BALUNS FOR RF SIGNAL CONVERSION AND IMPEDANCE MATCHING

Abstract

A magnetically-coupled structure is integrated with an integrated circuit in back end-of-line (BEOL) digital CMOS fabrication processes. A differential primary (or secondary) coil is formed by patterning a thick copper (Cu) metal layer, and a single-ended secondary (or primary) coil is formed by patterning a thick aluminum (Al) top metal bonding layer. Crossovers and/or cross-unders are formed using thin metal layers. One embodiment provides a stacked balun with a differential primary input winding defined in the copper layer, directly underneath a single-ended spiral winding defined in the aluminum layer. The spiral forms the single-ended secondary output of the balun and is rotated by 90° to prevent metal shorting for its cross-under connections. Another embodiment provides a transformer with one differential primary (or secondary) coil defined in the copper layer and another differential secondary (or primary) coil defined in the aluminum layer and adding a center tap. The position of the tap is selected to compensate for phase differences and provide desired balance.

Description

BACKGROUND

This disclosure relates to magnetically coupled primary and secondary coil structures such as balun and transformer devices incorporated together with other components of integrated circuit and similar devices.

Radio frequency (RF) circuitry is implemented using differential signaling in order to eliminate common mode noise and increase dynamic range. However, real life signals like those received by an antenna or transmitted by an antenna are single-ended. So, somewhere in the receiver (RX) and transmitter (TX) circuitry, the signal is converted from single-ended to differential or from differential to single-ended. A device that utilizes magnetically coupled primary and secondary coil windings to convert radio frequency (RF) signals that are balanced about a common mode (viz., ground) reference (i.e., differential signals) to signals that are unbalanced (i.e., single-ended signals), and vice versa, is called a “balun.” This conversion is typically done right after the low noise amplifier (LNA) stage for the RX chain and right after the power amplifier driver (PA Driver) stage for the TX chain. It can also be done right before the LNA and right before the PA Driver, respectively.

In addition to providing signal conversion, baluns are also used to provide impedance transformation between preceding and following stages. This allows for maximum power and signal transfer. Impedance matching is accomplished by changing the turns ratio between the two sides (primary and secondary) of the balun. This provides the ability to either step up or step down the impedance. Further, the physical value of the balun inductance may be an important parameter in enabling resonance to be achieved at the operating frequency. Especially in low voltage, high current applications, the DC resistance of the primary side may be important because it translates into lost headroom, thereby impacting linearity.

It is generally desired to keep the loss of baluns to a minimum. However in many digitally-oriented advanced deep submicron processes, back end-of-line (BEOL) processing steps typically use very thin metal layers in order to maximize the efficiency of routing for digital applications. And, with continuing downscaling of process dimensions, the thickness of metals used in BEOL layer stacks continues to decrease. Such thin metals, however, have very high RF loss (poor quality factor Q). In many situations, only one thick copper (Cu) metal layer may be made available for power routing. A thick aluminum (Al) redistribution layer (RDL) is also usually available for bonding (for wirebond or stud-bump flip-chip packaging). Copper is generally not currently used for bonding because of its high malleability. Aluminum, however, has higher resistivity and very poor electromigration characteristics.

In general, higher frequency baluns, such those implemented for wireless local area network (WLAN) applications at 2.5 GHz and 5.8 GHz frequencies, may be easier to realize on chip because of smaller area (need for small inductance) and higher Q with higher frequency. Low frequency cellular band baluns are generally more difficult to realize and are typically implemented in a single in-line package (SIP) configuration in an integrated product development (IPD) process or on a printed circuit board (PCB). See, J. Mondal, et al., “Design and Characterization of an Integrated Passive Balun for Quad Band GSM Applications,” IEEE Electronic Components and Technology Conference, pp. 534-540, 2006, incorporated herein by reference.

