Basic examples of a sound level meter (slm) coded python3 with pyaudio and numpy (thanks people!). It allows to compute equivalent levels in dB, dBA and dBC (with linear filter) which is fastest than using 1/3 octave values and ponderations.
Features :
- Equivalent Level (Leq) in dB.
- dbA and dBC ponderations (for 44100Hz and 48000Hz sampling frequency).
- Level Calibration (in dB).
- Soundcard listing with possible sampling frequency.
- Easy soundcard selection.
Tested on a Raspberry Pi 4 / Raspbian Buster / Python 3.7. Microphone used NSRT MK3 by Convergence Instruments.
-
Make sure the raspberry pi is up to date
sudo apt-get update & sudo apt-get upgrade -y -
Go to the HuBBuB-SLM folder (git clone it or download and unzip it)
cd HuBBuB-SLM -
Install required packages : pip, portaudio lib (pyaudio is a python portaudio binding), atlas lib (for numpy)
sudo apt-get install python3-pip portaudio19-dev libatlas-base-dev -
Install virtualenv and create a virtual environment
python3 -m pip install virtualenvpython3 -m virtualenv hubbub -
Install python required librairies
python -m pip install -r requirements.txtcython numpy scipy pyaudio nsrt-mk3-dev
The rpi doesn't have an audio input by default. You can any usb soundcard, usb webcam with builtin microphone, a digital mems microphone, there also hats with audio input capability, etc... If you use a laptop you'll probably have a builtin microphone.
List all recording soundcard detected.
arecord -l
For me outputs :
**** List of CAPTURE Hardware Devices ****
card 1: NSRTmk3Dev [NSRT_mk3_Dev], device 0: USB Audio [USB Audio]
Subdevices: 1/1
Subdevice #0: subdevice #0
python hubbub/soundcards.py
Outputs :
-- 0 --
Name : bcm2835 Headphones: - (hw:0,0)
Default Fs : 44100.0
Max Input Channels : 0
Max Output Channels : 8
__________________________________________________________
-- 1 --
Name : NSRT_mk3_Dev: USB Audio (hw:1,0)
Default Fs : 48000.0
Supported Fs : [32000, 48000]
Max Input Channels : 1
Max Output Channels : 0
__________________________________________________________
-- 2 --
Name : sysdefault
Default Fs : 44100.0
Max Input Channels : 0
Max Output Channels : 128
__________________________________________________________
-- 3 --
Name : default
Default Fs : 44100.0
Max Input Channels : 0
Max Output Channels : 128
__________________________________________________________
-- 4 --
Name : dmix
Default Fs : 48000.0
Max Input Channels : 0
Max Output Channels : 2
When used as a usb soundcard, the NSRT MK3 microphone outputs a weighted signal. This is a really nice feature ! The NSRT has a serial communication port to change theses settings and thanks to Xander Hendriks there is a python package called nsrt-mk3-dev that allow us to communicate easily with it.
Two possibilities now :
- For multiple weighted measure (i.e dBA, dBC, and dBZ), we need to setup it as zero weighting (a.k.a Z weighting). like the first 3 examples
- For only one measure, we can change it to the wanted weighting. see the soundlevels_nsrt example
-
Find the microphone serial port (from the nsrt-mk3-dev documentation)
dmesg | grep -i usband look for :
[22339.508035] usb 1-1.2: Product: NSRT_mk3_Dev [22339.508044] usb 1-1.2: Manufacturer: Convergence Instruments [22339.520711] cdc_acm 1-1.2:1.1: ttyACM0: USB ACM device -
Open a python console (in the virtualenv)
import nsrt_mk3_dev nsrt = nsrt_mk3_dev.NsrtMk3Dev('/dev/ttyACM0') weighting = nsrt.read_weighting() print(weighting)Outputs :Weighting.DB_C# FOR DBA nsrt.write_weighting(nsrt.Weighting.DB_A) # FOR DBC nsrt.write_weighting(nsrt.Weighting.DB_C) # FOR DBZ nsrt.write_weighting(nsrt.Weighting.DB_Z)
Don't overflood the microphone ! Here is a warning from the NSRT-MK3 documentation :
The Flash memory that is used to contain these values can sustain approximately 10 000 write cycles
over the lifetime of the instrument. Even though that is a large number, the instrument is not designed
to sustain constantly changing the values in rapid succession. For instance, switching the weighting
function back and forth in an attempt to read both A and C levels all the time will quickly exhaust the
number of cycles guarantied for that Flash memory.
