Hacker News new | past | comments | ask | show | jobs | submit login
The Hummingbird Clock: date videos by background mains hum (hummingbirdclock.info)
537 points by wodow on Oct 9, 2022 | hide | past | favorite | 148 comments



There's an artist called Øystein Wyller Odden that works with mains hum. I find it pretty fascinating:

Kraftbalanse is a musical translation of the hum from the mains; i. e. the frequency of the alternating current. The piece is based on the fact that this frequency is not stable, it fluctuates subtly around 50 Hz as a direct result of supply and demand in the power market.

The composition consists of a self-resonating piano that is tuned to resonate on 50 hz and overtones of 50 hz (100 Hz, 150 Hz, 200 Hz etc.) The piano is fitted with vibration-elements – transducers – plugged directly into the electrical grid, causing the resonance and timbre of the piano to change with the fluctuations on the power market.

The piano is accompanied by a string octet. The musicians are equipped with voltmeters that measure the frequency of the current in real time, as well as a score of instructions on how to respond to changes in this frequency.

https://vimeo.com/370554138


I was expecting one of those music YouTubes where they take some sample and process it so much that it doesn’t really matter what the sample even was. But I’m pleasantly shocked (ha!)

They really evoke the feeling of electricity. It reminded me of the soundtrack to the Chernobyl miniseries. It’s… powerful and eerie and calmly, effortlessly, sluggishly intimidating.


This is awesome. I've been kicking around the idea of a notation/composition system that uses external forces (temperature, moon phase, time, etc) as live input but prior research is a bit hard to come by.

Feeding signal into e.g. a guitar pedal is easy enough, but mutating an entire orchestra in real time is a different beast. If anyone knows of similar works I'd be grateful for links.


The ease of feeding it into something like an orchestra varies depending on the granularity of your input. Moon phase for example can be divided into four weekly states which can be rehearsed prior to the event, whereas ambient humidity for example might change much more.

The only thing that comes to my mind as prior art would be the Tierkreis by Karlheinz Stockhausen, which has a variable order of composite parts depending on the time of year. I'm sure that's the world of music you could look towards to find further inspiration.

Please share any results you come up with here, it sounds very interesting indeed!


The Weather Warlock at Epsilon Spires in Brattleboro, Vermont uses weather sensors to drive an analog audio synthesizer:

https://www.epsilonspires.org/permanent-installations

There are a few other Weather Warlock installations around the USA.


Ugh. Login to watch. Hard no. Too bad, seems interesting.


Odd, I was able to watch it without logging in. (It is!)


I wonder if it's based on geolocation. I'm being forced to log in, in the UK. I've noticed sites increasingly doing this to UK audiences, despite the fact that, iiuc, the draconian age-gate legislation hasn't actually landed yet.


I've been wondering this, didn't they age restrict porn years ago? How do you prove your age?


They've just been talking about it to get us used to the idea, but the plans have been shelved until they can answer that question.


I thought you still had to phone your ISP and say "I want the porn, please"?


I'm in the UK and don't have to login to view.


I saw the "unrated video, login to watch" screen for a split second then it went away by itself and I could watch the video.


youtube-dl https://vimeo.com/370554138

works for me!


Umm, I simply ignored the "log in with Google" popup and clicked play. Does that not work for you?


> Video is not rated. Log in to watch.


Same.


You can also hear it around 3:00 and towards the end, while the artist talks about the work: https://youtu.be/I_BanULj8wk


The record is being released in just a few days, for those that enjoyed it: https://www.sofamusic.no/album/kraftbalanse


> Video is not rated. Log in to watch.

Is there a way to watch the video without logging in? I tried multiple browsers and still hit the login wall.


Weird, I'm logged out and lets me play it. On related videos I got that prompt, refreshing the page solved it.


Why is Vimeo always the chosen venue for projects like this? Frankly this might be the most grounded thing I've ever seen on Vimeo.


More control over your content with the caveat being you have to pay them for the service.


Also probably fewer people in the comments telling you your art project that breaks eggs whenever the stock market falls is "stupid and gay".


This 5 minute video from Tom Scott on YouTube is a good overview without that annoying hum:

https://www.youtube.com/watch?v=e0elNU0iOMY


It's still there, just much quieter


I much prefer mains hum to Tom Scott. Call me crazy, I guess.


