Improved self-oscillating class-d amplifier device

Abstract

To provide a self-oscillating Class-D amplifier device which implements the improvement of acoustic performance, the reduction of distortion and the improvement of stability.SOLUTION: An amplifier device 300 for amplifying an input signal Sx at a device input X to an output signal Sy at a device output Y, for driving reactive load such as a loudspeaker, comprises a self-oscillating Class-D power amplifier stage K with transfer function KA(s). The amplifier stage includes a comparator, a power stage, a second-order low-pass output filter and a time delay unit. The time delay unit provides a delay determined by the comparator, the power stage, and a first feedback filter Hosc. The output signal Sy is fed back to the comparator via a first feedback path. A correction stage comprises a second feedback path that feeds the output signal Sy to a virtual ground input of a transimpedance amplifier. A third feedback path for further enhancing correction includes a passive network which applies a first transfer function H2 to the output signal Sy and applies a second transfer function H3 to the input signal Sx.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] FIELD OF THE INVENTION The present invention relates to an amplifier device, and more particularly to an amplifier device with improved sound quality. [Background technology]

[0002] Background of the Invention In recent years, the field of audio amplification has seen significant advances with the development of high-performance Class D amplifiers. Class D amplifiers, also known as switching amplifiers, are electronic amplifiers in which the power stage semiconductors, usually transistors such as MOSFETs, operate as electronic switches rather than as linear gain devices like older amplifiers. They operate by generating a train of fixed-amplitude but variable-duty-cycle rectangular pulses that represent the amplitude changes of the analog audio input signal. The output of a modulator is used to alternately turn the output transistors on and off. Because the transistors are either fully on or fully off, they consume very little power, a typical advantage of Class D amplifiers. A low-pass filter, consisting of an inductor and a capacitor, provides a path for audible low frequencies in the audio signal and removes most of the high-frequency components from the audio signal.

[0003] Class D amplifiers have been known for quite some time and are known to offer several advantages over other amplifier classes. Among them, Class D amplifiers are known to be consistently more efficient and relatively compact in size compared to Class A and Class AB amplifiers. However, they also present certain challenges, such as potential EMC-related issues.

[0004] Class D amplifiers can present challenges in design and implementation due to the complexity of their control circuitry. Achieving a balance between efficiency, size, and maintaining low distortion requires careful engineering and often involves trade-offs. Additionally, the sensitivity of Class D amplifiers to load variations can be a concern. Reactive loads can affect performance, so it is important to design the amplifier to perform well under widely varying load conditions.

[0005] However, modern high-performance Class D amplifiers perform well in most applications and often feature high-fidelity characteristics that overcome to some extent many of the above-mentioned shortcomings of typical Class D amplifiers.

[0006] However, even modern Class D amplifiers are not always without compromise and often still leave room for improvement.

[0007] Over the years, several improvements have been made to enhance the fidelity and robustness of Class D amplifiers, including improved feedback control techniques and component integration. Despite these advances, many commercially available Class D amplifiers still have room for further improvement, especially in maintaining high sound quality under real-world dynamic conditions and reactive loads.

[0008] Balancing the trade-offs between efficiency, distortion control, and control loop stability remains a design challenge, especially as higher fidelity is demanded from these amplifiers in high performance audio applications.

[0009] It is therefore an object of the present invention to provide an improved self-oscillating Class D amplifier arrangement which provides improved acoustic performance, reduced distortion and improved stability whilst retaining the inherent efficiency advantages of Class D operation. Summary of the Invention [Means for solving the problem]

[0010] The above object is achieved, at least in part, in a first aspect of the present disclosure by an amplifier apparatus for amplifying an input signal Sx at a device input X to an output signal Sy at a device output Y for driving a speaker or equivalent reactive load. The amplifier apparatus has a first transfer function K AThe amplifier includes a self-oscillating class-D power amplifier stage (10) having a comparator (01), a class-D power stage (03), a second-order low-pass output filter (04), a time delay unit (02), and a first feedback path from a device output Y. The comparator (01) compares the comparator output signal Sy with a local input voltage SKin via a first feedback filter Hosc (06). The class-D power stage (03) is coupled to the output of the comparator (01) and driven by the output of the comparator (01) to convert the comparator output voltage to a higher voltage level with high current capability. The second-order low-pass output filter (04), which includes an inductor and a capacitor, is coupled between the power stage (03) and a speaker or equivalent load to remove at least a portion of high-frequency switching noise from the output signal of the class-D power stage. The time delay unit (02) is configured to provide a time delay according to one or more of the comparator, Hosc, and the power stage. The first feedback path couples the output signal Sy back to the comparator (01) via Hosc (06). This first feedback path provides control of the oscillation frequency and at least partially provides low-frequency feedback to reduce signal distortion in the audio frequency band. The correction stage has a second feedback path. The second feedback path couples the output signal Sy from the device output Y to the feedback input of the correction stage via H1 (104) to summing node S2, which is the virtual ground input of transimpedance amplifier HZ1 (102), to further correct errors in the output signal Sy generated by the amplification stage. This further correction is performed by pre-distorting the power amplifier input signal SKin of the amplification stage (10) with the inverse of the isolated and filtered distortion components caused by the power amplification stage, output filter, and speaker or equivalent load. The isolated distortion components are obtained by passing the input signal Sr through tracking filter KT (101) and then subtracting the result from the feedback signal from the output of H1 (104), which is a scaled version of the output signal Sy. The correction stage further includes a third feedback path.The third feedback path couples the output signal Sy to an additional feedback input through the correction stage HZ2 via H2+H3 to further enhance the correction of the output signal Sy by the second feedback path. The third feedback path is configured as a higher-order sum to enhance the selectivity of the feedback path and provide additional feedback gain. The third feedback path is composed of a network implementing a first transfer function H2 applied to the output signal Sy and a different second transfer function H3 applied to the input signal Sx. These H2 and H3 functions are combined within a single passive network including resistors and capacitors. The respective outputs of H2 and H3 are summed and fed as current inputs to the transimpedance amplifier HZ2, whose output defines the input signal Sr.