The conventional approach to integrating baluns together with other components on a same chip has been to use a stacked arrangement of primary and secondary coil windings of two superposed differential inductors, with one coil formed using a first thick copper metal layer and the other coil formed using a second thick copper metal layer. At least two additional metals are needed to implement the cross-unders/crossovers for the two differential coils, respectively. The accomplishment of such approach in an integrated circuit fabrication flow, such as a modern CMOS process flow, that otherwise requires only a single copper layer is, however, disruptive and expensive because additional metal layer and masking steps have to be added that would not otherwise be necessary.

To avoid the need for adding another copper layer and masking step into a chip fabrication process with only one thick copper metal layer available, on-chip baluns have been implemented using coplanar configurations wherein the windings of primary and secondary coils of two differential inductors run side-by-side and magnetically couple in a coplanar fashion. In the coplanar implementation, the single thick copper layer is used for the side-by-side placement of the coil windings, with the overlying thick aluminum and underlying thinner copper layers being used for the crossover and the cross-under connections of the interwoven windings, respectively. The magnetic coupling of coplanar coil windings is, however, inferior to that of stacked windings and, because the coplanar coupling factor (k) becomes severely degraded for higher turns ratios. Further, for coplanar arrangements, a higher turns ratio implies larger area and longer crossovers. This can degrade the loss of the balun further since the longer crossovers are implemented typically using the poorer conductivity aluminum. Because of the coplanar coupling, a 2:1 turn ratio is typically the highest that is implemented.

SUMMARY

In one aspect, the invention provides a balun comprising a differential primary (or secondary) inductor winding and a single-ended secondary (or primary) inductor formed in a stacked configuration. In advantageous configurations, the balun is formed using two thick layers of metal available in processing steps utilized for the fabrication of integrated circuits or other devices fabricated with semiconductor or similar processing steps.

In an example implementation, the balun is formed using metal layers available in, for example, a digital CMOS fabrication process, with the differential primary (or secondary) inductor formed at the level of the single available thick copper (Cu) metal layer, the single-ended secondary (or primary) inductor formed at the level of the single available thick aluminum (Al) top metal bonding layer, and crossovers (cross-unders) formed using available underlying thin metal layers. An illustrative embodiment described below provides a high magnetic coupling, low loss, stacked balun using the thick aluminum bonding metal layer to provide one multi-turn metal winding over another multi-turn metal winding provided using the thick copper layer. The thick copper is used to realize a differential primary input winding that resides directly underneath a single-ended spiral winding using the aluminum. The spiral forms the single-ended secondary output of the balun and is rotated by 90° so as to prevent any metal shorting for its cross-under connections. Occupying an area of 0.078 mm² on a digital 65 nm process, a five-turn primary with a three-turn secondary has a measured coupling of 0.94 and a total balun loss of 1.55 dB at 1845 MHz.

In another aspect, the invention provides a transformer comprising a differential primary (or secondary) inductor winding and a differential secondary (or primary) inductor formed in a stacked configuration. In advantageous configurations, the transformer is formed using two thick layers of metal available in processing steps utilized for the fabrication of integrated circuits or other devices fabricated with semiconductor or similar processing steps, following an approach similar to the approach used to form the balun. In an example implementation, the transformer is also formed using metal layers available in, for example, the digital CMOS fabrication process, with one differential primary (or secondary) inductor formed at the level of the single available thick copper (Cu) metal layer, and the other differential secondary (or primary) inductor formed adding a center tap in a structure similarly formed at the level of the single available thick aluminum (Al) top metal bonding layer by adding a center tap to a structure such as used in the single-ended secondary (or primary) inductor formed for the balun. The position of the tap is selected to compensate for phase differences and provide the desired balance.

In another aspect, the invention provides methods of fabricating baluns, transformers, and other devices having metal windings.

