This application is based upon and claims the benefits of the priorities of Japanese patent application No. 2011-026439, filed on Feb. 9, 2011 and Japanese patent application No. 2011-241620, filed on Nov. 2, 2011, each disclosure of which is incorporated herein in its entirety by reference thereto. The present invention relates to a semiconductor device. In particular, it relates to a semiconductor device in which a power supply oscillation noise generated during switching of an output signal is suitably reduced.
Along with the increase in signal transmission rate, assurance of a noise margin and a timing margin in designing a semiconductor device is becoming increasingly difficult year after year. Particularly, in the case of a DRAM (Dynamic Random Access Memory) or the like, to realize high-speed single-ended signal transmission over Gbps (Giga-bits per second), for example, it is essential to reduce the SSN (Simultaneous Switching Noise) generated when output signals from a plurality of output circuits (output buffers) are simultaneously switched. When output signals from a plurality of output circuits are switched simultaneously from one logic level (a low level or a high level) to the other logic level, a large switching noise is generated in high- and low-potential power supplies (VDDQ) and (VSSQ).
The following describes the switching noise briefly based on an example in which output signals from N output circuits are simultaneously switched from a low level to a high level. Assuming that an inductance of a power supply network between a high-potential power supply (VDDQ) and high-potential power supply nodes of the output circuits on a semiconductor LSI (Large Scale Integrated circuit) is LDD, a power supply current is IDD, and currents flowing from the power supply VDDQ to load capacitances Cj (j=1 to N) connected to the output nodes of the N output circuits are I (IDD=ΣIi), a voltage V′DDQ at each high-potential power supply terminal of the output circuits on the semiconductor LSI is represented by the following expression.
Likewise, an example in which output signals from the N output circuits are simultaneously switched from a high level to a low level will be described. Assuming that an inductance of a power supply network between a low-potential power supply (VSSQ) and low-potential power supply nodes of the output circuits is LSS, a power supply current is ISS, and currents flowing from load capacitance Cj (j=1 to N) connected to output nodes of the N output circuits to the low-potential power supply (ground) VSSQ are (ISS=ΣIj), a voltage V′SSQ at each low-potential power supply terminal of the output circuits is represented by the following expression.
Thus, by the SSN switching noise, high- and low-level voltages (VOH) and (VOL) of the output signals from the output circuits on the semiconductor LSI are reduced from VDDQ and VSSQ by LDD×dIDD/dt and LSS×dISS/dt, respectively, resulting in a decrease in noise margin. In addition, for example, because of such signal distortion, signal integrity is deteriorated and timing jitter is increased. In practice, depending on impedance characteristics of the power supply network, the SSN is accompanied by a power supply oscillation noise.
Patent Document 1 discloses a method for switching a pulse period of a pulse generation circuit, in which a first oscillator outputs a first pulse signal having a square wave, a control unit outputs a first control signal for controlling a period of the first pulse signal, a second oscillator outputs a second pulse signal having a square wave, and the control unit outputs a second control signal for controlling a period of the second pulse signal, so that the period is set to be constantly shorter than that of the first pulse signal, and a switching unit receives the first and second pulse signals and selectively outputs one of the first and second pulse signals based on a switching signal supplied from the control unit. According to the method for switching the pulse period of a pulse generation circuit disclosed in Patent Document 1, it is possible to realize a stable pulse generation circuit that generates no pulse-shaped noise at a switching point.
Patent Document 2 discloses a noise reduction circuit for reducing a noise generated by a fluctuation of a reference voltage by simultaneous change of outputs from output circuits on an LSI configured by MOS transistors. The noise reduction circuit includes: a processing circuit that is arranged in a preceding stage of an output buffer that outputs data and that delays input data and supplies the delayed input data to the output buffer; and a delay means that controls an operation timing of the processing circuit based on an input control signal. Under the control of the delay means by an external operation, the processing circuit is switched to operate at a normal timing (output timing 1) or at a delay timing (output timing 2).