Tom Scott did a video about this topic in 2021: https://www.youtube.com/watch?v=e0elNU0iOMY


In 2018, we went with school to the national forensics institute where they (among other things) presented their research on this. Not sure if it's also online somewhere; their website is nfi.nl


That's weird.

I came here to check that someone had linked to that video, but in my mind it feels a lot longer ago than Dec 2021. I feel like that video should be 3 or 4 years old.

It's doubly weird because that normally happens the other way around, where I think something happened about a year ago, and it was actually 3 or 4 years ago. Or I think something happened 5 years ago, and it was actually 10-12.


People seem to be reporting anecdotally that the pandemic has messed with their sense of time. I wonder if there are any serious research projects underway to study this phenomenon.


Maybe you should submit the video to The Hummingbird Clock and find out for sure! hehe


I find it personally hard to date stuff that came out or that I watched during the pandemic. To me it felt like I consumed 4 or 5 years of content in the span of each year. My sense of time is completely twisted.


I would have said 3-4 years as well, and I would have been certain beyond any doubt that it was more than 2 years. Crazy.


I too could swear I saw that video much earlier. But I guess it was actually this one: https://youtu.be/bij-JjzCa7o


I was excited to possibly see this provided as an open source tool or at least an automated service, but the site offers a form to request dating, which looks like a manual process. That's not really making it available to everyone.


It's super fishy. Why should I trust them to process potentially sensitive material? If you want to level the playing field, just put the collected data out in the open so anyone can download it and correlate any video they like with it, no need to upload the video to Mr. Anonymous and hope he's being truthful about his pure and noble intentions to support human rights.


I suppose you could calibrate Mr. Hum with some known-date video material, but I agree that the data should simply be made public. Further, even if the data is released, you would be assuming that Mr. Hum captured his electrical data with the correct time attached to it. So, hopefully, he's either got a Cesium clock connected to his capture device or he's using NTP, or the like. This would further assume that he has maintained this same data capture regimen continuously since he began his capture in 2016.

This definitely raises some questions and shows that its likely not something that should be left to entities outside of standards-setting agencies (that could offer some assurances and certifications).


Yeah it's kind of surprising that nobody has set up a properly open one with downloadable history. This "fill in a form" one is lame.


Warning: site autoplays a loud (presumably 50Hz) buzz...


@dang Maybe you could add a [Loud sound] tag on the title. I clicked on it without knowing and blasted a huge noise from external speakers at midnight.

Wife and neighbors are not happy.


Why do (some?) browsers still let JS autoplay audio but not plain <audio> elements?


Didn't do it for me. To test my headphones and ears, I tested a 50hz waveform on Youtube and could hear it.


Didn't work for me at first, but I moved on the website pages and it did.


According to http://hummingbirdclock.info/static/js/hum.js the base frequency is 200 Hz, but has a gain of 4, which seems to push it more into the higher harmonic range as the waveform is clipped:

    // hum.js
    
    // toggle html audio on/off
    // with cookie to maintain state
    // and use html5 audio synthesis
    // ios html5 audio must be triggered
    // call init_hum() after DOM loaded
    
    // 0. create audio context
    // 1. create oscillator, define
    // 2. connect oscillator to gain
    // 3. create gain
    // 4. connect gain to output
    // 5. start oscillator (on click)
    
    var audio;
    var audio_context;
    var control;
    var vco, vca;
    var hum_delta = 1;
    var hum_base = 200;
    var hum_min = 100;
    var hum_max = 400;
    
    /* init */
    
    function init_hum() {
        audio = get_cookie("audio");
        console.log("audio = " + audio);
        audio_context = get_audio_context();
        console.log("audio_context = " + audio_context);
        control = get_control();
        if (audio_context != "false") {
            if (audio != "off") {
                set_hum();
                hum_on();
            }
        } else {
            set_cookie("audio", "off");
        }
    }
    
    function get_audio_context() {
        var which_audio_context = window.AudioContext || // default
                                  window.webkitAudioContext || // safari
                                  false;
        this_audio_context = new which_audio_context;
        return this_audio_context;
    }
    
    function get_control() {
        var this_control = document.getElementById("control");
        this_control.addEventListener("click", hum_on_off);
        return this_control;
    }
    
    function set_hum() {
        vco = audio_context.createOscillator();
        vco.type = 'sine';
        vco.frequency.value = hum_base;
        vca = audio_context.createGain();
        vca.gain.value = 4.0;
        vco.connect(vca);
        vca.connect(audio_context.destination);            
    }
    