BRIEF DESCRIPTION OF THE DRAWINGS

Example embodiments are described for illustrative purposes with reference to accompanying drawings, wherein:

FIG. 1 is a block diagram of a radio frequency (RF) transmitter circuit of a type to which the principles of the invention may find application;

FIG. 2 is a schematic representation of layers in an example digital CMOS integrated circuit fabrication process flow;

FIGS. 3A and 3B are a top view and an enlarged partial view of a differential inductor implemented in a integrated circuit fabrication flow process;

FIGS. 4A-4C are top, cross-sectional and schematic layer views of a balun implemented in a stacked differential inductor configuration;

FIGS. 5A and 5B are top and schematic layer views of a balun implemented in a coplanar differential inductor configuration;

FIGS. 6A-6B are top and schematic layer views of a balun implemented in a stacked single-ended on differential inductor configuration;

FIGS. 7-9 are graphs helpful in understanding parameters of the implementation of FIGS. 6A-6B;

FIG. 10 is a micrograph of a balun constructed as shown in FIGS. 6A-6B;

FIGS. 11A-11B are top views of a transformer implemented in accordance with the principles applied in the fabrication of the balun of FIGS. 6A-6C; and

FIG. 12 is a schematic diagram of showing two coupled differential inductors

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

FIG. 1 is a block diagram of a typical wireless communication transmitter (TX) arrangement 10 of a type to which embodiments of the invention find application. A digital input signal of voice and/or data information codified into in-phase (I) and quadrature (Q) components (viz., digital baseband signal) is received by a digital transmitter (TX) circuit block 12 (such as from the output of a digital baseband integrated circuit or DBB chip) which does filtering and other conventional digital signal processing functions. The filtered and processed digital data is then passed to a digital-to-analog (D/A) converter block 14 for conversion to an analog signal which is then filtered by a low pass filter block 16. The filtered analog data signal is then passed to a mixer for modulation of a transmission frequency signal provided by a local oscillator path. In the case of a typical cellular telephone frequency band wireless transmitter (e.g., GSM Edge, Wideband CDMA, ETSI LTE, or similar), the local oscillator path may have a local oscillator running at a given (e.g., 4 GHz) frequency, a /N divider for dividing (e.g., by 2 or 4) to provide a divided transmission frequency (e.g., 2 or 1 GHz) at a local oscillator buffer (viz., amplifier) which provides the carrier signal for modulation. Following the mixer, the modulated analog signal is passed to a first stage variable gain amplifier (VGA) which connects to a transformer. The transformer is used to match the impedance between two stages of amplification, the output impedance of the first stage VGA (VGA Stage 1) and the input impedance of a second stage VGA (VGA Stage 2). As indicated by dual signal lines in FIG. 1, the transformer receives the modulated analog signal at two inputs (generally designated as + and −) referenced to a common mode (viz., some reference voltage, e.g., ground or supply) in differential input signal form, and provides the impedance balanced modulated analog signal at two outputs (+ and −) referenced to the common mode (ground) in differential output signal form. The differential output signal is then converted by a balanced signal-to-unbalanced signal converter (commonly referred to as a “balun”) from its differential signal form to a single-ended output signal (viz., a single + or − output signal referenced to ground). The balun, in effect, performs a signal subtraction between its + and − inputs. The single-ended output signal is then passed to a power amplifier (PA) for amplification and wireless transmission (at, e.g., 2 watts power) from an antenna. For the illustrated configuration, the power amplifier is located off-chip (i.e., not located on the same integrated circuit chip as the other described transmitter components) and is indicated schematically as an RC load (parallel resistor and capacitor). An off-chip arrangement may, for example, be useful where the power amplifier is fabricated using one semiconductor fabrication process (e.g., GaAs process) and the other components are fabricated using another fabrication process. (e.g., Si and/or SiGe CMOS, etc., process)

The location of the balun may be varied to suit individual preferences and circumstances. For the configuration illustrated in FIG. 1, the balun is located after the power amplifier driver (PA Driver) circuitry at the output of the VGA Stage 2. The balun may also be located elsewhere in the transmitter chain, such as before the VGA Stage 2. (And, although shown incorporated into a same chip with other transmitter components formed, e.g., in a digital CMOS fabrication process, baluns according to principles of the invention are also candidates for formation off-chip together with the power amplifier itself, e.g., in a GaAs fabrication process.)

Baluns have been integrated with other components on the same die using stacked arrangements of primary and secondary coils formed in superposed relationships in different thick copper (Cu) layers, with readily usable other metal layers available for crossovers and cross-unders. The integration of baluns with other components on the same chip in many modern device fabrication processes is, however, hampered by restrictions in the number and thicknesses of available metal layers. For example, a typical modern digital CMOS process flow used for fabricating wireless communication transmission/receiver channel components may have only two thick metal layers (viz., thicknesses greater than or equal to 1 μm) available: a single thick interconnect layer in copper and a top metal layer of typically aluminum used for bonding. All other metal layers are thin lower layers (viz., thicknesses less than 1 μm) used for digital routing and having typical thicknesses being in a range of about 0.1 to 0.3 μm, or less. Examples of available metal layers/thicknesses are given in Table 1, below, with optional levels/thicknesses shown in parentheses. The layers of a digital CMOS C014 (DRPu) process are shown schematically in FIG. 2. This process flow, for example, has available a single copper (Cu) metal layer (M7) of about 2.4 μm thickness, a top metal aluminum (Al) bonding layer (METTOP or ALCAP) of about 1.26 μm thickness, and six thin metal layers (M1-M6) of about 0.15 μm thickness or less, separated by interlevel dielectric layers of typically about 0.11 μm thickness.

	C027(DRP2)	C021(DRPe)	C014(DRPu)	C28	Comp.
	(90 nm)	(65 nm)	(45 nm)	(28 nm)	(130 n)
M1	0.25 μm	0.14 μm	0.095 μm	0.1 μm	0.32 μm
M2	0.25	0.16	0.15	0.1	0.32
M3	0.25	0.16	0.15	0.1	0.32
M4	0.25	0.16	0.15	0.1	0.32
M5	0.25	0.16	0.15	0.1	0.8
M6	1	1.6	0.15	0.1	3
M7			2.4	1.6 (2.4)
(M8)			(2.4)	(2.4)
ALCAP/	1	1	1.26	1.26	1.2
METTOP

FIG. 12 is a schematic diagram showing two coupled differential inductors which are magnetically coupled and usable for differential-to-single-ended conversion and impedance matching. Each of the primary and secondary coil windings have positive and negative input/output terminals (IN_P+, IN_P− for the primary coil and OUT_S+, OUT_S− for the secondary coil), and each has a center tap (CT_P for the primary, and CT_S for the secondary) which, for differential signals, is connected to the common mode voltage (typically ground or supply). For balun operation, the single-ended side is connected with the center tap left open and one of the input/output terminals (typically the negative terminal) grounded. Thus, with the primary connected to receive the differential input signal and the secondary connected to deliver the single-ended output signal (no DC current), the secondary center tap CT_S is left open and the secondary negative output terminal OUT_S− is connected to ground. In example implementations, self-resonance=>capacitance=>tuning range, frequency coverage; and DC resistance impacts Pout and ACLR (voltage headroom) 50 mA on 2 ohms=>1.445V/1.55V=>0.3 dB Pout hit, 0.6 dB ACLR hit.

FIGS. 3A and 3B illustrate a differential inductor implementation for a single primary or secondary coil integrated within the flow of an integrated circuit fabrication process. The illustrated configuration shows a coil winding having three turns. At least two metal layers are used to implement the windings. One metal layer (viz., thick copper layer M7) is used for implementation of the three turns. The turns are coplanar and concentric, evenly-spaced with uniform widths, with each successive turn having a greater radius and being spaced by a same interval spacing from the next adjacent turn. As shown, the turns are implemented in octagonal form with increasing lengths of sides of the octagons. Other polygon or curved configurations or uneven spacings and widths are possible. Another metal layer (viz., one of the overlying thick aluminum bonding layer ALCAP or underlying one or more thin copper or other metal layers M1-M6) are used to make crossover or cross-under connections between the turns. The example differential inductor has an odd number (three) of main turns (windings) formed using an M7 copper layer of 2.4 μm thickness, identical 6 μm widths and 2 μm spacings, with crossovers between the main turns provided by a METTOP aluminum layer of 1.2 μm thickness. Crossover areas of non-copper material are preferably minimized in order to minimize deleterious impact on inductor quality factor Q. A center tap lead is formed in the thin metal layers, to cross under the middle and outside turns to provide contact with the center of the inside turn (which is also the center of the three-turn coil). Contacts between the main turns and the crossovers and cross-under are established by forming via openings in the interlevel dielectric (ILD) layers that electrically separate the metal layers, and filling the via openings with conductive material (e.g., copper or tungsten). The use of an odd number of terms provides a symmetrical device. For an odd number of turns, the total numbers of crossovers and cross-unders are identical.

