Floating, balanced output circuit

[0002] This application claims benefit of provisional application 60/164,359, filed Nov. 9, 1999.

[0003] The present application relates to a floating, balanced output circuits, and more specifically to an improved floating, balanced output circuit that maintains control of the common-mode output current in both output legs of the circuit even when the differential voltage output is clipped and the circuit is driving a ground-referred load.

[0004] Professional audio equipment often employs electronically-balanced output circuits intended to mimic the behavior of output transformers as closely as possible. Such circuits are designed to accept a single-ended input voltage and to produce a differential output voltage with a low differential output impedance. They are further designed to possess a substantially higher common-mode output impedance (common-mode output impedance being defined as the impedance from either leg of the differential output to the ground or reference potential). This allows the differential output voltage to “float” with the common-mode voltage of the load, thus allowing the circuit to properly drive both balanced and ground-referred loads. This behavior is similar to that of an output transformer, wherein the differential output impedance is determined by the source impedance driving the primary reflected to the secondary (output) winding, while the impedance from either leg of the secondary winding to ground is quite high, being determined primarily by the stray capacitance from the secondary winding to ground. A consequence of this arrangement is that the output currents exiting the two legs of the balanced output are substantially equal in magnitude and opposite in polarity regardless of the load configuration.

[0005] A widely used circuit described in 1980 (T. Hay, “Differential Technology in Recording Consoles and the Impact of Transformerless Circuitry on Grounding Technique.” Presented at the 67th Convention of the Audio Engineering Society, Journal of Audio Engineering Society (Abstracts), vol. 28, p.924 (December 1980)) is shown in FIG. 1. It accepts a single-ended input voltage v_inwith respect to ground at terminal IN. It produces a differential output voltage (equal to twice the input voltage) between nodes OUT+ and OUT−. This circuit accomplishes the desired goals with respect to differential and common-mode output impedances. Under normal operation, the differential output impedance is substantially determined by the sum of output resistances R_O1and R_O2, as negative feedback around the operational amplifiers OA1 and OA2 substantially reduces their internal closed-loop output impedances. R_O1and R_O2are typically between 10 and 100 ohms in order to keep the differential output impedance relatively low. The common-mode output impedance is quite high, and can be infinite if the ratios of the resistances labeled R and 2R in the schematic are precisely maintained. It should be noted that mismatches in these resistor ratios can either reduce the common-mode output impedance if the mismatches are in one direction, or can lead to instability if they are in the other direction. This requirement for precise resistor-ratio matching is a drawback to this circuit.

[0006] It should be clear that the common-mode behavior of the circuit of FIG. 1 is governed by both opamps, OA1 and OA2. When driving a single-ended load, as in FIG. 2, the combined common-mode feedback forces the output currents to be equal and opposite (assuming exact resistor ratios around the opamps). This behavior is one of the most desirable properties of such circuits. However, if an input signal is applied to terminal IN that causes the output signal at the ungrounded output (in this case, OUT+), to exceed the maximum permitted by the power supply voltage, both the differential and common-mode feedback loops are broken. As is expected, the differential output voltage waveform at the OUT+ output would be “clipped” at the opamp's maximum output voltage. Its output current will be the output voltage divided by the load resistance. What is not as obvious is that, while clipping is occurring, the output current of the grounded OUT− output will be quite high, typically limited only by any protective current limiting circuit in the opamp, or by the maximum opamp output voltage divided by the value of the 10-to-100 ohm output resistor. This current must flow through an indeterminate path through the ground structure of the load device to return to the output stage, which can lead to disturbances on the audio waveform that are more audible than simple clipping.

[0007] An alternative approach to a floating balanced output circuit was described in 1990 by Chris Strahm in U.S. Pat. No. 4,979,218. Strahm's circuit includes separate feedback loops for differential and common-mode output signals. The differential loop is configured to force the differential output voltage to substantially equal the input voltage multiplied by some desired gain, and the common-mode feedback loop is configured to force the two output terminal currents to be equal and opposite. This at least opens up the possibility of preventing the clipping behavior and the audio waveform disturbances described above. Also, as described in the Strahm patent, precise resistor ratios are not necessary to maintain stability of the circuit.