    /* on off */
    
    function hum_on() {
        set_hum();
        vco.start(0); 
        control.innerHTML="&times;";
        set_cookie("audio", "on");
        audio = get_cookie("audio");
        console.log("audio = " + audio);
    }
    
    function hum_off() {
        vco.stop(0); 
        control.innerHTML="+";
        set_cookie("audio", "off");
        audio = get_cookie("audio");
        console.log("audio = " + audio);
        cleanup();
    }
    
    function hum_on_off() {
        if (audio == "off")
            hum_on();
        else
            hum_off();
    }
    
    function cleanup() {
        vco.disconnect(0); 
    }
    
    /* cookies */
    
    function set_cookie(cname, cvalue) {
        document.cookie = cname + "=" + cvalue;
    }
    
    function get_cookie(cname) {
        var name = cname + "=";
        var ca = document.cookie.split(';');
        for(var i = 0; i <ca.length; i++) {
            var c = ca[i];
            while (c.charAt(0)==' ')
                c = c.substring(1);
            if (c.indexOf(name) == 0)
                return c.substring(name.length,c.length);
        }
        return "";
    }
    
    function check_cookie(cname) {
        if (getCookie(cname) != "")
            return true;
        else
            return false;
    }
The audio spectrum analyzer on my phone (Spectroid) records it most strongly around 1000 Hz.


At 50Hz it would be inaudible on most computers.

https://szynalski.com/tone#50,v1


True, but cheating a bit with a triangle wave would solve it thanks to the harmonics.


This is widely known and allegedly used by law enforcement forensics etc.

I have often thought about building a SaaS service that strips or adds the correct hum and harmonics to videos if you wanted to plant some false evidence or make a good deepfake. If it failed to make much cash I could put it on github for free.

I have never quite found the time and openly doing that kind of thing is probably likely to attract undesirable attention.


>If it failed to make much cash I could put it on github for free.

Silk road was closed in 2013


Markets are alive and well. Not as prolific, but not dead.

Not really sure what your point is, though? They don’t typically have audio recordings so not sure I see the relevance.


You advertise it as a tool, but not the 'nefarious' uses. Even uTorrent, for example sells a pro version.


I'm surprised when it says "the exact same buzz can be heard at the exact same moment nationwide from Land's End to John o' Groats", I didn't realise it would be the same across the UK.

I played a little with a transformer to take 230VAC -> 12VAC, and then using a potentiometer to feed the waveform into a soundcard. But I'm not sure the sample rate of the soundcard was high enough to accurately measure the times between 0 crossings.

Photo of the very messy setup - https://www.anfractuosity.com/files/hum.jpg


Explicitly watching for zero crossing seems like a rather limiting approach to quantifying a sine wave. Even then, at 48Khz you should get 400 samples between zero crossings, so you should do better than 0.5% accuracy.

There's a lot of information held within the time series data of an audio signal, and something like an FFT can tap into it. For a 60Hz signal the Nyquist frequency is 120Hz, which is well below any audio interface. It seems like accuracy of any processing would be limited by a crystal oscillator somewhere more than anything, and perhaps any unintentional signal aliasing if there's no low pass filter before the audio interface's ADC.


That's a good point re. the crystal oscillator.

Maybe I'm wrong re. the sample rate then, I found this paper - http://www.forensic.to/ENF%20processed.pdf , where they mentioned 8kHz as the sample rate.

I think what I was thinking is commercial systems such as https://synectic.co.uk/product/sd037-frequency-monitor-html/ mention specs such as 'The frequency monitor units are supplied pre-calibrated to an accuracy of +/-0.0001Hz'


I was mainly responding to the hobby experiment of connecting a 12V transformer to an audio input on a computer.

Even then, it seems like 8Khz sampling would be plenty to fingerprint a 60Hz signal. 8Khz sampling can capture information up to 4Khz, per the Nyquist frequency.

That frequency monitor spec sheet mentions 90ms response and filtering, which would imply it averages information across 5 periods. Also, because it's a purpose built device, it's likely not periodically sampling the waveform with an ADC, but rather has a dedicated zero-crossing detection circuit that feeds into a capture input on a digital timer with at least microsecond resolution.


Well 60/50Hz isn't the only relevant frequency here. Lets say you plug in ethernet-over-power (available off the shelf). You'd need a very fast interface to notice that.


