Method and apparatus for a dynamically self-bootstrapped switch

[0001] The present invention relates, in various embodiments, to switches and systems and methods of dynamically self-bootstrapping the same.

[0002] Switches, such as field-effect transistor (“FET”) or micro-electro-mechanical system (“MEMS”) switches, are used extensively in analog, mixed-signal, and radio-frequency (“RF”) applications. RF applications may use switches for signal routing (e.g., as transmit/receive switches), as switched attenuators, switched resonators, switched capacitors, or as capacitor band-switched voltage-controlled oscillators (“VCO”). A switch may also be used as, for example, a mixer.

[0003] One example of a FET switch is an n-type metal-oxide-semiconductor (“NMOS”) transistor pass-gate, wherein the input signal is received at the transistor's source, the output is generated by the transistor's drain, and the gate and back gate of the transistor control the state of the switch.

[0004] This FET switch may exhibit poor performance at high frequencies, such as RF frequencies. The NMOS transistor may be sized such that its on-resistance (“R_on”) provides a low-resistance signal path from the switch's input to its output. At this size, however, the transistor may have large parasitic capacitances connecting the source and drain to the transistor's gate and back gate. At high frequencies, the gate and back gate may have low impedances, and the parasitic capacitors may short the switch's input signal to these low impedance nodes and thereby limit the bandwidth of the switch. Furthermore, a varying signal applied to the transistor's input may modulate the conductance (G_ds) of the transistor, which may produce intermodulation distortion (“IMD”) and reduce the linearity of the switch. A similar effect may be observed with any switch that capacitively links an input signal to a control node, such as a MEMS switch.

[0005] Resistors may be added in series with the control node(s) to mitigate these problems. The resistors may limit the effects of the parasitic capacitances by raising the impedances of the control node(s), which may prevent the input signal from shorting to the control node(s), thereby reducing signal loss and improving the bandwidth of the switch. In addition, by raising the impedance of the control node(s), the resistors may “free” the control node(s) to track changes in the input voltage applied to the transistor's source—a method of self-bootstrapping. For example, as the FET switch's gate-source voltage (“V_gs”) changes in response the input voltage, the resultant reduction in G_ds variation may reduce IMD.

[0006] These resistors, however, provide a means for the switch's input to couple capacitively to the switch's output, even when the switch is in an off-state. Thus, while the resistors may reduce the on-state loss of the switch, they also undesirably affect the off-state isolation of the switch. A need exists to obtain the benefits of series-connected control node resistors without their drawbacks.

[0007] A circuit in accordance with one embodiment of the present invention connects resistors in series with the control node(s) of a switch, and, by placing bypass switches in parallel with the resistors, may selectively short out the resistors from the circuit. This system and method of dynamic self-bootstrapping disables the bypass switches when the switch is in an on-state, allowing the resistors to self-bootstrap the switch, and enables the bypass switches when the switch is in an off-state. The bypass switches thereby reduce or eliminate the coupling of the voltage at the input of the switch to the control node(s) when the switch is in an off-state. The circuit therefore reduces signal loss and improves the signal bandwidth and linearity of the switch, while also maintaining the off-state isolation of the switch.

[0008] In general, in a first aspect, a dynamically self-bootstrapping circuit for a switch includes a first resistor connecting a control node of the switch to a switch control input. A first bypass switch connects the control node of the FET switch to ground. The first bypass switch is disabled when the switch is in an on-state and enabled when it is in an off-state. An input signal for the first bypass switch may be derived from the switch control input.

[0009] The switch may be a MEMS switch. In an alternative embodiment, the switch may be a FET switch, and the control node may be a gate node of the FET switch. A second resistor and a second bypass switch may connect a back gate node of the switch to ground. The second bypass switch, like the first, may be disabled when the switch is in an on-state and enabled when it is in an off-state.

[0010] The FET switch may be a pass-gate switch, such as an NMOS transistor, a PMOS transistor, or an NMOS transistor in parallel with a PMOS transistor. The FET switch may include a plurality of cascaded transistors. In another embodiment, resistors in series with the gate and back-gate nodes of the bypass switches may self-bootstrap the bypass switches.

