BACKGROUND
In many wireless communication systems, a mobile wireless device such as a cellular telephone, mobile internet device (MID), personal digital assistant (PDA), laptop, personal computer and so forth communicates wirelessly at a radio frequency (RF). To enable communication in a given wireless communication system, such devices are required to output an RF signal having a given power level. Different wireless systems have different requirements for power levels, and furthermore in a given communication system, power levels may vary depending on communication mode. To generate output power needed for transmitting information, typical mobile devices include a power amplifier (PA) or other such amplifier that receives RF signals from processing circuitry of the device, such as a baseband processor and transceiver or so forth, and amplifies this RF signal to a required power level.
Due to power control requirements in many transmitter systems, average output power can vary over a wide range, for example, several tens of decibels (dBs). This makes a power amplifier's efficiency change dramatically over the power range. Most PA's have maximum efficiency at maximum output power, and efficiency drops rapidly as power is reduced.
SUMMARY OF THE INVENTION
According to one aspect, the present invention includes an apparatus with multiple gain stages and an output network coupled to the gain stages. Each of the gain stages can be independently controlled to amplify a radio frequency (RF) signal to an output power level for transmission from a mobile wireless device such as a cellular handset or so forth. When controlled to be inactive, at least one of the gain stages can be placed into a low impedance state. Each of the gain stages can have substantially different drive point impedances when they are active, to accommodate different output power levels. In one implementation, the output network may include a transformer.
Another aspect of the present invention is directed to a power amplifier that has multiple stages each including pairs of amplifiers and a transformer coupled to each of the stage outputs, where the transformers have different turns ratios. The stages may be controlled such that when a given stage is inactive, it can be placed into a low impedance state, and a net inductance is switched to be coupled in parallel to an output coil of the opposite transformer.
Yet another aspect of the present invention is directed to an integrated circuit that includes at least three gain stages and at least three transformer, each coupled to a corresponding one of the gain stages. At least two of the transformers may have substantially equal turns ratios that in turn are greater than a turns ratio of another of the transformers. The gain stages can be individually enabled, and at least two of the stages may correspond to a high power gain path, while another stage corresponds to a low power gain path.
A still further aspect of the present invention is directed to a power amplifier that has multiple differential complementary amplifier stages and an output transformer. The output transformer may be coupled to receive an amplified output signal from any of the amplifier stages, depending on which is enabled. A turns ratio of the output transformer may also be dependent upon which of the amplifier stages is enabled. In some implementations, the non-enabled amplifier stage(s) can be placed into a high impedance state.
Still further, an aspect of the present invention is directed to an integrated circuit that includes, in addition to multiple gain stages, multiple transformers. The transformers may each have substantially equal turns ratios and include coils formed of inductors integrated on a common complementary metal-oxide-semiconductor (CMOS) die with the gain stages. More specifically, the inductors can be sized to manage inductance variation of a CMOS process, and the transformers can enable the gain stages to see a drive point impedance that is substantially less than a load impedance of a load to which the circuit is coupled.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a conceptual schematic diagram of a power amplifier output stage and output network in accordance with the present invention.
FIG. 2 is a conceptual schematic diagram of an output network in accordance with one embodiment of the present invention.
FIG. 3A is a conceptual schematic diagram of a multi-tap transformer implementation in accordance with one embodiment of the present invention.
FIG. 3B is a conceptual schematic diagram of another multi-tap transformer implementation in accordance with an embodiment of the present invention.
FIG. 3C is an equivalent schematic of the circuit of FIG. 3B in a high power mode of operation.
FIG. 3D is an equivalent schematic of the circuit of FIG. 3B in a low power mode of operation.
FIG. 3E is an implementation of a tuning circuit that can be used in connection with a multi-tap transformer.
FIG. 4A is a conceptual schematic diagram of parallel gain stages and an output network in accordance with an embodiment of the present invention.
FIG. 4B is another conceptual schematic diagram of parallel gain stages and an output network in accordance with an embodiment of the present invention.
FIG. 4C is a conceptual schematic diagram of an alternate input coupling to a gain stage in accordance with one embodiment of the present invention.
FIG. 5A is a conceptual schematic diagram of a reconfigurable output network in accordance with another embodiment of the present invention.
FIG. 5B is an equivalent schematic of the circuit of FIG. 5A in a high power mode of operation.
FIG. 5C is an equivalent schematic of the circuit of FIG. 5A in a low power mode of operation.
FIG. 6A is a conceptual schematic diagram of a reconfigurable output network in accordance with another embodiment of the present invention.
FIG. 6B is an equivalent schematic of the circuit of FIG. 6A in a high power mode of operation.
FIG. 6C is an equivalent schematic of the circuit of FIG. 6A in a low power mode of operation.
FIG. 6D is a conceptual schematic diagram showing a single ended implementation of a reconfigurable output network in accordance with one embodiment of the present invention.
FIG. 7 is a conceptual schematic diagram of a reconfigurable output network in accordance with another embodiment of the present invention.