[0011] A key element in the research that led to the possibility of new amplifier designs, and ultimately to the amplifier design of the present invention, was the fact that, even more than in previous designs, the actual perception of sound quality is a critical factor: the insight that traditional Fourier-based distortion measurements cannot be considered the only way to judge amplifier performance.

[0012] Throughout the design phase, the inventors had the insight to update known amplifier designs and subject them to extensive listening tests. Their work led to different insights into feedback loop design, where the objective was not just to achieve the highest possible overall feedback loop gain, but also to evaluate the audible effects of phase response and gain distribution within the loop system.

[0013] Key elements such as operational amplifiers have been extensively tested, resulting in slight differences in perceived sound quality that cannot be clearly explained from the associated datasheets. This, among other things, led to the proposal to use discrete operational amplifiers in one example of the present invention.

[0014] The amplification device according to this first embodiment is shown in FIG. 2 and can be represented by three different blocks, as indicated by the reference numeral 300, identified as follows: - self-oscillating power amplifier (10), -Power amplifier enhanced with first generation correction loop (100), -Power amplifier further enhanced with second generation high order correction loop system (200).

[0015] Block 10, the self-oscillating class D power amplifier stage, has a first transfer function K A The power amplifier includes a comparator 01, a class-D power stage 03, a second-order low-pass output filter 04 having an inductor and a capacitor, a time delay unit 02, and a first feedback path from the device output Y. The comparator 01 compares the comparator output signal Sy with a local input voltage SKin via a first feedback filter Hosc 06. The class-D power stage 03 is coupled to the output of the comparator 01 and is driven by the output of the comparator 01 to convert the comparator output voltage to a higher voltage level with high current capability. The second-order low-pass output filter 04 is coupled between the power stage 03 and a speaker or equivalent load to remove at least a portion of high-frequency switching noise from the output signal of the class-D power stage. The time delay unit 02 is configured to provide a time delay according to one or more of the comparator, Hosc, and power stage. The first feedback path couples the output signal Sy back to the comparator 01 via Hosc 06. This first feedback path provides control of the oscillation frequency and at least partially provides low frequency feedback to reduce distortion of signals in the audio frequency band.

[0016] The first feedback path, in one example, can periodically update the output level of comparator 01, forcing output filter 04 to integrate Sy close to the exact value it should have when switching. Those skilled in the art will understand that because the actual output level at the output of 03 is only updated at specific instants defined by the switching frequency, the voltage Sy that is always integrated at the device output Y will still have a small error.

[0017] It should be noted that 10 denotes the earliest generation of self-oscillating class D amplifier manufactured and sold by the applicant under the trade name UcD. and disclosed in EP3721553B1, which is incorporated by reference into the present disclosure to form part of the present invention. The second generation of these self-oscillating amplifiers significantly improved performance by adding additional error correction in the audio frequency band with an additional feedback loop and relatively high loop gain, while still effectively eliminating the oscillatory feedback.

[0018] To achieve this goal, a correction stage has been added to block 10. The correction stage has a second feedback path. The second feedback path couples the output signal Sy from the device output Y to the feedback input of the correction stage via H1 (104) to summing node S2, which is the virtual ground input of transimpedance amplifier HZ1 (102), to further correct errors in the output signal Sy generated by the amplifier stage. Further correction is performed by pre-distorting the input signal SKin of the power amplifier of amplifier stage (10) with the inverse of the isolated and filtered distortion components caused by the power amplifier stage, the output filter, and the speaker or equivalent load. The isolated distortion components are then filtered by a tracking filter K. T It is obtained by passing the input signal Sr through (101) and then subtracting the result from the feedback signal from the output of H1 (104), which is a scaled version of the output signal Sy.

