Inductor-less RF/IF CMOS buffer for 50Ω off-chip load driving

[0002] 1. Field of the Invention

[0003] The invention relates to CMOS integrated semiconductor circuits, and more particularly to a 50Ω analog buffer amplifier operating in the RF/IF frequency range which avoids using external inductors.

[0004] 2. Description of the Related Art

[0005] Despite the many buffer amplifier designs of the related art, CMOS equivalents of bipolar buffer amplifiers, like the Darlington or the class AB circuit, tend to be inferior. We will demonstrate this point by examining a CMOS equivalent of the “Darlington pair” buffer amplifier in FIG. 1, and a CMOS equivalent class AB buffer amplifier in FIG. 2.

[0006] FIG. 1 shows an NMOS transistor M1 coupled in series with a current source I₁coupled between a positive voltage supply V_DDand ground (GND). The gate of M₁is coupled via capacitor C₁to input v_iand via resistor R₁to another voltage supply V_R. Similarly coupled between V_DDand ground is a second NMOS transistor M₂in series with a second current source I₂. The gate of M₂is coupled to the source of M₁, and the source of M₂is coupled via a second capacitor C₂to output V_o. C_pis the total parasitic capacitance of the pad, pin and printed circuit board (PCB) and R_Lis the external load, typically 50 ohm.

[0007] However, the CMOS version does not work as well as the bipolar one. The expression for voltage gain of the circuit in FIG. 1, where g_m2is the transconductance of M₂, is given by $\begin{array}{cc}\mathrm{Av}=\frac{v_o}{v_i}\cong \frac{g_{\mathrm{m2}}{R}_L}{1+g_{\mathrm{m2}}{R}_L}& \left(1\right)\end{array}$

[0008] From (1) we can easily see that to achieve Av=1, it is necessary to have g_m2R_L>>1. This is a very difficult requirement—even in sub-micron CMOS—since g_m2must be very high as R_Lis equal to 50Ω only. Very large sizes for M₂and very large values for I₂are required to achieve this. Such transconductances, however, are much more easily achievable with bipolar devices. In CMOS technology this circuit usually has a voltage gain that is substantially less than unity while driving 50Ω.

[0009] FIG. 2 shows a four transistor CMOS equivalent of a bipolar class AB push-pull buffer amplifier. The section from input v_ito M₁, I₁is identical to that of FIG. 1. However, M₂is a PMOS transistor which has its source coupled via I₂to +V_DD. The drain of M₂is connected to ground and its gate it coupled to the gate of M₁. M₁and M₂drive a CMOS output stage comprising NMOS transistor M₃and PMOS transistor M₄coupled between +V_DDand ground. The gate of M₃connects to the source of M₂and the gate of M₄connects to the source of M₁. The junction of M₃and M₄couples via C₂to output v_o. Coupled between the output and ground are C_pand the load R_Las described in FIG. 1 above.

[0010] In these and subsequent figures, like parts are identified by like numerals or characters.

[0011] This circuit is superior to that in FIG. 1 in some respects. The quiescent current can be low, the sizes of the output transistors M₃and M₄can also be smaller. However the disadvantages are as follows. Firstly, the signal swing is limited at the output to V_DD−(V_GS4+V_GS3) compared to V_DD−V_GS2in FIG. 1. Since V_GSis of the order of 1V, this is a substantial loss with V_DD=3V which is quite common for deep sub-micron CMOS technology. Secondly, the DC voltage at the sources of M₃, M₄, designed to be equal to V_R=V_DD/2, may not be realised in practice because of V_GSmismatches of M₁, M2, M3 and M₄. This further degrades the output signal swing of the buffer. Thirdly, neither can this circuit achieve unity voltage gain, when realised with source followers similar to the circuit in FIG. 1.