I can't imagine folks' ethernet-over-power signals are propagating across the national power grid, and even if they were, they aren't getting picked up in any audio streams.


> I didn't realise it would be the same across the UK.

It has to be, that's the definition of a power grid. The relative phase can have minute variations and you get massive power flows as a result, which result in the frequency across the whole grid being the same.

I believe the whole UK is a single grid. The US is divided into three: The eastern interconnect, the western interconnect, and Texas. (Insert meme here.) There are a few DC-links that allow power to flow between them, but due to the expense of rectifying and inverting all that power (necessary for the frequency and phase to vary), they are rare. It's much cheaper to connect directly at AC, even though this requires that everyone on the system run at the same frequency.

Amusingly, this also means that if you could (theoretically arrange enough long shafts and bearings and neglect the curvature of the earth and) orient all the generators on the grid in a line, a single shaft could run through all of them, and the shaft would not twist -- they all spin in precise synchrony since they are, effectively, a single machine.

When a new generator is brought on-line, it has to be phase-matched to the grid. These days there are probably fancy PLCs to handle it, but the simplest instrument is called a synchroscope, which is a motor driven by the phase relationship between two AC waves. You can find these on YouTube, they're pretty neat. Basically if the isolated generator is running slower than the grid, the synchroscope needle spins one way, and if it's running faster, the needle spins the other way. You adjust the throttle on the generator until it's running _ever so slightly_ faster, like 60.00Hz on the grid and 60.10Hz on the isolated generator, (50.xx in the UK) which will make the 'scope revolve once per ten seconds. Then you wait until the needle is pointing straight up which indicates zero phase shift, and slam the switches to connect it to the grid, and that's that!

If you did it just right, there's a minimal shudder and the generator now begins sending power into the grid, its frequency constrained to 60.00Hz and the extra throttle showing up as positive power flow. If you did it wrong, there's a tremendous amount of torque exerted on the rotating machinery by the grid power, which may simply force it into compliance, or in more extreme cases, may shred millions of dollars of equipment in an instant. Bigger synch setups have lockouts that won't even allow the button to be pressed unless the phase is _pretty close_. The most colossal kind of cockup is when someone wired the synchroscope wrong, so the rotation still works (it has to, that's impossible to screw up) but in-phase would be represented by some needle angle other than straight up. Joining the grid 120 degrees out would be a career-limiting event for someone.


> a single shaft could run through all of them, and the shaft would not twist

I think this assumes that every AC generator has the same number of poles per revolution.

> If you did it wrong, there's a tremendous amount of torque exerted on the rotating machinery by the grid power, which may simply force it into compliance, or in more extreme cases, may shred millions of dollars of equipment in an instant.

Indeed. When I read news articles during the Texas blackout of 2021, the authorities were watching in horror the system frequency dropping and saying they had to shed load before the generators experienced permanent damage. It took me several days and reading Reddit comments to understand that the entire AC grid is one big synchronized machine, and if generators fall out of frequency or phase with each other, they will drive each other instead of the load, thus damaging the generators.


> I think this assumes that every AC generator has the same number of poles per revolution.

Oh yeah. An oversimplification, and one that need not exist (all the generators _could_ be wound with the same number of poles), and I think would've complicated the point beyond GP's level. But yes, you are right.

> and if generators fall out of frequency or phase with each other, they will drive each other instead of the load, thus damaging the generators.

That's one reason, another is that some of the prime movers simply can't exert full torque at significantly-wrong RPM. I'm not a turbine engineer but it's my understanding that even in a standalone situation with no grid at all, if you apply full power to a turbine and then slow the rotor to a stop, somewhere before you even reach a full stop, some of the blades will break, because they can only handle that gas pressure when moving and experiencing the rotating flows of a running machine. The point where that occurs depends on the engineering of each individual machine, but a highly-optimized design may have a tolerable operation range that is narrow indeed. (When a new unit is started up, it does so unloaded, so it only takes a relative whisper of fuel to get it moving. Only once synchronized and loaded do they really hit the gas.)

In my mind this seems very similar to the back-EMF developed by a rotating motor, and the phenomenon of locked-rotor current burning out the windings when the shaft isn't allowed to turn at some appreciable fraction of rated RPM. Just the back-pressure is fluid rather than voltage. Any actual turbine engineers here?


