2651: Air Gap
Air Gap |
Title text: You can still do powerline networking, but the bitrate does drop a little depending on the lightbulb warmup and cooldown delay. |
Explanation
This is another one of Randall's Tips, this time an Energy Tip. The comic conflates the concepts of computer network security and home electrical power safety to comical effect, resulting in a deeply impractical and ineffective proposed solution. In computer security, air-gapping is a measure used to secure sensitive computers or networks of computers by isolating them from the broader internet, since computers are often breached through the internet.
Randall suggests increasing the security of your home power supply by air-gapping it, using the light from a powered lightbulb to power a solar panel which then supplies power to the home, such that there is no physical wired connection between your house and the public electricity network. This is a large and very inefficient version of an opto-isolator, but would protect equipment behind the solar panel from power surges such as lightning strikes (which in an improperly grounded home could blow out the light bulb, but not so easily risk frying the equipment beyond the photovoltaic cell and its inverter). Due to its inefficiency, this approach would waste substantial amounts of energy. Optical power beaming is being investigated to recharge drones in flight.[1]
The title text mentions that a computer can still be connected to the internet via the power supply by using powerline networking, but that the bandwidth would be reduced by the lightbulb's warmup and cooldown delay, which would reduce the signalling rate the lightbulb could accomplish to no more than hundreds of bits per second, if that, for incandescent bulbs. However, as the light bulb cannot pick up any signals (possibly) emitted from the solar panel, the unidirectional link would be useless for traditional networking, because essential requests and acknowledgments would be unable to travel from behind the solar panel to the lightbulb. Early communication satellite systems for data networking used high-bandwidth unidirectional downlinks paired with low bandwidth ground telephone lines for outbound transmission, but such network configurations remain very uncommon.[citation needed]
Randall's solution is of course a joke. But in reality he could have used isolation transformers, which serve to allow the transfer of power via changing electromagnetic fields without an electrically conductive path. Most transformers, including "wall wart" power adapters, provide this form of isolation and protect devices from noise, voltage transients, most surges, and shock hazard, using fuses and other circuitry. They also limit powerline networking bandwidth by filtering out high frequencies. One relatively obscure way this comic is funny involves the relationship of the two concepts being conflated. Power analysis in computer security is a form of side-channel attack where the attacker observes and/or manipulates the power use by a device for some reason — for example, to gain insight into an otherwise protected process, or to exfiltrate information without having to use a monitored network connection. Power analysis in fire safety means measuring the power factor, watts, resistance, inductance, capacitance, volts, and amps of electrical circuits.
The look and subject of this comic is reminiscent of the Cursed Connectors series. But without the numbered cursed connector in the comic, this is not one of those connectors.
Why this would be inefficient and impractical
- Even energy-efficient LED lightbulbs are only about 35% efficient at turning electricity into light, with the rest emitted as heat.
- The air gap is inefficient at passing light from the bulb to the panel, causing some of the light from the lightbulb to be lost to places other than the solar panel, such as to the eye of the observer. A rough guess might be that in the configuration shown less than 60% of light produced will reach the panel, even assuming a perfect reflector.
- Solar panels are generally around 20% efficient at converting light into electricity, with claims at the world record from a single light source at around 40%.
All these efficiency-reducing factors, and others, multiply together. Therefore, only a small fraction of energy would be transmitted between the two ends of the air gap, making the circuit require much more electricity and be much less cost-efficient. For instance, the generous assumptions above lead to 96% of the power being lost. The solution as illustrated shows a single apparently-normal lightbulb, which typically draw no more than 250 watts, and usually much less power. Given the above efficiency issues, it would provide less than a tenth as much power.
