China
Africa
Explore chapters and articles related to this topic
Applications of Machine Learning in VLSI Design
Published in Sandeep Saini, Kusum Lata, G.R. Sinha, VLSI and Hardware Implementations Using Modern Machine Learning Methods, 2021
Sneh Saurabh, Pranav Jain, Madhvi Agarwal, OVS Shashank Ram
Another problem that can utilize the capabilities of ML is the post-silicon validation. Before production, we carry out post-silicon validation to ensure that the silicon functions as expected, under on-field operating conditions. For this purpose, we need to identify a small set of traceable signals for debugging and state restoration. Traditional techniques such as simulation take high runtime in identifying traceable signals. Alternatively, we can employ ML-based techniques for efficient signal selection [30]. We can train an ML with a few simulation runs. Subsequently, we can use this model to identify beneficial trace signals instead of employing time-consuming simulations [30].
A low-power, low-offset, and power-scalable comparator suitable for low-frequency applications
Published in International Journal of Electronics, 2023
Riyanka Banerjee, M. Santosh, Jai Gopal Pandey
The comparator designed by Lan et al. (2011) operates at 0.8 V, giving a 12-bit resolution, but FoM has not been provided. The comparator designed in 40 nm technology by Baradaranrezaeii et al. (2015), operates at a frequency of 6 GHz, and the FoM is 57.65fJ/conversion. In the proposed work, the inverter-based comparator has a low input offset voltage, that is, 256 μ V and the kickback noise is less than the architectures of Lan et al. (2011), Fayomi et al. (2009), Chin et al. (2010), Baradaranrezaeii et al. (2015) in Table 2. To measure the kickback noise during simulation, a 100 ohm resistor (picked from the basic library) has been added in series at the input side and the voltage has been measured across the resistor. The selected value matches the configuration of the output channel settings of the arbitrary waveform generator (Model No. Tektronix AFG-3022B), so that the post-silicon validation data match the simulation data. Such low kickback noise makes them suitable for data converter applications, that is, digital dominant SAR architecture.