FIG. 8A is a conceptual schematic diagram of a reconfigurable output network in accordance with yet another embodiment of the present invention.
FIG. 8B is a conceptual schematic diagram of alternate implementation of the reconfigurable output network of FIG. 8A.
FIG. 9A is a conceptual schematic diagram of a power amplifier in accordance with one embodiment of the present invention.
FIG. 9B is a conceptual schematic diagram of a portion of a power amplifier in accordance with yet another embodiment of the present invention.
FIG. 10 is a block diagram of a wireless device in accordance with an embodiment of the present invention.
DETAILED DESCRIPTION
Embodiments may provide an output network that can provide a controllable load impedance. In this way, the output power/efficiency reduction problem can be overcome. By increasing load impedance as the desired output power decreases, efficiency can be kept near a maximum value.
In various embodiments, different implementations of an output network coupled to an output gain stage of a power amplifier or other such amplifier device may be provided. Such output networks may aid in controlling output power while allowing for operation in high efficiency modes. That is, output power (Pout) of an amplifier is related to load resistance and output voltage according to the following equation:
P
out
=V
RMS
2
÷R
L [EQ. 1]
where VRMS corresponds to a RMS output voltage level and RL is the load resistance. The maximum output power, and the maximum efficiency, occurs when VRMS is maximized. The maximum value of VRMS is proportional to the supply voltage, with the actual maximum value determined in part by various performance requirements such as linearity or harmonic level. In different implementations, various manners of controlling power by controlling load resistance can be realized using an output network in accordance with an embodiment of the present invention. For example, an output network may provide high efficiency at two or more power levels. In these cases, the output network can provide optimal (or approximately optimal) loads for the gain stage amplifiers, thereby enabling high efficiency operation at the two or more power levels. Also, the networks provide a means by which different power levels can be selected electronically. Thus, the network can be called a reconfigurable output network. In various embodiments, the impedance presented by disabled gain stages can be used to control, or partially control, the state of the reconfigurable output network.
To enable load variations in accordance with an embodiment of the present invention, a transformer-based technique can be used to electrically reconfigure a transformer of an output network to change its turns ratio. This reconfiguration has the effect of changing the load impedance, and is done so that the transformer's RF performance is maintained.
Referring now to FIG. 1, shown is a conceptual schematic diagram of the final stages of a power amplifier with a reconfigurable load in accordance with one embodiment of the present invention. A final gain stage 100 may include two or more parallel gain stages, A(x) generically, which connect to an output network 101. The gain stages may be of differential or single-ended topology. One or more of the gain stages may be controlled by signals, C2(x). Similarly, the output network may have various control signals, C1(x), that modify the response of the output network. The control signals C2(x) may enable or disable the gain stages and may also control the output impedance presented by the gain stages, Ro(x)(C2). For example, the signal C2(1) may cause the output impedance Ro(1) to be at a high impedance level when gain stage A1 is disabled. Since the various signal paths are connected to network 101, the drive-point impedance of each gain stage is a function of all control signals, hence the name Ri(x)(C1,C2). By changing the control signals C1 and C2, the effective load impedance presented to the various enabled gain stage(s) can be changed, allowing for high efficiency operation at various output power levels. The output network 101 may also provide an impedance match to an external load 102.
Referring now to FIG. 2, shown is a conceptual schematic diagram of an output network in accordance with one embodiment of the present invention. As shown in FIG. 2, output network 10, which may be part of a PA, includes a plurality of different paths or slices. Specifically, multiple amplifiers 20a-20n (generically amplifier 20) each of which may be a gain stage formed of one or more transconductance amplifiers may be present. In one embodiment, gain stages 20 are complementary metal oxide semiconductor (CMOS) stages each including a pair of complementary metal oxide semiconductor field effect transistors (MOSFETs), namely a p-channel MOSFET (PMOS) and an n-channel MOSFET (NMOS). The gain stages may have differential outputs coupled to a primary coil of a corresponding transformer 30a-30n (generically transformer 30). In turn, the secondary coil of transformers 30 are coupled to a load resistance RL. As will be described further below, depending upon a mode of operation, e.g., a high power mode or low power mode, one or more of amplifiers 20 may be inactive. Note that individual transformers can have different turns ratios, or all the same. Unused amplifiers can be forced to be in a high impedance (Hi-Z) or a low impedance (Lo-Z) mode. In different implementations primaries can be strongly coupled to each other or not. If the primaries are strongly coupled, the unused amplifiers are placed in a Hi-Z state or large losses may ensue. If there are different turns ratios, then some unused amplifiers (with low turns ratio transformers at their output) will see large voltages when in Hi-Z state. In other embodiments, the transformer can be weakly magnetically coupled, or not coupled at all. As used herein, with such weakly coupled primaries, the unused amplifier may be placed in a Lo-Z state. In such a case, the Lo-Z state the high voltages may be eliminated, while not causing high signal losses. The primaries can be each made with a center tap, allowing independent control of bias voltages for each amplifier, such that turning off one or more amplifier can reduce quiescent current.
