How to protect super-speed interfaces against damage caused by ESD strikes

Traditionally, ESD protection devices have been valued for a combination of low capacitance and robust operation. The introduction of the SuperSpeed USB connectivity standard, however, saw the development of transceivers that are highly sensitive to the remaining clamping voltage. As a result, the emphasis shifted from the question of how robust an ESD protection device was, to the strength of its protection of the system. 

SuperSpeed USB signals have fundamental frequencies as high as several GHz, which means that a device’s S-parameters become a more appropriate attribute to study than capacitance when evaluating the RF performance of an ESD protection device. In addition, the use of retimers or redrivers shortens the trace lengths between the ESD protection device and the IC it protects, which increases the peak clamping voltages for the same ESD diode. 

And as this article shows, there is much else to consider when designing-in protection for very fast data lines, such as those using the USB 4 or Thunderbolt protocols. 

 

Failure modes of high-speed transceivers

Due to its sensitivity, a transceiver generally fails ESD tests before the limit of an external ESD protection device’s operating parameters is reached, so a low clamping voltage is more important than the absolute robustness of this device.

The protected IC can fail in one of two ways: 

• Its on-chip ESD protection can be damaged due to thermal stress
• The transceiver itself can fail due to peak voltage overstress

The first failure mode, overstressing the on-chip ESD protection, is normally revealed by increased leakage current, an indication of failing on-chip protection. The main cause of damage here is the energy content of the ESD pulse. 

If the transceiver is damaged due to voltage overstress, the leakage of the system might stay low, but the transceiver simply stops working. The main cause of this kind of damage is generally the steep rise time of the ESD pulse, which causes high peak voltages in CMOS gate oxides.

For testing ESD performance, engineers use the IEC 61000-4-2 standard. This specifies an ESD pulse consisting of two elements: a first peak with a steep rise time of 0.6 to 1 ns, and a slower second peak with a higher energy content, shown in Figure 1.

Fig. 1: A standard IEC 61000-4-2 ESD test pulse

Due to its high accuracy, repeatability and reproducibility, the transmission line pulse (TLP) has become the industry standard for analyzing the ESD behavior of single components or complete systems. These very fast TLP signals also allow for accurate peak voltage measurements.

So how should design engineers respond to a failure of an IEC 61000-4-2 test? In fact, it is common to overlook the contribution that the inductance and resistance of the board traces can make to system-level robustness. The best position for an external ESD protection device is immediately behind the connector. This best takes advantage of the inductance and resistance of the signal line between the external ESD protection and the protected IC, shown in Figure 2. 

Fig. 2: An EMI scan of two transceiver boards. On the left, the ESD protection device is in the middle of the board; on the right, it is immediately behind the connector. EMI radiation, which is proportional to ESD stress, is significantly lower when the protection device is immediately behind the connector.

With the external ESD protection unchanged, an increase in line inductance will decrease the peak voltage at a protected IC. Just 10 mm of differential SuperSpeed lines adds between 3 nH and 3.5 nH of extra inductance. 

In USB 4 systems, retimer or redriver devices are a popular way to maintain signal integrity. Since these devices are typically positioned close to the connector, the trace inductance between the ESD protection device and the protected IC is reduced. While this improves signal integrity, the new position needs ESD protection with better peak pulse suppression.

This calls for a study of the quasi-static I(V) curves in the data sheet of an ESD protection device, representing the clamping voltage for each current in a TLP test. While standard TLP pulses have a duration of 100 ns to show the same energy content as a corresponding IEC 61000‑4‑2 pulse, these values are ‘static’ from the point of view of a modern ESD protection device, which will have reached its full clamping performance after a few nanoseconds. Unfortunately, these curves do not show the peak voltage, which becomes obvious when looking at voltage behavior over time, as shown in Figure 3.

Fig. 3: Measurement of the voltage of a 100 ns TLP in the time domain. The I(V) diagram in datasheets is typically sampled between 70 ns and 90 ns, but will not register the actual peak voltage.

Figure 4 compares two ESD protection devices that have the same RF performance. Looking at the quasi-static I(V) curves on the left, it would be easy to assume that the device represented by the red line would offer better peak pulse suppression, due to its lack of a trigger voltage. The actual peak voltages in very fast TLP measurements with a 600 ps rise time, however, show that the trigger voltage is actually not the dominant factor. 

Fig. 4: The quasi-static 100 ns TLP diagram on the left compares two protection devices with similar clamping behavior. The right I(Vpeak) diagram from 0.6 ns to 5 ns in very fast TLP measurements shows that the peak voltage is not dominated by the trigger voltage but by the switching speed of the protection devices. 

To assess the ability of an ESD protection device to protect a given system, fully dynamic System-Efficient ESD Design (SEED) simulations are becoming standard. The term ‘fully dynamic’ means that the dynamic turn-on behavior of the protection device is taken into account. SEED simulations can substantially reduce time to market.

The peak pulse suppression performance of an ESD protection device is influenced by the switching speed and inductance of an ESD protection device. While this reinforces the view that low inductance in an ESD protection device helps to improve ESD clamping, it is less obvious that this inductance, which might for example be attributable to bond wires, also weakens its RF performance. 

The inductance and capacitance of an ESD protection device form a band-stop filter are shown in Figure 5.

Fig. 5: The parasitic inductance of an ESD protection device turns it into an LC band-stop filter

Figure 6 compares the calculated insertion loss of an ideal capacitance, with a dashed line, to the measured insertion loss of a wire-bonded device, with a solid line, that has the same capacitance at 10 GHz.

 Fig. 6: Comparing a calculated pure capacitance, with a dashed line, to a measured device, with a solid line, shows that inductances turn ESD protection devices into band-stop filters with reduced RF performance compared to pure capacitances.

As can be seen clearly, the RF performance of the real device is notably lower due to the band-stop behavior of the added inductance. This means that the RF performance of ESD protection devices should be evaluated over S-parameters rather than just the device capacitance, since parasitic inductances can significantly reduce RF performance in the GHz range.

Conclusions

• Always maximize the inductance between an ESD protection device and the protected IC
• Fully dynamic SEED simulations allow the engineer to evaluate complete systems against peak voltage and thermal overstress failure modes, and can help to reduce time-to-market
• The RF performance of ESD protection devices should be evaluated using S-parameters, since the capacitance and inductance of a protection device both have a marked influence in the GHz range