How this could have a theoretical benefit
In electrical engineering, galvanic isolation is a measure to prevent an electric current flow between two circuits, instead, signals and energy are exchanged through indirect methods, e.g. magnetically, optically, or wirelessly. This is used to isolate a dangerous high-voltage circuit from the rest of the device, ensuring equipment and personal safety, it's also used to isolate sensitive measurement instruments from external noise, interference, and surges. Isolation transformers have several inherent limitations, and must be used together with other filtering and surge protection devices. The first problem is voltage rating, it's difficult to find a mains-voltage isolation transformer rated beyond a few kilovolts. Secondly, a transformer offers strong protection in steady-state DC and low-frequency 50/60 Hz AC faults, but only limited protection from differential-mode transients and surges. If an electrical surge has significant energy that happens to overlap with the transformer's working frequency (for a switched-mode power supply, this is around several kilohertz), the surge can partially bypass the transformer and enter supposedly-isolated sensitive equipment. Parasitic capacitance is another problem. A capacitor is formed whenever two conductors are separated by an insulator, and the insulated windings inside transformers are no exception. At 100 MHz, the impedance of even a tiny 20 pF capacitance is 79.5 jΩ. As a result, even though the DC impedance across a transformer is several megaohms, but it quickly deteriorates at high-frequency, allowing noise and interference to bypass the transformer and getting into sensitive measurement instruments. Worse, the primary and secondary sides of the transformer can radiate strong electromagnetic interference, since a dipole antenna is formed by two metal plates at different electric potentials. The radiation is suppressed by bridging the transformer with capacitors, forcing the electric potential to be the same at both sides at high frequency. The drawback is a further increase of capacitance, and a possible reduction of the isolation voltage rating, since capacitors are often the weakest part of the barrier.
Thus, there are exotic situations where an electrical connection must be avoided at all costs regardless of its efficiency, when safety or electromagnetic interference problems are critical. Power over Fiber (PoF) technology has been developed to address these needs. Using lasers, photovoltaic cells, and an optical fiber in between, the isolated load can be placed at a long distance away, allowing high voltage rating and extremely low parasitic capacitance. One example is high-voltage isolation at utility-grid scale, when the voltage can be 10 kV or higher. Electronic Design magazine reported an early 2006 product, with the lasers in the transmitter consume about 48 watts of power, in order to deliver about 720 milliwatts at the receiver - an efficiency of 1.5%. More recently, Avago (now Broadcom) also commercialized this technology, with receivers available for sale at $710. "With 1.5 W of laser light incident [...] up to 120 mA of current can be extracted at an operating voltage of 5.0 V and a total power delivery of 600 mW." Typical applications include "high voltage current sensors and transducers", "E-field and H-field probes", and "MRI/RF imaging coils and patient monitoring equipment".
Transcript
- [A solar panel and a lamp are pictured together, with the lamp pointed at the solar panel, and electronic equipment connected to the solar panel. Lines point outward from the bulb, indicating that it is shining.]
- [Caption below the panel]
- Energy tip: Increase the security of your home power supply by installing an air gap.
Discussion
Worth noting that this is a large and inefficient version of an opto-isolator 108.162.221.79 05:37, 28 July 2022 (UTC)
- Incandescent light bulb (assuming it the lamp does not use LED in the shape of light bulb) is not only less efficient than diode, but also much slower to warm up and cool down - it usually is much more sensitive to rapid switching, and has shorter life counted in the number of on/off cycles. --JakubNarebski (talk) 07:45, 28 July 2022 (UTC)
- There’s not even any indication that the bulb is shaped like an incandescent bulb. Only that the front of the light (either fixture or bulb) is a convex curve. For all we know that could be a lens or diffuser in front of a flat LED. Whoever wrote that needs to go back and walk, because the claim that an incandescent bulb is depicted is quite simply false.172.71.142.89 10:35, 28 July 2022 (UTC)
- The warmup and cooldown delays mentioned in the title text must imply an incandescent bulb, mustn't they? 172.70.210.145 03:29, 29 July 2022 (UTC)