[0019] Therefore, the signal coming from S2 is filtered by the transimpedance amplifier HZ1 (102), which serves to significantly amplify the low frequency error while maintaining the oscillatory feedback.

[0020] Those skilled in the art will appreciate that this modification of Class D amplifier 100 with this second feedback path included in block 100 represents the second generation self-oscillating Class D amplifier manufactured and sold by Hypex Electronics under the brand name NCore and disclosed in EP 2221964 B1, and as such, in the form disclosed therein or in modified form, is part of and incorporated by reference into this disclosure. Further details regarding the modifications and their effects that result in the present invention are set forth in the Summary and Detailed Description sections later in this specification. Throughout the figures and block schematics, this modification is designated 100.

[0021] While Block 100 proved to be a significant improvement in both measured and perceived audio reproduction quality, research and testing concluded that there was still room for improvement in sound quality. Further research led to the insight that this further improvement could be achieved by adding additional low frequency gain in the feedback loop, and that a higher order transfer function would make this possible.

[0022] The correction stage has further configuration added to the feedback loop structure in the form of an additional signal path entering the correction stage via H2+H3 and HZ2.

[0023] H2 and H3 are two different transfer functions combined into a single network. H2 is a feedback transfer from the device output Y, and H3 is a feedforward transfer from the device input X, which in one example comes from input buffer 301. The summed output of both signals Sy and Sx via transfer functions H2 and H3, respectively, is fed as a current to transimpedance amplifier HZ2 to form signal Sr, which is the input signal of block 100. Those skilled in the art will understand that the expanded loop structure formed by H2+H3 and HZ1 should not be considered a separate loop, but rather a higher-order expansion of the correction stage.

[0024] This is reflected in the features of the first aspect of the present invention, where the correction stage further includes a third feedback path. The third feedback path couples the output signal Sy to an additional feedback input via H2+H3 and HZ2 of the correction stage to further enhance the correction of the output signal Sy by the second feedback path. The third feedback path is configured as a higher-order additional path to enhance the selectivity of the feedback path and provide additional feedback gain. The third feedback path is configured with a network implementing a first transfer function H2 applied to the output signal Sy and a different second transfer function H3 applied to the input signal Sx. H2 and H3 are combined within a single passive network including resistors and capacitors. The respective outputs of H2 and H3 are summed and provided as current inputs to the transimpedance amplifier HZ2, the output of which defines the input signal Sr.

[0025] The amplifier device 200 is arranged to amplify an input signal Sx, such as an analog audio signal at a device input X, into an output signal Sy (e.g., an amplified audio signal that is a representation of the analog audio signal), which is sent to a load (05) at an output of the device, also referred to as device output Y.

[0026] The class D amplifier stage 10 is at the heart of the device. The amplifier stage consists of several elements: a comparator (01), a power stage 03, a low-pass output filter 04, and a first feedback path that returns the device output signal Sy to the input of the comparator via Hosc 06. The intrinsic delay 02 is actually a distributed function and is configured by design choices in all parts of 10.

[0027] An amplifier K, 10 is provided to generate an amplified output signal, which is an amplified representation of the amplifier's local input signal SKin. The amplification is performed by a first transfer function K A It can be defined according to (s).

[0028] The output filter is a low pass filter made up of an inductor and capacitor 04, and this LC filter is placed between the output of the power stage 02 and the actual output of the device (Y). This LC filter removes most of the high frequency switching noise from 03 and provides a filtered output that is an amplified representation of the input signal SKin of the amplifier section 10. This representation is the actual output signal Sy at the device output Y.

[0029] The first feedback path via O6 is primarily responsible for the oscillation frequency and also for the primary feedback of the LF audio, even if the loop gain is limited. The second and third feedback paths are specifically intended to significantly improve the loop gain in the audio frequency band while maintaining the positive feedback required for the power stage oscillation.

[0030] The operation of the feedback path is based on the principle of feedback error correction: it compares the input signal of the amplifier with a scaled output signal (Sy). Any difference between these two signals is considered an error. This error is then isolated, filtered, amplified, and used to correct the output signal (Sy) to ensure an accurate representation of the input signal.

[0031] Isolation of the error can be achieved by subtracting a scaled version of the output signal Sy from the local input signal Sr. For a perfect power amplifier, the result of the subtraction should be zero. This scenario is valid if the signal being subtracted at summing point S2 is first subjected to the same transfer function. Since the output signal Sy is multiplied by the transfer function of (10), the reference signal from Sy must be multiplied by a replica of the same transfer function. This replica is then filtered by the tracking filter K. T Since the transfer function H1 has no other purpose than to attenuate the output signal Sy, we now know that the signal coming from summing point S2 must be distorted, at least in the audio frequency band.