[0012] A patent search revealed the following U.S. Patents relating to buffer amplifiers: U.S. Pat. No. 6,052,028 (Heaton), U.S. Pat. No. 4,495,471 (Barrett), U.S. Pat. No. 4,471,319 (Metz), and U.S. Pat. No. 4,390,852 (Addis), describe Bipolar and FET buffer amplifiers for different applications. U.S. Pat. No. 5,111,157 (Komiak) describes a buffer with MOS devices but uses a complicated circuit with several inductors. U.S. Pat. No. 5,959,475 (Zomorrodi) and U.S. Pat. No. 5,668,500 (LeFevre) describe CMOS buffers but they are complicated and for a different application. U.S. Pat. No. 5,877,634 (Hunley) discloses a CMOS buffer with a controlled slew rate at the output. The circuit uses a differential circuit for feedback but is otherwise different.

[0013] None of the related art satisfies the requirement to provide a 50Ω analog buffer amplifier operating in the RF/IF frequency range (typically 100 MHz-2 GHz) which can be integrated in a CMOS technology because it is difficult to integrate inductors for such frequencies when external inductors are to be avoided. The only off-chip components are the 50Ω load resistor and its coupling capacitor.

[0014] It is an object of at least one embodiment of the present invention to provide a method and circuits for an analog buffer amplifier operating into a 50 ohm load and in the radio frequency/intermediate frequency (RF/IF) range of typically 100 MHz to 2 GHz (MHz=10⁶hertz, GHz=10⁹hertz).

[0015] It is another object of the present invention to integrate this analog buffer amplifier in a CMOS technology.

[0016] It is yet another object of the present invention to eliminate inductors from the integrated circuit of the analog buffer amplifier.

[0017] It is still another object of the present invention to eliminate the need for external inductors.

[0018] It is a further object of the present invention is to achieve a voltage gain of one or better.

[0019] It is yet a further object of the present invention is to avoid using very large transistors and large current sources.

[0020] It is still a further object of the present invention is to provide a large signal swing.

[0021] These and many other objects have been achieved by providing buffer amplifier comprising a source-follower plus common drain stage with a feedback path from the output of the common drain stage to the input gate of the source follower stage. The feedback circuit is designed such that the output of the common drain stage can be guaranteed to be at a voltage midway between the positive and the negative voltage supply of the circuit. This is the optimum operating point since it allows the largest signal swing. The transconductance of the feedback amplifier must be small, because the output coupling capacitor cannot be made any smaller to ensure stability of the feedback loop. A small transconductance is realized by biasing the transistors of the feedback amplifier with very low currents; such as by operating them in the weak inversion region. In addition, the transistors of the feedback amplifier can be dimensioned in such a way as to reduce the transconductance even further. Another important fact is that the feedback through the feedback amplifier is only present at DC (direct current) and a very low frequencies. This stabilizes the DC voltage at the drain of the common drain transistor, which, via an output capacitor, is also the output of the buffer amplifier.

[0022] These and many other objects and advantages of the present invention will be readily apparent to one skilled in the art to which the invention pertains from a perusal of the claims, the appended drawings, and the following detailed description of the preferred embodiments.

[0023] FIG. 1 is a circuit diagram of a bipolar Darlington pair buffer amplifier of the prior art.

[0024] FIG. 2 is a CMOS equivalent of a bipolar class AB push-pull buffer amplifier of the prior art.

[0025] FIG. 3 is a high level circuit diagram of the preferred embodiment of the buffer amplifier of the present invention.

[0026] FIG. 4 is a high level circuit diagram of the preferred embodiment of the differential buffer amplifier of the present invention.

[0027] FIG. 5 is a more detailed circuit diagram of the preferred embodiment of the buffer amplifier of FIG. 3.

[0028] FIG. 6 is a circuit diagram of a preferred embodiment of the operational transconductance amplifier (OTA) of the present invention as used in FIGS. 3, 4, and 5.

[0029] FIG. 7 is a block diagram of the preferred method of the present invention.