- There’s not even any indication that the bulb is shaped like an incandescent bulb. Only that the front of the light (either fixture or bulb) is a convex curve. For all we know that could be a lens or diffuser in front of a flat LED. Whoever wrote that needs to go back and walk, because the claim that an incandescent bulb is depicted is quite simply false.172.71.142.89 10:35, 28 July 2022 (UTC)
- I don't think it's less or more efficient than an opto-isolator, it essentially is just an opto-isolator. But an opto-isolator isn't supposed to be energy efficient to begin with; it's only designed to transmit data between circuits, not power. So the output side only needs to generate enough voltage/current to change the state of a transistor, and the input side only needs to generate enough light for the output side to do that. The voltages and currents involved aren't comparable to power circuits. --NeatNit (talk) 08:14, 28 July 2022 (UTC)
- By the way, wikipedia links can be written like this:
[[wikipedia:opto-isolator|]]
result: opto-isolator (the final|
automatically gets expanded to the article title without thewikipedia:
prefix). --NeatNit (talk) 08:26, 28 July 2022 (UTC)- Or more often here on ExplainXkcd,
{{w|article}}
or{{w|article|anchor text}}
. 172.70.206.213 08:35, 28 July 2022 (UTC)
- Or more often here on ExplainXkcd,
- Yes, thanks :) although there is a tiiiiiny advantage to the direct link without the template (the way I said),
[[wikipedia:Pipe (computing)|]]
becomes Pipe whereas{{w|Pipe (computing)}}
becomes Pipe (computing). The pipe trick strips out the disambiguation parts of the title according to some rules. --NeatNit (talk) 12:30, 28 July 2022 (UTC)- Why not
{{w|Pipe (computing)|pipe}}
? 172.70.214.95 15:18, 28 July 2022 (UTC)- Yes we always use the
{{w|xkcd|xkcd comic}}
format here on explain to make xkcd comic these links. --Kynde (talk) 09:16, 29 July 2022 (UTC)
- Yes we always use the
- Why not
- Yes, thanks :) although there is a tiiiiiny advantage to the direct link without the template (the way I said),
It is missing that air-gapping the power supply would protect your home from voltage surges in the power network caused by lightning strikes. Depending where the lightning hits the power network, there may be no fuses protecting your home or single fuses may fail to protect you. --172.70.246.115 07:57, 28 July 2022 (UTC)
- That is true. But the suggestion that this might have anything to do with general energy security (as is currently very prominent in the explanation) is entirely unconvincing to me.
- I also originally thought this was the main joke, until seeing the title text about bit rate. Certainly it's worth mentioning, even if this isn't the main joke, since it would actually work, with a wide enough gap, ideally with a vacuum in between.
Incandescent light bulbs convert most of their energy to infrared light. There are solar cells that work in this infrared spectrum, so this might not be all that inefficient as stated. This should in fact be a lot more efficient than any LED+visible spectrum based panel, as incandescent bulbs are very efficient in converting electricity into infrared light, much more than LEDs most likely will ever be. The (mostly) omnidirectionality of the light source might be an even bigger loss, as most of the light (however efficient) does not even reach the panel. And regarding sending data over this construct: As soon as there's a 0V state (which will be the case as soon as the transmission starts, due to some form of manchester coding, regardless of it being a 0 or 1 bit) the PC behind the solar panel would not only have a data transmission problem :) (With incandescent bulb, that is. A LED 0V might be short enough for capacitors in the PC's power supply to buffer it, if it is only at 50%(+PSU conversion loss) load max, as manchester coded signals per definition have a duty cycle of 50% to keep the DC bias at 0V) 172.68.51.204 08:26, 28 July 2022 (UTC)
What is the highest wattage commercial opto-isolator, and how can I get one mounted from the ceiling in my bedroom? 172.70.207.8 14:10, 28 July 2022 (UTC)
- Ordinary florescent bulbs or neon tubes are close, the electrical circuit being isolated to flow through a gas which only becomes ionized and thus conductive in a way which causes it to emit light. But the signal stays as electrical without any intervening photons. You probably don't want a laser BBQ on your bedroom ceiling, but then again, the military gear referred to below isn't commercial, as it would never pass laser safety certifications. 172.70.210.145 01:01, 30 July 2022 (UTC)
I wonder if it's worth noting the significant understatement within the title text, where it says "the bitrate does drop a little" in contrast to the severe and drastic drop in bitrate that would actually occur, especially in light of today's typical Internet speeds. It might not be worth mentioning, but it struck me as a humorous understatement of the true impact. Ianrbibtitlht (talk) 14:25, 28 July 2022 (UTC)