[0011] In still another embodiment, the FET switch is an RF switch, such as a transmit/receive switch, a switched attenuator, a multipole/multithrow (MPMT) switch, or a capacitor band-switched VCO. The FET switch may be built in a P-doped well within an N-doped well, and the N-doped well is formed on a P-doped substrate. A first reverse-biased diode may be formed between the P-doped well and the N-doped well, and a second reverse-biased diode may be formed between the N-doped well and the P-doped substrate.

[0012] In general, in a second aspect, a method of dynamically self-bootstrapping a switch includes connecting a first resistor between a control node of the switch and a switch control input. A voltage on the control node is allowed to track an input signal voltage when the switch is in an on-state. The voltage on the control node is prevented from tracking the input signal voltage when the switch is in an off-state by bypassing the first resistor with a first bypass switch.

[0013] The switch may be a MEMS switch. In an alternative embodiment, the switch may be a FET switch, and the control node may be a gate node of the FET switch. A second resistor and a second bypass switch may connect a back gate node of the switch to ground. The second bypass switch, like the first, may be disabled when the switch is in an on-state and enabled when it is in an off-state.

[0014] The FET switch may be a pass-gate switch, such as an NMOS transistor, a PMOS transistor, or an NMOS transistor in parallel with a PMOS transistor. The FET switch may include a plurality of cascaded transistors. In another embodiment, resistors in series with the gate and back-gate nodes of the bypass switches may self-bootstrap the bypass switches.

[0015] In still another embodiment, the FET switch is an RF switch, such as a transmit/receive switch, a switched attenuator, a multipole/multithrow (MPMT) switch, or a capacitor band-switched VCO. The FET switch may be built in a P-doped well within an N-doped well, and the N-doped well is formed on a P-doped substrate. A first reverse-biased diode may be formed between the P-doped well and the N-doped well, and a second reverse-biased diode may be formed between the N-doped well and the P-doped substrate.

[0026] In general, the present invention pertains, in various embodiments, to devices, systems, and methods for dynamically self-bootstrapping a switch. To provide an overall understanding of the invention, certain illustrative embodiments are described, including devices, systems, and methods for dynamically self-bootstrapping a MEMS switch and an NMOS pass-gate switch by using control node resistors and bypass switches.

[0027] FIG. 1A illustrates a circuit 100 in accordance with one embodiment of the present invention. The circuit 100 features an NMOS pass-gate transistor 102 with a signal input 104, an output 106, and a switch control input 108. One of skill in the art will understand that a pass-gate switch is but one possible implementation of a FET switch, and that many other implementations exist. For example, the pass-gate switch may alternatively feature a PMOS transistor, or NMOS and PMOS transistors in parallel. A gate resistor 110 is in series with the gate 112, and a back-gate resistor 115 is in series with the back gate 114. Bypass switches 116, 118 are connected to the gate 112 and back gate 114, respectively, and are controlled by a bypass switch control input 120. In one embodiment, the bypass switches 116, 118 are connected in parallel with the resistors 110, 115. In an alternative embodiment, the bypass switches 116, 118 connect the gate 112 and back gate 114 to low-impedance nodes, such as power or ground. The bypass switch control input 120 may be derived from the switch control input 108 or may be separately generated. In alternative embodiments, the transistor 102 may not have a back-gate connection 114, or the back-gate connection may not be bootstrapped, self-bootstrapped, or dynamically self-bootstrapped.

[0028] The circuit 100 may be used in many applications; for example, the circuit 100 may be used as an RF switch. As an RF switch, the circuit 100 may be used as a transmit/receive switch, a switched attenuator, a multipole/multithrow (MPMT) switch, or a capacitor band-switched VCO. In addition, the circuit 100 may be used as, for example, a mixer.

[0029] In general, MOS transistors, for example the NMOS transistor 102 of the present circuit 100, may exhibit parasitic capacitances 122, 124 between the source 126 and gate 112 of the transistor and between the gate 112 and drain 128 of the transistor. The parasitic gate-source capacitance 122 may allow a signal applied to the signal input 104 to couple to the gate 112. If the gate 112 is a low-impedance node, the voltage on the gate 112 may not change significantly in response to coupling from the signal input 104, and, furthermore, the gate 112 may short out a portion of the voltage on the signal input 104. Adding the gate resistor 110, however, raises the impedance of the gate 112, prevents shorting to the signal input 104, and “frees” the voltage on the gate 112 to track the voltage on the signal input 104. The back-gate resistor 115 similarly prevents shorting to the back gate 114 and frees the back gate 114 to also track the voltage on the signal input 104.

[0030] When the switch 100 is in an on-state, the resistors 110, 115 thus allow the voltage on the signal input 104 to modulate the voltage on the gate 112 and back gate 114, and hence V_gs, thereby reducing the variation in G_ds caused by the varying input signal 104. The reduction in G_ds variation may reduce the IMD distortion caused by the transient signal swing on the source 126 of the pass-gate transistor 102. For example, a transient change in the source 126 voltage of the transistor 102, which would ordinarily cause a nonlinear response in the transistor 102, may be offset by the induced change in the voltage on gate 112. The induced change in V_gs may bring the device back into linear operation.

[0031] When the switch 100 is in an off-state, however, the high-impedance gate 112 and back gate 114 may introduce a deleterious effect. A changing voltage on the signal input 104 may couple to the gate 112 and back gate 114 through the parasitic source-gate capacitance 122, and then to the output 106 through the parasitic gate-drain capacitance 124. The off-state isolation of the switch 100 may therefore suffer. The bypass switches 116, 118 may mitigate this effect by changing the impedance of the gate 112 and back gate 114 while the switch 100 is in an off-state. The bypass switch control input 120 may activate the bypass switches 116, 118 when the switch 100 is in an off-state. The bypass switches 116, 118 may then provide a low-impedance path between the gate 112 and back gate 114 and ground, effectively removing the resistors 110, 115 from the circuit. The low-impedance gate 112 and back gate 114 may therefore prevent a voltage on the signal input 104 from coupling to the output 106.

[0032] FIG. 1B illustrates a circuit 126 in accordance with an alternative embodiment of the present invention. A MEMS switch 128 may be used in a switching application, as discussed above, and may exhibit a first capacitance 122′ between a signal node 104′ and a control node 112′ and a second capacitance 124′ between the control node 112′ and an output node 106′. The capacitances 122′, 124′ may be deliberately added to the circuit 126, or may be intrinsic side-effects of aspects of the design of the MEMS switch 128.

[0033] A resistor 110′ may be connected in series with the control node 112′, between it and a switch control input 108′. As described above, the resistor 110′ may allow the control node 112′ to track changes in the voltage of the signal input 104′ while the MEMS switch 128 is in an on-state. The simultaneously changing signal input 104′ and control node 112′ may thus improve the linearity and bandwidth of the MEMS switch 128 by reducing the variation in the signal-control voltage. When the MEMS switch 128 is off, a bypass switch control input 120′ may activate a MEMS bypass switch 130, providing a low-impedance path between the control node 112′ and ground. Like the circuit of FIG. 1A, the MEMS bypass switch 130 may prevent the control node 112′ from tracking the voltage on the signal node 104′ when the switch is in an off-state, thereby preventing the signal node voltage from coupling to the output node 106′ through the capacitors 122′, 124′.

[0034] FIG. 2 illustrates another embodiment 200 of the present invention. In this embodiment, the bypass switches 116, 118 are self-bootstrapped with resistors 202, 206 in series with their gates and resistors 204, 208 in series with their back gates. The bypass switch resistors 202, 204, 206, 208 may modulate the G_ds of the bypass switches 116, 118 and improve their bandwidth and linearity, just as the resistors 110, 115 do for the transistor 102, as described above.

[0035] The switch 200 may be formed inside of a substrate well in order to isolate the switch 200 from, for example, electrical noise from nearby circuits. In one embodiment, the switch 200 may be formed inside a P-doped well. In another embodiment, the P-doped well may itself be formed inside an N-doped well. Both wells may be formed on a P-doped substrate. In an alternative embodiment, the switch 200 features PMOS transistors and is formed inside a N-doped well.