Basic Fourier theory from the 1800s explains that your sound card is good enough to record the changes in frequency.

Chances are internally it uses a much higher sampling rate and digitally downsamples.


I don’t see how this can be used to “authenticate” a video. If I’m already editing the video anyway, could you filter the original hum out and replace it with the correct frequency for the time you wanted to fake? There’s got to be a margin of error anyway, so would seem very possible to fake unless I’m missing something.


You could. The point is to disprove the validity of the video. Correct signature is not a proof the video is valid but incorrect one could be proof that it is not.

In practice, most people who manufactured videos before very recently were simply unaware of this possibility. I learned about it many years ago but I still forgot about it and probably would not try to do anything about it if I was manufacturing a video.


i am curious how/where you learned about this years ago if it was not common knowledge until recently?


I learned about it at least a few years ago watching a history of video game speedrunning video (don't remember the specifics from the video or which video it is, sadly). At least 6 years ago, the speedrunning community used the mains hum to prove that a video was spliced together, rather than a continuous stream.

Also, clearly the creator of this page has been recording the mains hum for at least 6 years, so presumably they also knew about it.


Yes, in the same way you could leave fake fingerprints and DNA at a crime scene.

This was used in a court case in the UK where the defence claimed the police had tampered with a recording - editing together separate recordings into a single one to make their case. An academic 'expert' analysed the recording and corroborated the claimed recording times and the accused were found guilty.


Only if you know to do that. Nothing is ever 100%.


You can use it as evidence either way, especially evidence that something is _not_ genuine. It's not proof though.


That clock graph is fantastic.

I haven't let it run long enough to see if the minute and hour hands do the same line, but if so what a great way to show minute, hour, and day (24 hour clock) trends.


This would be good to use for syncing clips from multiple cameras in video editing packages


Pluraleyes software does this using the audio recorded by each camera:

https://www.maxon.net/en/red-giant/pluraleyes

If you filtered the audio to exclude all but the 50Hz/60Hz signal then it might also work based just on the power hum, but why make it's job more difficult?


I think this is useful in many ways, I hope one can get access or build one.


Filtering out 50Hz/60Hz noise as a band stop filter is a common audio post processing step, precisely because it's noise from the power system.

I wouldn't be surprised if there were phones that automatically applied this filter.


Of course, and a lot of microphones don't pick up that low anyway.

That's immaterial because of harmonics.

While the AC wave itself is nearly a perfect sinusoid, various devices introduce all sorts of nonlinear properties, and produce higher frequencies that are harmonics of the base powerline frequency, and thus vary right along with it. These are insidiously hard to remove. Your best bet would be to encode all the voices with a vocoder, ditch the background sounds entirely, and play the coded voices back. Which of course makes it obvious that the audio has been altered...

I suspect that similar artifacts will be visible in lights too, since the brightness of an incandescent or neon source varies with the AC wave. Some fluorescents too, though LEDs all have DC drivers so they'd make it more difficult.


Your point still stands but the noise here is effectively a higher frequency modulation of the base AC oscillation.

In other words, the base oscillation of 50 or 60 Hz is not stable, but instead varies a little above and a little below at higher rates.

The magnitude is significantly lower than the base mains frequency. So possibly not audible above any other noise in the recording, but present nonetheless.


These filters will take that into account since it never has exactly been 50/60Hz. It's just the expected mean.

Maybe it's present in subsequent harmonics (e.g., 100/150/200Hz for 50Hz) as a systematic deviation from additive white noise. But, this would be difficult in practice to reliably isolate these signals via some form of component analysis.


How far do I or can I transmit into the grid? How far do load changes from my house transfer into the grid? Surely my neighbors could detect 1us spikes not big enough to pop mains at say 800hz?


There were some small, low power AM radio stations that "transmitted" via power grid in mountain towns in North Carolina, a while back. I've been told that they had real trouble getting through transformers and they were using the main distribution lines and broadcast AM frequencies. That's with decent levels of power applied.

I expect the incidental signals your usage generates are going to be hard to detect past your power meter and almost certainly un-resolvable past the distribution step down transformer.


Using ethernet-over-powerlines within my house, different circuits within the same building have trouble talking to each other, so I wouldn’t recommend that for cross-building communications...


I remember consumer devices from the 80s that allowed transmitting audio using mains. They were already outdated by the time I saw them in the 2000s.

