1/f and RTS noise in submicron devices: faster is noisier

The origin of 1/f-like noise in devices is still under discussion. There is one school of thought explaining low-frequency noise as a surface effect due to trapping and suggesting number fluctuations as the origin. The number of (surface) traps is the key-parameter. According to another school, pure 1/f noise is considered as a bulk phenomenon due to mobility fluctuations. There the 1/f noise parameter α plays an important role. Low frequency noise is often due to a mixture of conductance fluctuations due to number fluctuations and due to mobility fluctuations. A good strategy to improve the analysis is as follows. First step: make a decomposition of the observed noise into pure 1/f noise (1/fγ with 0.9<γ<1.1) and, if possible, a Lorentzian stemming from e.g., random telegraph signal noise (RTS). Second step: investigate how the different noise components depend on bias and device size. We explain why: i) for N<1/α in sub-micron devices we can expect RTS on top of 1/f noise, ii) faster devices are noisier; iii) 1/f and 1/f like noise are often confused. For silicon resistors there is a possibility of RTS noise on top of the 1/f noise if the number of carriers is less than 1/α. The RTS criterion is normally fulfilled for sub-micron MOSFETs with an area LW<1 μm2. With N=CgateVG*/q<1/α, the RTS noise on top of the 1/f noise can become dominant. RTS on top of the 1/f noise is to be expected in forward biased diodes at low currents if I 0 < I < IRTS ≈ q/ατ; where I 0 is the saturation current, q the elementary charge and τ the minority carrier life time.