Referring now to FIG. 3A, shown is a conceptual schematic diagram of an output network in accordance with another embodiment of the present invention. Specifically, as shown in FIG. 3A, output network 100 includes a multi-tap transformer 120 having a primary coupled to outputs of a plurality of differential amplifiers 110a-110c (generically amplifier 110). Each gain stage may be a CMOS gain stage formed of a complementary pair of transconductance amplifiers. While shown with only three amplifier paths or slices in the embodiment of FIG. 3A, understand that in other implementations more or fewer such paths are possible. As shown in FIG. 3A, each amplifier 110 has outputs coupled to a different tap location of primary 120. In this way, a turns ratio between primary and secondary of transformer 120 can be adjusted by selecting which of amplifiers 110 are active. Specifically, amplifier 110a, which may correspond to an amplifier of a low power path, is coupled to the full extent of the primary of transformer 120, represented by tap connections at the top and bottom of the primary of transformer 120, and exemplified by the width of spacing 130. Similarly, amplifier 110b may be a medium power amplifier and is coupled to mid-tap connections represented by a width 132. In turn, amplifier 110c, which may correspond to an amplifier of a high power path, may be coupled to tap locations that provide the smallest number of turns, thus providing a largest turns ratio (between secondary and primary) for transformer 120. The width of the primary to which amplifier 110c is coupled is exemplified by width 134.
By selecting one or more of the amplifiers for activation, the turns ratio may be adjusted and accordingly so too is the output impedance, thus controlling output power by switching in the selected one or more amplifiers 110. Note that in the embodiment of FIG. 3A, all amplifiers share a single center-tap. In implementations in which primary taps are highly coupled (i.e., have a coupling coefficient k close to 1), “shorting” any taps to a ground potential or another tap reflects a low impedance across all other taps, resulting in a change of impedance as seen by “active” amplifiers, and high losses in the network. For this reason, unused amplifiers may be put in the Hi-Z mode. Other taps (in Hi-Z) will see a voltage that is “gained up” from a driven path of inner taps, with a gain corresponding (approximately) to the turns ratio between the taps. Note that, while not shown in the figures, the transformers may have capacitors in parallel with the windings to provide an impedance match at a desired operating frequency.
In various implementations, the size (i.e., transistor area) of amplifiers 110b may be related to the area of amplifier 110a according to:
A
110b=(Nlow/Nmed)2×A110a [EQ. 2]
where Nlow is proportional to the number of turns of the primary coil corresponding to the low power path, and likewise for Nmed. Of course, other sizes and configurations of a transformer and gain stages may be implemented in other embodiments.
Referring now to FIG. 3B, shown is a conceptual schematic diagram of another multi-tap transformer implementation in accordance with an embodiment of the present invention. In this embodiment, the differential amplifiers are formed of two approximately identical single-ended amplifiers. As shown in FIG. 3B, output network 100′ includes a multi-tap transformer coupled to multiple amplifiers 110a-110d, pairs of which form differential gain stages. Thus, the pair of single-ended amplifiers 110a and 110d form one differential amplifier. By coupling different ones of these amplifiers 110 to the primary of transformer 120, different turns ratios can be established to effect different output impedances. Each amplifier 110 may be a CMOS amplifier formed of a complementary pair of transconductance amplifiers. As shown in FIG. 3B, amplifiers 110a-110d have corresponding areas A1-A4. In various implementations, the size of amplifiers 110a and 110d may be substantially equal. Similarly, amplifiers 110b and 110c may also be substantially equal. The windings shown in FIG. 3B corresponding to L11, L12 and L13, which form the primary of transformer 120, may have losses and are coupled by a coupling coefficient, k. Assume for purposes of discussion that transformer 120 can have two states: a high turns ratio state, and a low turns ratio state. For a high turns ratio state, amplifiers 110a and 110d are biased “off” so that their output impedance is high. As shown in FIG. 3C, amplifiers 110a and 110d are biased off and are controlled such that their output impedances are high (i.e., Hi-Z). Gain stages 110b and 110c are biased on and are driven differentially. Accordingly, the primary inductance of transformer 120 is L12. Thus the impedance ratio of the transformer (which is approximately the square of the turns ratio) is L2/L12. In various embodiments, L12 may be relatively small such that the turns ratio in this state, which may correspond to a high power state, Nhigh, may be large (4 or higher for example). Note that the mutual inductance in this arrangement between L2 and L12 is maximized. Amplifiers 110b and 110c are “on” and are driven differentially.
In contrast, for the low turns ratio state, as shown in FIG. 3D amplifiers 110b and 110c are biased “off,” i.e., with Hi-Z output impedances. Amplifiers 110a and 110d are “on” and driven differentially. Accordingly, the primary inductance L1 corresponds to the total inductance of the primary of transformer 120 shown in FIG. 3D, namely L1+L12+L13+M, where M is the composite mutual inductance seen by the various primary coils. In various implementations this total inductance L1 may be substantially equal to L2 such that the turns ratio is one or substantially equal to one. In other embodiments, the turns ratio may less than one. Note in this implementation the mutual inductance between L1 and L2 is maximized.
