When the noise becomes the signal

In the world of engineering it is common to separate what we want to measure, the signal, from whatever makes the measurement imprecise, the noise. For instance, trying to weigh yourself in a digital scale, the signal is your weight (technically is something like a month-average of your weight) and the noise are the variations of your weight due to your bodily functions, the effects of temperature and humidity in the performance of the scale, etc. This nomenclature mimics the act of trying to listen to a conversation in a noisy room: the conversation we are trying to listen to is the signal, whereas all the other conversations, the clinking of the glasses, and the background music are all noises. This leads to the definition of the signal-to-noise ratio (SNR), which describes the relative intensity of both of them: if the room is silent, the conversation is much louder than the noise, so the SNR is high and the conversation can be understood. Conversely, in a dance club the background noise is very loud, so the SNR goes down to the point that very frequently you are unable to retrieve the content of the message. However there are situations where the noise is, precisely, the signal you are looking for.

The post on Friday reminded me of a different kind of "music" that also came from a cassette: the games for my Sinclair ZX Spectrum + computer. Back in 1985 when I got it the storage systems were still fairly primitive. A few computers already had disk drives, but most were content with saving the programs to tape. In fact, one of the most difficult parts of my initiation into programming was indeed getting a cassette player with recording capabilities. For months and months I had not choice but to write my programs by hand on paper, type them into the computer, test them and debug them only to see them vanish when the computer was powered off. Having a cassette player to load the few commercial games I had bought was easy enough, but a recorder was a completely different game.

Photo: Bill Bertram

One of the peculiarities of the ZX Spectrum was that it showed a varying pattern of horizontal lines as it was loading the games from tape. Of course, the sound was pure digital noise to my ears but still after watching the loading sequence (sometimes up to 10 minutes) run several times I ended up being able to tell if the program was loading properly or not. In this case, the noise was the signal.

Another instance where the noise also became the signal happened a few years later, around 1991 when we got our first modem at home. The connections process was excruciatingly long, from the dialing to the wait time until the internet provided answered (if ever) on the other side. Then came and exchange of electronic chirps which allowed the two computers first to communicate and the to confirm that we were allowed to use that internet access (it was a paid service, of course). Once the connection was established, the loudspeaker of the modem went silent, but in the intervening minutes one could hear (but not understand) the digital traffic. As in the case of the Spectrum, with some practice I got to be able to tell different kinds of successes and errors, in particular when the service provided was overloaded because the sound of the connection was significantly different.

It might seem that the noise-as-signal phenomenon is limited to the digital world, but it is not the case. For instance shortly after I finished my engineering degree I bought a small used Yamaha motorbike which, with a single-cylinder two-stroke engine, had a very distinctive sound. Whenever I heard a motorbike on the street I could determine without any doubt if it was the same model as mine or not (and it was rather popular at the time).

However, there are two impressive uses of noise as signal that I know of. Karen's late grandfather told us the first one, which was applied to railroad wheels: whenever a train had a relatively long stop at a station someone (he, the wheel tapper) would walk along the train hitting every wheel with a heavy wrench or a hammer, listening to the noise it made. If the noise was "clean" on every wheel, everything was right and the train was allowed to continue. However, sometimes one of the wheels would produce a strange sound, normally reflecting the fact that it was cracked. In those cases the mechanic would crawl under the train, identify the severity of the problem and decide what to do. And all thanks to the noise from the wrench.

The other example shocked me when I saw it in a roof tile factory. The plan was amazing from the extrusion of the mud and the molding and un-molding, to the drying and the firing processes. But in the end, every single tile went on a conveyor belt and an employee hit every single one with a big chain link as they passed by at a rate of approximately one tile per second. In that time he not only had to hit the tile and listen to the noise, but also decide if it was good enough and eventually press a pedal that would take it out of the conveyor belt. The idea of spending 8 hours a day seeing roof tiles pass in front of someone seemed haunting to me, but the percussion with the chain link was very smart.

As we have seen, the definition of what is signal and what is sound varies a lot with your perspective. Probably the passengers of the train found the wheel tapping quite annoying and labeled it as noise, but it was clearly the signal for the railroad mechanics. The same applies to people speaking in a foreign language or using some technical jargon: the fact that we do not understand them does not imply that it does not make sense. In the same way that one man's trash is another man's treasure, one man's noise can be another man's signal. Have a nice evening.