[0008] Although not mentioned in the Strahm patent, in order to prevent a grounded output of such a circuit from going into current limiting when the active output is driven into voltage clipping, the common-mode feedback loop must remain active even though the differential feedback loop is disabled. In fact, the integrated circuit device manufactured by the assignee (Audio Teknology Inc.) based on the Strahm patent is implemented in a way that does not preserve the functionality of the common-mode feedback loop when the differential feedback loop is broken due to voltage clipping into a grounded load. As shown in FIG. 3, a differential pair of transistors, Q₁and Q₂, accept the input signal and the differential feedback signal. Transistor Q₃provides the tail current I_tailfor the differential pair. Q₃is controlled by the common-mode feedback signal. In this case, the common-mode feedback signal is derived by sensing the sum of the output currents from the device, as described in the Strahm patent. Thus, the common-mode output voltage is adjusted via feedback through I_tailuntil the two device output currents sum to nearly zero, and, are thus nearly equal and opposite. When voltage clipping occurs at either amplifier output, one of Q₁or Q₂will saturate while the other will be cut off. If the circuit is driving a ground-referred load from the output amplifier that is driven by the cutoff transistor, then there is no way for Q₃to affect the output voltage and common-mode feedback is also disabled. Without common-mode feedback to maintain control over the output currents, the grounded output amplifier conducts as much current as permitted by other aspects of the amplifier design, such as protective current limiting.

[0009] FIG. 1 is a schematic drawing of a prior art circuit that uses positive and negative feedback to emulate a floating voltage source;

[0010] FIG. 2 shows a schematic drawing of the prior art circuit of FIG. 1 connected to drive a single ended load;

[0011] FIG. 3 shows a schematic drawing of a prior art circuit for implementing a common mode feedback loop;

[0012] FIG. 4 shows a schematic drawing of an improved circuit that uses separate differential and common-mode feedback loops to emulate a floating voltage source and controls the output common-mode current under clipping conditions while driving a ground-referred load;

[0013] FIG. 5 shows a schematic drawing of the preferred transconductance amplifiers used in the FIG. 4 circuit;

[0014] FIG. 6 shows a schematic drawing of the differential-input, dual-output transconductance amplifier shown in FIG. 5 modified to include an additional gain stage;

[0015] FIG. 7 shows a schematic drawing illustrates a further modification to minimize output common-mode voltage; and

[0016] FIG. 8 shows a schematic drawing illustrates another modification to minimize output common-mode voltage.

[0017] FIG. 4 shows a schematic drawing of one embodiment of an improved circuit that uses separate differential and common-mode feedback loops to emulate a floating voltage source and controls the output common-mode current under clipping conditions while driving a ground-referred load. As shown, transconductance amplifier 1 is a circuit that accepts a differential input voltage and delivers as its outputs a pair of differential output currents such that:

[0018] i₁=−i₂=g_m1•v_diff  (1)

[0019] This transconductance amplifier, along with identical inverting high-gain voltage amplifiers 2 and 3, identical buffer amplifiers 4 and 5, and identical compensation capacitors 6 and 7 form a two-stage fully-differential operational amplifier. Resistors R₁and R₂are connected from the input voltage terminals IN+ and IN−, respectively, to the non-inverting and inverting terminals of transconductance amplifier 1, respectively. Resistors R₃and R₄are connected from the outputs of buffer amplifiers 4 and 5, respectively, to the non-inverting and inverting inputs of transconductance amplifier 1, respectively, to provide differential negative feedback. As long as the forward gain provided by transconductance amplifier 1 and voltage amplifiers 2 and 3 is large compared to the desired closed loop gain, then the differential closed loop gain A_cldwill be substantially: $\begin{array}{cc}{A}_{\mathrm{cld}}=\frac{v_{\mathrm{out}+}-v_{\mathrm{out}-}}{v_{in}}=\frac{R_3+R_4}{R_1+R_2}& \left(2\right)\end{array}$