Note that when an unused amplifier is placed into the Hi-Z state, the output capacitance does not significantly change, and thus the unused tap is still tuned. In some implementations, switchable capacitors may be used on other taps to shift tuning while unused.
Referring now to FIG. 3E, shown is an implementation of a tuning circuit that can be used in connection with a multi-tap transformer. Specifically, as shown in FIG. 3E, transformer 120 may have capacitances coupled in parallel on both its primary and secondary sides. Specifically, parallel capacitances C1, C2 and C3 may each be coupled between a pair of nodes that are coupled to intermediate tap points of the primary. In some implementations, capacitors C1-C3 may be implemented using adjustable capacitors. Furthermore, such capacitors may be partially comprised of the output capacitance of the corresponding amplifier (not shown in FIG. 3E).
Referring now to FIG. 4A, shown is a conceptual schematic diagram of parallel gain stages and an output network in accordance with an embodiment of the present invention. As shown in FIG. 4A, circuit 700 includes driver stages 710a and 710b that provide drive signals to a plurality of gain stages 720a-720d (generically gain stage 720). As shown in FIG. 4A, the driver outputs of driver stages 710a and 710b are coupled to gain stages 720 through AC coupling capacitances CACL and CACH, which are coupled in series to the parallel input lines to gain stages 720. Note further, a parallel inductance LD that is center tapped with a common mode voltage VCM is coupled to the output lines of stages 710a and 710b along with a matching capacitance, CMI. As further shown, gain stages 720 may each be biased by bias voltages VQPL and VQNL and VQPH and VQNH for the respective low power gain stages and high power gain stages. By providing individual bias voltages to each input of each gain stage, each gain stage can be individually disabled (i.e., placed in a Hi-Z state using these bias voltages). In various embodiments, CACL and CACH can be adjusted so that the RF signal amplitudes on all inputs to gain stages 720 are equal. The bias voltages may be fed with large-valued resistors which, in one embodiment may be approximately on the order of 10 K Ohms.
Still referring to FIG. 4A, the amplified outputs of gain stages 720 may be coupled through a reconfigurable transformer 730 and provided to an output load RL. As described above, in a low power mode of operation, gain stages 720a and 720d are activated and gain stages 720b and 720c are placed in a Hi-Z state, thus causing a turns ratio of transformer 730 to be LH+LL+MHL:L2, which may be approximately 1:1. (As before, MHL is the mutual of the primary coils which enhances the total L of the primary). Instead in a high power mode, gain stages 720a and 720d are disabled (i.e., in a Hi-Z state) and gain stages 720b and 720c are activated, thus providing an impedance ratio of LH:L2. Note the presence of matching parallel capacitances CML and CMH at the output of gain stages 720, and a corresponding matching capacitance CL coupled in parallel to the secondary of transformer 730. While shown with this particular implementation in the embodiment of FIG. 7A, the scope of the present invention is not limited in this regard.
For example, in other implementations a similar output network may be provided with an input transformer to provide the drive signals to the gain stages. Specifically, as shown in FIG. 4B, circuit 700 may be adapted similarly to that of FIG. 4A. However, note that a first transformer 715 may be provided that couples signals from driver stages 710a and 710b to gain stages 720. Note in this embodiment, the primary of transformer 715 is center tapped with a common mode voltage VCM and the multiple secondaries of transformer 715 are center tapped with bias voltages VQP and VQN, respectively.
Note that in the embodiment of FIG. 4B, the need for coupling capacitances on the input lines to gain stages 720 can be avoided, providing for higher gain in the gain stages. Furthermore, by controlling the turns ratio of input transformer 715, voltage gain can be realized. Furthermore, as described above, bias implementation can be eased by providing a single bias voltage for all NMOS and PMOS devices. Note that switches S1-S4 are provided to switch on or off the high power gain stages 720b and 720c when switching between power modes. Similarly, switches S5-S8 may switch on or off the low power stages 720a and 720d.
To avoid switch losses due to these switches, in another implementation, cascoded output stages may be provided. Referring now to FIG. 4C, shown is a conceptual schematic diagram of the high power gain stages 720b and 720c, which are provided with cascode devices that are driven by the outputs of the two secondary coils of input transformer 715. Thus as shown in FIG. 4C, circuit 700″ shows only the two high power gain stages 720b and 720c. However, similar cascoding may be realized on the low power gain stages. Specifically as shown in FIG. 4C, gain stage 720b includes a NMOS device M1 and a PMOS device M2 coupled as a complementary transconductance amplifier. As further shown, each of these devices has a corresponding cascode device M3 and M4 coupled thereto, each of which is driven with the corresponding bias voltage VBN and VBP, respectively. Similar cascode devices M7 and M8 are cascoded to the complementary amplifier of gain stage 720c, formed of NMOS and PMOS devices M5 and M6, respectively. By controlling the bias voltages VBP and VBN, the corresponding cascode devices can be disabled (i.e., in a Hi-Z state) while the shared output center tap voltage VQP and VQN remains on for the other paths (i.e., the low power path not shown in FIG. 4C). In this way, gain stages 720 can tolerate higher swings, and the need for a voltage regulator can be avoided. Furthermore, other transconductance devices, such as thin oxide devices can be used. Further, while not shown in FIG. 4C, understand that the center taps on the secondaries of input transformer 715 can be shared by the other gain stage paths.