Following describes the analysis of the related techniques.
There are two types of power supply noises due to the SSN, as follows:
(A) a spike-shaped noise which is generated when output signals are switched; and
(B) an oscillation noise which is generated by an impedance characteristic of a power supply network of an output circuit.
In the related techniques, a spike-shaped noise having a large noise amplitude is reduced by a noise cancellation circuit. However, along with the increased transmission rate of a signal, the reduction of the oscillation noise (B), which has a great effect on a jitter, has become indispensable.
The above Patent Document 1 is directed only to the pulse-shaped noise (A) generated simultaneously with a switching and hence the oscillation noise (B) cannot be cancelled.
According to Patent Document 2, timings of a plurality of outputs from the LSI are divided into two types, and the switching timings are shifted. However, no matter how the switching timings are shifted, the power supply noise generated when the output buffers are switched cannot be cancelled. In addition, Patent Document 2 does not disclose a means of cancelling the oscillation noise (B).
According to the present invention, which seeks to solve at least one of the above described issues there may be provided the following configuration, but not limited thereto.
According one aspect of to the present invention, there is provided a semiconductor device, comprising:
first and second power supply wirings connected to first and second power supplies, respectively, having power supply potentials different from each other;
an output circuit connected between the first and second power supply wirings, receiving an input signal, and outputting, based on the input signal, an output signal having a logic level corresponding to the first or second power supply voltage to an output node; and
a noise cancellation circuit connected between the first and second power supply wirings and producing a power supply oscillation for a power supply oscillation noise generated in at least one of the first and second power supply wirings when a logic level outputted to the output node of the output circuit is switched, the power supply oscillation being delayed from the power supply oscillation noise by a predetermined time and having a direction opposite to that of the power supply oscillation noise, to make the power supply oscillation and the power supply oscillation noise counteract each other.
According to the present invention, an SSN oscillation noise of output circuits can be reduced. According to the present invention, a single-end output signal can be suitably transmitted at high speed.
Still other features and advantages of the present invention will become readily apparent to those skilled in this art from the following detailed description in conjunction with the accompanying drawings wherein only exemplary embodiments of the invention are shown and described, simply by way of illustration of the best mode contemplated of carrying out this invention. As will be realized, the invention is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the invention. Accordingly, the drawing and description are to be regarded as illustrative in nature, and not as restrictive.
A power supply noise ascribable to the SSN (Simultaneous Switching Noise) which is generated when output signals from a plurality of output circuits (output buffers) on a semiconductor LSI are simultaneously switched, includes an oscillation noise having a grievous effect on a jitter. According to the present invention, a semiconductor device is configured to include a circuit that produce an oscillation that is delayed from the power supply oscillation noise by a predetermined time period and that oscillates in almost exactly in the direction opposite to that of the power supply oscillation noise, as a result of which these oscillation noises cancel out each other. In this way, the deterioration of signal integrity, such as distortion of a signal waveform ascribable to the SSN, can be prevented. For example, a signal can be transmitted at high speed in a single-ended form.
More specifically, a semiconductor device according to the present invention includes a noise cancellation circuit that is connected between first and second power supply wirings and that receives a signal switching an output signal of an output circuit. The noise cancellation circuit produces a damped oscillation (power supply oscillation) that is delayed, from a power supply oscillation noise ascribable to the SSN generated immediately after switching of output signals from output circuits, by half a period of the power supply oscillation noise and that has a phase opposite to that of the power supply oscillation noise ascribable to the SSN. The damped oscillation and the power supply oscillation noises can counteract each other. As a result, the power supply oscillation noise ascribable to the SSN can be reduced.
The present invention is effective in reducing a power supply noise (power supply oscillation noise) generated when multiple-bit parallel output signals are simultaneously switched, as will be described in some of the following exemplary embodiments. However, of course, the present invention is also effective in reducing a power supply noise generated when a single output signal is switched.
