Pushing a AD7476 on a pHAT looks promising. 12 bits, 1Msps, let's see !
Testing with a SPI_CLK at 1MHz, with a 1khz beacon, getting 9.5khz from the ADC:
Humm.. not the speed I'd be expecting. Let's explore the speed of sampling (Fech, in orange) vs the speed of the SPI_CLK (in blue). Seems it's stuck at 10khz for the RPi:
That would come from the RPi SPI overheads, with large intervals between captures. The below has a SPI_CLK at 18MHz, with large overheads between words:
Now: just need to find out where to accelerate the spi =)
Some time has passed, but I'm willing to go back to some hardware for going fishing ultrasound beacons! I'm gathering some info about these beacons - so I've been walking around with my phone to fish some interesting sounds. Some examples include a beacon in a starbucks close to montparnasse : interesting central frequency at 20khz, and two side freqs at +-250hz The spectrogram shows it continues:
On another hand, I was in a coffee (called Ajami possibly), and I found a nice peak around 20khz:
A small proof of concept has been done so far. Some extra items are missing:
Shifting back to the 32u4 to benefit from the SD card and standalone FFT -> I need to have the ADC go faster for the acquisition, and to get the data compressed a way or another for longer acquisitions, but that should be feasible.
Developing a more simple GUI for the OLED screen: I should be able to know when the signals are being recorded, and, if possible, how the beacon is encoded.
Saving: when a signal is detected, the data should be recorded to the onboard SD card.
Mapping: with a timestamped data, I'd be able to map on a geographical map the location of the beacons, and why not have a collaborative map ?
Spoofing: when a beacon is detected, why not send a louder signal saying "HEY GUYS I KNOW YOU'RE LISTENING SO HEY THERE!", to outpower the actual beacon?
Playing around with the STM32, I wanted to test if the electret was
capable of listening to 18 to 20kHz sounds... Indeed! I've created 18kHz and 20kHz wav files on the repo, then played those and tested with the FFT capacity: the two peaks are very well resolved ! You can see the jupyter notebook for the experiment.
The peaks correspond the beacons are using. This proves that it'd definitely be possible to have a small device tracking the ultrasound beacons on the road!
Next steps.. having a better looking GUI on the small system OLED =)
The
first step is to get a microcontroller, a micro, and we're set. The setup looks like this:
First,
I've tried with a 32u4.. the limited RAM onboard was problematic. So I
took something a bit more powerfull - a feather WICED - which I used to
acquire bits of 160ms (see below) - ~ 8k points at 50kHz.
Now, let's try and see if we can isolate a clear 8kHz signal : indeed - documented on github !
The only thing that bugs me is that the FFT libs for the WICED are not yet well developped .. si I can't definitely use the WICED as a stand alone so far. Dang.
kelu124
Now: just need to find out where to accelerate the spi =)
The only thing that bugs me is that the FFT libs for the WICED are not yet well developped .. si I can't definitely use the WICED as a stand alone so far. Dang.
That's freaky. The worst thing is that having a cross tracking device can also help you in being tracked even though you're using anonymizing tools:
