Mitigation of Single-Event Effects in SiGe-HBT Current-Mode Logic Circuits

It has been known that negative feedback loops (internal and external) in a SiGe heterojunction bipolar transistors (HBT) DC current mirrors improve single-event transient (SET) response; both the peak transient current and the settling time significantly decrease. In the present work, we demonstrate how radiation hardening by design (RHBD) techniques utilized in DC bias blocks only (current mirrors) can also improve the SET response in AC signal paths of switching circuits (e.g., current-mode logic, CML) without any additional hardening in those AC signal paths. Four CML circuits both with and without RHBD current mirrors were fabricated in 130 nm SiGe HBT technology. Two existing RHBD techniques were employed separately in the current mirrors of the CML circuits: (1) applying internal negative feedback and (2) adding a large capacitor in a sensitive node. In addition, these methods are also combined to analyze the overall SET performance. The single-event transients of the fabricated circuits were captured under the two-photon-absorption laser-induced single-event environment. The measurement data clearly show significant improvements in SET response in the AC signal paths of the CML circuits by using the two radiation hardening techniques applied only in DC current mirrors. The peak output transient current is notably reduced, and the settling time upon a laser strike is shortened significantly.


Introduction
Silicon-germanium (SiGe) heterojunction bipolar transistors (HBTs) have received extensive attention for implementation in extreme-environment applications due to their excellent total-ionizing-dose (TID) tolerance, high-speed operation (i.e., high unity gain frequency, f T ), superb cryogenic performance, and ease of integration in a complementary metal oxide semiconductor (CMOS) platform [1][2][3][4]. However, studies have reported that SiGe HBTs are susceptible to single-event phenomena, especially single-event upset (SEU) and single-event transient (SET) [5][6][7], and thus, it is highly critical that the sensitivity of SiGe HBT circuits to single-event effect (SEE) be improved, either via changing the underlying device profile, i.e., radiation hardening by process (RHBP) or circuit and external negative feedback [4]. Since these SET reduction techniques in DC current mirrors are devised to regulate (or suppress) the abrupt AC transient due to single-event phenomena, it is expected that these design techniques within the current mirrors can also improve the SET response through the AC signal path in CML circuits by preventing or suppressing the crosstalk at higher frequencies. In order to validate these claims, four SiGe CML circuits were investigated, as shown in Figure 2. As SET hardening, two hardening techniques utilized in the fabricated CML circuits include an internal negative feedback loop and the addition of a large capacitor. Negative feedback has been proven to lessen the impact of SETs in electronic systems, thereby helping their rapid recovery because it decreases the sensitivity of a system to abrupt changes. Capacitors can be used in forming a low-pass filter, which mitigate undesirable high-frequency transients. In fact, it is a common practice to add large capacitors on supply rails in order to filter out abrupt transients so that remaining electronics do not malfunction. Another way of looking at this from the time-domain perspective is that since the voltage across a capacitor cannot change instantaneously, it mitigates unwanted rapid transients; thus, low-pass filters are also referred to as smoothing filters.
The internal negative feedback is created by adding the degeneration resistor RE (100 Ω) at the emitters of the current mirror ( Figure 2b). The capacitor (12 pF) is connected between the base of the reference device (Q1) of the current mirror and ground ( Figure 2c). Since the voltage across a capacitor cannot change instantaneously, as shown in Equation (1), the rapid transient of the base voltage of the reference device due to SEE can be effectively reduced. Here, iCap(t) is the current through the capacitor, vCap(t) is the voltage across the capacitor, and C is the capacitance.
(1) Figure 1. Current-mode logic (CML) circuit consists of an AC signal path (differential input pair) and a DC current source.
The AC signal path in CML circuits is minimally affected by the tail source at a specific DC bias condition at low frequencies where the impact of impedance variations is negligible (typically, < 10 GHz or higher in advanced technology nodes). However, as the operating frequency increases, the AC differential pairs and the DC current mirrors begin to interact with each other through parasitic components (both capacitive and inductive). It has been shown that the SET response can be significantly improved in DC current mirrors by applying SET reduction techniques such as internal and external negative feedback [4]. Since these SET reduction techniques in DC current mirrors are devised to regulate (or suppress) the abrupt AC transient due to single-event phenomena, it is expected that these design techniques within the current mirrors can also improve the SET response through the AC signal path in CML circuits by preventing or suppressing the crosstalk at higher frequencies.
In order to validate these claims, four SiGe CML circuits were investigated, as shown in Figure 2.
As SET hardening, two hardening techniques utilized in the fabricated CML circuits include an internal negative feedback loop and the addition of a large capacitor. Negative feedback has been proven to lessen the impact of SETs in electronic systems, thereby helping their rapid recovery because it decreases the sensitivity of a system to abrupt changes. Capacitors can be used in forming a low-pass filter, which mitigate undesirable high-frequency transients. In fact, it is a common practice to add large capacitors on supply rails in order to filter out abrupt transients so that remaining electronics do not malfunction. Another way of looking at this from the time-domain perspective is that since the voltage across a capacitor cannot change instantaneously, it mitigates unwanted rapid transients; thus, low-pass filters are also referred to as smoothing filters.
The internal negative feedback is created by adding the degeneration resistor R E (100 Ω) at the emitters of the current mirror ( Figure 2b). The capacitor (12 pF) is connected between the base of the reference device (Q 1 ) of the current mirror and ground (Figure 2c). Since the voltage across a capacitor cannot change instantaneously, as shown in Equation (1), the rapid transient of the base voltage of the reference device due to SEE can be effectively reduced. Here, i Cap (t) is the current through the capacitor, v Cap (t) is the voltage across the capacitor, and C is the capacitance.

Measurement Setup
The single-event transient (SET) measurements were conducted at the U.S. Naval Research Laboratory using a two-photon-absorption (TPA) pulsed-laser single-event effect through the wafer technique [3,4]. The configuration diagram is depicted in Figure 3, together with monitoring terminals. For the present investigation, four CML circuits with hardened and unhardened DC bias blocks (current mirrors) were fabricated using a commercial 130 nm SiGe HBT technology. The devices under test were packaged using a high-speed custom-designed printed circuit board, highfrequency connectors, low-loss cables, DC power supplies, and 67 GHz-capable bias tees. Resulting SETs were captured through a Tektronix high-speed DPO71254 12.5 GHz real-time oscilloscope. In addition, Figure 3 shows a photograph of the actual measurement setup.

Results and Discussion
The current mirror in the reference CML circuit, as shown in Figure 2a, does not utilize any radiation-hardening techniques. The internal negative feedback created by the emitter degeneration resistor RE is employed in the current mirror of the second CML circuit (Figure 2b). The mirror in the

Measurement Setup
The single-event transient (SET) measurements were conducted at the U.S. Naval Research Laboratory using a two-photon-absorption (TPA) pulsed-laser single-event effect through the wafer technique [3,4]. The configuration diagram is depicted in Figure 3, together with monitoring terminals. For the present investigation, four CML circuits with hardened and unhardened DC bias blocks (current mirrors) were fabricated using a commercial 130 nm SiGe HBT technology. The devices under test were packaged using a high-speed custom-designed printed circuit board, high-frequency connectors, low-loss cables, DC power supplies, and 67 GHz-capable bias tees. Resulting SETs were captured through a Tektronix high-speed DPO71254 12.5 GHz real-time oscilloscope. In addition, Figure 3 shows a photograph of the actual measurement setup.
Sensors 2020, 20, x FOR PEER REVIEW 5 of 10 third CML circuit (Figure 2c) utilizes the capacitor between the base of the reference device Q1 and ground. In the fourth CML circuit, both the internal feedback and the capacitor are used in order to improve the SET response of the circuit. In all cases, the supply voltage VCC was set to 2.2 V, the reference current Iref was set to 200 μA, and RC was chosen to be 50 Ω. First, the output transistor Q2 of the DC current mirrors of all four CML circuits in Figure 2 was struck by the TPA laser (Figure 4a), and the transient response at the differential output terminals was measured. The internal negative feedback created by the emitter degeneration resistor RE (100 Ω) in Figure 2b, as reported in Jung et al. [12], improves the SET response of the CML circuit. As a result, the peak transient of the output signal of the CML circuit (solid blue line in Figure 4b) is reduced by 50% compared to the control circuit (dotted black line). The capacitor (12 pF) connected between the base of the reference device of the current mirror (Q1) and ground in Figure 2c reduces the peak transient of the output signal of the CML circuit when the output transistor Q2 of the DC current mirror is struck by the TPA laser; the peak transient is decreased by about 50% (dotted magenta line in Figure 4) compared to that of the CML control circuit (dotted black line). This substantial reduction in SET occurs due to the fundamental nature of capacitors; that is, the voltage across a capacitor cannot change instantaneously. When device Q2 is struck by the TPA laser, the voltage at the base node of Q1 tends to undergo a strong current transient. However, the capacitor prevents this fast transient, endeavoring to maintain the original potential across itself. The best SET improvement is achieved by the CML circuit with the current mirror employing both the internal negative feedback and the capacitor, as shown by the solid red line of Figure 4b. The internal negative feedback together with the capacitor suppresses the peak transient of the output signal by 70% (red solid line in Figure 4).
In order to investigate the effects of these radiation hardening by design techniques implemented in the DC current mirrors on the AC signal path of the CML circuits, the differential input device of the CML circuits (Q4) was also struck by the TPA laser (Figure 5a). Figure 5b shows the measured transient responses of the output signal of the CML circuits when the input device Q4 is struck by the TPA laser. It is clearly shown that the internal negative feedback created by the emitter resistor RE reduces the peak output transient by 35% (solid blue line in Figure 5b) compared to the CML control circuit (dotted black line). The transient current induced by the laser strike in Q4 flows through Q2 and the degeneration resistor RE. This transient current increases the emitter voltage of Q2 (vE2), resulting in a decrease in the base-emitter voltage of Q2 (vBE2). Thus, the collector current of Q2 decrease, as does the collector current of Q4 based on the collector current Equation (2); IS is the reverse saturation current, VT is the thermal voltage (which is about 26 mV at room temperature), and

Results and Discussion
The current mirror in the reference CML circuit, as shown in Figure 2a, does not utilize any radiation-hardening techniques. The internal negative feedback created by the emitter degeneration resistor R E is employed in the current mirror of the second CML circuit (Figure 2b). The mirror in the third CML circuit (Figure 2c) utilizes the capacitor between the base of the reference device Q 1 and ground. In the fourth CML circuit, both the internal feedback and the capacitor are used in order to improve the SET response of the circuit. In all cases, the supply voltage V CC was set to 2.2 V, the reference current I ref was set to 200 µA, and R C was chosen to be 50 Ω.
First, the output transistor Q 2 of the DC current mirrors of all four CML circuits in Figure 2 was struck by the TPA laser (Figure 4a), and the transient response at the differential output terminals was measured. The internal negative feedback created by the emitter degeneration resistor R E (100 Ω) in Figure 2b, as reported in Jung et al. [12], improves the SET response of the CML circuit. As a result, the peak transient of the output signal of the CML circuit (solid blue line in Figure 4b) is reduced by 50% compared to the control circuit (dotted black line). The capacitor (12 pF) connected between the base of the reference device of the current mirror (Q 1 ) and ground in Figure 2c reduces the peak transient of the output signal of the CML circuit when the output transistor Q 2 of the DC current mirror is struck by the TPA laser; the peak transient is decreased by about 50% (dotted magenta line in Figure 4) compared to that of the CML control circuit (dotted black line). This substantial reduction in SET occurs due to the fundamental nature of capacitors; that is, the voltage across a capacitor cannot change instantaneously. When device Q 2 is struck by the TPA laser, the voltage at the base node of Q 1 tends to undergo a strong current transient. However, the capacitor prevents this fast transient, endeavoring to maintain the original potential across itself. The best SET improvement is achieved by the CML circuit with the current mirror employing both the internal negative feedback and the capacitor, as shown by the solid red line of Figure 4b. The internal negative feedback together with the capacitor suppresses the peak transient of the output signal by 70% (red solid line in Figure 4).
In order to investigate the effects of these radiation hardening by design techniques implemented in the DC current mirrors on the AC signal path of the CML circuits, the differential input device of the CML circuits (Q 4 ) was also struck by the TPA laser (Figure 5a). Figure 5b shows the measured transient responses of the output signal of the CML circuits when the input device Q 4 is struck by the TPA laser. It is clearly shown that the internal negative feedback created by the emitter resistor R E reduces the peak output transient by 35% (solid blue line in Figure 5b in a decrease in the base-emitter voltage of Q 2 (v BE2 ). Thus, the collector current of Q 2 decrease, as does the collector current of Q 4 based on the collector current Equation (2); I S is the reverse saturation current, V T is the thermal voltage (which is about 26 mV at room temperature), and V A is early voltage. This negative feedback effect reduces the peak transient as well as the transient time.
Sensors 2020, 20, x FOR PEER REVIEW 6 of 10 VA is early voltage. This negative feedback effect reduces the peak transient as well as the transient time.
(a) (b) Interestingly, however, adding the capacitor alone in the current mirrors does not improve the SET response of the AC signal paths in the CML circuits (dotted magenta line in Figure 5b) compared to the CML control circuit (dotted black line). This is because the transient current induced by the laser current in Q4 flows through Q2 directly into the ground. In other words, the capacitor can only mitigate the rapid transient in the base node of Q2. However, the induced current by the laser strike does not cause a significant transient (or fluctuation) at the base when it flows through the device Q2 (even if it increases the base current slightly). The capacitor does not form any negative feedback loop, either.   1 (2) Interestingly, however, adding the capacitor alone in the current mirrors does not improve the SET response of the AC signal paths in the CML circuits (dotted magenta line in Figure 5b) compared to the CML control circuit (dotted black line). This is because the transient current induced by the laser current in Q4 flows through Q2 directly into the ground. In other words, the capacitor can only mitigate the rapid transient in the base node of Q2. However, the induced current by the laser strike does not cause a significant transient (or fluctuation) at the base when it flows through the device Q2 (even if it increases the base current slightly). The capacitor does not form any negative feedback loop, either.
(a) (b) Figure 5. (a) Transistor Q 4 of the differential input pair of all four CML circuits in Figure 2 was struck by the TPA laser; (b) Transient responses at the positive differential output terminal (Q 4 collector) was measured.
Interestingly, however, adding the capacitor alone in the current mirrors does not improve the SET response of the AC signal paths in the CML circuits (dotted magenta line in Figure 5b) compared to the CML control circuit (dotted black line). This is because the transient current induced by the laser current in Q 4 flows through Q 2 directly into the ground. In other words, the capacitor can only mitigate the rapid transient in the base node of Q 2 . However, the induced current by the laser strike does not cause a significant transient (or fluctuation) at the base when it flows through the device Q 2 (even if it increases the base current slightly). The capacitor does not form any negative feedback loop, either.
As in the four cases in which the output device Q 2 of the DC current mirrors was struck by the laser, the best SET improvement in the AC signal paths was achieved also in the CML circuit with the current mirror that employed both the internal negative feedback and the capacitor, as shown in the solid red line of Figure 5b. The internal negative feedback, together with the capacitor, most effectively suppresses the SET in the AC path (by about 50%). The same negative feedback mechanism applies in the CML circuit with the degeneration resistor R E and the capacitor as in the CML circuit with only R E . The base voltage of Q 2 (v B2 ) tends to increase as the emitter voltage (v E2 ) rapidly increases due to the base-emitter parasitic capacitance. This is, again, because the voltage across a capacitor cannot change instantaneously (the fundamental nature of a capacitor). However, this transient at the base of Q 2 is suppressed by the much bigger capacitor connected between the base and the ground (C B in Figure 2d). This is possible because one end of the capacitor C B is connected to the ground (i.e., fixed DC voltage node); in other words, if the capacitor were floating, the further SET improvement shown in Figure 5 (solid red line) would be less effective.
The experimental results were also verified by the using double-exponential current-injection method [27,28]. The double-exponential current source was modeled following the SET response of 130 nm SiGe HBTs described in [29,30]. The simulated waveforms of the AC signal path due to a strike in the DC current-mirror transistor (Q 2 ) and the differential-pair transistor (Q 4 ) are shown in Figure 6. The simulation shows the similar trends observed from the experiment. When the excess current was injected at the Q 2 transistor, it showed a reduction of 55% in transient peak when applying both R E and C. The other two RHBD techniques also reduced transient peaks of 25% (R E = 100 Ω, without C) and 35% (C = 12 pF, without R E ), respectively. Similarly, when the current was injected in the Q 4 transistor, the simulation results agree with the experimental data. The RHBD techniques of (1) using both R E and C and (2) using R E only without C exhibited 40% and 30% reductions in transient peaks, respectively. However, employing C only did not reduce transient peaks effectively, which was consistent with the experimental results. It should be noted that there are differences in the peaks and duration between the experimental results and simulation data due to limited accuracy in a double-exponential model and the current-injection method.
Sensors 2020, 20, x FOR PEER REVIEW 7 of 10 Figure 5. (a) Transistor Q4 of the differential input pair of all four CML circuits in Figure 2 was struck by the TPA laser; (b) Transient responses at the positive differential output terminal (Q4 collector) was measured.
