Hacker News Reader

Comment by pasta | Hacker News Reader

Take a look at Chirp: http://www.chirp.io/

Edit: In Firefox it seems this website is constantly playing audio. I think it is using a WebSocket to do this. Since I don't hear a thing I'm not sure what is going on here.

View on HN

Since I don't hear a thing I'm not sure what is going on here.

Click on the gem and the web page will play a "chirp," which is loud and beepy and not ultrasonic or secret at all. It's an analog modem type system, except probably way slower than 300 or even 110 baud. Everything old is new again, I guess.

In any case, the point is: the Chirp system is distinct from the supposedly-possible secret systems the Ars story is fretting over.

I've seen several articles worrying about the use and abuse of data transmission via secret ultrasound, but I've never seen any claims that named names or that offered a way for readers to verify or check anything.

I don't think you can count on the average laptop/phone speakers, mics, and digital audio systems to deal with ultrasound.

If a computer's audio drivers and audio hardware are set up to play back at standard CD/DVD sample rates (44KHz or 48KHz), then ultrasonics are not happening there, period. The same principle applies to inputs.

Even if all the hardware and software on two devices is capable of and configured for producing/receiving ultrasonic frequencies, it seems pretty likely that loud ultrasonic screeches played on a 50-cent cell-phone speaker would produce audible harmonics, and the jig would be up.

I expect secret ultrasonic data transmission over commodity systems is the sort of thing a well funded state organization might turn to for a targeted operation. Because it might work under exactly the right circumstances, but it would be fiddly and terrible for general use.

I think that most people (not teenagers) cannot hear past 15 kHz, especially with other sounds playing. This is reproducible at standard sampling rates you mention.

Harmonics are integer multiples of frequencies. Harmonics of an ultrasonic signal are even more ultrasonic. If there's any aliasing, those harmonics might map back to audible frequencies. However, no matter how cheap a speaker is, no aliasing happens. Perhaps there's another reason why an ultrasonic waveform might be audible when played through a cheap speaker, but not because of harmonics, right?

It's not necessarily a function of the speaker but of the low-pass filter on the output. Unfiltered samples played back at 48 kHz will mirror the audio content at 0-24 kHz to 48-24 kHz, and repeat that pattern at every 48 kHz multiple. (This is not quite exact because DACs emit step functions, not delta functions, and slew rate comes into play, but it is a rough approximation.) This happens even if the sampled signal is a sine wave.

So, any output stage after a DAC necessarily incorporates a low-pass filter to eliminate the aliasing artifacts. A cheap low-pass filter will allow these aliasing artifacts through. You can simulate this e.g. by digitally upsampling a, say, 8 kHz sampled signal to 48 kHz without any interpolation.

Harmonics of an ultrasonic signal are even more ultrasonic.

You're right. I was thinking of resonances. Even in that case, I might be wrong.

Nowadays many standard cheap embedded audio chips included in phones and computers are capable of 92/96khz sample rate - so they can easily drive 40+khz ultrasonics.

Are there other devices besides the audio driver that could produce ultrasonic sounds when driven in a certain way?

I was curious about this too. My EE (particularly RF) studies didn't go far enough for me to chime in.

Is there power circuitry, etc that could likely (a) generate inaudible frequencies that overlap with mass-market microphone capability & (b) would be sufficiently controllable (albeit likely indirectly) for data transfer?

> If a computer's audio drivers and audio hardware are set up to play back at standard CD/DVD sample rates (44KHz or 48KHz), then ultrasonics are not happening there, period. The same principle applies to inputs.

What do you mean? All you need to do is sample the received audio at 44.1 kHz. Android at least has a low-level API which allows you to customize sampling rate.

According to Nyquist's work, you need a sample rate of 2f to represent a frequency f.

So with 44.1 kHz you could only represent audio up to 22 kHz. There are kids who can hear at these frequencies.

I am aware of the Nyquist rate. You're talking about very rare edge cases though. I believe the cut-off for an average young human is 20 kHz.

Yeah, I get that, just saying, if we are going to worry about people with disabilities, I believe that we should care about the psychoacoustic wellbeing of "edge case" humans who can hear ultrasonic tones.

So you're comparing disabilities, like not being able to walk or see, to a slightly annoying tone? Of course, it also depends on the estimated number of "edge cases" out there.

And even then, you need a few kHz bandwidth for the low-pass filter. Those high frequencies are greatly attenuated.

Perhaps Chirp is what Silverfish metamorphosed to?

Reading the blurb on their site, they state they're the only people doing this, oh, and they reference India too (something that happened a bit on Silverfish's site). Coincidence?

Reply on news.ycombinator.com