[0020] Preferably, R₁=R₂and R₃=R₄, so that the differential closed loop gain will be: $\begin{array}{cc}{A}_{\mathrm{cld}}=\frac{R_4}{R_1}& \left(3\right)\end{array}$

[0021] In a preferred embodiment, R₁=R₂=5 kΩ and R₃=R₄=10 kΩ. It should be noted that either terminal IN+ or IN− can be grounded, and the input signal connected in a single-ended fashion, with no loss of functionality.

[0022] Transconductance amplifier 8 is a circuit that accepts a differential input voltage and delivers as its outputs a pair of matched output currents i₃and i₄such that:

[0023] i₃=i₄=g_m2•v_cm  (1)

[0024] These output currents i₃and i₄respectively sum with the output currents i₂and i₁of transconductance amplifier 1. Note that the output currents i₃and i₄from transconductance amplifier 8 will cause both output voltages (V_out+ and V_out−) to move in the same direction, while the output currents i₁and i₂from transconductance amplifier 1 will cause the two output voltages (v_out+ and v_out−) to move in opposite directions. Resistors R₉and R₁₀are used to sense the individual output currents, and preferably are of equal value between about 10 and about 100 ohms in order to maintain low differential output impedance, although the values can be outside this range. Resistors R₁₁and R₁₂serve to establish a minimum common-mode load for the circuit, and are preferably between about 1 kΩ and about 100 kΩ, although the values can be outside this range. Resistors R₅through R₈form a bridge used to sense the common-mode output current. Preferably, R₅=R₆and R₇=R₈. In this case the voltage v₃at the junction of R₅and R₆will be: $\begin{array}{cc}{v}_3=\frac{v_1+v_2}{2},& \left(5\right)\end{array}$

[0025] and the voltage v₄at the junction of R₇and R₈will be: $\begin{array}{cc}{v}_4=\frac{v_{\mathrm{out}+}+v_{\mathrm{out}-}}{2}.& \left(6\right)\end{array}$

[0026] Noting that the currents through R₉and R₁₀are: $\begin{array}{cc}{i}_{\mathrm{out}+}=\frac{v_2-v_{\mathrm{out}+}}{R_{10}}\mathrm{and},& \left(7\right)\\ i_{\mathrm{out}-}=\frac{v_1-v_{\mathrm{out}-}}{R_9}& \left(8\right)\end{array}$

[0027] the input voltage to transconductance amplifier 8 will be: $\begin{array}{cc}{v}_{cm}=\frac{i_{\mathrm{out}-}·R_9+i_{\mathrm{out}+}·R_{10}}{2}& \left(9\right)\end{array}$

[0028] Assuming that R₉and R₁₀are equal, this voltage is: $\begin{array}{cc}{v}_{cm}=\frac{\left(i_{\mathrm{out}+}+i_{\mathrm{out}-}\right)·R_9}{2},& \left(10\right)\end{array}$

[0029] which is proportional to the common-mode output current, $\begin{array}{cc}\frac{\left(i_{\mathrm{out}+}+i_{\mathrm{out}-}\right)}{2}.& \left(11\right)\end{array}$

[0030] The large gain provided by the combination of transconductance amplifier 8 and voltage amplifiers 2 and 3 will tend to minimize the differential voltage at the transconductance amplifier's inputs via negative feedback. This will then tend to minimize the common-mode output current, leaving only differential (equal and opposite) currents.