[0030] FIG. 3 depicts the basic idea of the invention. The inductor-less RF/IF CMOS buffer comprises a source follower stage with a first NMOS transistor M₁and a common drain stage with a second NMOS transistor M₂. M₁is coupled in series with a first current source I₁, where the drain of M₁is coupled to +V_DD, the positive voltage supply. The other side of I₁is couple to ground (the negative voltage supply). The gate of M₁is coupled via a first capacitor C₁to input v_i. A series combination of resistor R₁and a second capacitor C₂is coupled between the gate of M₁and ground. A series combination of a second current source I₂and a second NMOS transistor M₂is coupled between +V_DDand ground. The gate of M₂is coupled to the source of M₁The drain of M₂, coupled to I₂, is also coupled via a third capacitor C₃to output v_o. C_pis the total parasitic capacitance of the pad, pin and printed circuit board (PCB). A₁is an operational transconductance amplifier (OTA) having a positive and a negative input. The positive input of A₁is coupled to the drain of M₂and the negative input of A₁is coupled to a voltage supply V_R. The output V_Fof A₁is coupled to the junction of R₁and C₂. The external load R_Lof typically 50 ohm is connected between output V_oand ground. The CMOS buffer can provide unity gain at a signal frequency response f_sof 100 MHz to 2 GHz (MHz=10⁶hertz, GHz=109 hertz) with R_Lranging from 40 to 60 ohm. C₁is the input coupling capacitor, C₃is the output coupling capacitor, and C₂is the compensating capacitor for the feedback path formed by M₁, M₂, A₁and R₁. The bias current of M₂and M₁is set by I₂through the drain and by I₁through the source, respectively. The feedback path adjusts the DC gate voltage of M₁such that the DC drain voltage of M₂is set precisely equal to V_R. I.e., the direct current (DC) voltage at the drain of transistor M₂is determined by OTA A₁to be essentially equal to voltage supply V_R. The best value for V_Ris exactly V_DD/2 (when the negative voltage supply is ground, else it is the midway point between the positive and negative voltage supply), as this will allow maximum signal swing at the output. The maximum output swing is V_DD−2V_Dsat which is a substantial improvement over the two prior arts, described above, since V_Dsat can be made as low as 0.3V.

[0031] Assuming that the voltage gain of the source follower M₁is nearly unity, the gain for the overall buffer is given by: $\begin{array}{cc}\mathrm{Av}=\frac{v_o}{v_i}\cong g_{\mathrm{m2}}{R}_L& \left(2\right)\end{array}$

[0032] For Av=1, from (2), we need to have g_m2R_L=1. This is a much less stringent requirement compared to that of FIG. 1 and FIG. 2 where g_m2R_L>>1 is required. In fact, it is not difficult to obtain some voltage gain to offset any loss at the source follower M₁or for other purposes.

[0033] The stability requirement for the feedback loop is given by: $\begin{array}{cc}\frac{C_2}{C_3}\ge \left[g_{\mathrm{m2}}{r}_{\mathrm{o2}}\right]\left[g_{\mathrm{ma}}\left(R_L+r_{\mathrm{o2}}\right)\right]& \left(3\right)\end{array}$

[0034] Here, r_o2is the equivalent resistance looking into the drain of M₂(without R_L), g_mais the transconductance of A₁. From (3) we see that, for reasonably small values of C₂, C₃and g_mamust be as small as possible. Since C₃cannot be made very small, g_mawill have to be made so by biasing the internal transistors of A₁(shown in FIG. 6 and discussed below) with very low currents—operating in weak-inversion, if required. In FIG. 6, usually, T₉, T₇, T₈and T₁₀are identical transistors. So are T₅, T₆and also T₁, T₂. However, by altering the circuit such that (W/L)₉: (W/L)₇=(W/L)₁₀: (W/L)₈=1: N, it is possible to reduce g_maby N times, N being an integer.

[0035] Referring back to FIG. 3, the values of the coupling capacitors C₁and C₃are determined from: $\begin{array}{cc}{C}_1>\frac{1}{2\pi R_1{f}_s}\mathrm{and}{C}_3>\frac{1}{2\pi R_L{f}_s}& \left(4\right)\end{array}$

[0036] where f_sis the signal frequency (100MHz-2 GHz).

