You Can’t Sing Over the Internet Because Nuclear War

These days I spend a lot of time helping people use video conferencing software over the web, like Zoom. Once people overcome the various strange myriad of settings, and learn how to mute, un-mute and turn their camera on and off, the conversation often turns to how un-conversational it is to have a group discussion. Delays and audio switching makes us create a whole new set of conversational cues, as we adapt to the strange jumps and gaps that characterize communication over the internet.

The limitations of internet conferencing are made particularly acute when people try to sing or make music together. The delays between singing and hearing other people are long and uneven making it impossible to sing in synch. This is pretty disappointing to choirs and musicians, and we all wish it worked as well as being in the same room. But the limitations of being able to synchronize multiple destinations over the internet isn’t the fault that any app can overcome, they are limitations of the internet itself.

Circuit-switching by hand

In the early days of the replacement of analog networks to digital, two models for communications were developed, one called circuit-switching and the other called packet-switching. It’s pretty easy to understand circuit switching. You’ve seen circuit-switching in action in these old photos of switchboards.

Circuit-switching is defined by providing a fixed channel between two end-points that remain in place for the duration of the connection. Even if no information is passing from one end to the other, the circuit is dedicated to that connection. The biggest circuit-switched networks in the world was the international telephone system. Until the 1980’s every phone call was made by switching circuits to provide a pair of copper wires that connected two telephones. The equipment to switch all those circuits was immense and each neighbourhood has large brick buildings without windows that also housed air conditioners and backup batteries. Circuit switching is inherently wasteful. It dedicates an entire circuit whether it’s being used or not. As the demand for global communications increased, and fax machines threatened to alone double the number of circuits needed, analog telephone engineers looked for more and more clever ways to get more calls onto a pair of copper wires that carried the conversation. In the early 1980’s they developed a way of listening for silences, or breaks in conversation over international phone lines and inserting other calls into those gaps, an approach called multiplexing.

But the circuit-switch world was already losing ground to its digital rival, packet switching. With packet switching, analog information is converted to digital thousands of times a second, collected into packages and sent over the internet to their destination. Packet switching makes more efficient use of the channel and many simultaneous conversations can share the same connection.

packet switching mixes information from different conversations together over a common channel

The protocol for how to send this digital packages over circuit-switches was designed by the telephone industry, and digital over circuit-switched networks had one benefit we didn’t miss until we didn’t have it anymore, and that is low-latency. Latency is important when you want to synchronize events. Many easy digital cameras had considerable latency, the delay between the time you pressed the “shutter” button and the time it took the picture. You can hear latency if you use you cell phone to call your partner in the same room and compare what they hear with when you said it.

Why are there delays in packet-switched networks? The answer lies in the origins of the internet. What we now call the internet was born on January 1, 1983 when the US Department of Defence Advanced Research Projects Agency Network (ARPANET) chose TCP/IP as their protocol to exchange digital files and messages between computers located far apart. The primary goal was to have reliability, and the design of packet routing systems provided high reliability that a digital packet sent from one computer would eventually get the the destination computer, by letting each relay point along the way determine the best way to get it there. When your computer sends a packet it doesn’t really know how it will get to the destination, it just sends it along to another computer that then relays it, like a bucket brigade where each person in the line only knows who they get the bucket from and where to hand it next. Dynamic routing created a very robust and reliable network, and at the time the importance to the DoD to have a communications system that might survive a nuclear war was a higher priority than a bit of unpredictable small delay along the way.

It’s the “small delay” that makes it impossible to sing in synchrony over the internet, because the delay varies for every packet. We don’t usually notice it when using email or browsing, and video conferencing works hard to minimize the latency of the audio so we can hold conversations, sacrificing video speed and quality if it needs to. But to perform or sing together over the internet is not a question of which video application is better than another, but a basic limitation of the underlying structure of the internet itself.

The pending and controversial 5G network which is supposed to replace the current LTE cellular network is designed to have ultra-low latency to the extent that in the never ending battle between internet service providers and cellular network provides, we might end up being to sing and perform together over our new cell phones better then we will be able to through our home networks.

Dave worked at Mitel, a company that designed and built one of the first digital PBX exchanges in 1984, around the time that the circuit-switched and packet-switched worlds started to diverge.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.