The parasitic devices that are formed in CMOS inverter are - Parasitic Resistance (R), Parasitic Capacitance (C), Parasitic Diode (D), Parasitic BJT, Parasitic SCR
$Q_1$ is the PNP transistor and $Q_2$ is the NPN transistor. This forms a regenerative forward bias Pair that represents SCR in forward blocking mode.
- $R_P$ is the P-well resistance and $R_S$ is the N-Si substrate resistance.
- Since nMOS & pMOS transistor must be fabricated on the same chip side by side, COMS process is more complex as compared to standard nMOS only.
- In particular the COMS process must provide a N-substrate for the pMOS transistor and vice versa.
- This can be achieved by building either n-type well and p-type wafer or vice-versa.
- However, the close proximity of nMOS & pMOS transistor may result in to formation of 2 parasitic BJTs causing a latch-up problem.
Latch up Mechanism
- Due to increase in ambient temperature, incident radiation electromagnetic interference power supply transients a small leakage current is drawn by N-Si substrate from supply i.e. $I_S$
- $I_S$ causes drop across $R_S$.
- If drop is sufficient enough to forward bias Base-Emitter junction of Q1 then Q1 conducts.
- Due to this P-well draws a current through Q1 and therefore potential drops across $R_P$
- If drop is high enough to forward bias Base-Emitter junction of Q2, then Q2 conducts.
- Q2 conducts and then Q1 & Q2 self-sustained ON state and then SCR turns ON.
- If SCR is ON then there exists low resistance path VDD to ground.
- Therefore current increases abruptly to large value and thus the device melts
- This destructive phenomenon is called COMS latch up.
Prevention of CMOS latch up
- One method of preventing latch up is to decrease the parasitic transistor current gain. If the product of two current gains β1.β2 < 1, then forward bias is not self-sustaining and hence device cannot latch.
- The current gain of PNP transistor is determined by the process design and can be decreased by using buried layer of $N^+$ materials.
- The NPN transistor current gain can be reduced by increasing the space between the $N^+$ source, drain & well.
- This will increase the width of transistor base and hence current gain will decrease.
- Second method is to keep VDD and ground contacts as shown in figure. By defining multiple contacts we reduce the resistance offered by well as well as the substrate.
- By using dielectric isolation between pMOS & nMOS transistor leakage current can be decreased and hence latch up can be prevented.