- We don't know if the light bulb is incandescent or LED, so we can't describe the bandwidth drop other than in very general terms. 172.70.211.88 15:34, 28 July 2022 (UTC)
- It's safe to say the warmup and cooldown delays mentioned in the title text imply an incandescent bulb. Is there a standard or average response time for household bulb incandescent filaments? 172.70.214.43 03:33, 29 July 2022 (UTC)
- Maybe this helps? https://ieeexplore.ieee.org/abstract/document/5625426 172.70.142.21 06:51, 29 July 2022 (UTC)
- Yes! Figure 1.10 seems to suggest that frequencies above 10 Hz are filtered, and that seems consistent with Figure 3.14, in that recovery time is a tenth of a second (though drop-off "cooldown" time is much shorter.) So if I remember my modem math, even with the most sophisticated coding, anything more than 640 bits per second should be impossible. 172.69.33.225 07:39, 29 July 2022 (UTC)
- 640 bits per second should be enough for anyone. But seriously, Fig. 1.10 shows a fall-off, not a sharp peak. How do you know there isn't an 8,192-dimensional analog to the Leech lattice which would allow kbps? 172.69.33.199 22:56, 29 July 2022 (UTC)
- Yes! Figure 1.10 seems to suggest that frequencies above 10 Hz are filtered, and that seems consistent with Figure 3.14, in that recovery time is a tenth of a second (though drop-off "cooldown" time is much shorter.) So if I remember my modem math, even with the most sophisticated coding, anything more than 640 bits per second should be impossible. 172.69.33.225 07:39, 29 July 2022 (UTC)
- Maybe this helps? https://ieeexplore.ieee.org/abstract/document/5625426 172.70.142.21 06:51, 29 July 2022 (UTC)
- It's safe to say the warmup and cooldown delays mentioned in the title text imply an incandescent bulb. Is there a standard or average response time for household bulb incandescent filaments? 172.70.214.43 03:33, 29 July 2022 (UTC)
I'm almost tempted to suggest that this should be an (honourary) addition to the Cursed Connectors comic-collection. 172.71.94.187 14:56, 28 July 2022 (UTC)
- I agree that it reminds me of some of those. But it could never be added to the category, as it misses the cursed connector name and number. But I have added a mention of the similarity here above, and also mentioned it on the Category:Cursed Connectors page. --Kynde (talk) 09:22, 29 July 2022 (UTC)
- It's a cursed disconnector. Barmar (talk) 15:24, 29 July 2022 (UTC)
- Dis-connectors are always cursed! ;) 162.158.159.95 17:11, 29 July 2022 (UTC)
Yes, solar panels can transform electrical into electromagnetic signals. It is more on the side of the incandescent bulb that the capability to receive and forward these signals is missing. My source: https://www.youtube.com/watch?v=TGUteH93xNo Torge (talk) 15:11, 28 July 2022 (UTC)
- Interesting! 172.70.142.21 06:51, 29 July 2022 (UTC)
- Good link for free electrons in the recent 2649: Physics Cost-Saving Tips. For this comic, Part 1 might be better.,172.69.33.245 00:46, 30 July 2022 (UTC)
Nobody caught on that powerline networking is about sending data through powerlines? or that a crude opto-isolation setup could effectively scrub it? Where my networking geeks at? I am disappoint 172.69.71.51 15:19, 28 July 2022 (UTC)
An electrical engineer here: perhaps we should mention that the box pictured after the solar panel must be an inverter? The lightbulb/solar panel pair will be acting as a rectifier, putting out purely positive voltage, and to get back to AC to run appliances on there would need to be an inverter. 172.69.62.49 01:10, 29 July 2022 (UTC)
- Can you please help out with the argument about rectifier efficiency at 2642? 172.70.214.95 03:24, 29 July 2022 (UTC)
Most of the bulbs in my house (before compact flourescent and LED) were 75 Watt or 100 Watt. 50 Watt would be a very dim bulb.
- Mice wouldn't agree. 172.70.214.43 05:05, 29 July 2022 (UTC)
There are actually usecases for optical power links. For example electric field probes use this: https://www.raditeq.com/products/electric-field-probes/ --Casandro (talk) 08:10, 29 July 2022 (UTC)
- Some drones can be recharged in the air by lasers, but that's a really obscure application only the military needs, and then only as a contingency backup in most cases, but there were flying demos by companies interested in the space over the past decade. 172.70.211.88 22:09, 29 July 2022 (UTC)
I think it might be possible to construct a scenario where this could have some security benefit, albeit not one that justifies such poor efficiency. Suppose you have an embedded system running off USB power to perform some cryptographic task, but the USB port the cable's plugged into is under the control of an attacker who has it instrumented to sample the power consumption at a high frequency. He can perform power_analysis if the load changes based on the operations performed by your device, but the setup in the comic should prevent information from leaking upstream in this manner. D5xtgr (talk) 19:44, 29 July 2022 (UTC)