Referring now to FIG. 5A, shown is a conceptual schematic diagram of a reconfigurable output network in accordance with another embodiment of the present invention. As shown in FIG. 5A, output network 200 includes multiple paths, including a first path having amplifiers 210a and 210b and a second path including differential amplifiers 220a and 220b. Generally, path 210 having amplifiers 210a and 210b may correspond to a low power path and path 220 including amplifier 220a and 220b may correspond to a high power path. As shown in FIG. 5A, these parallel paths are coupled to a primary of a transformer 230 that in turn has a secondary coupled to an output load, RL. In the embodiment shown in FIG. 5A, a turns ratio of transformer 230 may be 1:N. Note that the primary of transformer 230 may be center tapped.
Referring still to FIG. 5A, low power path 210 may have an inductance LL coupled in parallel across the outputs of amplifiers 210a-210b. Furthermore, series capacitances CX may couple low power path 210 to a primary of transformer 230. In turn, high power path 220 may have switchable capacitances CH coupled to the outputs of each of amplifiers 220. In addition, an input matching capacitance C1 may be coupled in parallel to the primary of transformer 230. In turn, an output matching capacitance C2 may be coupled to the secondary of transformer 230, in parallel with the output load RL.
Referring now to FIG. 5B, shown is an equivalent schematic in a high power mode of operation in which high power path 220 is active and low power path 210 in inactive. As shown in FIG. 5B in the high power mode, the switchable capacitances may be opened and low power path 210 may be placed in a Lo-Z power state, causing the effective capacitances CX to be coupled in shunt to the outputs of amplifiers 220a and 220b. In this way, a relatively low impedance, depending on the turns ratio of the transformer, may be realized by effectively placing low power path 210 into the Lo-Z state.
Referring now to FIG. 5C, shown is an equivalent schematic for the low power mode of operation. As shown in FIG. 5C, in a low power mode of operation, low power path 210 is activated, and high power path 220 may be placed into a Hi-Z state. Additionally, the switchable capacitances CH may be added in shunt with the primary of transformer 230 by closing the switches. As low power path 210 is activated, inductor LL is also switched in parallel, along with the series-connected capacitances CX, thus providing a further impedance transformation. In this embodiment, transformer 230 provides the same turns ratio in low power mode as in the high power mode. However, the presence of capacitors CX and inductor LL acts as a second impedance transformer to provide another impedance transformation to thus set a load impedance, as seen at the output of amplifiers 210a and 210b. In some embodiments, this load may be approximately 50 ohms, thus setting the maximum output power in the low power mode. Thus in low power mode, there are two impedance transformations, from a nominal 50Ω to an intermediate lower impedance and from the intermediate impedance to a higher final load impedance. This may cause higher loss in low power mode. In low power mode, the inductor LL typically dominates loss in the second transformation. While shown with this particular implementation in the embodiment of FIGS. 5A-5C, understand the scope of the present invention is not limited in this regard. For example, additional pull down NMOSs may be connected to the outputs of the low power amplifiers 210a and 210b to lower impedance. Still further, in other embodiments, the switchable capacitances CH may be controlled to be switched in or out when the high power mode is enabled. Additionally, the switchable capacitor may be combined with a fixed-valued capacitor to provide a controlled range of capacitance. Generalizing the concept illustrated in FIG. 5A, yet another parallel path can be provided for even lower power output connected through another transformation network to the transformer 230.
Referring now to FIG. 6A, shown is a conceptual schematic diagram of a reconfigurable output network in accordance with another embodiment of the present invention. As shown in FIG. 6A, output network 300 includes parallel low power and high power paths. Specifically, the low power path includes amplifiers 310a and 310b, which may be differential amplifiers coupled to a primary of a first transformer 330 having a turns ratio of 1:M. Note that matching capacitances C1 and C3 are coupled in parallel to the primary and secondary, respectively. As further shown in FIG. 6A, the secondary of transformer 330 is also coupled through serial capacitances Cx to a load RL. In turn, the high power path is coupled to this load. Specifically as shown in FIG. 6A, high power amplifiers 320a and 320b are coupled to a primary of a second transformer 340 having a turns ratio of 1:N. Note the presence of matching capacitances C2 and C4 on the primary and secondary, respectively. By controlling which signal path is active, different impedances may be seen at the output of amplifiers 310 and 320, thus enabling operation in different power modes. In an example embodiment, transformer 330 may have a turns ratio of 1:1, and transformer 340 may have a turns ratio of 1:4.
Specifically, when in high power mode the low power path is placed in a Lo-Z state. Referring now to FIG. 6B, shown is a conceptual schematic diagram showing an equivalent circuit of output network 300 in a high power mode. As shown in FIG. 6B, in the high power mode the high power path through transformer 340 is connected. In addition, a parallel capacitance Cx, representing the series capacitance shown in FIG. 6A, is present in series with an inductance LL′, which corresponds to the net inductance on the secondary side of transformer 330 when the low power path is in the Lo-Z state. Accordingly, in this mode the impedance seen at the output of amplifiers 320a and 320b may be low, for example approximately 4 ohms.
Referring now to FIG. 6C, shown is a conceptual schematic diagram showing circuit 300 in a low power mode of operation. As shown in FIG. 6C, in low power mode, the high power path may be placed in a Lo-Z state or a Hi-Z state. In either case, an impedance ZH′ may be in parallel with the load resistance and which corresponds to impedance on the secondary of transformer 340 when the high power path is off. Also note that the serial capacitances Cx are coupled between the secondary of transformer 330 and the load. Thus in low power mode, an intermediate impedance, as seen by the secondary winding of transformer 330, can be provided that is more unconstrained than that of FIG. 5C, discussed above. For example, in some implementations this low power mode intermediate impedance may be closer to the load resistance, e.g., 25Ω. While not shown in the embodiment of FIGS. 6A-6C, in some implementations switchable capacitances may be provided on the primary or secondary side of the transformers, similar to that shown in the high power path of circuit 200 described above with regard to FIG. 5A.
FIG. 6D shows a single-ended implementation. Specifically, circuit 350 as shown in FIG. 6B includes the same transformers and parallel matching capacitances. However, note that only a single series matching capacitance Cx couples the secondary of transformer 330 to the output load RL. While not shown for ease of illustration in FIG. 6D, understand that the multiple amplifiers, for example in FIG. 6A, may be coupled to the primaries of the two transformers.
Generalizing the concept illustrated in FIG. 6A, yet another parallel path for even lower power output can be connected through yet another transformation network to the transformer 340.
Referring now to FIG. 7, shown is another conceptual schematic diagram of an output network 400 including a low power path 410 and a high power path 420, configured similar to output network 300 of FIG. 6A. However, in the embodiment of FIG. 7, the series matching network (shown as series capacitances Cx in FIG. 6A) is not present.
Note that as paths are turned off, the amplifiers can be Hi-Z or Lo-Z. In one implementation transformer 440 may have a turns ratio to provide a low load impedance to amplifiers 420a and 420b, for example 1:4, and transformer 430 may have a turns ratio to provide a higher load impedance to amplifiers 410a and 410b, for example 1:1. In this case, the amplifiers 410a and 410b can be placed in Lo-Z mode when disabled to avoid damage. When a path is in Lo-Z, the transformer from the unused path becomes an inductor of lower value than the winding inductance. For example, if a transformer has primary and secondary windings each having an inductance of L with an unloaded quality factor Q, then with a Lo-Z state on the primary, the equivalent shunt inductance value is approximately L/2 with an effective quality factor of Q/2 (approximately). A similar implementation can be realized as a single-ended circuit.
This shunt inductance of the unused path can be tuned out by capacitors, which may be located on the secondary side, in various embodiments, although voltage concerns may cause the tuning capacitors to be switchable capacitors in the primary side, although switchable capacitors on the primary may increase losses. In some cases, the shunt inductance from the unused path can increase losses. This can be improved by increasing the winding inductance, since RP≈ωoLQ, where RP is the parallel equivalent loss of an inductor. By increasing L, the network losses are decreased. However, by increasing the winding inductance, the losses are increased when the transformer in question is “active.” Thus, there may be a trade-off between those losses. Generalizing the concept illustrated in FIG. 7, yet another parallel path can be provided for even lower power output connected to the secondary winding of transformer 440.
In other implementations, because transformers can become more difficult to implement on die as winding inductances become smaller, as will be described further below with regard to FIGS. 9A and 9B, at least the high power path may be split into a plurality of parallel paths. Referring now to FIG. 8A, shown is a conceptual schematic diagram of an output network in accordance with yet another embodiment of the present invention. As shown in FIG. 8A, circuit 500 may be configured similarly to that discussed above with regard to FIG. 7. However, in the implementation of FIG. 8A, the high power path is split into two parallel paths, namely a first high power path including differential amplifiers 520a and 520b having outputs coupled to a primary of a first high power transformer 540, and a second high power path having differential amplifiers 520c and 520d having outputs coupled to a second high power transformer 550. Note that in the embodiment of FIG. 8A, the winding inductances on both coils of the high power transformers 540 and 550 may have a value of 2LHp:LHs, where LHp is a primary inductance of a high power path (e.g., the single high power path shown in FIG. 7) and LHs is a secondary inductance. In turn, the coils of low power transformer 530 having a primary coil connected to the output of differential amplifiers 510a and 510b, and a secondary coil may have winding inductances of LLp and LLs, corresponding to the low power path winding inductances. As shown in FIG. 8A, the secondaries of each of transformers 530, 540 and 550 is coupled to the output load RL.
In this implementation the high power amplifier areas are each half that of a single high power path (such as that of FIG. 7). By halving the amplifiers, the output capacitance is halved, and the required drive-point (load) impedance is doubled. As such, the transformer windings are doubled. Typically in high power mode, all high power amplifiers 520a-520d are simultaneously enabled. High power/low power operation is otherwise the same as described above in FIG. 7. Splitting the high power path may also provide additional power mode(s), if one pair of high power amplifiers, 520a and 520b for example, are disabled while the other pair 520c and 520d are enabled. There are other alternatives that exist which may increase winding inductances in the high power path, while keeping the low loss topology for the low power path. For example, referring now to FIG. 8B, shown is an alternate implementation of parallel high power paths 520 and 525, in addition to a low power path 510. As shown in FIG. 8B, each path is coupled to an independent transformer 530, 540 and 550. Note however, that the secondaries of transformers 540 and 550 are coupled together in series, and the matching capacitance Cy is coupled in parallel across the secondary of transformer 530, while matching capacitance Cx is coupled in parallel across the combined secondaries of transformers 520 and 525. In other respects, circuit 500′ may operate the same as circuit 500, discussed above.
A transformer in accordance with an embodiment of the present invention may be fabricated in various manners. For example, the primary and secondary windings may each be formed on a metal redistribution layer (RDL). Typically, RDL is formed with relatively thick copper in a “post-processing” step. In contrast, the metal layers provided as part of the standard IC process are typically thinner and may be made of aluminum or copper. Alternatively, all winding may be formed in the standard IC metal layers. The windings may in one embodiment be physically configured as smooth spirals or substantial spirals, such as octagonal or square spirals, or another such shape. In one embodiment, the primary winding may be on a lower RDL layer, and the secondary (output winding) formed on an upper RDL layer, or a combination of RDL and IC metal layers.
As mentioned above, inductance values used in a PA for transformer coils, filters or so forth, can be small, which can be physically difficult to implement in a reliable manner. For example, depending on the supply voltage, desired output power and operating frequency, a winding inductance can be as small as 150 picoHenries (pH). To implement such small values robustly an output stage can be split into at least two equal halves. This doubles the inductor value that needs to be used, making for a robust implementation, making the network less sensitive to various parasitic elements.
By placing multiple inductors in parallel, each inductor has a larger value, e.g., double the value if a pair of inductors are in parallel and there is no mutual inductance. Such inductors may take up twice the amount of area and may be easier to accurately implement. For planar inductors placed next to each other, the magnetic coupling coefficient, k, is quite small (approximately 0.05 or less). A positive k tends to enhance the value of the final effective inductance (Leff), making this effective inductance, Leff greater than 150 pH for example. However, (especially in a differential PA), one could change the flow of current (i.e., clockwise to anti-clockwise) and make k negative, which will decrease Leff. With large geometries, a predictable decrease in inductance can be realized, and the achievable Q may be higher as well, depending on the implementation of the inductors.
Note that only parts of a transformer, e.g., only the primary may have multiple coils in parallel, while the secondary is a single coil. This serves the purpose of combining two parallel paths into a single (secondary) output port as well.
As frequencies get higher (i.e., multi-GHz operation) the inductance (L) and/or capacitance (C) values must get smaller for tuned (resonant) networks. The C is also constrained by the required output power that needs to be delivered (and the chosen technology). A specified output power will imply a transistor area, which in turns implies an amplifier output capacitance. This output capacitance sets a lower limit on C. Hence, as frequency increases, the inductances must become smaller for tuned networks.
Referring now to FIG. 9A, shown is a conceptual schematic diagram of a power amplifier in accordance with one embodiment of the present invention. As shown in FIG. 9A, power amplifier 800 includes a driver stage 810 coupled to receive an incoming signal, Vin. As shown, an input matching network 805, formed of an inductance LM and a capacitance CM is coupled to the input signal line. Driver stage 810 may condition the incoming signal and pre-amplify it to a given level.
Referring still to FIG. 9A, power amplifier 800 includes an output stage that is split into two paths, namely a first path P1 and a second path P2. In the embodiment shown in FIG. 9A, these paths may have components of substantially equal values, the outputs of which are provided to an external output load, e.g., of 50 ohms. The two paths may have insignificant mutual coupling between them, but low coupling is not a required condition.