Me and some friends tested them and tried transmitting to a neighbours house (connected to the same electrical post) and had no success at all. There was just too much noise, coming from other houses. However with digital technology I'm sure you can get a much better range.


Things like HomePlug GreenPHY are used for smart meters. (Also electric car charging for some inexplicable reason) I don't know how they're typically deployed but I'd think at least a few miles? It wouldn't make much sense to use if you'd have to have the receiver really close - just use wireless.


How does this auto-play sound without user interaction?


Synthesized sound (AudioContext) works a bit differently from playing videos/sound files in the web permissions model, for whatever reason. IIRC Chrome and Firefox both treat various forms of interaction as an OK to synthesize audio, so your browser probably decided you had met that criteria.

Chrome has a list of special websites that are given that access automatically, though I doubt this website is on that list...


Very interesting. I'm using Firefox and it started playing immediately when I switched to the tab. Thank you for the details!


It didn't for me on Chrome on OS X. I had to click one of the links before the sound played.


I wonder how much of this historic data could be crowdsourced from recordings of radio shows, VHS taped newscasts etc.

Audio dendrology, if you will.


How difficult would it be to defeat that completely or fake a different date? If people can find subtle noise that fools computer vision algorithms, can something similar not be applied to this?

A band-pass filter might be a little simplistic because it's never 100% attenuated and the harmonics persist, so given a long enough sample you would probably be able to correlate. But perhaps more advanced techniques exist, if you know exactly what the other party is looking for?


Does anyone know what resolution of frequency data I'd need to be recording to be able to date a recording myself?

After reading this, I remembered that I have an IoTaWatt throwing mains voltage and frequency into Influx every second or so, but I've got no idea if this would be high enough resolution to match against a recording (or even how one would go about doing it).


Interesting. What effect if any will my home being entirely on off-grid inverter 100% of the time have on this? I use Solar/Commercial/Generator -> DC Power Supplies / Charge Controllers -> Inverter -> Non-Resistive loads. My inverters are entirely isolated from commercial power. Can someone perhaps tell the make/model of my inverter?


> Can someone perhaps tell the make/model of my inverter?

Possibly, if someone went through and fingerprinted all the inverters.

If it's a plain simple inverter with no grid-tie capability (i.e. one that might be installed in a car), then it might not have any distinctive characteristics, particularly under load. Just plain simple 60Hz all the time. Lightly loaded, some inverters have a power-save mode where they alter the waveform to that it produces less wasted power in the MOSFET stage, but I don't know if that would show up as a distinctive sound.

If it's a grid-tie-capable inverter, even if you're not using it in that mode, it probably has some distinctive characteristics. These units try to wibble their frequency around a little bit to see if anything "pushes back", as a way of ascertaining whether there's grid power present. They may also shift their frequency as a way of communicating with AC-tied solar panels on microinverters, analogous to how the grid's own frequency varies with load, but not so smooth.

All of those behaviors would likely produce pretty distinctive sounds, and they may vary per model, and certainly per manufacturer.


I guess it depends on how far you are from mains electricity infrastructure.


How many videos, outside fixed webcams, come from devices plugged into the wall? I venture that 99.99% of videos where the timestamp is questionable come from battery-powered devices. As for mains hum, portable devices pick up so many hums from so many sources that i doubt much useful mains hum is ever recorded.


Mains hum is by far the most powerful source of electromagnetic radiation in most indoor places. And the device doesn't have to be connected to the mains to pick up EMI from it - turn up the gain on any battery-powered preamp and you'll hear it.


If you follow the "How it works" link:

> Digital recordings almost always have mains hum on them, either because the device was plugged in to the mains or because it inducts it off nearby cables, lights and appliances in a room.


This is why guitars often use a humbucker pickup to cancel the EMI from the mains.

https://en.wikipedia.org/wiki/Humbucker


You don't pick up the hum from being plugged into the mains. Nearby equipment generates sound at the frequency (and harmonics) which are picked up by the microphone. Various other equipment is sensitive to the EMF. As others mentioned its pervasive indoors and near power lines.


I've heard this fairy tale for decades, and quite frankly, I'm putting it into the same camp as claims of undeleting data from magnetic media or reading license plates from space --- By that I mean that such things may be theoretically possible, but from a pragmatic and practical perspective, they might as well be complete science fiction.