[0037] The DC gain A_DCand the unity-gain bandwidth f_oof the feedback loop is given $\begin{array}{cc}{A}_{\mathrm{DC}}\cong \left[g_{\mathrm{m2}}{r}_{\mathrm{o2}}\right]\left[g_{\mathrm{ma}}{r}_{\mathrm{oa}}\right]& \left(5\right)\\ f_0\cong \frac{\left[g_{\mathrm{m2}}{r}_{\mathrm{o2}}\right]\left[g_{\mathrm{ma}}\right]}{2\pi C_2}& \left(6\right)\end{array}$

[0038] where r_oais the output resistance of the A₁. Since g_mahas to be made low, r_oashould be made as high as possible to maintain a reasonably high value of A_DCfor accuracy of the feedback loop. However, this is easily achieved, even if the transistors of A₁are operated with very low currents, since the output impedance is inversely proportional to the current.

[0039] Another important fact to note is that the feedback through A₁is only present at DC and at very low frequencies. This stabilizes the DC voltage at the drain of M₂. However, at signal frequencies, A₁and C₂form a g_m-C low-pass filter—as given by (6)13 which essentially removes the feedback so that the required gain as in (2) can be obtained. The resistor R₁provides the input impedance.

[0040] FIG. 4 shows a differential CMOS buffer driving R_L=100Ω. In the figures like parts are identified by like designations. In this case, no coupling capacitor to the load is required. To operate at such high frequencies, it is inevitable that one will have to use the minimum channel length permitted by the technology for M₁and M₂. However, the current sources I₁and I₂will be implemented, as in FIG. 5 discussed below, with transistors having long channel length for accuracy of the bias currents. It is not difficult to see that for the circuit topologies of FIG. 3 and FIG. 4, the bias voltages and currents of M₁and M₂, and, therefore, the performance of the buffer is almost unaffected by the unavoidable inaccuracies in etching of the minimum channel length during fabrication.

[0041] The differential CMOS buffer of FIG. 4 comprises a first buffer 40 and a second buffer 41. Buffers 40 and 41 are both identical with the CMOS buffer of FIG. 3 and confer the same advantages to the differential CMOS buffer. The only difference is that the coupling capacitor C₃is omitted and that the load resistor R_Lis connected between the two outputs v_oand v_o′. Components of buffer 41 are identified by having their designations ‘primed’. The differential CMOS buffer of FIG. 4 provides unity gain at a signal frequency response f_sof 100 MHz to 2 GHz where R_Lis typically 100 ohm, but which may range from 80 to 120 ohm.

[0042] In FIG. 5 is the same circuit as that of FIG. 3, but current sources I₁and I₂are shown implemented with transistor as follows. Current source I₁, comprises NMOS transistors M₃and M₄. The gate of M₄is coupled to the gate and the drain of M₃, both sources of M₃and M₄are coupled to ground (negative voltage supply). The drain of M₃is coupled to current source IBIAS, and the drain of M₄is coupled to the source of M₁. The gate of M₅is coupled to the gate of M₃and the source of M₅is coupled to ground. Current source I₁comprises PMOS transistors M₅, M₆, and M₇. The gate of M₇is coupled to the gate and the drain of M₆, both sources of M₆and M₇are coupled to the positive voltage supply +V_DD. The drain of M₆is coupled to the drain of M₅, and the drain of M₇is coupled to the drain of M₂.

[0043] We now describe one possible and preferred implementation of the operational transconductance amplifier (OTA) A₁as illustrated in FIG. 6:

[0044] First and second PMOS transistor T₁and T₂are a differential transistor pair, where the gates of T₁, T₂are coupled to the negative and positive input IN−, IN+, respectively, and where the sources of T₁, T₂are joined.