First signal path P1 includes an interstage matching network 815a including an inductance 2i1 and a capacitance Ci1/2 that is coupled to a gain stage 820a. In this schematic representation, Ci1/2 may include the input capacitance of the amplifiers 820a, any parasitic shunt capacitance, and any required additional extrinsic capacitance. Note that the inductance value may be twice that if only a single path were present, while the capacitance value may be half that of a single path capacitance value. Furthermore, gain stage 820a may be sized with devices having half the size of those present in a single path implementation (and thus provide half the gain). Coupled to output gain stage 820a is an output matching capacitance CO1/2, which again is half the value that would be present in a single path implementation. Again, the output capacitor CO1/2 may include any amplifier output capacitance and any parasitic shunt capacitance, any required extrinsic capacitor.
Still referring to FIG. 9A, the output of gain stage 820a is coupled to a transformer 830a. Note that transformer 830a may be formed of inductors that are twice the value as would be present in a single path implementation (namely 2L11 and 2L22). The secondary coil of transformer 830a is coupled to an output load. Transformer 830a may enable gain stage 820a to have a drive point impedance that is substantially less than an impedance of the output load. In addition, an output matching capacitance CO2/2 is also coupled in parallel to the secondary coil of transformer 830a. Similar components exist in the parallel path P2. While shown with this particular implementation in the embodiment of FIG. 9A, understand the scope of the present invention is not limited in this regard.
Embodiments can be extended to more than two parallel paths, e.g., N paths. Assuming mutual inductances can be neglected, this would result in self-inductance values that are N times larger. If mutual inductances cannot be neglected, then the self-inductances would still be larger, but not exactly N times larger. The drivers and the capacitances are also N times smaller. The former (may) impact the area allocated to the inductors and it may be N times larger. A side benefit is that Q of the composite inductor is enhanced by providing more metal for the inductor. Yet another side benefit is the trace (parasitic) inductance is reduced by the same factor N. While the parasitic capacitance increases, this can be comprehended within the total capacitance needed (i.e., as part of the total capacitance that needs to be implemented).
As discussed above, in such a parallel combination of circuits, each gain stage can be individually controlled to enable/disable varying numbers of stages to scale the output power. As described above, one has to be careful about how the parallel stages are disabled. Ideally a disabled parallel stage should be an “open” or Hi-Z. One way to accomplish this in tuned bandpass systems is to turn off the output stage in a Hi-Z state, and simply resonate out the transformer with an appropriate capacitance.
Referring now to FIG. 9B, shown is a conceptual schematic diagram of a portion of a power amplifier in accordance with yet another embodiment of the present invention showing the details of a disabled gain stage. As shown in FIG. 9B, power amplifier 800′, which may have similar parallel output stages to that of FIG. 9A, shows operation when one gain stage is disabled. When gain stage 820b is disabled, as shown in FIG. 9B, there may be a small change in the output capacitance of the gain stage 820b, represented by ΔC. To compensate for this, ΔC can be added to path P2 so that the output network of the disabled path is properly resonant. Once the disabled path P2 is resonant, the net impedance of the disabled path is a loss term, Rloss. Note that in a parallel connection such as shown, the circuit being disabled is opened (or resonated out), while for a series connection, it may instead be shorted out.
As described above, a reconfigurable output network in accordance with an embodiment of the present invention can be implemented in a PA such as a CMOS PA. Such a device can be used in various wireless systems, including handsets, mobile devices, PDAs and so forth. Referring now to FIG. 10, shown is a block diagram of a wireless device 1000 in accordance with an embodiment of the present invention. As shown in FIG. 10, wireless device 1000 may include an applications processor 1010 which may be a microprocessor or other programmable logic to handle various system features, such as running application programs desired by a user. To perform its functions, applications processor 1010 may communicate with a memory 1015, which may be a flash memory or other non-volatile memory. Applications processor 1010 may further communicate with a display 1020, such as an LCD display of the system. For handling RF communications, e.g., wireless phone calls, wireless data transmissions and so forth, e.g., according to a communications protocol such as EDGE or W-CDMA, applications processor 1010 may communicate with a baseband processor 1030, which may handle baseband operations both for transmit and receive paths. In turn, baseband processor 1030 is coupled to a transceiver 1040, which may receive incoming baseband signals from baseband processor 1030, and perform processing to upconvert the signals to RF levels for transmission to a PA 1050. PA 1050 may be a power amplifier in accordance with an embodiment of the present invention that includes one or more gain stages coupled to a reconfigurable output network as described above. In turn, PA 1050 may be coupled to an antenna switch, duplexer or both 1055 which in turn is coupled to an antenna 1060, which radiates the amplified RF signal.
In a receive path, antenna 1060 couples through antenna switch 1055 and possibly through the duplexer or SAW filters and then to transceiver 1040, which may demodulate the incoming RF signals back to baseband for transmission to baseband processor 1030 for further processing. While shown with this particular implementation in the embodiment of FIG. 10, the scope of the present invention is not limited in this regard.
While the present invention has been described with respect to a limited number of embodiments, those skilled in the art will appreciate numerous modifications and variations therefrom. It is intended that the appended claims cover all such modifications and variations as fall within the true spirit and scope of this present invention.
Patent Application
20090273397