The noise cancellation circuit 13 includes an input node 131 for receiving the signal 111 from the preceding stage circuit 11 and an input node 132 for receiving a signal (parameters for setting an operation of the noise cancellation circuit 13: delay amount information and amplitude amount information) 141 from the phase and amplitude adjustment circuit 14. While
The noise cancellation circuit 13 adjusts a timing of an operation relating to the data signal 111 supplied from the preceding stage circuit 11 and a power supply current amount during the operation, based on delay amount information and amplitude amount information supplied from the phase and amplitude adjustment circuit 14. The delay amount information and amplitude amount information will be described with reference to
The resonance characteristic of the noise cancellation circuit 13 may be set by executing an SSN simulation in view of the impedance (inductance component in particular) of a power supply network of an LSI, a package, or a circuit board (PCB: Printed Circuit Board) and the chip capacitance (load capacitance of an output circuit or capacitance between the power supply and the GND, for example), so that the SSN oscillation noise will be canceled for the power supply network. The noise cancellation circuit 13 produces a power supply oscillation (an oscillation that damps exponentially, and an oscillation period is twice the delay time represented by the delay amount information), by delaying the signal 111 supplied from the preceding stage circuit 11 by a delay time represented by the delay amount information (½ of the oscillation period of the signal 111) and by adjusting an amplitude of the signal 111 based on the amplitude amount information. Bottoms of the power supply oscillation signal produced temporally match peaks of the power supply oscillation noise of the SSN generated when the output signal 17 from the output circuit 12 is switched and hence the peaks of the produced power supply oscillation temporally match bottoms of the power supply oscillation noise of the SSN. In this way, inclusive of the first peak (or bottom) of the oscillation noise of the SSN, the power supply oscillation and the power supply oscillation noise of the SSN counteract each other. As a result, the power supply oscillation noise of the SSN is cancelled. A capacitance C connected to the noise cancellation circuit 13 is implemented by a bypass capacitor connected to a power supply. Instead of such on-chip capacitance, the capacitance C may be connected to an external terminal.
In the present exemplary embodiment, the delay amount information and the amplitude amount information set in the noise cancellation circuit 13 may be supplied from the outside of the semiconductor LSI 1. In such case, for example, the delay amount information and the amplitude amount information is supplied from a Mode Resister Set (MRS) (not illustrated) provided in the semiconductor LSI 1. A DRAM includes a mode register for setting various modes and parameters. It is assumed that bit areas for holding the delay amount information and the amplitude amount information are allocated in advance in the mode register and the delay amount information and the amplitude amount information can be set by providing a mode register set command for setting parameters and the like in the mode register to the semiconductor LSI 1. The mode register is set, for example, during initialization when the power supply is turned on or during a reset operation. The mode register can be rewritten in an idle state. The phase and amplitude adjustment circuit 14 includes a storage unit (register) that holds the delay amount information and the amplitude amount information set in the mode registers (not illustrated) and supplies the delay amount information and amplitude amount information held in the storage unit to the noise cancellation circuit 13.
Alternatively, a measurement circuit (such as an on-chip sampling oscilloscope, not shown in the drawing) may be arranged in the semiconductor LSI to measure a power supply noise and to execute automatic adjustment (also termed as self-alignment) in the LSI. In case a measurement circuit (such as an on-chip sampling scope) for monitoring a power supply waveform is arranged in the semiconductor LSI, for example, a power supply noise waveform generated during switching of a DQ signal may be observed, and the period and the damped oscillation term of the waveform may be quantified by a CPU or the like provided in the semiconductor LSI. Alternatively, the semiconductor LSI may be operated by a tester or the like to monitor a power supply wave. In this way, the delay amount information and the amplitude amount information can be acquired based on a monitored SSN oscillation waveform, and the information can be set in the semiconductor LSI.
As illustrated in (i) of
The waveform of the power supply oscillation noise (B) is determined by the anti-resonance characteristic of the power supply network impedance (mostly, the LC parallel resonance (the impedance viewed from the outside is theoretically infinite at a resonance frequency (=½π√{square root over ((LC))})). The power supply oscillation noise (B) oscillates with a certain period (=Td) and damps exponentially (exp (−αt): where α is a damping coefficient (α>0) and t is time). In (i) of
The period Td and the damping coefficient α are dependent on the characteristic (inductance, capacitance, and the like) of the power supply network.
Thus, the period Td and the damping coefficient α each may be dealt as constant or unchanged unless at least one of the LSI (semiconductor), the package (QFP (Quad Flat Package), CSP (Chip Size Package), etc.), and the circuit board (PCB) is changed. Namely, if the LSI, the package, and the circuit board are unchanged, basically, the period Td and the damping coefficient α are identical.
In the present exemplary embodiment, the noise cancellation circuit 13 uses an oscillation noise generated in the power supply because of simultaneous switching of the output circuits 12, to cancel the oscillation noise. Namely, the noise cancellation circuit 13 produces an oscillation noise for cancelling out the power supply oscillation noise ascribable to the SSN, and makes the produced oscillation noise and the power supply oscillation noise ascribable to the SSN counteract each other.
More specifically, the noise cancellation circuit 13 includes a circuit (a circuit for manipulating a power) that produces a power waveform that corresponds to an oscillation noise generated in the power supply, as schematically illustrated in (ii) of
Namely, after the original power supply oscillation noise (B) and the power supply oscillation (C) produced by the noise cancellation circuit 13 are combined together, the power supply voltage exhibits a waveform (power waveform) which is controlled to be kept constant after Td/2, as illustrated in (v) of
Following describes the noise characteristic of the output circuit 12. In the output circuit 12 (which is also referred to as “a DQ output circuit” or “a DQ output buffer”), when an input signal (an output from the preceding stage circuit 11) supplied to a node In is switched from a low level to a high level or from a high level to a low level, a pass-through current (a current flowing from the power supply VDDQ to VSSQ via NMOS and PMOS transistors, termed also as a short-circuit current) flows through NMOS and PMOS transistors forming a CMOS buffer (not illustrated) in the output circuit 12. Namely, a current flows between VDDQ and VSSQ. The waveform of this current is determined by the impedance between the VDDQ node and the VSSQ node of the output circuit 12. More specifically, the shape of the current waveform is mainly determined by the LC parallel resonance caused by parasitic inductances Lv and Lg in the power supply wirings 15 and 16 respectively connected to the power supplies VDDQ and VSSQ of a package (PKG) and a circuit board (PCB) and an on-chip capacitance C in the chip. In addition, the current waveform is a damped oscillatory waveform. This current flows through not only the VDDQ wiring 15 and the VSSQ wiring 16 but also the output circuit 12. Thus, the current results in a noise that disturbs the waveform of the output signal from the output circuit 12.
A resonance angular frequency ωd of the oscillation noise caused by this current is determined as follows:
ωd=1/√{square root over ({(Lv+Lg)×C})}
Based on this resonance angular frequency w d, the delay time by which the noise cancellation circuit 13 delays the signal 111 supplied from the preceding stage circuit 11, that is, the delay time specified by the delay amount information, can be determined as follows:
Td=2π/ωd
In addition, the damping coefficient α in
α≈1(2×C×Rp)
Rp=(Rv+Rg)×Ronc/{(Rv+Rg)+Ronc}
In the present exemplary embodiment, this approximation to hold accurately and hence by executing an SSN simulation and determining the parasitic resistances Rv and Rg parasitized in the power supply wirings 15 and 16 connected to the power supplies VDDQ and VSSQ of the package (PKG) and the circuit board (PCB), the on-chip capacitance C in the chip, and the parasitic resistance Ronc of each of the power supply main wirings VDDQ and VSSQ in the chip, the damping coefficient α can be determined. Quantitatively, if a resistance component is inserted in parallel to the LC parallel circuit, the smaller the value of the resistance component is, the smaller the damping coefficient α will be. On the other hand, if a resistance component is inserted in series with the LC parallel circuit, the larger the value of the resistance component, the smaller the damping coefficient α will be.
By executing a simulation to determine the amplitude of the pass-through current I1 generated between the power supply main wirings VDDQ and VSSQ when the output circuit 12 is operated, the amplitude of the pass-through current I2 generated between the power supply main wirings VDDQ and VSSQ when the noise cancellation circuit 13 is operated can be determined as follows:
I2=I1×exp(−α×(Td/2))
Thus, the resonance characteristic of the noise cancellation circuit 13 can be determined by executing an SSN simulation and determining: the inductances Lv and Lg being parasitic respectively in the power supply wirings 15 and 16 connected to the power supplies VDDQ and VSSQ, of the package (PKG) and the circuit board (PCB); the on-chip capacitance C in the chip, the resistances Rv and Rg being parasitic respectively in the power supply main wirings VDDQ and VSSQ of the package (PKG) and the circuit board (PCB), the parasitic resistance Ronc of each of the power supply main wirings VDDQ and VSSQ in the chip, and the pass-through current I1 generated between the power supply main wirings VDDQ and VSSQ when the output circuit 12 is operated. The determined resonance characteristics are supplied as the delay amount information and the amplitude amount information to the noise cancellation circuit 13.
I2=I1×exp(−α×(Td/2))
where I1 is an amplitude of said pass-through current waveform flowing between said first and second power supply wirings at said output timing of said output circuit (the waveform (c) in
The pass-through current flows between the power supply main wirings VDDQ and VSSQ, when PMOS and NMOS transistors of a CMOS buffer connected between the power supplies VDDQ and VSSQ are simultaneously brought in an on-state.
The dummy buffer 13A shares the power supplies VDDQ and VSSQ and the logic signal input InD outputted from the preceding stage circuit 11 with the DQ output circuit 12A. The dummy buffer 13A has a dummy output node OD, a node SN receiving a serial number signal for setting parameters, and a dummy buffer enable signal node /DE (active in a low level). The dummy output node OD is connected to a capacitance element having a capacitance value of several picofarads (pF). This capacitance value is approximately equal to a load capacitance (not illustrated) of a wiring connected to an output node of the DQ output circuit 12A and of a receiver that receives a DQ signal. To activate (operate) the dummy buffer 13A, a low level is supplied to the node /DE. To inactivate the dummy buffer 13A, a high level is supplied to the node /DE.
The phase and amplitude adjustment circuit 14 outputs information (SN) about phase and amplitude adjustment amounts to the dummy buffer 13A. For example, the phase and amplitude adjustment circuit 14 supplies the delay amount (Td/2) illustrated in
The serial number signal SN in
The dummy buffer 13A includes a transmission gate (TG) 134 that controls a signal flow through the data input node InD, a delay adjustment circuit 133 that is connected to an output node of the transmission gate 134, and an amplitude adjustment circuit 136. The amplitude adjustment circuit 136 includes transmission gates (TG1 to TG3) 134_1 to 134_3 connected to an output node of the delay adjustment circuit 133, and CMOS buffers 135_1 to 135_3, each of which has an input connected to an output node of a corresponding one of the transmission gates (TG1 to TG3) 134_1 to 134_3, has an output connected to the dummy output node OD, and has a different gate width W.
The transmission gate (TG) 134 includes a PMOS transistor PM0 and an NMOS transistor NM0 connected in parallel. A control signal (dummy buffer enable signal /DE) is inverted by an inverter INV and the inverted signal is supplied to a gate of the NMOS transistor NM0. In addition, the control signal (dummy buffer enable signal /DE) is supplied to a gate of the PMOS transistor PM0. When the control signal (/DE) is at a low level, both the transistors PM0 and NM0 are turned on, and the terminal InD and an input node of the delay adjustment circuit 133 are electrically connected. When the control signal /DE is at a high level, both the transistors PM0 and NM0 are turned off, and the node InD is electrically disconnected from the delay adjustment circuit 133. As control signals, the transmission gates (TG1 to TG 3) 134_1 to 134_3 receive the respective bits of the control signal SN<2:0>. For example, when the LSB (Least Significant Bit) of the control signal SN 2:0> is ‘1’, the transmission gate (TG1) 134_1 is turned on. When the 1st bit of the control signal SN<2:0> is ‘1’, the transmission gate (TG2) 134_2 is turned on. When the 2nd bit of the control signal SN<2:0> the transmission gate (TG3) 134_3 is turned on.
The CMOS buffer 135_1 includes a PMOS transistor PM1 having a source connected to VDDQ, and an NMOS transistor NM1 having a source connected to VSSQ, a drain connected to a drain of the PMOS transistor PM1, and a gate connected to a gate of the PMOS transistor PM1. An input node of the CMOS buffer 135_1 is formed by the coupled gates of the PMOS transistor PM1 and the NMOS transistor NM1 and is connected to an output node of the transmission gate TG1. An output node of the CMOS buffer 135_1 is formed by the coupled drains of the PMOS transistor PM1 and the NMOS transistor NM1 and is connected to the node OD. In case the CMOS buffer 135_1 has a unit gate width W (×1), the CMOS buffer 135_2 and 135_3 have gate widths W twice (×2) and four times (×4) as large as the unit gate width W, respectively, though not limited thereto.
When the control signal /DE is at a low level, the transmission gate (TG) 134 is turned on to allow a signal to pass through the node InD and reach the delay adjustment circuit 133. The delay adjustment circuit 133 adds a delay based on a delay amount determined by the control signal SN_D<2:0> and transmits the signal to the next stage. In the example in
Next, the signal delayed by the delay adjustment circuit 133 reaches the transmission gates (TG1 to TG3) 134_1 to 134_3 in the amplitude adjustment circuit 136. The transmission gates (TG1 to TG3) 134_1 to 134_3 control transfer of the signal to the CMOS buffers 135_1 to 135_3, each of which has a different gate width W.
Namely, if the signal SN_A<2:0> activates any one of the transmission gates (TG1 to TG3) 134_1 to 134_3, the activated transmission gate transfer the signal to a CMOS buffer connected thereto. A plurality of the transmission gates (TG1 to TG3) 134_1 to 134_3 may be turned on simultaneously. For example, if all the transmission gates (TG1 to TG3) 134_1 to 134_3 are turned on simultaneously, a total gate width of CMOS buffers is increased to be seven times as large as the unit gate width (CMOS buffer 134_1). Namely, the amplitude adjustment circuit 136 charges and discharges the capacitance element at the node OD, as a COMS buffer having a current driving capability seven times as large (in this case, the amplitude 12 of the pass-through current waveform flowing between the power supplies VDDQ and VSSQ of the dummy buffer in
As illustrated in
As illustrated in
Based on the circuit configuration illustrated in
In case the signal supplied to the nodes S of the selectors A to C from the node SN_D<2:0> is ‘100’, the selector A selects the node b and the selectors B and C select the respective input nodes a. As a result, a signal supplied through the node IN is outputted to the node OUT via L stages of inverters INV and the selectors A to C. The delay time is set by the L stages of inverters.
In case the signal supplied to the nodes S of the selectors A to C from the node SN_D<2:0> is ‘110’, the selectors A and B select the respective nodes b, and the selector C selects the input node a. As a result, a signal supplied through the node IN is outputted to the node OUT via the L stages of inverters INV, a NAND gate (this NAND gate receives the selection signal S inputted to the selector A. Since the selection signal 5, which is supplied to one input of the NAND gate is, ‘1’, the NAND gate functions as an inverter), M stages of inverters, the node b of the selector B, and the node a of the selector C. The delay time is set by (L+M) stages of inverters. If a delay amount of the NAND gate and a delay amount α of the selector are considered, the delay time is set by (L+M+1) stages of inverters +α. Output signal from the selector A reaches the node a of the selector B, the selector B selects the node b and outputs an output signal therefrom.
In case the signal supplied to the nodes S of the selectors A to C from the node SN_D<2:0> is ‘111’, all the selectors A to C select the respective nodes b. As a result, a signal supplied through the node IN is outputted to the node OUT via the L stages of inverters INV, the NAND gate (this NAND gate receives the selection signal S inputted to the selector B. Since the selection signal S, which is supplied to one input of the NAND gate is, is ‘1’, the NAND gate functions as an inverter), the M stages of inverters, a NAND gate (this NAND gate receives the selection signal S inputted to the selector A. Since the selection signal S, which is supplied to one input of the NAND gate is, the NAND functions as an inverter), N stages of inverters, and the node b of the selector C. The delay time is set by (L+M+N) stages of inverters. If a delay amount of the NAND gate and a delay amount a of the selector are considered, the delay time is set by (L+M+2) stages of inverters +2α. Output signals from the selectors A and B reach the nodes a of the selectors B and C, respectively, the selectors B and C select the respective nodes b and output output signals therefrom.
The bit number of the control signal and the number of the selectors are not limited to 3. A greater number of bits and selectors may be used to realize more detailed delay adjustment. Alternatively, the delay adjustment circuit 133 may be configured by using a 2- or lower-bit control signal and two or less selectors.
The 0th bit (LSB) of the control signal SN<2:0> and a signal obtained by inverting the 0th bit by the inverter INV are supplied to a gate of the NMOS transistor NM1_2 and a gate of the PMOS transistor PM1_2 of the clocked inverter 137_1, respectively. If the value of the 0th bit is ‘1’, the clocked inverter 137_1 is activated to output an inverted signal of the input InD to the node OD. If the value of the 0th bit is ‘0’, the clocked inverter 137_1 is turned off.
The first bit of the control signal SN<2:0> and a signal obtained by inverting the 1st bit by an inverter INV are supplied to a gate of an NMOS transistor NM2_2 and a gate of a PMOS transistor PM2_2 of a clocked inverter 137_2, respectively. If the value of the 1st bit is ‘1’, the clocked inverter 137_2 is activated to output an inverted signal of the input InD to the node OD. In case the value of the 1st bit is ‘0’, the clocked inverter 137_2 is turned off.
The 2nd bit of the control signal SN<2:0> and a signal obtained by inverting the 2nd bit by an inverter INV are supplied to a gate of an NMOS transistor NM3_2 and a gate of a PMOS transistor PM3_2 of a clocked inverter 137_3, respectively. If the value of the 2nd bit is ‘1’, the clocked inverter 137_3 is activated to output an inverted signal of the input InD to the node OD. If the value of the 2nd bit is ‘0’, the clocked inverter 137_3 is turned off. If the clocked inverter 137_1 has a unit gate width (W) (×1), for example, the clocked inverters 137_2 and 137_3 are set to have gate widths (W) twice (×2) and four times (×4) as large as the unit gate width (W), respectively. The amplitude of the pass-through current of the dummy buffer is set by the 3-bit values of the control signal SN<2:0>.
The following describes a second exemplary embodiment of the present invention.
In the present exemplary embodiment, the delay circuit 18 delays the output signal 17 from the output circuit 12 and supplies the delayed signal to a dummy output circuit 19. The dummy output circuit 19 is connected between the first power supply wiring 15 (VDDQ) and the second power supply wiring 16 (VSSQ) and includes an output buffer having a configuration identical to that of the output circuit 12. However, the output node of this output buffer is not extended to the outside of the LSI package. While no control nodes for the delay circuit 18 and the dummy output circuit 19 are illustrated in
In case the configurations of the LSI, the package (QFP (Quad Flat Package), CSP (Chip Size Package), etc.), and the circuit board (PCB: Printed Circuit Board) have already been determined and a generated noise waveform can be evaluated in advance (or in case information about a noise waveform has already been obtained by a simulation or the like), a delay amount by the delay circuit 18 and a current amount by the dummy output circuit 19 are determined in advance. The delay circuit 18 can be configured by CMOS inverters cascade-connected in a plurality of stages (for example, an even number of stages). Each of the CMOS inverters is connected between the first power supply wiring 15 (VDDQ) and the second power supply wiring 16 (VSSQ). A variable delay circuit can be configured by changing the stage number of the CMOS inverters. Alternatively, a current source may be inserted in a power supply path of the CMOS inverter, and the bias voltage of the current source may be made variable. In this way, the unit delay time per CMOS inverter stage can be made variable.
