Extended experimental data

[daccordeon]

goals: improve amplitude and frequency limits

new kit to source from the labs:

• photo-diode/resistor for higher fps (faster response)
• horn/parabolic dish/reflector to amplify sound
• conversion for AC (red-black) to jack
• tin can and string
• suspension system (putty, water, strings) to make mirrors more sensitive

[daccordeon]

estimate noise level by adding in second signal can only do once second speaker is sourced

[daccordeon]

spoke to my JayCar friend, and he said a whole bunch of useful things:

• LDR hysteresis is too much, basically webcam response time, (at least a ~10ms response time, which looks to kill any signals above 100Hz)

• photodiode would be great, just find one in the right wavelength range i.e. https://au.rs-online.com/web/p/photodiodes/4975661/

• arduino vs raspberry pi is a consideration of cost vs usability both have the capability to capture the time series we want, however:

• arduino, although cheaper, works at a slow capture rate without clever work to improve it (which is possible)

• data output to SD card is annoying, don't need to stream but would like to just save the datafile to desktop on USB connected PC or whatever

• both have steep initial learning curves to implement, although the task we're interested in is sufficiently common for existing code to (hopefully) work https://www.jaycar.com.au/arduino-learning https://www.tinkercad.com/circuits

recommendation: https://www.jaycar.com.au/duinotech-arduino-starter-kit/p/XC3902 https://www.jaycar.com.au/photodiode-5mm-led/p/ZD1948

https://au.rs-online.com/web/p/photodiodes/4975661/

[daccordeon]

raspberry pi benefits: """ This is because you'll get much better SD read/writes, it already has the card reading in-built, and you can also process the data as it's recorded in real time and plot that data which is much more useful for checking to make sure you're getting the data you think you are. """

[daccordeon]

https://electronics.stackexchange.com/questions/73732/how-to-use-sfh235-ir-photodiode-correctly

[hannahm8]

here's the specs for our laser: https://www.thorlabs.com/thorproduct.cfm?partnumber=CPS532-C2

[daccordeon]

just as a note: """ The alternative is a phototransistor, which will give you an analogue signal out, but will require a more custom circuit to allow it to interface with the Arduino, unless you're able to find a purpose built module.

Just searching "arduino module ..." should likely exist somewhere already. """

[daccordeon]

also good to note: Blake has some custom code to direct capture live data from the arduino onto a USB connected PC, worth a try

[daccordeon]

things to do for week 3:

• add in microphone and second speaker, examine viterbi noise threshold (water noise in right band)
• attempt tin can to microphone capture
• set up Raspberry Pi (henceforth: R-Pi)
• when received, set up photo-diode circuit
• collect and process faster response data

[daccordeon]

audio commands for controlling internal mic: https://askubuntu.com/questions/123798/how-to-hear-my-voice-in-speakers-with-a-mic?noredirect=1&lq=1

[daccordeon]

problems interfacing with surface: microphone not recognised, likely unless combined with USB port or re-combiner with speakers

[daccordeon]

for noise estimation, see pink noise for noise over the relevant band https://mynoise.net/NoiseMachines/whiteNoiseGenerator.php

[daccordeon]

also, take spectrum of background room sound

[daccordeon]

raspberry pi set-up: https://projects.raspberrypi.org/en/projects/raspberry-pi-setting-up/3 https://www.raspberrypi.org/downloads/noobs/ http://qdosmsq.dunbar-it.co.uk/blog/2013/06/noobs-for-raspberry-pi/

[daccordeon]

pi doesn't have analogue inputs, may need these so need an analogue-digital converter (ADC): e.g. MCP3008 https://au.rs-online.com/web/p/general-purpose-adcs/7386651/ edit:acronym

[daccordeon]

better pins on this ADC: https://au.rs-online.com/web/p/general-purpose-adcs/6696064/

[daccordeon]

result from end of week 3: photodiode works, appears to read light level but is very sensitive week 4: begin measuring the interference pattern with the photodiode

[daccordeon]

if we can keep up a sampling rate of over 8kHz we should be able to capture all human speech https://en.wikipedia.org/wiki/Voice_frequency https://www.dpamicrophones.com/mic-university/facts-about-speech-intelligibility current fps is at least 8kHz, so we look good for it

---

also look into audio play back from photodiode

[daccordeon]

correction: current fps is 16kHz, so we're well in the clear for observing the maximum frequency

---

frequency resolution = fps / number of datapoints however, the human voice is not a slowly changing continuous source, so viterbi tracking is not useful. even if we tried, the window width is small and so the number of datapoints decreases, leading to higher frequency uncertainty(resolution)

---

audio play-back after lunch

[daccordeon]

better quality audio through increasing intensity resolution:

• replicate 10-bit audio, or more realistically to current op amp -> 100 channel, by ripping audio from yt, pushing to 100 channels (in numpy), then play back in wav
• use this to determine if the kit has any chance of recognisable play back

[daccordeon]

looking at an impasse now having simulated the audio (which showed that digitisation should work). tactical choice now is to reel back the project scope and write up the simple viterbi results.

[daccordeon]

try: threshold frequency noise

[daccordeon]

https://stackoverflow.com/a/19122540 F(f . g) = F(f) * F(g)

[daccordeon]

decided to keep this issue open until discussion is over on issue #11

[daccordeon]

actually, refreshing focus is good, this is moved to issue #11 from here