Please, can anyone find a single concrete example of ENF working? All the information I can find about the UK's supposed widespread and automated use of it starts to evaporate into the same kind of conjectural nonsense as polygraph tests.



I have occasionally wondered if the earths magnetic field signature, which is present in solidified rock, could be read from a magnetic tape on which something has been recorded. Never had the analogue skills needed to investigate.


The sound is very aggravating.


Exactly. Although it _is_ interesting, was not expecting auto-play.


This is fascinating. Does anyone know if this phenomenon is unique to Britain? I wonder if this could be replicated in other countries.


It should work any place that has an AC power grid with accurate records being kept by someone. Isolated off-grid settlements running their own generation would probably not have that. Isolated off-grid solar, ditto.

The UK is all one grid, Ireland (the island) is another, there's a Nordic grid, a Baltic grid, and one for the whole rest of continental Europe (plus Morocco and Tunisia). The US has two plus a separate one for Texas, Canada uses the same two plus a separate one for Quebec (the jokes just make themselves), and Japan weirdly is cut in half, with a 60Hz grid in the south and a 50Hz grid in the north.

https://en.m.wikipedia.org/wiki/Wide_area_synchronous_grid


Seems like you wouldn’t need to actually record and store the mains hum… you could just extract it from other videos with known datestamps.


Definitely wouldn't work. You'd be matching two very noisy signals instead of one clean and one noisy one (which already requires a long time period to match), and it would be really hard to get enough videos to cover all time and then why would you bother when recording the hum directly is easier and better?


And then add it to your newly made black and white video tape. Forgery 2022.


Except for the 10 other techniques we don't know about. Did you know about this before the HN post (or original post from 2021)? I sure didn't, which means there are other tricks hiding out there.


Exactly. Making good fakes is hard. Making good future-proof fakes is probably impossible.

But what you can do instead is make a half-decent fake, then accuse other things of being fakes, accuse your own fake of being fake with softball arguments, fake evidence for the other side, and in general throw up a wall of noise and hope it benefits you (the underdog propagandist's way).

Or just not talk about the evidence, and convince people to not look at or think about the evidence (the hegemonic propagandist's way).

That the truth can be found doesn't mean you can convince people of the truth.


I didn't.


you might have not, but many did


For this statement to make sense, the reader would need to assume that I think I'm the only one who didn't know this. Since I stated I didn't know this in my actual post, and since I assume you read it, I honestly wonder what motivated you to post this. If you're trying to take me down a peg, it makes no sense since I stated I didn't know. If you're trying to remind me that I didn't know and others did, well, the article that I literally commented on wouldn't exist if that was false. So again, why did you say this?


I think you're overcomplicating things. This is not some personal attack, nor it "only makes sense" if the reader assumes that "that [you] think [you're] the only one who didn't know this".

I'll break my intent down to excruciating detail:

Your comment can be understood as a strong claim that you one can't make a forgery by adjusting for such detections (as the grantparent suggested), because there would be "10 other techniques we don't know about", and to support this you give the case of you not knowing about this technique as an example.

In short, I read it as "Since this existed and I didn't know about it, who knows how many other such techniques are in use that people don't know about. One can't make a forgery then, because he'd run afoul of one he doesn't know about".

Well, my point was that it's not like those techniques are some "industry secret" that nobody outside a select group can know about. They're still publicly available information (like this was). They're just unknown to most who didn't have reason to give this space much thought or researched it (like most nonetheless common practices in any field which laymen won't typically know about, but can always find out).

So, to sum this up, I read your claim to the parent as "you'll always miss some techniques, as you can't know them all, like I didn't know about this one", and my claim is "nah, you can pretty much find what all techniques in use are".


Would be interesting to generalize this technique to surmise (independently) the location of, e.g. the Osama bin Laden videos.


Interesting topic.

I tried to mimic that sound design using Glicol but winded up getting something different.

For those who are interested:

1. go to (https://glicol.org)

2. run the following code:

// ----------------

o: sin ~freq >> mul 0.5;

~freq: sin 0.2 >> mul ~range >> add 50;

~range: sin 0.4 >> mul 1 >> add 2;

// ----------------

3. tweak the numbers to get different sound


That's a really fat side-channel if you can extract it with minimal noise.

You should be able to extract time and also location based up the waveform.

And I think that it should also be possible to use baseline expectations for known channels to reduce the noise of the extract.