The dummy output circuit 19 may be configured by CMOS buffer (CMOS inverter) circuits, as in the case of the output circuit 12. A current amount (power supply current) during an operation, in the dummy output circuit 19 may be adjusted as follows. A plurality (M) of CMOS inverters including transistors each having a smaller size (a smaller gate width W) than that of the CMOS buffer in the output circuit 12 are arranged in parallel, and the power supply current can be adjusted by connecting K-number (1≦K≦M) CMOS inverters out of the M-number of CMOS inverters to the feed system (that is by connecting K-number of CMOS inverters to the power supply path and operating the K-number of CMOS inverters). The gates of the K-number (1≦K≦M) of CMOS inverters in the dummy output circuit 19 may be configured to receive respective delay signals from the delay circuit 18, each having a different delay time.
In the present exemplary embodiment, the SSN oscillation noise is smoothed by causing the delay circuit 18 to delay a signal by (½) Td, where Td is a period of the SSN oscillation noise) and by causing the CMOS inverter of the dummy output circuit 19 that receives an output signal from the delay circuit 18 to make switching based on the delay time.
The following describes a third exemplary embodiment of the present invention.
In the present exemplary embodiment, since the normally-used output circuit 20 (a READ-time inactive output circuit) is used, the increase in the circuit area is suppressed, as compared with the case in which the noise cancellation circuit 13 in
During a READ operation, the switch 23 is turned off and the output is set in a High-Z state (high-impedance state). Alternatively, as the switch 23, a tri-state switch may be used to achieve a short circuit to the power supply of the output circuit 20 or to the ground.
The following describes a fourth exemplary embodiment of the present invention.
While depending on the layout of the package and the layout in the chip, the present exemplary embodiment is effective, for example, when circuits are connected and wirings are installed in the chip and signal wirings are arranged outside the chip, separately on per a byte lane. This is because the noise is saturated per byte lane. That is, since the SSN noise generated in a byte lane is basically independent of (uncorrelated to) other byte lanes, it is only sufficient to cancel the SSN noise per byte lane. In
The following describes a fifth exemplary embodiment of the present invention.
DQ7 I/O, and the noise cancellation circuit 13-2 is arranged between a 4-bit group of DQ8 I/O to DQ11 I/O and a 4-bit group of DQ12 I/O to DQ15 I/O. In
According to the above exemplary embodiments, since the SSN oscillation noise can be reduced, the signal deterioration such as distortion of a signal waveform can be prevented. As a result, high-speed single-ended signal transmission can be realized. The above exemplary embodiments are not limited to high-speed transmission systems and are as a matter of course applicable to relatively low speed operation systems and products. For example, the above exemplary embodiments are also applicable to consumer appliances such as digital consumer products requiring a low jitter.
The disclosures of the above Patent Documents 1 and 2 are incorporated herein by reference thereto. Modifications and adjustments of the exemplary embodiments and examples are possible within the scope of the overall disclosure (including the claims) of the present invention and based on the basic technical concept of the present invention. Various combinations and selections of various disclosed elements (including the elements in each of the claims, examples, drawings, etc.) are possible within the scope of the claims of the present invention. That is, the present invention of course includes various variations and modifications that could be made by those skilled in the art according to the overall disclosure including the claims and the technical concept.
Number | Date | Country | Kind |
---|---|---|---|
2011-026439 | Feb 2011 | JP | national |
2011-241620 | Nov 2011 | JP | national |