[0036] The substrate wells surrounding the switch 200 may form PN junction diodes. In one embodiment, a diode 210 may be formed from a P-doped well inside an N-doped well. A second diode 212 may be formed from the N-doped well and the P-doped substrate. Both diodes 210, 212 may be reversed-biased to prevent current flow through them.

[0037] FIG. 3 illustrates a third embodiment 300 of the present invention. In this embodiment, a second pass-gate transistor 302 has been cascaded with the first pass-gate transistor 102. This cascade configuration may be used in high-power applications, where a large difference between the voltages at the signal input 104 and output 106 may induce breakdown effects in a single switch transistor. By cascading the transistors 102, 302, the voltage differential may be shared across both transistors, thereby halving the voltage across a single transistor. In alternative embodiments, three or more transistors may be cascaded together. Each cascaded transistor 102, 302 may be self-bootstrapped with gate resistors 110, 310 and back-gate resistors 115, 315.

[0038] The transistors 306, 308 in bypass switch 304 may also be cascaded together to protect them from high voltages, and may also be self-bootstrapped with resistors. In one embodiment, the bypass switch 304 may be placed at node 310 to provide a low-impedance path between node 310 and ground while the switch 300 is in an off-state. The bypass switch 304 may also be used to bypass the resistors 110, 115 of transistor 102 and/or the resistors 310, 315 of transistor 302.

[0039] FIG. 4 is a flow chart of a process for dynamically self-bootstrapping a switch in accordance with one embodiment of the invention. In the first step 402, a resistor is connected between the control node of a switch and a switch control input. A voltage on the control node is allowed to track an signal input voltage when the switch is in an on-state in the second step 404. Finally, in the last step 406, the voltage on the control node is prevented from tracking the signal input voltage when the switch is in an off-state by bypassing the resistor with a bypass switch.

[0040] FIGS. 5-8 plot on-state loss and off-state isolation versus frequency for various embodiments of the present invention. FIG. 5 shows a graph 500 that plots the characteristics of a FET switch that has neither self-bootstrapping resistors nor bypass switches. A first curve 502 shows the on-state loss of the switch, and a second curve 504 shows the off-state isolation of the switch. At 2 GHz, the loss may be approximately −1.4 dB and the off-state isolation may be approximately −42 dB.

[0041] FIG. 6 shows the effect of adding self-bootstrapping resistors to a FET switch. The resistors may improve the on-state loss of a non-bootstrapped circuit, shown by a first curve 606, to a second curve 602 that represents the on-state loss of a self-bootstrapped circuit. At 2 GHz, the on-state loss may improve from approximately −1.4 dB to approximately −0.2 dB. As explained above, however, this boost in on-state loss may come at the expense of off-state isolation, as reflected in the lower curves 604, 608. The off-state isolation 608 of a non-bootstrapped circuit may fall, as reflected in the curve 604 representing the off-state isolation of a self-bootstrapped circuit. At 2 GHz, the off-state isolation may fall from approximately −46 dB to approximately −9 dB.

[0042] FIG. 7 shows the on-state loss 702 and off-state isolation 704 of a FET switch employing self-bootstrapping resistors and bypass switches, such as, for example, the switch 100 of FIG. 1A. The on-state loss at 2 GHz is approximately −0.3 dB, but the addition of the bypass switches may improve the off-state rejection at 2 GHz to approximately −37 dB.

[0043] FIG. 8 shows the on-state loss 802 and off-state rejection 804 of a FET switch with self-bootstrapping resistors and self-bootstrapped bypass switches, such as, for example, the FET switch 200 shown in FIG. 2. At 2 GHZ, the on-state loss is approximately −0.6 dB, and the off-state rejection improves further to approximately 42 dB.

[0044] It will therefore be seen that the foregoing represents a highly advantageous approach to a self-bootstrapped FET switch. The terms and expressions employed herein are used as terms of description and not of limitation and there is no intention, in the use of such terms and expressions, of excluding any equivalents of the features shown and described or portions thereof, but it is recognized that various modifications are possible within the scope of the invention claimed.