[0032] The function of HZ1 is to dramatically increase the gain in the audio frequency band while not interfering with the feedback necessary for the power stage to oscillate. This can be achieved with a filter of order 2 or 3. Several alternatives can be seen in Figures 6A and 6B. These alternatives differ from the well-known T-integrator, which has different characteristics, each with its own sound profile.

[0033] In the Class D amplifier of the present disclosure, the feedback system plays a key role in the amplifier's operating characteristics. The function of the feedback system in a self-oscillating amplifier such as the amplifier of the present invention is two-fold: it determines the HF switching frequency, while at the same time significantly reducing the LF distortion compared to classical Class D amplifiers.

[0034] The amplifier stage 10, known per se, achieves a simple yet robust amplifier arrangement capable of providing high quality and high fidelity audio signals. However, due to the basic configuration of its internal feedback path via the Hocs, the stage 06 also has a relatively low loop gain, resulting in non-optimal distortion cancellation.

[0035] For this reason, a correction stage is further provided in the amplification stage.

[0036] This correction stage provides an additional, second feedback path designed to further improve the accuracy of the output signal Sy. It does this by subtracting the input signal from a scaled version of the output signal to isolate the nonlinear distortion components of the output signal. The transfer function K T (s) is the linear transfer function K of the power stage in the LF audio frequency band. A is inserted into the input signal to replicate (s). That is, for a perfectly distortion-free power stage, the result of the subtraction will be equal to 0, and for a real power stage with a load, the result of the subtraction will only contain error or distortion components.

[0037] The distortion components are then filtered in HZ1 and subtracted from the input signal Sr to the power amplifier, which is therefore "pre-distorted" with the inverse of the distortion components due to the output stage and the (reactive) load. HZ1 is a transimpedance amplifier with high gain at frequencies in the audio range.

[0038] This correction stage therefore includes a feedback path that couples the scaled output signal Sy to the input of the correction stage via H1 104. This allows the system to continuously monitor the output signal and make any necessary adjustments to further correct for errors introduced by amplifier stage K.

[0039] The second feedback path provides a substantial improvement over amplifiers with only a power amplifier stage, however listening tests have shown that such designs still have room for further improvement.

[0040] Therefore, the amplifier device of the present disclosure includes a third feedback path to enhance the correction stage and further improve performance, reduce distortion, and improve loop gain compared to known designs based on an amplifier stage and a first correction stage.

[0041] This enhancement adds another layer of error correction to further improve the accuracy of the output signal Sy. This stage includes an additional feedback path that couples the output signal Sy to the input of the correction stage. This additional third feedback path is located between the device's output Y and the input stage. Thus, from input to output, the amplifier comprises a correction stage with the third and second feedback paths, an amplification stage with the first feedback path, followed by an LC filter.

[0042] The third feedback path is actually an enhanced version of the second feedback path described above. The higher-order transfer function of the combined outer loop allows the loop structure to operate with a higher loop gain compared to a lower-order solution. Only one correction signal is obtained from the loop structure. While the entire system appears to have multiple, completely independent loops at first glance, it actually operates as a single, higher-order loop, with a narrowly tuned loop response resulting in improved loop gain and a more accurate correction signal. The third feedback path therefore provides further enhancement or refinement of error correction, further increasing the accuracy and fidelity of the output signal. This makes it a distinctive feature of the amplifier, especially in applications requiring high precision and signal quality. The higher-order loop filter configuration of the additional feedback path allows for higher loop gain and more advanced error correction, making this amplifier highly effective and accurate.

[0043] As an example, the HZ1 filter features at least three capacitors and is configured to allow for a higher-order transfer function compared to previous designs. This allows for higher loop gain compared to lower-order configurations with fewer capacitors in the feedback network, improving distortion cancellation and output signal fidelity. The output signal therefore more faithfully represents the input signal than previous versions, thereby improving the overall sound quality of the amplifier.

[0044] In one example, the third feedback path features a passive second-order network H2 and H3 at the amplifier input, preferably including two capacitors and four resistors, feeding a first-order transimpedance amplifier HZ2. This passive network has two distinct transfer functions: one seen from the amplifier input X and the other seen from the feedback signal Sy. This improves the stability of the overall feedback structure and maximizes sound quality.

[0045] In one example, the input signal (Sx) is provided by an input buffer with a very low source impedance, the low source impedance buffer being provided within the amplifier device.

[0046] This arrangement ensures that the input signal is presented to the amplifier with minimal impedance. This is necessary because input X features a passive filter network (H2 and H3) whose transfer function depends on the source impedance from the signal source (preferably close to 0 ohms). The H2 and H3 networks are added to improve the loop gain and can therefore be considered a characteristic feature.