[0045] A first current source CS1 comprises NMOS transistors T₇and T₉. The gate and drain of T₇are connected to the gate of T₉. The source of T₉is coupled to the negative voltage supply. The drain-source path of T₇is coupled between the drain of T₁and the negative voltage supply, respectively.

[0046] A second current source CS2 comprises NMOS transistors T₈and T₁₀. The gate and drain of T₈are connected to the gate of T₁₀. The drain-source path of T₁₀is coupled between output V_Fand the negative voltage supply, respectively. The drain-source path of T₈is coupled between the drain of T₂and the negative voltage supply, respectively.

[0047] A third current source CS3 comprises PMOS transistors T₃and T₄. The gate and the drain of T₄are connected to the gate of T₃. The source-drain path of T₃is coupled between the positive voltage supply (+V_DD) and the junction of the sources of T₁and T₂, respectively, and the source of T₄is connected to the positive voltage supply.

[0048] A current source IBIAS is coupled with one terminal to the drain of T₄and with the other terminal to the negative voltage supply.

[0049] A fourth current source CS4 comprises PMOS transistors T₅and T₆. The gate and the drain of T₅are connected to the gate of T₆. The source-drain path of T₅is coupled between the positive voltage supply an the drain of T₉, respectively, and the source-drain path T₆is coupled between the positive voltage supply and output V_F, respectively.

[0050] With reference to FIG. 7, Blocks 1 to 6, we describe below the preferred method of providing an inductor-less CMOS buffer amplifier with feedback, using the steps of:

[0051] a) providing a source follower stage with a first NMOS transistor, its gate in communication with the input of the CMOS buffer amplifier, its source coupled to a first current source,

[0052] b) providing a common drain stage with a second NMOS transistor, its gate coupled to the source of the first NMOS transistor, the drain of the second NMOS transistor in communication with an output,

[0053] c) coupling the drain of the second NMOS transistor via a second current source to a positive voltage supply,

[0054] d) forming a feedback loop by coupling the positive input of an operational transconductance amplifier (OTA) to the drain of the second NMOS transistor and coupling the output of the OTA via resistive means to the gate of the first NMOS transistor,

[0055] e) coupling the negative input of the OTA to a reference voltage, coupling the output of the OTA via capacitive means to a negative voltage supply, where the capacitive means together with the OTA form a low pass filter,

[0056] g) adjusting via the feedback loop the direct current drain voltage of the second NMOS transistor such that the voltage at the drain of the second NMOS transistor is precisely equal to the reference voltage, and

[0057] h) preferably selecting the reference voltage to be midway between the positive and negative voltage supply to allow for a maximum signal swing at the output.

[0058] The advantage of this method is that the feedback loop is only present at direct current (DC) or low frequencies. Transistors of the OTA are preferably operated in their weak inversion region. The above described method provides a CMOS buffer amplifier which can drive a 50 ohm load at unity gain with a signal frequency response ranging from 100 MHz to 2,000 MHz (MHz=10⁶Hertz).

[0059] Using some numerical values from simulation at 400 MHz, I₁=250 mA, I₂=3 mA, R₁=15 KΩ, C₁=1 pF, C₂=10 pF, C₃=100 pF, g_m2=20 mS, r_o2=1 KΩ. g_ma=1 μS, r_oa=500 MegΩ. With these values, we can see that (3) and (4) are satisfied and (5) and (6) yield A_DC=80 dB and f_o=320 KHz which are close to the respective simulation values of 75 dB and 270 KHz. The phase margin in simulation is 70°.

[0060] In summary, the advantages of the present invention are:

[0061] 1) an inductor-less buffer amplifier which can be integrated in CMOS

[0062] 2) capable of driving at unity gain a 50 ohm off chip load, or a 100 ohm load for the differential version

[0063] 3) capable of operating in the RF/IF frequency range of 100 MHz to 2 GHz

[0064] 4) providing a large signal swing (rail to rail minus about 0.6 Volt)

[0065] While the invention has been particularly shown and described with reference to the preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention.