Sampling the signal

The signal from the VLBI antenna is extremely noisy. The noise comes from cosmic background radiation and also from the electronics and circuitry used to process the signal. The circuitry at the antenna itself is cooled almost to absolute zero, as this minimize analog component noise. The signal to noise ratio is 1/1000, meaning one part signal and 999 parts noise. Moreover, the signal itself can be characterized as white noise. So we have noise buried in noise.  This is often sampled at 1 bit resolution, meaning a signal that is either zero or one. It is still a mystery to me how we can make any sense at all out of this frenzy of noise. But we can.

As the signal approaches us, it needs to pass through earth’s ionosphere , where it is also distorted. Commonly, the ionosphere introduces a delay that depends on how violent it’s turbulence is. This again comes as a result of sunstorms and other external disturbances. To counteract these disturbances, the signal is sampled at several different frequencies simultenaously. The idea being that some frequencies are more prone to delay in the ionosphere than others. So we use some sampling frequencies to figure out the ionospheric delay, and then we can use another frequency to actually capture the signal and then compensate for the delay.

Raw bitstream


In the installation, we can hear the raw signal used as a pattern for rhythmic pulses.  It is layered in 16 parallell streams, according to the different sampling frequencies. These streams is then played back at different pitches, and each stream is given a dedicated position in the sound image (12 channel image at the physical installation, 2-channel stereo image in the online streaming version). The playback rate is very much slowed down in relation to the original recording rate (of several gigahertz), this is done purely of musical reasons, and similar freedom of tempo has been taken with all data in the installation. Here is an example, where channels are turned on and off in a pattern, to show the different characteristics of the channels:

The broadband noise from the quasar is recorded at different frequency bands. The 2Gz band (S-band) is used to calculate the atmospheric delay, as it is less prone to this kind of disturbances than the 8Gz band (which is the band we are actually more interested in).


As we are trying to reconstruct a broadband signal, but we only have space (data size) to capture parts of the spectrum, we record different frequency bands that are spread out at inharmonic ratios. If were recording at harmonic ratios, we would capture less new information with each new band (redundancy), so the frequency bands are carefully chosen to be at nonharmonic intervals.


Here are two sounds created by transposing the observation frequencies down into the audible range. In and of thems selves, they sound kind of "outer space-y" to my ears.