The illustrated configuration has two differential input/output terminals (left + and − terminals in FIG. 2) made at the ends of the outer turn, and a common mode terminal (center tap) made by a crossover or cross-under connection to the center of the inner turn. The shown configuration also has crossover or cross-over connections respectively joining inner and intermediate turns and intermediate and outer turns to provide a single coil (spiral-like) electrical path between the differential terminals. One of the connections between adjacent turns is made using the same (thick copper) metal layer used for the turns. The other is made using one of either an overlying (thick aluminum) metal layer (making a crossover) or an underlying (one or more of underlying thin copper layers) metal layer (making a cross-under). The common mode terminal is shown brought out from the center of the inner turn to make it accessible for connection to other circuitry on the same die. The leads to the differential input/output terminals are made using the same (thick copper) metal layer used for the turns. The lead to the center tap is made using the other of either the overlying (thick aluminum) metal layer (making a crossover) or an underlying (one or more of underlying thin copper layers) metal layer (making a cross-under). The center tap metals in this area need to use different metals than the inductor windings and crossings to prevent the center tap from shorting the inductor. To achieve a good balance and minimize losses, the number and areas of crossovers and cross-unders are preferably minimized, and they are preferably symmetrically placed relative to other parts of the structure. The use of an odd number of turns leads to a preferred symmetrical device.

One solution to getting lower loss involves stacking two differentially wound inductors on top of each other (primary winding on top of secondary winding). Since the intermetal dielectric (IMD) layer is thinner than the minimum metal pitch, the stacked implementation has a much better coupling factor, as shown in FIG. 4. The stacked implementation allows for H (desired) and E (undesired) field confinement; so, higher turn ratios may be easily implemented without negatively impacting k, or the area, or the loss due to longer crossovers. Because the stacked implementation involves two differential windings, both windings need cross-unders (additional metal for crossings) as shown in FIG. 4. This means that additional metals are needed to implement this structure. This solution then needs at least two thick metals in addition to the thick aluminum layer already present.

A stacked approach to balun (and transformer) implementation using two superposed coils is illustrated schematically in FIGS. 4A-4C. This approach uses two differential coils, each having a form similar to that shown in FIG. 3. The primary and secondary (which could alternatively be reassigned as the secondary and primary) windings are each formed using a thick copper metal layer. The turns of the secondary coil are formed in the M7 thick copper metal layer, for example, and the turns of the primary coil are formed in another M8 thick copper metal layer which, for processes having only one thick copper layer (viz., M7), is added to the normal flow of the integrated circuit fabrication process (together with a corresponding additional masking step). Cross-connections between turns and for leads to the coil terminals are provided by the overlying aluminum bonding layer (ALCAP) and the underlying thin metal layers (M1-M6) (or, for locations where no conflict is present, by the other of the two thick copper layers M7 or M8). As illustrated in FIG. 4B, the magnetic coupling (k) for the superposed coils is good for different turns ratios, for preferred configurations which minimize the numbers and areas of crossings and which maintain a relationship whereby the lateral area occupied by the turns and spacings of the primary coil matches the area occupied by the turns and spacings of the second coil, i.e.:

N1×W1+(N1−1)×S1=N2×W2+(N2−1)×S2;  (1)

where N1, N2=number of turns, W1, W2=width of the turns, and S1, S2=spacing between turns. A main disadvantage of this approach, therefore, is the time and expense needed to add process steps (i.e., additional thick copper metal layer and masking steps) to a flow which would otherwise only utilize a single thick copper metal.

Another approach to balun (and transformer) implementation is illustrated in FIGS. 5A and 5B. Here, primary (or secondary) and secondary (or primary) differential coil windings are formed in a coplanar arrangement using a single thick copper metal (viz., M7) layer, wherein turns of one coil winding are alternated with adjacent turns of the other coil winding. Crossings for connecting adjacent turns of each same coil winding and crossings for leads to/from internal terminals of the coil windings are made using the overlying thick aluminum top layer (crossovers in ALCAP layer) and in one or more of the underlying thin metal layers (cross-unders in M1-M6 layers). Because of their side-by-side relationships, the 1:1 (upper) and 1:2 (lower) turns ratio coplanar arrangements depicted schematically in FIG. 5B will have poorer magnetic coupling (k) (1:2 even poorer than 1:1) than similar turns ratio structures implemented using the previously discussed stacked arrangements of FIGS. 4A-4C. Moreover, the side-by-side coil windings of coplanar implementations will require greater area (at least twice the die surface area) than coil windings of corresponding stacked implementations. Large pitch (i.e., spacing) between turns provides better process control, but leads to smaller k.

In the absence of a second thick copper layer, a stacked balun can be realized by placing a single-ended winding (secondary) on top of the differential primary winding. The single-ended inductor may be built with the thick aluminum and only a single cross-under. The single-ended winding is rotated by 90 degrees, thereby allowing the cross-under for the single-ended winding to be implemented using the same metals as the cross-unders used for the differential winding. This is illustrated in FIGS. 6A-6C. In order to implement this, the innermost single-ended winding is first slightly jutted inwards and then routed down through a via to the cross-unders, which are then routed to the outside. This point of the secondary winding of the balun is typically the ground connection and can potentially be directly connected to a ground core pad (in the case of a flip-chip assembly) in the center of the balun. In this configuration, there is no need for any cross-unders. Because of the small size for internal core pads versus the inner diameter, this option has minimal impact on the balun loss.

The primary differential winding is implemented using the single thick copper layer with two thin lower metal layers being used as cross-unders, as is typically done for a differential inductor. The lower metals are 10 times thinner than the copper of the main winding and can be a substantial Q bottleneck. The inner radius, width and spacing of the two windings are optimized to provide lowest loss. The total dimension of the five primary winding widths plus spacings is kept equal to the three secondary winding widths plus spacings in order to maximize k (see FIG. 6 and Equation (1)). An odd number of turns on the differential winding provides a fully symmetric structure since the numbers of crossovers and cross-unders are equal around both sides of the center tap. The primary winding is implemented using the single thick copper layer with the thin lower metal layers being used as cross-unders as is typically done for a differential inductor. The single-ended secondary winding is implemented using the Al RDL (ALCAP or METTOP).

For the balun at the transmitter (TX) output, the primary winding typically carries heavy DC current and should thus preferably be wound using copper in order to mitigate electromigration. Further, a wide center tap is preferably used to handle the current. Simulations can be performed to ensure that electromigration will not be an issue, particularly in the hot temperature and max R (thinnest metals) cases. For copper, a general rule of 10 mA/um² of cross-sectional area can be used as a first order approximation. The secondary winding does not carry DC current and so can be implemented using aluminum. If the balun is placed before the final preamplifier (PA) driver, the same considerations apply, because the gate of the preamplifier driver does not need any DC bias current. The same applies for a balun located at the input of the low noise amplifier (LNA), since the LNA gate does not draw any DC current. As a first order approximation, aluminum has less than 40% of the current carrying capacity of copper.

The described embodiment uses only a single thick metal layer. The stacked metal maximizes the high coupling factor because of the confinement of both the H and E fields. The described stacked implementation is area efficient—utilizing approximately the same amount of area as a single inductor. It enables use of the shortest crossovers, thereby providing the lowest loss. Because the secondary inductor is wound in a single-ended configuration, it can be implemented using a single thick Cu metal. Since the secondary inductor does not carry DC current in both the case of a balun on the preamplifier (PA) stage driver output and the case of a balun on the preamplifier (PA) stage driver input, the single-ended secondary inductor is implemented using aluminum with minimal electromigration impact. The single-ended winding for the secondary allows for half or quarter (or even ⅛th) turns to be implemented. Hence a 5:2.5 turn balun can be readily implemented. This allows for great flexibility in impedance matching.

A four port network analyzer is used to measure the balun with the center tap of the primary AC grounded. An open and short deembedding process is used to calibrate out the balun feeds and obtain 4 port Z-parameters. Ports 1 and 2 are the plus and minus of the primary and ports 3 and 4 are the plus and minus of the secondary. Port 3 is ground connected as is the case in the actual balun. The following key balun parameters are extracted:

$\begin{array}{cc}{\mathrm{Ld}}_p=\frac{1}{2\pi f}\mathrm{Im}\left[z\left(1,1\right)-z\left(1,2\right)-z\left(2,1\right)+z\left(2,2\right)\right]& \left(2\right)\\ {\mathrm{Qd}}_p=\frac{\mathrm{Im}\left[z\left(1,1\right)-z\left(1,2\right)-z\left(2,1\right)+z\left(2,2\right)\right]}{\mathrm{Re}\left[z\left(1,1\right)-z\left(1,2\right)-z\left(2,1\right)+z\left(2,2\right)\right]}& \left(3\right)\\ {\mathrm{Ls}}_s=\frac{1}{2\pi f}\mathrm{Im}\left[z\left(4,4\right)\right]& \left(4\right)\\ {\mathrm{Qs}}_s=\frac{\mathrm{Im}\left[z\left(4,4\right)\right]}{\mathrm{Re}\left[z\left(4,4\right)\right]}& \left(5\right)\\ \mathrm{Md}=\frac{1}{2\pi f}\mathrm{Im}\left[z\left(4,1\right)-z\left(4,2\right)\right]& \left(6\right)\\ \mathrm{kd}=\frac{M_d}{\sqrt{{\mathrm{Ld}}_p\times {\mathrm{Ls}}_s}}& \left(7\right)\end{array}$

FIG. 7 shows the extracted Ldp (differential inductance on primary) and Lss (single-ended inductance on secondary) for the high band balun. FIG. 8 shows the normalized extracted Qdp (differential Q on secondary) and Qss (single-ended Q on primary). FIG. 9 shows kd, the coupling factor between the primary and secondary inductors. The flatness of the k curve shows that up to 1 GHz, the balun coupling is entirely magnetic. At higher frequencies, undesirable capacitive coupling kicks in increasing k. Having a single-ended winding over a differential structure can impact the amplitude and phase imbalance of the differential portion. This will impact the balun loss and dynamic range of the amplifier. The loss of the balun is obtained with an ideal balun at the input and a matched input and matched output. The loss is composed of losses due to metal Q, coupling, and phase and amplitude imbalance.

Table I gives measured performance for the low band balun at 870 MHz and the high band balun at 1845 MHz.

TABLE I
Measured Balun Performance
	Band	Low Band	High Band
	Frequency	870 MHz	1845 MHz
	Area (mm2)	0.096	0.078
	Ldp (nH)	8.58	7.46
	Lss (nH)	3.16	2.82
	Qdp	5.1	6.12
	Qss	3.1	4.84
	Kd	0.94	0.94
	Loss (dB)	2.23	1.55
	Phase imbalance	1.5°	1.1°
	Amp. imbalance	0.78 dB	1.4 dB

The coupling for both baluns is very high at 0.94 compared to a measured value of 0.74 for the 2:1 balun in FIG. 2. The low band balun exhibits higher loss due to the poor Q on both the primary and secondary. This can be substantially improved by thickening both the Cu and the Al layers without impacting the coupling. The higher loss in low band is however not problematic because low band generally has higher output power (due to lower parasitics).

A high magnetic coupling, low loss, area efficient balun with one thick Cu primary winding and one Al secondary winding for a PA driver output has been presented. The stacked metal provides the highest coupling factor possible because of the confinement of the H field. The stacked implementation is an area efficient implementation utilizing the same area as a single inductor. It has the shortest crossovers possible, implying lowest loss. Because the secondary inductor is wound in a single-ended configuration, it can be implemented using the single thick Al. Since the secondary does not carry DC current in both the case of a balun on the PA driver output or PA driver input, it has minimal electromigration concerns. The single-ended winding for the secondary allows for half or quarter (or even ⅛) turns to be implemented. Hence a 5:2.5 turn balun can easily be implemented. This allows for flexibility in impedance matching. The presented balun is well-suited for integration in today's advanced nanometer digital processes. An example implementation is in the form of an integrated cellular band balun designed for a transmitter output operating from 824 MHz to 915 MHz (low band TX). With an output power of 12 dBm, a five-turn primary with a three-turn secondary (5:3) balun enables the use of a single external capacitor for impedance matching. Implemented in a standard digital 65 nm CMOS fabrication process with no additional mask requirements, the low band balun occupies an area of 0.096 mm². A second balun is designed for high band (1710 MHz to 1980 MHz) transmitter (TX) operation and occupies an area of 0.078 mm². A micrograph of the low band balun is shown in FIG. 10.

The same principles can be applied in the construction of a transformer. In contrast to a balun, a transformer is a balanced structure both on the primary and secondary, with center taps on both sides. FIGS. 11A and 11B show an approach for introducing a center tap into the single-ended wound secondary. The position of the center tap is displaced angularly until the unbalanced single-ended structure is again balanced. If the center tap on the single-ended winding (realized using the top metal layer METTOP) is to be connected to an inner turn, appropriate metal configurations are used in order to prevent conflicts (metal shorts). In the example implementation in FIG. 11 A, this is accomplished by breaking the outer turn metal at the point at which the center tap is to go into the inner turn (and connect it to the inner turn). The break of the outer turn is then shorted using the lower thick metal (M7 metal layer in this case) to realize the primary coil that is under the secondary coil (or vice versa). In order to prevent shorts with the primary (lower) coil, this connection is offset from the lower coil by routing the top metal layer METTOP first outwardly and then shorting it using the M7 metal layer. At this point, the magnetic coupling is diminished because the primary and secondary coils do not overlap, but are instead coplanar. However this occurs for only a small portion of the length of the winding. The two (primary and secondary) coils remain stacked together for the majority of the winding lengths. So the overall reduction impact on the magnetic coupling is minor.

Those skilled in the art to which the invention relates will appreciated that modifications may be made to the described example embodiments and other embodiments may be implemented without departing from the scope of the claimed invention.

Claims

1. A magnetically coupled structure incorporated on a chip with an integrated circuit, comprising: a first coil defined by a first generally planar multi-turn metal winding formed by patterning a copper metal layer utilized for the manufacture of the integrated circuit;a second coil formed in a stacked configuration with the first coil, the second coil defined by a second generally planar multi-turn metal winding formed by patterning a top metal bonding layer formed for the manufacture of the integrated circuit; andcrossings between turns of the respective coils formed by patterning one or more other metal layers formed for the manufacture of the integrated circuit, the one or more other metal layers being thinner than the copper metal layer and top metal bonding layer.

2. A method of manufacturing a magnetically coupled structure incorporated on a chip with an integrated circuit, integrated within a process for fabrication of the integrated circuit, comprising: forming a first coil comprising a first generally planar multi-turn metal winding by patterning a copper metal layer formed for the manufacture of the integrated circuit;forming a second coil in a stacked configuration with the first coil, the second coil comprising a second generally planar multi-turn metal winding, by patterning a top metal bonding layer formed for the manufacture of the integrated circuit; andforming crossings between turns of the respective coils by patterning one or more other metal layers formed for the manufacture of the integrated circuit, the one or more other metal layers being thinner than the copper metal layer and top metal bonding layer.