Compressive Sensing system for recording of ECoG signals in-vivo

Appendix CC.1. Analog ImplementationThere is a large range of applications in which random generation is necessary in order to beused in signal processing blocks. Radiofrequency communications signal mixing, test systemsand more recently biosignal compression based on CS are the most relevant ones. By means ofan analog implementation of such randomness, a true random generation is achieved at theexpense of a larger circuit complexity of the integration of more exotic circuitry. The maintechniques have been developed are: direct amplification of noise, high-frequency oscillatorsampling and randomness generation based on chaotic systems, and they are described below.C.1.1. Direct Amplification of NoiseThis method uses a device noise to make a decision of a bit 1 or 0 depending if the signal isgreater than certain threshold or vice versa. This can be done by amplifying the white noise of aresistor and then using a comparator to obtain a random bit. These schemes need a high-gainbandwidthlow-offset amplifier due to the fact that if a low-bandwidth amplifier is used, theoutput noise spectrum can be concentrated around DC and so, noise samples are correlated.Thus, they are not suitable for Radio Frequency Identification, RFID, systems. Animplementation has been presented in [29], and the main architecture is shown in Fig.C.1.1.1.This approach is not feasible to an on-chip application, due to area constraints.Figure C.1.1.1. Random generator based on direct amplification of noise [29].C.1.2. High-Frequency Oscillator Sampling [29]A TRNG can be implemented by combining a low-frequency jitter clock and using it to sample ahigh-frequency oscillator. The timing jitter is a stochastic phenomenon caused y thermal noisepresent in the transistors of a ring oscillator. The drawback of this procedure is that themegahertz generation is power hungry, so it is not an option for an on-chip CS design.81