[0031] Both transconductance amplifiers must be designed to have a maximum possible output current that is achieved when the input voltage exceeds a predefined level. (This is a natural consequence of the preferred implementations, as will be illustrated below). In order to ensure that the common mode feedback loop will remain active when the differential loop has been disabled due to clipping, the maximum output currents from transconductance amplifier 8 must be made greater than the maximum output currents from transconductance amplifier 1. As an example, assume that R₁₁is a short circuit, such that R₁₂serves as a ground-referred load, and that the input voltage v_inis sufficiently positive to drive v₂to the maximum possible positive voltage allowed by the circuit power supplies. The negative feedback path via R₄is now broken, as the voltage v₂no longer responds to any change in the input voltage. Under such conditions, i₁and i₂will be at their maximum possible values (in opposite directions). Current i₁will tend to drive voltage v₁negative, and, if left unchecked, will cause a large current to flow through low-valued resistor R₉. However, if transconductance amplifier 8 has sufficient output current capability to sink the maximum value of i₁, then the common-mode feedback loop will act to minimize the common-mode output current. Under these conditions, i₄will be substantially equal to −i₁, and i_out− will be substantially equal to −_out+.

[0032] It should be clear that output stages 9 and 10, consisting of voltage amplifiers 2 and 3, buffer stages 4 and 5, and compensation capacitors 6 and 7 may take many preferred forms without departing from the scope of the invention. As an example, when utilizing bipolar transistors, voltage amplifiers 2 and 3 may consist of current-source-loaded common emitter amplifiers, and buffer amplifiers 4 and 5 may consist of complementary common-collector amplifiers. Other devices, such as MOS transistors, could also be substituted with no loss of essential functionality. Also, differential feedback resistors R₃and R₄could alternately be connected directly to the OUT+ and OUT− terminals, rather than to the outputs of buffer amplifiers 4 and 5. Such an arrangement would result in lower differential output impedance, but would require more elaborate frequency compensation in order to maintain stability into capacitive loads.

[0033] A preferred embodiment of transconductance amplifier 1 is shown in FIG. 5. This structure comprises differential pair transistors Q₁and Q₂, current sources I₁, I₂, and I₃and optional equal-valued emitter degeneration resistors R₁₃and R₁₄. The differential input to the transconductance amplifier is applied to the bases of Q₁and Q₂. The differential output currents are taken from the collectors of Q₁and Q₂. Preferably, the values of current sources I₂and I₃are each equal to one half of the value of current source I₁. In this case, the maximum current available in either direction from the collectors of Q₁and Q₂is equal to I₁/2. The transconductance from the voltage between the bases of Q₁and Q₂to either of the collectors of Q₁or Q₂is: $\begin{array}{cc}{g}_{\mathrm{m1}}=\frac{1}{\frac{2\mathrm{kT}}{{\mathrm{qI}}_1}+R_{13}+R_{14}}=\frac{1}{\frac{2\mathrm{kT}}{{\mathrm{qI}}_1}+2R_{13}}.& \left(12\right)\end{array}$

[0034] A preferred embodiment of transconductance amplifier 8 is also shown in FIG. 5. It comprises transistors Q₃, through Q₈, current source I₄, and optional emitter degeneration resistors R₁₅through R₁₇. Preferably, transistor Q₃has an emitter area twice that of Q₄and Q₅. Also, if included, the value emitter degeneration resistor R₁₇is half the value of identically-valued resistors R₁₅and R₁₆. Thus, with no differential voltage applied between the base of transistor Q₃and the common bases of transistors of Q₄and Q₅, Q₃will operate at a collector current equal to I₄/2, and transistors Q₄and Q₅will each operate at a collector current equal to I₄/4. Similarly, transistor Q₆has an emitter area twice that of transistors Q₇and Q₈. Thus, ignoring base currents, the collector current of Q₆will be mirrored to the collectors of Q₇and Q₈with a gain of 0.5, such that each will operate at a collector current equal to one half of Q₆'s collector current. The differential input voltage to transconductance amplifier 8 is applied between the base of transistor Q₃and the common bases of transistors Q₄and Q₅. Identical output currents are taken from the collectors of Q₇and Q₈. The transconductance from the input differential voltage to either of the identical current outputs will be: $\begin{array}{cc}{g}_{\mathrm{m2}}=\frac{1}{\frac{2\mathrm{kT}}{{\mathrm{qI}}_4}+\frac{R_{15}+R_{16}}{2}+R_{17}}=\frac{1}{\frac{2\mathrm{kT}}{{\mathrm{qI}}_4}+2R_{17}}.& \left(13\right)\end{array}$

