Fairchild's Process Enhancements Eliminate the CMOS SCR Latch-Up Problem in 74HC Logic

Category: Logic
Manufacture: Fairchild Semiconductor
Datasheet: Download this application note

Enhancements Eliminate the CMOS SCR Latch-Up Problem In 74HC Logic
Fairchild's Process Enhancements Eliminate the CMOS SCR Latch-Up Problem In 74HC Logic
INTRODUCTION SCR latch-up is a parasitic phenomena that has existed in circuits fabricated using bulk silicon CMOS technologies. The latch-up mechanism, once triggered, turns on a parasitic SCR internal to CMOS circuits which essentially shorts VCC to ground. This generally destroys the CMOS IC or at the very least causes the system to malfunction. In order to make MM54HC/MM74HC high speed CMOS logic easy to use and reliable it is very important to eliminate latch-up. This has been accomplished through several layout and process enhancements. It is primarily several proprietary innovations in CMOS processing that eliminates the SCR. First, what is "SCR latch-up?" It is a phenomena common to most monolithic CMOS processes, which involves "turning on" a four layer thyristor structure (P-N-P-N) that appears from VCC to ground. This structure is formed by the parasitic substrate interconnections of various circuit diffusions. It most commonly can be turned on by applying a voltage greater than VCC or less than ground any input or output, which forward biases the input or output protection diodes. Figure 1 schematically illustrates these diodes found in the MM54HC/MM74HC family. Standard CD4000 and MM54C/ MM74C logic also has a very similar structure. These diodes can act as the gate to the parasitic SCR, and if enough current flows the SCR will trigger. A second method of turning on the SCR is to apply a very large supply voltage across the device. This will breakdown internal diodes causing enough current to flow to trigger latch-up. In HC logic the typical VCC breakdown voltage is above 10V so this method is more uncommon. In either case, once the SCR is turned on a large current will flow from VCC to ground, causing the CMOS circuit to malfunction and possibly damage itself. CMOS SCR problems can be minimized by proper system design techniques or added external protection circuits, but obviously the reduction or elimination of latch-up in the IC itself would ease CMOS system design, increase system reliability and eliminate additional circuitry. For this reason it was important to eliminate this phenomena in Fairchild's high speed CMOS logic family. Characterization of this proprietary high speed CMOS process for latch up has verified the elimination of this parasitic mechanism. In tests conducted under worst case conditions (VCC = 7V and TA = 125C) it has been impossible to latch-up these devices on the inputs or on the outputs.
Fairchild Semiconductor Application Note 339 November 1987
In testing for latch-up, caution must be exercised when trying to force large currents into an IC. As with any integrated circuit there are maximum limitations to the current handling capabilities of the internal metalization, and diodes, and thus they can be damaged by excessive currents. This is discussed later in the test section. To enable the user to understand what latch-up is and how it has been eliminated, it is useful to review the operating of a simple discrete SCR, and then apply this to the CMOS SCR. Since most latch-up problems historically have been caused by extraneous noise and system transients, the AC characteristics of CMOS latch are presented. Also various methods of external and internal protection against latch-up is discussed as well as example test methods for determining the latch up susceptibility of CMOS IC's. SIMPLE DISCRETE SCR OPERATION To understand the behavior of the SCR structure parasitic to CMOS IC's, it is first useful to review the basic static operation of the discrete SCR, and then apply it to the CMOS SCR. There are two basic trigger methods for this SCR. One is turning on the SCR by forcing current into its gate, and the second is by placing a large voltage across its anode and cathode. Figure 2 shows the basic four layer structure biased into its forward blocking state. The SCR action can be more easily understood if this device is modeled as a cross coupled PNP and NPN transistor as shown in Figure 3. In the case of latch-up caused by forward biasing a diode, if current is injected into the base of Q2, this transistor turns on, and a collector current beta times its base current flows into the base of Q1. Q1 in turn amplifies this current by beta and feeds it back into the base of Q2, where the current is again amplified. If the product of the two transistors' Beta becomes greater than one, B(NPN)xB(PNP) > 1, this current multiplication continues until the transistors saturate, and the SCR is triggered. Once the regenerative action occurs a large anode current flows, and the SCR will remain on even after the gate current is removed, if enough anode current flows to sustain latch-up. However, if the transistor current gains are small no self sustaining positive feedback will occur, and when the base current is removed the collector current will stop. In a similar manner the SCR can be triggered by drawing current by forward biasing the base of Q1.
AN005346-2 AN005346-1
FIGURE 1. Schematic Diagram of Input and Output Protection Structures
FIGURE 2. Simplified SCR Structure
1998 Fairchild Semiconductor Corporation
The second case, the SCR may also be triggered without injecting any gate current. In the forward blocking state the small leakage current that is present does not trigger the SCR, but if the voltage is increased to a point where significant leakage currents start conducting, these currents could also trigger the SCR, again forming a low impedance path through the device. The same requirement that the Beta product of the PNP and NPN be greater than one in order for the SCR to trigger applies here as well. This leakage current trigger is characteristic of Schottkey diode operation. THE CMOS SCR: STATIC DC OPERATION For discussion purposes CMOS SCR latch up characteristics can be divided into two areas. One is the basic operation of the SCR when static DC voltages are applied, and the second is the behavior when transients or pulses are applied. First looking at the device statically, the parasitic SCR in CMOS integrated circuits is much more complex and its triggering is somewhat different than the simple SCR already discussed. However, the regenerative feedback effect is basically the same. Figure 4 shows a simplified P- well CMOS structure illustrating only the diffusions and the resultant parasitic transistors. The NPN transistor is a vertical device whose emitter is formed by n+ diffusions. The P- well forms the base and the N- substrate forms the collector of the NPN. The PNP transistor is a lateral device. Its emitter is formed by p+ diffusions, its base is the N- substrate, and its collector is the P- well.
FIGURE 3. Cross-Coupled Transistor Model of SCR
Figure 5 illustrates a cross section of a simplified N- well process and its corresponding parasitic bipolar transistors. In this process the NPN is a lateral device and the PNP is vertical. Essentially the description of the P- well SCR is the same as the N- well version except the NPN is a low gain lateral device and the PNP is a high gain vertical transistor. Thus the following discussion for the P- well also applies to the N- well with this exception.

0-C     D-L     M-R     S-Z