Is this protocol significantly different from what 56K modems do over an analog phone line? How would the bandwidth compare to taking the 56K modem protocol and upping the frequency to whatever band is used here?
It would be fun to try this with a 96kHz audio card (audiophiles, this is your moment!) and trying this with a large band like 20kHz-40kHz (you would need an ultrasound transducer and amplifier, of course)
As an example, I just picked the first microphone from mouser who's datasheet includes ultrasonic spectral response. In this case it's a roughly 321 mm package with the acoustic port in the middle of the annular ground pad, likely requiring a hole/via below.
https://www.mouser.de/ProductDetail/PUI-Audio/AMM-2738-B-R?q...
It is not nearly as flat above 8 kHz as below, but it's between -35 and -28 dB (V/Pa) in the 20 to 80 kHz range. Nominal sensitivity is -38(+-1) dB (V/Pa) between 20 Hz and 8 kHz. What this means, is that a 24bit 192 kHz ADC with relatively flat response and no extra pre-amp should recover most of the SNR (specified at 64 dB(A) SNR 1kHz 94 dB @ 50cm).
I expect it to work well with a normal TRRS "smartphone" jack, just connecting the output via a decoupling capacitor to the sleeve and using the phantom power. The TRRS compatibility guess is based on the numbers given in this comment: https://www.epanorama.net/blog/2014/09/15/android-device-ext...
Assuming one could exploit 60 dB SNR over the 60 kHz wide band, the Shannon–Hartley theorem says there'd be a ceiling of 1.2 Mbit/s. Spectral efficiency would be 20 bit/s / Hz, and for the SNR range above ~10 dB, (dB SNR)/(bit/s /Hz) equals 3.
Your 20 kHz to 40 kHz band would actually give a nice 400 kbit/s with 60 dB SNR, or 200 kbit/s with 30 dB SNR. The ultrasonic speaker for these frequencies can be just a normal piezoelectric horn with either a crude class-A audio amplifier who's low-pass filter was suitably adjusted/removed, or a more efficient class-D device that can handle the relatively high frequencies. Considering the rather high lower end of the required band, a simple digitally-fed bipolar current source with pulse-density modulation fed to the tweeter should be sufficient.
Networking that doesn't use ethernet cables or Wi-Fi is always fascinating. I'm reminded of the way some labs in the pre-wifi days used infrared beacons and transceivers on laptops to create local wireless networks.
Oh gosh, I remember in college, playing with PCMCIA cards in laptops, and you'd have to kinda point them at each other. Can anyone confirm if that could be correct, or that I'm actually misremembering something?
Still you wouldnt be aiming whole computers with pcmcia cards in them, there was a dongle, most likely with a clip or velcro to slap in on the back of laptop screen.
Can't confirm because what we had instead was a task to implement a custom protocol over an RS-485 linear bus.
The device in question was only available on appointment, so we only had a few shots at this.
Eventually we went with adding a checksum and resending the whole "packet" (one byte) when it didn't match, because we couldn't get the damn thing to send stuff correctly 100% of the time.
Dave Jones from EEVBlog made a debunking video on uBeam. They apparently tried to to do ultrasonic wireless charging, which seems to fit your description. Here's the video, it's mostly about efficiency/coupling loss (beam forming has it's limits) and IIRC also about being potentially harmful for humans: https://www.youtube.com/watch?v=A8dqzVlhFkA