[0047] In one example, the third feedback path further includes a transimpedance circuit HZ2, which includes an operational amplifier, a capacitor, and one or two resistors. This transimpedance circuit, along with the passive networks H2 and H3 at the input X, is an integral part of the loop structure, enabling higher loop gain than conventional designs. This effectively enhances the high-order transfer function of the feedback system, making the feedback path more effective in terms of distortion cancellation. This arrangement therefore enhances the amplifier's ability to accurately reproduce the input signal with minimal distortion, making it suitable for demanding audio applications where high-quality sound reproduction is paramount.

[0048] In one example, the amplifier device further comprises a clamp circuit having one input coupled to the input of HZ1, another input coupled to the input of HZ2, and an output coupled to the output of HZ1.

[0049] The soft limiter (103) block compares the isolated error signal from (102) to a predetermined fixed voltage. Because the error signal is expected to be small under normal operation, this is a viable way to assess whether the amplifier is operating within its normal operating limits. The error signal may exceed the expected maximum level due to, for example, amplifier clipping and extremely low loads. In such cases, the feedback gain of the correction loop is scaled back proportionally. This approach was evaluated by listening tests against more extreme approaches to disabling the loop under abnormal operating conditions and was found to be the best-sounding solution.

[0050] In another aspect, an amplifier device is provided for amplifying an input signal (Sx) at a device input (X) to an output signal (Sy) at a device output (Y) to drive a speaker or equivalent reactive load. The amplifier device includes a self-oscillating class-D power amplifier stage (10) having a first transfer function K(s) and a correction stage (100). The self-oscillating class-D power amplifier stage (10) includes a comparator (01), a class-D amplifier stage (03), a second-order low-pass output filter (04), a delay unit (02), and a first feedback path from the device output (Y). The comparator (01) is configured to compare a scaled version of the output signal (Sy) received via the first feedback path, which includes a filter Hosc (06), with a local input signal (SKin). The class-D power stage (03) is coupled to the output of the comparator (01) and is driven by the output of the comparator (01) to convert the output of the comparator to a high-voltage switching output. The second-order low-pass output filter (04), including an inductor and a capacitor, is coupled between the power stage (03) and the load and configured to attenuate high-frequency switching components of the output signal Sy. The delay unit (02) is configured to implement a delay associated with one or more of the comparator (01), the power stage (03), and the filter Hosc (06). This delay is characteristic of self-oscillation. A first feedback path couples the output signal Sy to the comparator (01) via Hosc (06) to determine the oscillation frequency and provide at least low-frequency feedback for distortion reduction. The correction stage (100) has a second feedback path and a third feedback path. The second feedback path couples the output signal Sy to a feedback input of the correction stage via a scaling element H1 (104) to a summing node (S2), which is the virtual ground input of the transimpedance amplifier HZ1 (102). The correction stage is configured to pre-distort the input signal SKin based on the distortion signal. The distortion signal is obtained by applying a tracking filter KT (101) to a reference signal (Sr) and subtracting the filtered reference signal from the scaled output signal from H1 (104). The third feedback path includes a passive network.The passive network includes at least a resistor and a capacitor and utilizes a first transfer function H2 to an output signal Sy from the device output (Y) and a second transfer function H3 to an input signal Sx from the device input (X). The outputs of H2 and H3 are summed in the current domain and fed to a virtual ground input of a transimpedance amplifier HZ2. The transimpedance amplifier is configured to generate a reference signal Sr.

[0051] In one example of the first or second embodiment, the third feedback path includes a complex feedback network configured to utilize a first transfer function H2 for the output signal Sy from the device output Y and a different second transfer function H3 for the input signal Sx from the device input X.

[0052] In one example of the first or second embodiment, the transfer functions H2 and H3 are defined to differ in magnitude and / or phase over at least a portion of the audio frequency range.

[0053] In one example of the first or second embodiment, a feedback network is provided to supply a current representing the sum of the signals processed through transfer functions H2 and H3 to the virtual ground input of transimpedance amplifier HZ2.

[0054] In one example of the first or second embodiment, the output of the transimpedance amplifier HZ2 is coupled to the input of a tracking filter KT configured to emulate the transfer function of the amplifier stage.

[0055] In one example of the first or second embodiment, the feedback network implementing H2 and H3 has passive components including at least resistors and capacitors.

[0056] In one example of the first or second embodiment, the feedback network includes four resistors and two capacitors arranged to define transfer functions H2 and H3.

[0057] In one example of the first or second embodiment, transfer function H2 includes a first RC filter defined by a first resistor and a first capacitor connected to receive output signal Sy. Transfer function H3 includes a second RC filter defined by a second resistor and a second capacitor connected to receive input signal Sx. The outputs of the first RC filter and the second RC filter are summed in the current domain and fed to a virtual ground input of transimpedance amplifier HZ2. [Brief explanation of the drawings]

[0058] The present invention will be further described by specific, non-limiting embodiments thereof with reference to the accompanying drawings. [Figure 1] FIG. 1 shows a schematic representation of a self-oscillating class D amplifier device known from the prior art. [Figure 2] FIG. 2 is a schematic diagram of a self-oscillating class-D amplifier according to one embodiment of the present disclosure. [Figure 3] FIG. 3 shows details of H2 / H3 according to an embodiment of the present disclosure. [Figure 4A] 4A and 4B show details of two alternatives for HZ2 according to one embodiment of the present disclosure. [Figure 4B] 4A and 4B show details of two alternatives for HZ2 according to one embodiment of the present disclosure. [Figure 5] FIG. 5 shows details of H1 according to one embodiment of the present disclosure. [Figure 6A] 6A and 6B show details of two alternatives for HZ1 according to one embodiment of the present disclosure. [Figure 6B] 6A and 6B show details of two alternatives for HZ1 according to one embodiment of the present disclosure. [Figure 7A] 7A and 7B show details of two alternatives for the soft limiter. [Figure 7B] 7A and 7B show details of two alternatives for the soft limiter. DETAILED DESCRIPTION OF THE INVENTION

[0059] Detailed Description of the Drawings 1 shows the basic structure of a self-oscillating class D amplifier known from the prior art and sold by the applicant of the present disclosure as NCore. This amplifier consists of three main elements, shown as transfer functions K(s), H(s), and K'(s). K(s) is the transfer function of the self-oscillating output stage. K'(s) is the transfer function of a linear tracking filter that contains a replica of the transfer function K(s) of the output stage K, but without the nonlinear distortion of K.

[0060] The input signal Sx is input at node X and is conducted via a direct forward path to the summing node W, where it is combined with the signal coming from the filter H(s) and fed to the output stage K. The output stage has a transfer function K(s) and applies a (usually small) disturbance ε to the real output signal Sy. This real device output signal Sy enters a control loop from the output node Y to a first summing unit located between K' and HZ, i.e., from the device output Y back to the first summing unit. This control loop, together with the forward filter(s) H, significantly increases the (low-frequency) loop gain of the amplifying device.

[0061] The lead-lag compensation, together with the zero-crossing detector and output filter Hout, creates a self-oscillating amplifier unit with a well-defined gain and transfer function K(s). Those skilled in the art will appreciate that the internal configuration and operation of the power amplifier stage (10) is known per se and does not form part of the inventive concept described herein. While a schematic representation is shown in block (10) of Figure 2, further details can be found in the amplifier systems shown in US 7,113,038 B2 and EP 2 221 964 B1.

[0062] In the amplifier arrangement as shown, a tracking filter K' with transfer function K'(s) followed by a summing unit serves to separate the distortion component ε coming from the power stage K and its load.

[0063] K' receives an input signal from the device input node X. The transfer function K'(s) mimics the transfer function K(s) of the power stage K. Therefore, the output of the tracking filter K'(s) is an idealized version of Sy without the error term ε. The first subtraction unit (far left) subtracts the output signal Sy from an idealized version coming from the detection filter of the tracking filter, which is derived from the signal received via the control loop. Thus, the forward filter H(s) receives only the (inverted) disturbance ε (i.e., the difference between the signal received via the control loop and the signal received from the detection filter). The disturbance ε is expected to be small under normal conditions, and therefore the output of the forward filter is also small. H(s) is a precisely configured low-pass function and amplifier, designed and configured to maximize gain in the audio frequency band while simultaneously maintaining the appropriate conditions for oscillation. At the summing node W, the isolated and inverted error is added to the input signal of K(s). This pre-distorts the input signal with the inverse of the distortion components of K and its load. In one embodiment, the tracking filter K' is implemented as a simple RC filter.

[0064] 2 shows an amplifier apparatus (300) according to one embodiment of the present disclosure. The apparatus (300) can be shown in three distinct blocks identified as a self-oscillating power amplifier (10), a power amplifier enhanced with a first generation correction loop (100), and a power amplifier further enhanced with a second generation higher order correction loop system (200).

[0065] This diagram is best interpreted and explained from right to left, and best illustrates the evolution of technology. Three red boxes can be identified, each with a title that roughly encompasses a particular technological development stage: UcD10, NCore100, and Nilai200. Newer generations are continually stacked on top of the previous generation, with each technological stage forming a subset of the most recent developments.

[0066] As for the amplifier stage UcD, while this is still not a truly top-of-the-line amplifier, it can achieve surprisingly good performance with relatively simple technology. The entire amplifier can be thought of as a comparator capable of supplying a large current, followed by a simple LC low-pass filter and feedback mechanism. LC filter 04 significantly attenuates the high-frequency components of the pulse-width signal from 03, leaving a low-frequency signal for the speaker. This LF signal is fed back to the input of the comparator (via HOsc) and scaled before being compared with the input signal. Based on the result of this comparison, the comparator's state is adjusted, integrating the output signal in the correct direction and reducing the difference between the scaled output signal Sy and the local input signal SKin. Using imagination, the operation of comparator 01, LC filter 04, and feedback can be thought of as a linear amplifier with feedback. The output LC filter (with load), the feedback network, and the comparator itself introduce a certain delay into the described control loop. This delay determines the switching frequency of the self-oscillating system (approximately 500 kHz without an input signal). Furthermore, the switching frequency varies depending on the actual state of the audio signal, decreasing at higher excursions. The advantages of the UcD10 are primarily its simplicity and robustness, although the simple control loop has limited gain and is therefore not without compromise in terms of sound quality.

[0067] In the next block (Ncore 100), a control loop is added via H1 (104). Blocks 101 and 102 are also added. The strategy here is to isolate the low-frequency distortion component of the UcD section (also called K10 in the block diagram) and subtract it from the input signal (from R) to cancel the distortion. Therefore, the input signal of the power stage 10 is "pre-distorted" with the inverse of its own distortion. The output signal Y is fed back via H1 (104) and summed with the inverse of the signal from 101 at summing point S2. Therefore, the local input signal Sr is subtracted from the scaled output signal Y after being filtered in 101. 104 is simply a resistor responsible for the correct attenuation of Sy, while 101 is a filter network that attempts to mimic the transfer function of the UcD section 10 as accurately as possible. Therefore, if 10 provided perfect transfer (i.e., no distortion), the result of this subtraction would always be zero. However, in reality, the isolated distortion component of 10 appears here. 102 is a precisely configured transfer function and gain stage that provides high gain in the audio band while maintaining the oscillation condition of 10. We want to subtract only the low frequency portion of the error from the input signal, optimizing the loop gain in the audio passband. The "HZ" indicates that this is a transimpedance circuit, with a virtual ground as its input, making it easy to use as a summing point.

[0068] Next, in S1, the input signal to power stage 10 is pre-distorted with the inverse of the isolated, filtered error by subtracting the error from the input signal Sr. How successful it is in accurately isolating the distortion components depends on how well 101 tracks with 10 and the loop gain of the outer loop that can be obtained without sacrificing stability. In theory, the higher the gain, the better. The 102 function already existed in NCore, but it has undergone modifications for Nilai, making it slightly more complex than the variant present in NCore. See Figures 6A and 6B in the drawings section for details.

[0069] While NCore is a significant improvement over UcD, further improvements were deemed desirable and possible to achieve the ultimate in performance. This is achieved by further increasing the loop gain of the control loop. For this purpose, 201 and 202 have been added. 202 is again a transimpedance amplifier. H2 and H3 (201) could be interpreted as separate blocks. However, since they are connected by the same passive network, they are shown as a single block with two inputs. One of the two inputs is the actual amplifier input, and the other receives the amplifier output signal for feedback. The circuit features different transfer functions for both signal paths.

[0070] The open-loop response is that of a finely tuned low-pass filter; the closed-loop response in the audio band is all-pass, but phase margin is added to the control loop, allowing for greater gain without risking the entire circuit oscillating at undesired frequencies. The resulting structure functions as a complex control loop and must be designed and evaluated as a whole. The output of 102 is the point where a correction signal is available in the form of isolated and inverted distortion components. The circuitry up to the loop input of S1 can be considered a single outer loop with a single loop transfer function. While the block diagram shown in Figure 2 suggests separate second feedback loops via 104 and third feedback loops via 201, the overall structure can be viewed as a single higher-order correction loop, with summing point S1 as the actual point where the distortion components of 10 are canceled. Circuits 201 and 202 adjust the overall feedback in the amplifier and add additional feedback gain. Listening tests have shown that careful allocation of the loop gains of 102 and 201 has a significant impact on the perceived sound quality. Therefore, sizing these blocks was an important part of the design process.

[0071] The amplifier block 200 includes a soft limiter circuit 103. Two different embodiments of this level detection circuit are shown in Figures 7A and 7B. This circuit is useful to prevent the high gain feedback loop system from overreacting to undesirable operating conditions such as clipping of the power stage 10 and extreme underloading.

[0072] Some amplifiers employ hard-clip detection in the power amplifier, disabling the feedback system entirely. In (200), we instead chose to monitor the isolated error at the output of 102. Since the signal level at this point is small during normal operation, a viable solution is to compare the error signal to a predefined fixed voltage. If the error voltage suddenly exceeds this expected maximum, the circuit limits the gain of HZ1 and HZ2 proportionally in a tracking manner. Listening tests conducted at high sound levels found this type of limiting to sound more "relaxed" than a hard-limiting approach.

[0073] Figure 3 shows a filter network 201. This circuit, together with a transimpedance amplifier 202, adjusts the overall feedback of the amplifier and adds additional low-frequency loop gain to effectively cancel distortion components. This network consists of two capacitors and four resistors, allowing two transfer functions in the same circuit: one in the feedback loop and one fed from the input X. The signal Sx is fed from the input buffer 301 and has low impedance.

[0074] 4A shows one embodiment of a transimpedance amplifier 202. This configuration is a first-order low-pass filter with a virtual ground current input and is intended to work in conjunction with 201.

[0075] FIG. 4B is another embodiment of 202, featuring a shelving filter function.

[0076] Figure 5 is a schematic diagram of H1 (104), which is simply a resistor intended to properly scale the feedback signal from the device output Y to summing point S2, which is a virtual ground.

[0077] 6A shows one embodiment of the transimpedance amplifier HZ1 (102), which includes an operational amplifier, three resistors, and three capacitors.

[0078] Figure 6B shows another embodiment of the same circuit 102, this time with the middle capacitor removed. In practical circuits, the effect of this capacitor was very small and only significant at very high frequencies. Ultimately, it was discovered that this section could be omitted entirely without significantly affecting the overall loop performance. The key to this was tuning the circuits 201 and 202.

[0079] 7A shows one embodiment of the level detector of 103. When the input signal coming from the output of 103 exceeds a predetermined voltage level, this circuit sources current in a proportional manner to the virtual ground inputs of 102 and 202, limiting the gain of the loop.

[0080] FIG. 7B shows a second embodiment of the same circuit 103, but with an added resistor at the emitter of the transistor to provide further control of the soft-limit operation of the circuit.

Claims

1. 1. An amplifier device for amplifying an input signal Sx at a device input X to an output signal Sy at a device output Y to drive a speaker or equivalent reactive load, comprising: The amplifier device has a first transfer function K Ａ (s) and a correction stage, The self-oscillating class D power amplifier stage (10) comprises a comparator (01), a class D power stage (03), a second-order low-pass output filter (04), a time delay unit (02), and a first feedback path from the device output Y; The comparator (01) compares the output signal Sy of the comparator with the local input voltage SKin through a first feedback filter Hosc (06); the class D power stage (03) is coupled to the output of the comparator (01) and is driven by the output of the comparator (01) to convert the output voltage of the comparator to a higher voltage level with high current capability; the low-pass output filter (04) includes an inductor and a capacitor (04) coupled between the power stage (03) and the speaker or equivalent load for removing at least a portion of high-frequency switching noise from the output signal of the class-D power stage; the time delay unit (02) is configured to provide a time delay according to one or more of the comparator, Hosc and the power stage; the first feedback path couples the output signal Sy back to the comparator (01) via Hosc (06) and provides control of the oscillation frequency, at least in part providing low frequency feedback to reduce distortion of the signal in the audio frequency band; the correction stage has a second feedback path and a third feedback path; The second feedback path couples the output signal Sy from device output Y to a feedback input of the correction stage via H1 (104) to summing node S2, which is the virtual ground input of transimpedance amplifier HZ1 (102), to further correct errors in the output signal (Sy) generated by the amplification stage; said further correction being performed by pre-distorting the input signal SKin of the power amplifier of said amplifier stage (10) with the inverse of the separated and filtered distortion components caused by said power amplifier stage, said output filter and said loudspeaker or equivalent load; The separated distortion components are filtered by a tracking filter K Ｔ (101) and then subtracting the result from the feedback signal from the output of H1 (104), which is a scaled version of the output signal Sy, the third feedback path couples the output signal Sy to an additional feedback input through HZ2 of the correction stage via H2+H3 to further enhance the correction of the output signal Sy by the second feedback path; the third feedback path is configured as a higher order sum to enhance selectivity of the feedback path and to provide additional feedback gain; the third feedback path is formed by a network implementing a first transfer function H2 applied to the output signal Sy and a different second transfer function H3 applied to the input signal Sx; H2 and H3 are combined in a single passive network including a resistor and a capacitor; The respective outputs of H2 and H3 are summed and fed as a current input to a transimpedance amplifier HZ2, the output of which defines the input signal Sr. Amplification device.

2. the second feedback path comprises a minimum order 2, preferably 3 feedback network including at least two capacitors, preferably three, more preferably four capacitors, and three resistors, more preferably four resistors, configured to enhance error correction and further increase low frequency loop gain; 2. The amplifier device according to claim 1.

3. the third feedback path further includes an input / feedback summing function via a summing network H2+H3 including a capacitor and a resistor, configured to sum the input signal Sx and the output signal Sy feedback from the output (Y) in a frequency dependent manner; The summed output is fed as a current to the virtual ground input of the transimpedance amplifier HZ2.

3. An amplifier according to claim 1 or 2.

4. the summing function further comprises a transimpedance circuit HZ2 including an operational amplifier, one or two resistors and one capacitor; 4. The amplifier device according to claim 3.

5. a soft limiter circuit having an input coupled to the output of (102), an output coupled to the input of HZ1(s), and another output coupled to the input of HZ2(s); The soft limiter circuit effectively proportionally limits the gain of the correction loop structure if the voltage level at the output of HZ1 (102) exceeds a predetermined voltage level.

5. An amplifier device according to claim 1.

Images