As in the four cases in which the output device Q2 of the DC current mirrors was struck by the laser, the best SET improvement in the AC signal paths was achieved also in the CML circuit with the current mirror that employed both the internal negative feedback and the capacitor, as shown in the solid red line of Figure 5b. The internal negative feedback, together with the capacitor, most effectively suppresses the SET in the AC path (by about 50%). The same negative feedback mechanism applies in the CML circuit with the degeneration resistor RE and the capacitor as in the CML circuit with only RE. The base voltage of Q2 (vB2) tends to increase as the emitter voltage (vE2) rapidly increases due to the base-emitter parasitic capacitance. This is, again, because the voltage across a capacitor cannot change instantaneously (the fundamental nature of a capacitor). However, this transient at the base of Q2 is suppressed by the much bigger capacitor connected between the base and the ground (CB in Figure 2d). This is possible because one end of the capacitor CB is connected to the ground (i.e., fixed DC voltage node); in other words, if the capacitor were floating, the further SET improvement shown in Figure 5 (solid red line) would be less effective.
The experimental results were also verified by the using double-exponential current-injection method [27,28]. The double-exponential current source was modeled following the SET response of 130 nm SiGe HBTs described in [29,30]. The simulated waveforms of the AC signal path due to a strike in the DC current-mirror transistor (Q2) and the differential-pair transistor (Q4) are shown in Figure 6. The simulation shows the similar trends observed from the experiment. When the excess current was injected at the Q2 transistor, it showed a reduction of 55% in transient peak when applying both RE and C. The other two RHBD techniques also reduced transient peaks of 25% (RE = 100 Ω, without C) and 35% (C = 12 pF, without RE), respectively. Similarly, when the current was injected in the Q4 transistor, the simulation results agree with the experimental data. The RHBD techniques of (1) using both RE and C and (2) using RE only without C exhibited 40% and 30% reductions in transient peaks, respectively. However, employing C only did not reduce transient peaks effectively, which was consistent with the experimental results. It should be noted that there are differences in the peaks and duration between the experimental results and simulation data due to limited accuracy in a double-exponential model and the current-injection method.
(a) (b) Figure 6. (a) Transient responses at the output terminal (Q4 collector) of the differential pair when an excess current was injected in the output transistor Q2 of the DC current mirrors of all four CML circuits in Figure 2; (b) Transient responses at the output terminal (Q4 collector) when an excess current was injected in transistor Q4 of the differential input pair of all four CML circuits in Figure 2. The experiment and simulation results clearly demonstrate that employing RHBD such as applying negative feedback and adding a capacitor in the DC bias blocks can significantly improve SET response even in the AC signal paths. This is a very encouraging result for circuit designers because it is extremely challenging to apply RHBD techniques to AC signal paths in high-speed applications without degrading their overall AC performance, especially as their operating frequencies increase. On the contrary, applying these hardening techniques to a DC bias block has a minimal impact on high-frequency AC performance. Due to the use of relatively large capacitance (12 pF), an obvious penalty induced by the use of these RHBD techniques is increased chip area. The overall performance metrics of the CML with and without the RHBD techniques are summarized in Table 1.

Conclusions
This work demonstrates that the SET response in AC signal paths in SiGe HBT CML circuits can be significantly improved by applying RHBD techniques in DC bias blocks such as current mirrors. This is an important finding because it shows that designers can avoid AC performance degradation while simultaneously radiation hardening their circuits. The internal negative feedback created by the 100 Ω emitter degeneration resistor in the current mirrors reduces the peak output transient of the CML circuit by 50% when the output device Q 2 of the current mirror is struck and by 35% when one of the differential input device Q 4 is struck by the TPA laser. The 12pF capacitor connected between the base of the reference device Q 1 of the current mirror and ground decreases the peak output transient of the CML circuit by 50% when the output device Q 2 of the current mirror is struck by the TPA laser. However, the capacitor alone does not improve the SET response of the output signal of the CML circuits when Q 4 is struck by the TPA laser. The best result in terms of SET response improvement in the CML circuits is achieved by the combination of the negative feedback and the capacitor together in the cases where the DC device Q 2 is struck and when the AC device Q 4 is struck by the TPA laser; the peak output transient of the CML circuits is reduced by 70% when Q 2 is struck and by 50% when Q 4 is struck by the TPA laser, in circuits with both internal negative feedback and the capacitor utilized in the DC current mirrors.