I managed and coordinated the IIT Ropar's very first multi-project tapeout with 100% functional success. I designed the 60-pin padframe using the Faraday library, performed floor planning and pin distribution, power distribution, integrated all designs, and performed verification & sign-off.
Thanks to team for timely execution of each project's layouts, and the maker-checker approach. Of course, it would not have been possible without my supervisor's guidance.
Skills: PDK, Cadence Virtuoso, ADE Explorer, Assembler, layout, Calibre, DRC, LVS, Antenna, PEX, PVT, dummy, ESD, DFM, etc.
Inventory used: RF Source, Bias Tee, High precision power supply and SMU, Mixed signal oscilloscopes, function generator, 6.5 digit digital multimeter, SMA cables and connectors, BNC cables;
<to be updated...>
Electromagnetic or radio-frequency interference (EMI or RFI) that reaches the amplifier in the front-end circuit will be rectified by the amplifier into a DC output offset voltage. It may saturate in the worst case, in high-gain configurations. Even a small, unknown DC offset is enough to make the output unreliable. This EMI susceptibility is measured in Unity Gain Config.
A DC bias of 0.9V is given as input to Vin1, and a high-frequency EMI is superimposed momentarily. The output Vo1 is measured using a standard OTA in UGC, whereas Vo2 is measured for an EMI-immune OTA fabricated in the aforementioned tapeout.
A crossbar array is widely used in neural networks for weighted SOPs. A crossbar interface circuit comprising summing amplifiers at the end of each column provides the output, which may be further evaluated to make a decision. The EMI induces an unknown, distinct DC offset in the output for each column, which may flip the decision. Critical neuromorphic applications with large crossbar arrays may lead to incorrect inference due to EMI. (Link to article)
<to be updated ...>
<to be updated . . . >