3295: Semaphore
| Semaphore |
 Title text: The speed of light in air is 50% faster than in fiber, but it's a challenge to get the arms to move fast enough to realize the latency advantage. |
Explanation[edit]
|
This is one of 50 incomplete explanations: |
This comic has some similarity to 2651: Air Gap. An air gap may be usable for security if information cannot move through it (for example, keeping an AI in a "box"), but in this case the purpose is to transmit information between separated systems while presumably protecting from rogue power-spikes, in the manner of an opto-isolator component.
Semaphore is a signaling system involving the representation of messages, with flag semaphore being a subset that used patterned flags held at specific angles and commonly associated with line of sight ship-to-ship/ship-to-shore messaging. In the comic, the connection between two components is a system based on the flag-based system instead of a typical cable (electrical or fiber-optic cable) connection. In order to send information, the sending component repeatedly moves the flags such that the flags spell out the information in semaphore, then the receiving component recognizes the semaphore with a camera and (presumably) translates the semaphore back into the requisite electrical signals.
Semaphore towers have historically been used for long-distance communication, with optical telegraph chaining towers capable of displaying semaphore-style signals. Before electric communication technologies were available, a message relayed between towers allowed cross-country communication faster and further than it was possible to physically travel.
The title text notes that the speed of light is faster in air than in an optical fibre. This is correct: in air, light travels at close to the same speed as in a vacuum (about 300,000,000 m/s), but optical fibres are made of glass that is optically dense, specifically to guide the optical signal, and refraction causes the effective speed of light in glass to be much slower. This is known as the latency of the connection. In principle, therefore, it is quicker to transmit a signal by light directly through air than by light through a fibre. However, the rate of transmission of the signal is limited by how fast the flags can move between positions in a manner that can be understood by the other station, and this can only add to the point-to-point transmission time.
We can make a rough approximation. The speed of light through fibre works out to around 30% slower than through air. In the illustration, the distance between the cameras/flags appears to be about 1 meter. Light takes 3 nanoseconds to cross that distance in air, and 4.5ns in a fibre-optic cable, so it would save about 1.5ns per bit. It would require the processing, flag movement, camera scan, and image processing on the other side, to take less than 1.5ns per bit in order to gain any bitrate increase. Assuming flags move about 10cm to produce a bit, and that it can take that entire 1.5ns (you can process one bit while moving the flags for the others), that requires turning flags at 0.22c. Relativistic problems require relativistic solutions.
The setup could be improved by adopting a more efficient encoding scheme, such as directly encoding octets to send eight bits at a time, and compatible datastream compression techniques. More can be gained by positioning the ends of the cables farther apart, so long as it is not so far as to make the signal arms indistinguishable, making the same work to encode/decode the message transmit over a longer distance. Assuming we are transmitting octets over 100 meters, every signal can take up to 1200 nanoseconds and still arrive faster than the bit would through the cable. 10 cm arms would only have to be actuated at 83 km/s; 'only' 251 times the speed of sound in air.
Wireless communications using radio waves (like wi-fi, bluetooth, 5G, etc.) do travel at the speed of light in air, making them technically faster than fibre optics. Also, since they can operate by line-of-sight between strategically located relays, data is transmitted over the smallest practical distance, where fibre optic cables may need to be routed along a less direct route. This marginal advantage is used, for example, in high-frequency trading between Frankfurt and London, where microwave links can provide a small but lucrative latency advantage over fellow traders restricted to fibre-optic connections.
There are communication modules using lasers through air, without using fiber. They are quite fast, but have two main problems. Firstly, a slight misalignment is enough for the laser to miss the detector on the other side, so they may need frequent re-alignments or a deliberately diverging beam that is inefficient and may compromise security. Secondly, they are sensitive to atmospheric conditions and may be severely impaired in the case of fog, rain, etc.
Transcript[edit]
|
This is one of 34 incomplete transcripts: |
- [Two identical machines sit facing each other. Each one consists of a box with a wire connecting it offscreen, a webcam, and a machine with 2 flags. Each webcam is pointing at the other machine's 2 flags. Motion lines indicate that the machines are able to move the 2 flags independently.]
- [Caption below the panel:]
- Networking tip: Separate sensitive components with a semaphore gap.