Whenever Tau, Fs or Weighting are modified, the instrument’s correction filters are reset and that
creates a transient spike in the indicated levels. In order to read valid levels after changing one of these
parameters, a delay of at least the largest of 1 second, or 10 times the value of Tau should be
observed.
Even if it is not bundled as a python package, all useful function are located in the hubbub/ folder :
- soundcards.py : Useful functions to list the detected soundcard and find the index of the wanted soundcard by name.
- slm.py : Sound level computation and weighting.
There is 4 examples :
- soundlevels_simple.py : Basic pyaudio example with blocking thread. It is not very optimised as you can easily overflow the input and miss frame.
- soundlevels_and_recording.py : Basic pyaudio example to compute sound levels and record the output as the same time.
- soundlevels_callback.py : The more advanced example which give the more accurate result.
- soundlevels_nsrt.py : This example, based on the soundlevels_callback, is only intended to be used with the nsrt microphone. It outputs only one weighted measure ('A', 'C' or 'Z') by changing the internal configuration of the NSRT MK3.
In each example you change the config in the main() function.
### CONFIG ###
# Microphone device name
soundcard_name = 'NSRT_mk3'
# Sampling Frequency (for dBA and dBC : can be either 44100 or 48000)
sampling_rate = 48000
# Equivalent level window time (in seconds)
leq_time = 1
# Calibration levels (in dB)
calibration = 0
calibration_A = 0
calibration_C = 0
-
List the soundcards detected :
python hubbub/soundcards.py-- 1 -- Name : NSRT_mk3_Dev: USB Audio (hw:1,0) Default Fs : 48000.0 Supported Fs : [32000, 48000] Max Input Channels : 1 Max Output Channels : 0 -
Change the
soundcard_namedepending on your configuration (you don't need to put the full name). In my config :NSRT -
Change the
leq_timedepending on your need. In my config : 1 second -
Change the
sampling_ratedepending on your card. CheckDefault FsandSupported Fs. In my config : 48000 Hz -
Be sure to be in the same folder as the example script.
-
Launch it !
python soundlevels_callback.pyStream Opened - Soundcard : 1 NSRT_mk3 - FS : 48000Hz - Leq Time : 1sec - Leq Time (Real) : 1.0sec - Buffer Size : 48000 - Calibration : 0dB(z) 0dB(a) 0dB(c) __________________________________________ 2021-09-29 17:04:10.90: dB=42.08 dBA=42.05 dBC=41.77 2021-09-29 17:04:11.89: dB=46.49 dBA=46.15 dBC=45.94 2021-09-29 17:04:12.89: dB=41.9 dBA=41.64 dBC=41.65
PyAudio logs a lot of errors at the beginning, it is OK. Never succeeded to make it quiet, so if you have a hint please share in the issue comment!
The example that started all : http://stackoverflow.com/questions/4160175/detect-tap-with-pyaudio-from-live-mic
Theses examples are the result of several years of development of theses project :
-
Was first used for the Musikiosk project (2015). www.musikiosk.org / Project Page / Scientific papers
-
Hubbub is a fully automated low-cost noise monitoring station based on Raspberry Pi. Featuring an user interface (control/configuration), a telegram bot, visualization screen, scheduler, wifi hotspot and much more ; It is a perfect tool for long term sound level measurement without expensive equipment. It was used in several research studies. If you interested in this or if you want more informations please contact me.