Super interesting. I notice that this is only about the UK, but I assume the same principle should work in every country, right? Are there any similar projects doing this for other countries?


I have never heard mains hum as wobbly as on this site. It seems exaggerated.


It's not real. It's obviously too high pitched.


Just curious, are there any software hacks (not audio/video at all) that use any aspect of mains hum for say info leakage? (Could be impossible.)


This is brilliant. Anything that helps to disambiguate the authenic from the inauthentic is most welcome from my perspective.


Interesting how foreign this sounds to me, but how familiar it is to millions. 60 Hz must sound the same way to Britons.


Fun fact: You can hear the harmonics of the grid and IP switching if you open the site in two separate browser windows.


Fun fact: The source contains: rand = (Math.floor(Math.random() * 3) - 1); vco.frequency.value += rand * hum_delta;


I bet there was a movie or a documentary around this if anyone have the link please share it.


How can the whole nation's grid have the same variations everywhere?


Has anyone tied this technique into deepfake analysis/synthesis?


Time to get your own isolated grid huh? Maybe with solar and inverters?


An on-line UPS like those used in datacenters or music studios is probably enough. They basically have rectifiers and inverters in them.

Some of them have phase-locked loops to sync the output frequency with the mains, but probably won't stray as far from 50Hz (or 60Hz) as the mains themselves seem to do. If they do, unplugging the mains and using the battery is probably enough to get pure 50Hz back!


> An on-line UPS like those used in datacenters or music studios is probably enough.

Only if you don’t have mains power anywhere near the recording device. You don’t need to be plugged into mains power to have it be present in a recording.


True! And that applies for every case. Even if you have 100% isolated mains with solar or battery, you'll get 50hz from neighbours (especially apartments) and power lines close by.

I remember having this issue when working in a recording studio during college, even when we switched to battery and turn the mains off, we'd still get hum in some guitars.

Funny enough I was recently reading an interview with producer Michael Beinhorn and he mentioned having some "mystery EMF/RFI event" happening in New York around 1997, coming from a specific block, and he had to relocate a recording session to Los Angeles because of how strong it was interfering with the guitar amps and other equipment [1].

[1] https://gearspace.com/board/interviews/1385579-interview-mic...


I really wish this story included which half block area it was strongest.


That would be very interesting to know! Especially considering the two studios he mentioned aren't exactly close to each other:

https://goo.gl/maps/MLw3twvRi9GumYeX8


You'd need to make sure there is no grid nearby. If you were completely off-grid and away from power lines or anything then you'd have no signal at the mains frequency. Certainly possible. Probably easier to just use a filter to mask the frequency component.


Probably better to run a decoy signal (or two) in the background with similar characteristics.


To what end? What do you want to obfuscate by doing this?



That's not really an answer beyond "Because FUD".

FWIW, I've been an ACLU member for quite some time.


FUD is a perfectly valid precaution where there's reason to be fear, uncertainty, and doubt.

FUD is only a bad thing where there's no motive, precedent, and reason for those three - which is hardly the case here.

Like, "This new roommate of yours might be a killer. Sure looks like one. And who can trust a man that doesn't have a Facebook page?", is bad FUD.

But "Two ex-girfriends of this new roommate of yours have been found murdered in the woods, each a few years after they split. And I saw rope, duct tape, a set of hunting knives, plastic gloves, and a bottle of chloroform in the trunk of his car. Plus, when he got drunk he went on and on about how he likes this Dahmer tv-series. I think he might be a serial killer and you should be careful around him", is very reasonable FUD.


This is so fucking cool.


You may also like to read about:

- Yellow dots in printers.

- Eurion constellation.

- EXIF


Not just exif but camera sensor noise artefacts eg https://ia.binghamton.edu/publication/FridrichPDF/full_paper...


What I don't understand is how much technique assumes of the local oscillators on the recording device, which also vary.


This is absolutely fascinating.


It's super shitty to just autoplay that sounds as soon as you open the page. Firefox on Android.


Yet another reason that the world’s electric grids should be replaced with DC power.

AC power was only a good idea when we didn’t know how to make good DC-to-DC voltage converters.

https://www.electricaltechnology.org/2020/06/advantages-of-h...


Or when we decided to live more than 100m away from the power plant? DC doesn’t travel well.


Incorrect. DC power travels better than AC at the same voltage. The problem, historically, is that we didn’t have good technology for changing DC voltage. That has changed in modern times.





Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: