SWITCHING CIRCUIT

Field of the invention

The present invention is generally related to the field of switching circuits suffering from charge injection, as encountered in digital transmitters, high speed DACs and the like.

Background of the invention

Nanoscale CMOS has brought many high speed applications to the consumer thanks to the computing power that can be squeezed into modern signal processors. Unfortunately, the intrinsic analog properties of CMOS transistors do not follow the improvements of their digital counterparts. CMOS transistor parameters like output impedance, supply versus threshold ratio or intrinsic gain typically worsen with the advance of the CMOS technological nodes.

One example of an important application field relates to digital radio transmitters. Modern communication schemes impose tough requirements on radio transmitters. Transmitters operating at RF have to combine hard requirements such as RF bandwidth, linearity and out-of-band noise while maintaining a high efficiency. As a result, the porting of an analog RF transmitter from one technological node to another is complicated and thus slow and costly. Thus, transmitters need to have the least possible analog circuitry. In addition, it is desirable for radio transmitters to be easily scalable with the advancement of CMOS technologies.

To address the problem of analog RF transmitters, a new family of RF transmitters, digital transmitters (also referred to as RF-DACs or Direct Digital RF Modulators, DDRM), has been adopted. The digital transmitters feature a predominantly digital circuitry which is better suited for advanced CMOS technology and which scales much better with the various CMOS technological nodes. In contrast to their analog counterpart, the performance of digital transmitters intrinsically improves with the scaling of CMOS technology.

The first digital transmitters were based on a polar architecture, in which a phase modulated local oscillator (LO) signal is fed to a multitude of DDRM units and amplitude modulation is performed by enabling or disabling (switching on or off) these DDRM unit amplifiers and then combining their output power to form a modulated RF analog signal. Later, Cartesian DDRM architectures consisting of two such digital amplitude modulators, for modulating the in-phase (I) and the quadrature (Q) signals with the respective LO phases, were also adopted in digital transmitters. The outputs of these two digital amplitude modulators are summed before being fed to the antenna for transmission.

An RF DAC is created by summing the outputs of a multitude of switched current sources. In a traditional high speed DAC a differential operation is typically applied and the current of the unit current source is switched to either the positive or the negative side in order to keep the current in the current source constant. To achieve a low instantaneous amplitude during modulation at the RFDAC output a number of unit cells (current sources) is effectively turned off in order to avoid power waste and to keep a high efficiency of the transmitter.

In the paper "A linear 28nm CMOS digital transmitter with 2 x 12bit up to LO baseband sampling and- 58dBc C-IM3" (M. Ingels et al., Eur. Solid State Circuits Conference (ESSCIRC), ESSCIRC 2014, pp. 379-382) a current source based DDRM is disclosed, wherein a very high output impedance is targeted, even at maximal code and high output power. Each unit DDRM cell comprises a current source transistor, which allows achieving operation with linear response. The output current of the DDRM is given by the number of active units cells multiplied by the unit current of the active cells. Fig.1 illustrates an implementation of a unit cell. A switch transistor SW1 is connected in series to a transistor M1 which determines and delivers the current provided by the cell. M1 is biased by the voltage Vb at its gate and acts as a current source. The current provided by this current cell can be turned off by the MOS transistor SW1 which acts as a switch and which has a digital signal at its input. This signal will typically occur at a very high speed. In the paper this switching occurs at an (intermittent) LO speed. According to the switching on or off of SW1, the conversion cell is enabled or disabled. When SW1 is enabled, the current source transistor M1 generates an analog signal at the output of the cell. In other words, the switch transistor SW1 determines whether a current is flowing through the current source transistor M1. The value of that current is mainly determined by the bias voltage Vb applied at the gate terminal of M1 and the dimensions of transistor M1. By adjusting the bias voltage Vb, the value of the current flowing, i.e. the analog signal amplitude can be adjusted. Consequently, digital modulation is achieved by turning on more or less of the cells. Vb controls the level of the digitally modulated signal. It can be seen as controlling the digital transmitter gain. The switch transistor resistance results in some extra power dissipation when a current is flowing through the activated unit cell. Its resistance value should thus be chosen low enough to limit this dissipation. As the bias voltage of transistor M1 and its dimensions determine the current of the active unit cells, it determines the transmitter output power. While the current in the given unit cell is constant when the cell is in operation (and thus contributing to the modulation), the current may be adjusted to control the RMS output power of the transmitter and thus its gain.

Switching activity of the switch SW1 in an RF-DAC unit positioned as in Fig.1, causes the source voltage of the current source transistor M1 to vary rapidly and inject charge into the bias node of the current source through the transistor's gate-source capacitance. This affects the bias voltage of the current source in the unit and, thus, also the current provided by the current source. As a result, the current provided by the current source transistor is no longer the ideal nominal value and the sum of the currents of all the active unit cells is not the correct expected value. When the bias voltage is disturbed by charge injection, it start settling back to its nominal value right after the disturbance, but it does not reach its nominal value within one switching cycle (see Fig.2). This voltage settling on the bias node is determined by the time constant of that bias node, which is typically much longer than the switching speed of the unit cells. The time constant of the bias node is defined by the output impedance at the bias node and the decoupling capacitor Cd, which in turn form a low-pass filter. Indeed, the time constant of the current cell is typically in the frequency range of the baseband signal (MHz range) and is determined by the output impedance of the bias circuit (which may be a simple diode or a more elaborate circuit) and the decoupling capacitor, while the switching speed of the units typically occurs at the rhythm of the output RF frequency (GHz range). When a second charge injection occurs at the next switching event, the bias voltage is still settling and is further deviated from its nominal value (as it was already deviated from the first occurrence and was not able yet to regain its nominal state). If the unit cells had a continuous activity, the bias voltage would eventually settle to a certain equilibrium point (as the drive towards the nominal value gets stronger as the bias voltage deviates more). However, in a DDRM the activity of the various units is depending on the baseband data and activity periods are alternating with periods of inactivity. As a result, periods of switching activity are alternating with cycles with no switching activity. As a result, the current provided by a unit cell after a long inactivity is different from the unit cell current provided after a period of more intense switching activity (in the first case it is closer to the nominal value, in the second case it is further off). As mentioned, the switching activity in a DDRM depends on the baseband data, so the current provided by the unit cells is dependent on the 'history' of the baseband data. This is a memory effect. The current of the unit cells depends on the data applied to the cell before. This results in distortion of the output signal of the DDRM and is difficult to predistort. In a DDRM it is even tougher as the bias voltage is typically shared between many or all the DDRM units. As a result, the bias voltage for a given unit is affected by the activity in other cells, which makes the effect and the eventual predistortion even tougher. Furthermore, the partitioning of the units into binary and thermometer coded groups of units add even more to the distortion (as there is no simple relation with for example the signal amplitude and the amount of charge injection).

In order to avoid the need for predistortion, it is essential to keep the output impedance variation small relative to the load impedance. Therefore, the unit cell further comprises an additional transistor M3 in cascode to the current source transistor M1, as illustrated in Fig.3. As the output impedance of the DDRM can be considered in parallel with the load to the DDRM, the actual load seen by the DDRM is a parallel circuit of the effective load and its own output impedance. The higher this output impedance, the lower its contribution, and the lower the impact of variations (due to the modulation) of this output impedance. Consequently, the increased output impedance further improves the linear operation of the DDRM so that pre-distortion is not (or at least much less) needed. In an RF DAC the transistor M2 in cascode is typically a thick oxide device which protects the fast thin oxide core from a large output swing.

In the implementation shown in Fig.1 and Fig.3, the switch SW1 is placed below (i.e. the drain terminal of the switch is connected to the source terminal of the current source transistor) the current source in order to maximize the overdrive voltage of the switch when it is turned on and so to reduce its on-resistance. As the unit amplifier is turned on and off, the voltage at the source terminal of the current source moves up and down at the rhythm of the switching, with a voltage swing which is close to the overdrive voltage of the current source transistor M1. As the source terminal and the gate of the current source transistor M1 are coupled through its gate-source capacitance, a charge is injected from the source node to the gate node at each edge of the swinging voltage. This creates a voltage spike at the gate of the current source transistor M1, whose amplitude depends on the size of the gate-source capacitance, on the decoupling capacitor used to decouple the gate node and on the source impedance of the circuit providing the bias voltage to the current source. The latter circuit can be as simple as a diode connected transistor followed by a low-pass filter, but it may be more complex. The low-pass filter may consist of a single decoupling cap together with the non-zero output impedance of the (diode) source and is typically provided to stabilize the bias voltage of the current source. After each voltage spike, the bias voltage tends to evolve towards the nominal value, with a settling time determined by the time constant of the low-pass filter.