[0035] The maximum output current available in either direction from the collectors of Q₄and Q₅is equal to one half the value of current source I₄. Thus, as described above, current source I₄should be made greater in value than current source I₁in order to ensure that the common-mode feedback loop will remain active after the differential feedback loop is disabled by clipping.

[0036] In an embodiment of the invention such as that shown in FIG. 5, when the differential feedback loop has been broken due to output voltage clipping, either transistors Q₇and Q₈are sinking collector currents equal to at least I₁/2, or transistors Q₄and Q₅are sourcing collector currents equal to at least I₁/2. Under these conditions, a current imbalance equal to the value of current source I₁will exist between the collector current of Q₃and the sum of the collector currents of Q₄and Q₅. This current imbalance will cause an input offset voltage (in addition to that caused by random transistor and resistor mismatches) at the inputs of transconductance amplifier 8 equal to I₁/g_m2. This additional input offset voltage will degrade the matching of the magnitudes of the currents in resistors R₉and R₁₀under the aforementioned conditions. If this degradation of performance is unacceptable, an additional gain stage can be added to transconductance amplifier 8 as illustrated in FIG. 6. Differential pair transistors Q₉and Q₁₀, optional identical emitter degeneration transistors R₁₈and R₁₉, current mirror transistors Q₁₁and Q₁₂, and current source I₅make up a differential amplifier with a single-ended current output. The input voltage to transconductance amplifier 8 is applied between the bases of Q₉and Q₁₀. The output current from the collectors of Q₁₂and Q₁₀is applied to the base of Q₅of the previously described differential-input, dual-output transconductance amplifier. The common bases of Q₃and Q₄are tied to an appropriate bias voltage source, preferably far enough below Vcc to ensure proper operation of current source I₄and transistors Q₃through Q₅. The dual output currents are taken from the collectors of Q₇and Q₈as described above. With this modification to dual-output transconductance amplifier 8, the additional offset voltage created between the base of Q₅and the common bases of Q₃and Q₄when the differential feedback loop is disabled due to clipping is reduced by the gain of differential amplifier stage 11. This results in very little change in input offset voltage at the bases of Q₉and Q₁₀.

[0037] It should be understood that the functions of the circuits above can be implemented in different ways without departing from the scope of the invention. For instance, the current mirrors composed of Q₆through Q₈and Q₁₁through Q₁₂could be any of a number of improved current mirrors known in the art such as the Wilson current mirror, the cascoded current mirror, or the emitter-follower-augmented current mirror. Additionally, each mirror could have emitter degeneration resistors added to increase the output impedance. Further, the differential inputs of transconductance amplifiers 1 and 8 could have emitter follower buffers, and/or bias current cancellation circuitry added to minimize input bias current. As mentioned above, all of the circuits could be implemented with a different transistor technology, such as MOS transistors.

[0038] Referring to FIG. 4, there will always exist some finite input offset voltage at the differential inputs of transconductance amplifier 8 due to transistor and resistor mismatches. Such offset voltages will give rise to a common-mode offset current flowing in resistors R₉and R₁₀equal to the input offset voltage divided by the resistance value of R₉and R₁₀. These currents will then be converted to a common-mode offset voltage across resistors R₁₁and R₁₂(and any external load resistance). As R₉and R₁₀are preferably low-valued, as mentioned above, and R₁₁and R₁₂are preferably higher valued, a small input offset voltage at the input to transconductance amplifier 8 can result in a substantially larger common-mode output offset voltage at the OUT+ and OUT− terminals. FIG. 7 illustrates one preferred method to minimize this effect.

[0039] In FIG. 7, capacitor C₁is inserted between the junction of resistors R₇and R₈and the inverting input of transconductance amplifier 8. Resistor R₂₀is added from the inverting input of transconductance amplifier 8 to ground. R₂₀is preferably chosen to be large enough in value so as not to significantly load R₇and R₈, preferably, although not necessarily 1 MΩ or larger if R₅through R₈are all about or within a small range of 10 kΩ. C₁is chosen so that the high-pass filter formed by C₁and R₂₀has a pole frequency substantially lower than the operational frequencies of interest. For example, a value of 100 nF for C₁and 1 MΩ for R₂₀will result in a pole frequency 𝑓_phpof: $\begin{array}{cc}{f}_{\mathrm{php}}=\frac{1}{2\pi \left(100\mathrm{nF}\right)\left(1M\Omega \right)}=1.6\mathrm{Hz},& \left(14\right)\end{array}$

[0040] which is well below the band of interest for audio applications. Thus, in the audio band, the common-mode feedback loop will minimize the output common-mode current, forcing equal and opposite currents in R₉and R₁₀. At DC, the common-mode feedback loop will tend to force the junction of R₅and R₆(and thus the output common-mode voltage) to the ground potential, plus or minus any input offset voltage at transconductance amplifier 8's inputs.

[0041] One of the primary applications for floating, balanced output circuits in the professional audio industry is to drive audio signals over cables of up to 1000 feet long. Such cables represent a reactive load on the circuit, with resonant frequencies that may coincide with the unity-gain frequency of the common-mode feedback loop. Such resonances can cause peaks in the loop transmission that will compromise the stability of the loop. The common-mode feedback loop can be isolated from these loading effects with the addition of C₂, also shown in FIG. 7. C₂is preferably chosen so that the lowpass filter that it forms with the parallel combination of R₇and R₈is substantially higher than the operation frequencies of interest, but below the unity-gain crossover frequency of the common-mode feedback loop. In a preferred embodiment, with R₃through R₈all equal to about 10 kΩ, C₂is equal to about 10 pF. This results in a pole frequency of: $\begin{array}{cc}{f}_{\mathrm{plp}}=\frac{1}{2\pi \left(10\mathrm{pF}\right)\left(5k\Omega \right)}=3.2\mathrm{MHz},& \left(15\right)\end{array}$

[0042] which is well above the audio band. Thus, the common-mode feedback loop will continue to minimize the common-mode output current, while at frequencies substantially above 3 MHz, C₂will shunt the inverting input of transconductance amplifier 8 to ground, isolating it from the response peaks due to resonant loads.

[0043] FIG. 8 illustrates an alternative and preferred method of minimizing the output common-mode voltage. Capacitor C₃is inserted between the OUT− terminal and resistor R₇. Likewise, capacitor C₄is inserted between the OUT+ terminal and resistor R₈. Resistor R₂, is added from the junction of C₃and R₇to ground, and resistor R₂₂is added from the junction of C₄and R₈to ground. In a one implementation of this preferred embodiment C₃and C₄are each about 10 μF, and R₂₁, and R₂₂are each about 20 kΩ. Like the circuit shown in FIG. 7, this circuit will minimize the common-mode output current in the audio band, but force the output common-mode voltage to the ground potential at DC. The circuit in FIG. 8 will maintain a superior match between the output current magnitudes when driving a ground referred load compared with the circuit in FIG. 7. This is due to the absence of any loading on R₇and R₈. However, this comes at the expense of an additional capacitor and an additional resistor.

[0044] The embodiment and practices described in this specification have been presented by way of illustration rather than limitation, and various modifications, combinations and substitutions may be effected by those skilled in the art without departure either in spirit or scope from this disclosure in its broader aspects and as set forth in the appended claims.