69B!=<?=!=>6=!@=!@?!HEAA79@H6=@9I!
ORbR^R!38<??!r$=68=!K6C=78<s!=E!4<I@9!8<HE8B@9I!A<6?78<A<9=?!
ORbR^ROR #9=<8!5E78!;EH6=@E9!6?!%,#:!
!
OR^R +C<9!=><!AE9@=E8!48EL?<8!DE8!=><!;EL[HE?=!AE9@=E8!=>6=!5E7!68<!<J6;76=@9I!!
OR^ROR !KE99<H=!=E!=><!r(@8F76;@=5s!L@8<;<??!9<=LE8Y!

^M !
!
 OR^RMR!+C<9!K>8EA<!69B!<9=<8!=><!%3!6BB8<??!DE8!=><!AE9@=E8!6?!ONRO^MR^^ROVVS!
D@;;@9I!@9!=><!;6?=!=LE!B@I@=?!L@=>!=><!97A4<8!%"!DE8!=><!AE9@=E8!5E7!68<!
=<?=@9I!
OR^RbR!38<??!?=68=!H6C=78<!=E!4<I@9!8<HE8B@9I!MRa!?<HE9B!B6=6!B@8<H=;5!D8EA!=><!
AE9@=E8!
!
;: (56!#B8V^K>D!KM!%K<H8!.8IHK?!
MROR 37=!=><!'-@!-<?=!$@I96;!K"!@9=E!=><!HEAC7=<8R!$<;<H=!r3;65!67B@E!K"!7?@9I!
X@9BEL?!G<B@6!3;65<8s!D8EA!=><!67=EA6=@H!CEC[7C!L@9BELR!!
MROROR K>EE?<!-86HY!_!E9!=><!G<B@6!3;65<8[!3@9Y!'E@?<!cH8<?=!D6H=E8S!3(,‚k!bRa!
cONR`!B*dd!

!
1@I78<!ONQ!-<?=!K"!69B!CEC[7C!L@9BEL!L><9!5E7!C7=!K"!@9=E!HEAC7=<8!
!
MRMR -789!=><!6AC;@D@<8!E9!69B!A6Y<!?78<!=>6=!=><!=<?=!?@I96;!@?!C;65@9IS!=><9!C8<??!
C67?<!
MRbR ,<HE8B!=><!=@A<!69B!?=68=!5E78!D@8?=!bN[?<H!=<?=!C<8@EBR!->@?!@?!5E78!r(A4@<9=s!
=<?=!;<J<;R!$<<!1@I78<!OO!DE8!<V6AC;<!=64;<!DE8!8<HE8B@9I!?=68=!69B!?=EC!=@A<?R!

!
1@I78<!OOQ!#V6AC;<!8<HE8B!ED!?=68=!69B!?=EC!=@A<?!DE8!GE9@=E8!a!
!
MR^R +9H<!=><!D@8?=!=<?=!C<8@EB!>6?!HE9H;7B<BS!79[C67?<!=><!=<?=!?@I96;!
MRaR 0?<!=><!JE;7A<!HE9=8E;?!E9!=><!HEAC7=<8!=6?Y!468!=E!6B\7?=!=><!JE;7A<!79=@;!
=><!$)G!8<6B?!6CC8EV@A6=<;5!ja!B*(!
MRjR ,<HE8B!=><!=@A<!69B!?=68=!=><!bN[?<H!=<?=!C<8@EBR!->@?!@?!5E78!rja!B*(s!=<?=!;<J<;R!

^b !
!
 MRPR ,<C<6=!$=<C?!MRa[!MRj!DE8!=<?=!;<J<;?!6=!Pa!B*(!69B!_a!B*(R!.E7!A65!9<<B!=E!
8<?=68=!=><!=<?=!?@I96;!=86HYS!6?!=><!B786=@E9!@?!bQNN!A@9!69B!5E78!=<?=@9I!A65!
=6Y<!;E9I<8R!!
MR_R -789!=><!$)G!K6;@486=E8!E9!=E!=><!`^!B*!?<==@9I!69B!@9?=6;;!=><!H6;@486=E8!E9=E!
=><!AE9@=E8!A@H8EC>E9<R!c1@I78<!OMd!

!
1@I78<!OMQ!$)G!K6;@486=E8!6DD@V<B!=E!=><!AE9@=E8!A@H8EC>E9<!
!
MR`R ,<HE8B!=><!=@A<!69B!?=68=!=><!bN[?<H!=<?=!C<8@EBR!->@?!@?!5E78!r`^!B*(s!=<?=!;<J<;R!
MRONR!,<AEJ<!=><!$)G!H6;@486=E8!D8EA!=><!AE9@=E8!A@H8EC>E9<!
MROOR!+C<9!=><!AE9@=E8!48EL?<8!69B!C8<??!?=EC!
MROOROR KEC5!6;;!E7=C7=!J6;7<?!69B!C6?=<!@9=E!6!9E=<C6B!D@;<!
MROORMR $6J<!D@;<!6?!r'E@?<K><HYE7=T....GG""TGE9@=E822s!
MROMR!$<;<H=!=><!9<V=!AE9@=E8!DE8!H><HYE7=!69B!8<C<6=!$=<C?!OR^[MROO!
MRObR!+9H<!5E7!>6J<!=<?=<B!6;;!=><!AE9@=E8?S!C8<??!r$=EC!K6C=78<s!E9!=><!3,&!
C8EI86AR!-><!D@;<!L@;;!4<!?6J<B!@9!=><!DE;B<8!@9B@H6=<B!@9!r"6=6!36=>s!
MRO^R!$6J<!6;;!5E78!@9B@J@B76;!AE9@=E8!D@;<?!69B!=><!$)G!D@;<!=E!6!D;6?>!B8@J<!DE8!;6=<8!
C8EH<??@9I!
!
! !

^^ !
!
 Appendix  B  
 

Appendix  B  includes  the  SAS  codes  and  output  for  the  t-­‐tests,  mean  bias  by  test  level,  and  the  
boxplot  of  differences  by  test  level.  

T-­‐Test  by  Test  Level  

Code  
proc ttest data=noise;
title "T-test by Test Level";
paired reference*monitor;
by level;
run;

Output  

45  
 
Mean  Bias  by  Test  Level  
Code  
Data noise;
set noise;
bias= (monitor-reference)/reference;
run;

proc means data= noise;

title "Mean Bias by Test Level";
variable bias;
by level;
run;

46  
 
Output  

Box  Plot  of  Difference  at  each  Test  Level  

Code  
proc sort data=noise out=noise2 (keep=level difference);
by difference;
run;

proc sort data=noise2 out=noise3;

by level;
run;

47  
 
title 'Box Plot for Noise Levels';
proc boxplot data=noise3;
plot difference*level;
run;  

Output  

 
 

48