ESD at the Circuit Board
Most ESD problems occur when an excessive voltage appears across two or more pins on the chip, usually between signal and ground. Connector pins that are directly connected to a signal pin are especially troublesome. Indirect field magnetic or electric field paths coupling to a chip or a signal loop also exist, but these are usually of much lower intensity.
Let’s take a closer look at the connector pins, and how to deal with them.
Direct vs Indirect ESD
ESD directly to a connector pin is the most devastating (Figure 1A). Signal pins are the most vulnerable – unless the chip pin is well protected, damage is probable. Effects of discharge to a ground pin is less straightforward. ESD currents traveling along the ground path will result in ground bounce somewhere, depending on ground impedance. Ground bounce is usually negligible on a plane, but if the ESD currents cross a gap in the plane or flow out another cable, ground bounce is inevitable. Remember, the ESD currents have to go somewhere – either out another cable or capacitively to a nearby metallic member, often passing through a chip on the way.
If you have a two-layer board (no ground plane), trace impedance is high, and ground bounce becomes a major issue – you must keep ESD currents off a two-sided board at all costs. If you must use two layer boards, you must divert the currents before they get to the board.
Indirect discharge is much less of a damage threat – signal error is the usual result. Figure 1B shows magnetic and electric field coupling paths.
Figure 1. Electrostatic discharge paths
Whichever the situation, we need to keep the voltage difference at the chip pins from getting out of range – minor glitches, as from indirect discharge, usually result in signal error. Major glitches as from direct discharge usually result in damage to the chip, so the currents must be diverted or blocked. Our goal is to select protection methods to minimize adverse voltage across selected pins.
Transient Suppressers vs Filters
Transient suppressors are nonlinear devices (usually two-port) designed to have minimal impact during normal operation, but to clip the excess voltage during an ESD transient. That means that they don’t go into action until the pin-to-pin voltage goes out of normal range, so they are ineffective for preserving signal integrity. Basically, these devices are there to protect vulnerable circuits from damage.
Low pass filters are intended to divert the high frequency ESD transient while passing the desired signal. If the signal band width is low enough, you can preserve signal integrity even during the ESD event.
Series elements, including resistors and ferrites, insert impedance to limit current. This is a good idea in principle, but problematical in practice. The problem is that the device, as mounted, must be able to withstand the over voltage occurring in an ESD event. The device must have an adequate path length so the arc doesn’t simply go around it, pretty much excluding the common surface mount resistors like the 0603 and 0402. If you do find a resistor to meet this criteria, you still need to be concerned about the surface properties on the circuit board – ESD can travel a long ways on the surface.
There are several possible candidates for transient suppression. Your suppression device must respond fast enough to pick off the leading edge of an ESD pulse, which has a rise time of about 1 nS and a pulse width of about 20 nsec. Forget about the gas discharge devices – they work well for the very slow lightning surge, but are far too slow for ESD. Similarly, TVS devices (e.g. SCR), common for handling power surges, are typically too slow for ESD.
The old standby is the SAD (silicon avalanche device) – basically a Zener diode with a sufficiently large cross section to handle the ESD current pulse of the order of 10 amps. Large geometries are accompanied with high capacitance, limiting the signal bandwidth. This device switches very fast, but must be mounted with a low series inductance.
The other option is the MOV (metal oxide varistor), basically a highly nonlinear resistor. These devices are more robust than the TVS, so they take up less room. They don’t switch quite as fast as the TVS, but multilayer MOVs designed for circuit board mounting are fast enough, again, if mounted with low series inductance. These devices don’t have as sharp a knee as the Zener, so they don’t protect the voltage as precisely. Also, the MOV is bipolar, meaning they don’t protect well in the reverse direction.
Figure 2. Voltage and current characteristics of SAD and MOV
Figure 2 compares the current/voltage characteristics of a SAD and an MOV. Note that the SAD is a polar device with quite a sharp breakdown voltage in the reverse direction and nearly zero in the forward direction, making it well suited for protecting typical logic levels or DC voltages. If you want to protect AC, you mount two devices head to foot, making it a bipolar device.
Also, while your semiconductor device may have built in transient protection, it is not very robust – do not rely on it for primary protection.
Sizing Capacitor Filters
We like to use capacitors where feasible – they work for voltage supplies and low frequency analog, but usually not for digital, as they slow the circuit too much.
The capacitor needs to have a large enough value to absorb the available charge. A worse case is easily calculated, as shown in figure 3. In the worst case, all the charge on the capacitor is dumped directly onto the capacitor. As an example, let’s assume a charge voltage of 15 kV and a typical source capacitor value of 150 pF, for a total charge of Q = CV = 2.25 micro coulombs . If we need to limit the voltage on the capacitor to 50 v, then we need a total C = 2.25e-6/50 = 45 nF. Now, this is really a worst case, as the discharge will typically be divided somewhat, but it does illustrate that a small (say, 100 pF) capacitor won’t do the trick.
Where to Ground the Protection Device
Whether we use a transient suppressor or a capacitor, we need to decide how to mount it. Your goal is to divert the current away from the vulnerable circuit, so one leg needs to be connected to the signal or power line, but where does the other end go to? Your choice is usually circuit ground or enclosure ground.
Figure 3. Discharge to a circuit board capacitor
You can protect the chip connecting the device from signal pin to circuit ground. The question then becomes, once you dump the current into circuit ground, where does it go next? In the unlikely case that there are no other paths, then your only concern is with the small self-capacitance of the board, and you won’t have much trouble. If there are other paths, such as another cable, or a capacitive path from the circuit board to a metal member (such as the enclosure), you need to be concerned.
If the ESD current goes out the ground wire in another cable, you will have substantial voltage bounce on the cable. If the ESD current couples to a nearby metal member, you need to be worried about the chip closest to that member – adverse voltages inside the chip can cause upset, even damage.
If you have a metal enclosure, then you should consider grounding the element there. You need to have a very low inductance path for the capacitor or suppressor, otherwise the device won’t perform. Remember that the series inductance in the shunt path needs to be kept as low as possible, difficult to accomplish unless the device is mounted immediately at the connector ground. There is a possible downside, the device works both ways – ESD hits to the enclosure can be diverted to the pin.
Circuit protection is needed to divert ESD currents away from vulnerable devices, mostly semiconductors. Use capacitors if your bandwidth requirements are low, typically not much above the audio range, but you may still benefit up to perhaps one MHz
SADs and multilayer MOVs are the transient devices of choice, as they are fast enough to handle the one nsec ESD rise time.