Add comment 
Create topic (use sparingly) 
Refresh
Discussion
It's not directly to the speed of light that the speed of moving flags should be compared (in terms of signal transmission, as Randall rightly points, semaphore is faster), it is rather the rate at which a new semaphore symbol can be dialled (typically one of the half-a-dozen to a dozen patterns a given semaphore can present). If calculating the information flow, that dialling rate (thus speed of moving the flags divided by a typical size of semaphore, ie about 10 per second if moving a 10 cm semaphore at 1 m/s) times the number of possible patterns (let's say 8) should be compared to the baud rate (typically $10^11$ per second) times 2 to the power of the number of lanes (let's say 3). For the data chosen, we're comparing approximately 80 bits per second for semaphore to $8\cdot10^11$ for the optical fibre. This is the rate at which a permanent flow of information can be transmitted (if we consider that deciphering the semaphore pattern is faster than dialling it, so in the situation where the operator or software getting the image from the webcam is at least as fast as the dialling). However, this doesn't tell us about the latency time: how much time does it take for an elementary 8-bit signal to reach destination. That time is obtained by summing the time of "dialling" (or modulating the numerical signal for the optic fibre), plus the time it travels in air or fibre (faster in the former, that's where semaphore can try to catch up!), plus the deciphering/demodulating at reception. --2001:1C04:2F0B:5300:3C9D:4D7A:7E6C:66E7 20:37, 7 September 2026 (UTC)
- I think I addressed what you just said there, before spotting that you'd been making this point. There's elements of latency, lag, bandwidth and so on in the basic transmission rate. I think I managed to make a brief but salient edit to sort of cover that, but YMMV. 92.23.7.14 23:14, 7 September 2026 (UTC)
Basically, Opto-isolators writ large. --79.16.224.49 22:15, 7 September 2026 (UTC)
Propagation speed through whatever medium (air, fiber optic glass, copper, etc.) doesn't practically contribute to latency over short distances. Local latencies are dominated by processing at either end (or dropped packets). 75.230.130.169
What are sensitive components, in this context? Student Driver (talk) 01:33, 8 September 2026 (UTC)
- Electrically sensitive I would assume. If there is a power problem in on system (say some type of ground fault or a lighting strike) larger than design current can flow in the network wiring. There is a rule you never run copper between two buildings for this reason. Lordpishky (talk) 03:39, 8 September 2026 (UTC)
The explanation should probably explain what air gaps are in networking. An architect (talk) 01:45, 8 September 2026 (UTC)
- Only if it explains that this example is not an actual networking air gap, though. 92.23.7.14 12:47, 8 September 2026 (UTC)
The explanation should probably explain what "semaphores" usually refer to in computing as I believe that's an intended allusion/pun here; controlling access to common resources which could be construed to be "sensitive". 136.226.154.84 16:41, 8 September 2026 (UTC)
Randall is offering a ludicrous suggestion but the current explanation fails to a) discuss why it is useful to separate sensitive components and b) why this suggestion is impracticable. Inquirer (talk) 02:04, 8 September 2026 (UTC)
I guess I'm just waiting for the relevant implementation on GitHub to link in the explanation now. GreyFox (talk) 02:28, 8 September 2026 (UTC)
- While not as widely implemented as IP over Avian Carriers, there still is an "official" way to send IP diagrams using the Semaphore Flag Signaling System. See RFC 4824. 2A01:36D:1300:2A53:132B:E6E3:5CB5:54C9 04:35, 8 September 2026 (UTC)
I have tried to calculate the speed at which the flag arms need to move in order for this setup to save any time. My first edit on the wiki - hope I did it right. --KeizerHarm (talk) 16:11, 8 September 2026 (UTC)
Is it worth mentioning that some electronic connections, especially MIDI, use optical isolation with a LED and light sensitive resistor to prevent problems with earth loops and voltage differences.80.41.5.152 16:52, 8 September 2026 (UTC)
I added a paragraph about laser links. At my university we had/have such a system. When it worked it was great (I am not sure but I think it was 100Mb/s across 2kms). The main problem was that slightly movements of the buildings (probably due to thermal expansion/contraction) were enough to lose the signal (I just checked and an error of only 1 degree gives a ~35 meters error at the other side). Also, in case of fog or rain it also lost signal. Not sure what was the wavelength, probably red or near-infrared. Rps (talk) 18:05, 8 September 2026 (UTC) Add comment