Unfortunately, as the rising and falling edges at the source terminal of the current source transistor are not equal and as the settling of the bias node is typically much slower than the switching frequency of the switch, the amounts of charge injected onto and extracted from the decoupling capacitance within one switching sequence are not equal and some residual charge will persist. This residual charge is compensated by the bias source, but due to its limited bandwidth, it may not occur within one switching clock cycle (see Fig.2). As a result, when the unit amplifier is switched on a second time, the residual charge, which is still present, causes a slight difference in the bias voltage of the current source and the current at the output will be slightly different. This eventually results in distortion at the RF-DAC output. Moreover, as the residual charge is built up from the subsequent active cycles and as the recovery takes multiple switching cycles, the resulting distortion in a given unit amplifier is dependent on the past activity of this unit amplifier. In other words, there are memory effects in the distortion. The resulting distortion is even more significant when a common bias and/or decoupling is used for multiple unit amplifiers, as the charge injection of one amplifier disturbs the other amplifiers. This is typically the case in a DDRM. However, as predistortion algorithms having to deal with memory effects are complex and limited in effectivity, especially when there is interaction between the various units. this effect should be avoided in the design already.

US application US2010/254490 discloses various implementations of a mixer for a high gain range transmitter. A differential circuit is proposed to implement the mixer. Due to the switching activity, charge is injected through the gate-drain capacitances of the transistors at the local oscillator inputs. In order to compensate for this charge injection additional cross-coupled transistors are provided acting as capacitances. In this way the charge injection from the gate-drain capacitance of a first switching transistor receiving one LO phase can be cancelled through the gate capacitance of the extra crosscoupled transistor used as a capacitor and receiving the opposite LO phase signal and vice versa. In this way the direct charge injection into the signal path due to the switching activity in the LO signal is cancelled.

Hence, there is a need for modifying the RF DAC unit cell so that charge injection is reduced and thus any need for providing complex predistortion is avoided.

Summary of the invention

It is an object of embodiments of the present invention to provide for a RF DAC unit cell that by design is capable of limiting the need for predistortion or even of completely avoiding predistortion.

The above objective is accomplished by the solution according to the present invention.

In a first aspect the invention relates to a conversion circuit according to independent claim 1.

The proposed solution indeed allows for avoiding complex predistortion as the switching circuit for a unit cell now contains two branches so conceived that charge injection into the bias of the first branch is compensated by an inverse charge injection in the second branch.

Other advantageous embodiments of the invention are provided in accordance with the dependent claims.

For purposes of summarizing the invention and the advantages achieved over the prior art, certain objects and advantages of the invention have been described herein above. Of course, it is to be understood that not necessarily all such objects or advantages may be achieved in accordance with any particular embodiment of the invention. Thus, for example, those skilled in the art will recognize that the invention may be embodied or carried out in a manner that achieves or optimizes one advantage or group of advantages as taught herein without necessarily achieving other objects or advantages as may be taught or suggested herein.

The above and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.

Brief description of the drawings

The invention will now be described further, by way of example, with reference to the accompanying drawings, wherein like reference numerals refer to like elements in the various figures.

• Fig.1 illustrates a conventional implementation of a unit cell for a digital amplitude modulator comprising a plurality of such cells.
• Fig.2 illustrates a voltage swing generating glitches on current source bias.
• Fig.3 illustrates an implementation of a unit cell with a cascode transistor.
• Fig.4 illustrates an embodiment of the circuit according to the present invention.
• Fig.5 illustrates an embodiment of the circuit according to the present invention comprising a cascode transistor.
• Fig.6 illustrates various voltage swings in the circuit and the resulting stable bias voltage.
• Fig.7 illustrates an embodiment of the circuit according to the present invention.
• Fig.8 illustrates an embodiment of the circuit according to the present invention.
• Fig.9 illustrates an embodiment of the circuit according to the present invention wherein a bleeder current is added.

Detailed description of illustrative embodiments

The present invention will be described with respect to particular embodiments and with reference to certain drawings but the invention is not limited thereto but only by the claims.

Furthermore, the terms first, second and the like in the description and in the claims, are used for distinguishing between similar elements and not necessarily for describing a sequence, either temporally, spatially, in ranking or in any other manner.

It is to be noticed that the term "comprising", used in the claims, should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. It is thus to be interpreted as specifying the presence of the stated features, integers, steps or components as referred to, but does not preclude the presence or addition of one or more other features, integers, steps or components, or groups thereof. Thus, the scope of the expression "a device comprising means A and B" should not be limited to devices consisting only of components A and B. It means that with respect to the present invention, the only relevant components of the device are A and B.

Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment, but may. Furthermore, the particular features, structures or characteristics may be combined in any suitable manner, as would be apparent to one of ordinary skill in the art from this disclosure, in one or more embodiments.

Similarly it should be appreciated that in the description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.

Furthermore, some embodiments described herein include some but not other features included in other embodiments, as would be understood by those in the art. For example, in the following claims, any of the claimed embodiments can be used in any combination.

It should be noted that the use of particular terminology when describing certain features or aspects of the invention should not be taken to imply that the terminology is being redefined herein to be restricted to include any specific characteristics of the features or aspects of the invention with which that terminology is associated.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In other instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

In the present invention it is proposed to split the single branched traditional unit cell into two branches, which inject charges of opposite sign into the bias node and which are dimensioned (weighed) in such a way that the injected inverse charges effectively neutralize each other. This preferably occurs at an optimal ratio of the two branches.

By placing (see Fig.4) the switch transistor SW2 on top of the current source transistor M2 (i.e. the source terminal of the switch is connected to the drain terminal of the current source transistor), the function of the unit 10 remains the same: it is a current which can be turned on or off by adding a switch in series with the current source, but the voltage at the source terminal of the current source transistor now remains constant. However, now the activity occurs on the drain node of the current source transistor M2. As the drain-gate capacitance is typically smaller than the source-gate capacitance of the same transistor, the charge injection due to this activity is smaller. It is not zero though and a big decoupling capacitor is still needed to keep the bias voltage stable, while memory effects are still present.

The charge injected from the switching at the drain terminal has the opposite sign of the charge injected from switching at the source terminal of the active transistor. It is this observation that has led to the present invention. As this is a serious problem for the DDRM linearity, it is aimed to reduce the charge injected into the bias node by any given unit in the DDRM. Switching activity is inherent to a DDRM and cannot be avoided. This invention reduces the net charge injected into the bias node during the switching activity of the unit by injecting a negative charge into the bias node to compensate for any positive charge injection (and vice versa). This is performed in a clever way by splitting the unit into two sub-units (or two switching branches), which essentially have the same total dimensions, while no extra digital signals (LO or modulated LO) have to be created or provided, thus creating no overhead in area or power consumption.

In the proposed solution the unit cell 10 is divided into two weighted smaller sub-units, such that the sum of the sizes of both sub-units correspond to the original unit size. The first sub-unit (in branch 1) is designed in a configuration of a switch transistor SW1 connected to the source terminal of a current source transistor M1, while the other sub-unit in branch 2 has a switch SW2 connected at the drain terminal of the current source transistor M2. The outputs of both sub-units are combined at node RFout, as shown in Fig.4. Both sub-units are in first order functionally the same: they comprise a current source transistor (M1, M2) in series with a switch (SW1, SW2). Their output currents are combined at the outputs of the branches, which allows achieving the same function as the original cell as presented in Fig 1. However, as discussed above, when the unit cell is turned on or off, the voltages inside the sub-units, i.e. the voltages at nodes SW1out and SW2out, vary in a different way. In branch 1, with the current source below, the source terminal of the current source transistor M1 sees a relatively large voltage swing at its terminal, while in the second branch the voltage swing is present at the drain terminal of the current source transistor M2. Both swings (SW1out, SW2out) have an opposite sign. Hence, the charge injected into the (common) bias node through the Cgs1 of the first sub-unit and the Cgd2 of the second sub-unit has opposite sign as do the voltage swings at the corresponding nodes SW1out and SW2out.

By sizing the ratio of both sub-units correctly, the charge injected by one sub-unit is compensated by the charge injected by the other sub-unit. This ratio is mainly dominated by the ratio between Cgs1 and Cgd2.

The gate-source capacitance Cgs of a MOS transistor is typically larger by a factor of approximately 3 compared to the gate-drain capacitance Cgd of the same transistor. This ratio mainly determines the ratio of the sub-units. If the swing at the source terminal of the current source transistor M1 of the first sub-unit (SW1out) is different from the swing at the drain terminal of the current source transistor M2 of the second sub-unit (SW2out), this also has an impact on the sizing of the sub-units in such a way that a bigger swing requires a smaller capacitor to the gate for a certain charge injection into the bias node. When the sub-units are ideally weighted, opposite charges are injected through the respective Cgs1 and Cgd2 of both sub-units into the bias node, at the same instant in time, resulting in a net zero charge injection. As a result the bias voltage Vb stays constant and the current in the current sources as well (as shown in Fig.6). Furthermore, as the bias voltage does not have to settle back to its nominal value, which typically occurs at a relatively slow time constant of the bias source-decoupling capacitor combination, memory effects caused by bias voltage variations due to past activity of the units are avoided.

In another embodiment a transistor M3 in cascode to the drain terminal of M1 and the source terminal of SW2 may be provided, wherein the cascode transistor receives the output current of the two sub-units. Herein, the output current, RFout, of the unit cell is now at the drain terminal of the cascode transistor. Fig.5 provides an illustration. The output impedance of the digital transmitter is in parallel with its load. As already discussed above, the higher this output impedance, the lower its contribution. In this way the output impedance variation is kept small relative to the load impedance. The cascode transistor also protects the proposed circuit 10, which is typically realised in high speed but low voltage CMOS, from the high voltage swings at the output of the transmitter.

The skilled person will readily understand that it is not essential to have a single output terminal RFout. The output terminal may in certain embodiments of the invention be implemented as two separate terminals, one in each branch.

In a typical use case, for example in polar-based transmitters, the digital signal is a modulated local oscillator signal.

A digital radio transmitter normally contains a multitude of conversion circuits 10 connected in parallel as described above. Usually the digital radio transmitter also comprises a bias circuit 20 to generate the bias voltage signal. The bias circuit comprises a decoupling capacitance Cd at its output. The bias circuit is connected with one or more of the unit cells. Various options can be envisaged. In one embodiment each unit cell is provided with a bias circuit. Alternatively, some of the unit cells or all of them can be connected to a separate bias circuit.

Some alternative unit cell implementations are shown in the remaining figures. In Fig.7 an extra switch in each branch, i.e. SW1' and SW2' respectively, is added in the signal path (i.e. in between the switch transistor SW1,SW2 and the current source transistors M1,M2) which may be activated by a digital baseband signal BB. In Fig.8 an extra switch in each branch is added as well, but at a different position, i.e. below the switch transistor SW1, SW2. These architectures may for example be used when the LO signal is not yet modulated by the digital baseband signal BB so that the actual modulation occurs inside the unit cell by AND-ing the BB and LO signals using switches SW1 and SW1' and switches SW2 and SW2', respectively. Note that in the figures the BB signal time scale can be different from the LO's time scale. Such implementations may be used in quadrature-based transmitters, wherein a first digital signal is a non-modulated local oscillator signal and a second digital signal is a digital BB signal.

Other improvements can be implemented into the unit cell 10 that are also compatible with the idea. For example, in Fig.9 a unit cell is shown, wherein a bleeder current source 30 adds a bleeder current in the signal path branch 1 to keep a small amount of current through the current source M1 and through the cascode M3, if present. The bleeder current source 30 may be implemented as a transistor operating as a current source. The bleeder current increases the settling speed at the source terminals of M1 and, if present, M3 and improves the unit cell performance further. Indeed, even when the bias voltage Vb has been stabilized by nulling the net charge injection, the settling of the internal circuit nodes, i.e. SW1out and SW2out, may still result in memory effects and distortion if this settling is slower than the period of the high frequency signal (i.e. RFin or LO) controlling the switching of SW1 and SW2, for the same reason as discussed with the bias settling. If the settling is not complete before the next switching, the start position of the second switching is different from the first and the behaviour of the cell is slightly different in the second switching than in the first. So, following the same reasoning as before, this results in distortion and memory effects. Keeping the unit cell 10 on with a small current speeds up the internal nodes. Note that in the Fig.9 the bleeder 30 is placed in the first branch and does only pull current through M1 and the cascode M3 (if present) in the off-state of the unit. The second branch does not yield bleeding when turned off. This is done to save some current as a bleeder is less important in the second branch (with the switch at the top), as the nodes in this branch are faster. Indeed, in the second branch the Vgs of the current source (M2) is always present, and so are the charges in the channel of that transistor, both in on and off state. This is not the case in the first branch (and the cascode) where the channel charge disappears when the unit is turned off (without the bleeder). As building up charge is a relatively slow process, keeping the transistors M1 and M3 on (even during the off state of the unit) speeds up the cell. Note that this obviously results in a small DC power consumption. Similarly, a bleeder current source can be provided at the source terminal of M1 in the implementations of Fig.7 and Fig. 8. Additionally, in Fig.8 a bleeder current source can be added at the source terminal of M2 to keep a bleeder current flowing through M2 even when the unit cell is off. Note that in this case an additional current source is needed in parallel to SW2 to provide the current to M2 when the switch SW2 is closed.

While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. The foregoing description details certain embodiments of the invention. It will be appreciated, however, that no matter how detailed the foregoing appears in text, the invention may be practiced in many ways. The invention is not limited to the disclosed embodiments.

In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. A computer program may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.