功放的负反馈与无负反馈之争

筑明

推荐一本好书，比道格拉斯的《音频功率放大器设计手册 》各有所长，链接：https://pan.baidu.com/s/1ddUaqUJAKLduEF_8Pmu0gQ
提取码：maby


其中有关于负反馈的解说值得学习，我翻译出来供大家参考。



The Negative Feedback Controversy
负反馈之争
The use of negative feedback has been controversial for many years in the hi-end audio community. Some argue that none should be used, while others argue that only small amounts should be used. Some argue that wide open-loop bandwidth is required to achieve the best sound quality. Some argue that global negative feedback is bad but that local negative feedback in IPS and VAS stages is OK.

多年来，负反馈的使用在高端音响界一直存在争议。有人认为不应使用负反馈，也有人认为只应使用少量负反馈。有些人认为，要获得最佳音质，就必须有较宽的开环带宽。有些人认为，全局负反馈是不好的，但 IPS 和 VAS 阶段的局部负反馈是可以的。



How Negative Feedback Got Its Bad Rap
负面反馈的恶名是如何得来的
Negative feedback has gotten a mostly undeserved bad rap. Much of this is because poorly designed solid-state amplifiers of the 1970s happened to use large amounts of negative feedback. These were poor-performing designs in the first place, but negative feedback got the blame. Some designers carelessly believed that negative feedback could be used to linearize a design that was inherently not very linear to begin with.
Amplifier Limitations of the 1970s
Audio power amplifiers of the 1970s, being part of the early era of solid-state power amplifier technology, suffered many problems. These are some of the problems that contributed to poor sound quality.
•        Slow power transistors with inadequate SOA
•        Frequent use of quasi-complementary output stages
•        Aggressive protection circuits that misbehaved
•        Excessive crossover distortion
•        Input stages with little dynamic range, leading to inadequate slew rate
•        Poor stability margins and occasional parasitic oscillations
•        Output coils wrapped around aluminum electrolytic power supply capacitors
•        Poor capacitor choices

•        负反馈的坏名声大多是当之无愧的。这在很大程度上是因为 20 世纪 70 年代设计拙劣的固态放大器碰巧使用了大量负反馈。这些设计本来就性能不佳，但负反馈却成了罪魁祸首。一些设计师粗心大意地认为，负反馈可用于线性化设计，而这种设计本来就不是很线性化。
•        20 世纪 70 年代放大器的局限性
•        作为固态功率放大器技术早期的一部分，20 世纪 70 年代的音频功率放大器存在许多问题。以下是导致音质不佳的一些问题。
•        - 慢速功率晶体管的 SOA 不足
•        - 频繁使用准互补输出级
•        - 行为失常的强力保护电路
•        - 过度的分频失真
•        - 输入级动态范围小，导致压摆率不足
•        - 稳定性差，偶尔出现寄生振荡
•        - 输出线圈缠绕铝电解电源电容器
•        - 电容器选择不当

Guilt by Association
连带罪责
Most of these early designs achieved decent distortion measurements by using negative feedback, but nevertheless did not sound good. In a sense, the negative feedback allowed designers to make bad choices or cut corners. A very good example of such poor choices was the use of undegenerated differential input stages in the misguided belief that the resulting higher gain would provide more negative feedback and lower distortion, at least in the midband. Those designers did not realize that they were crippling the amplifier’s ability to deliver high slew rate. All of this was compounded by the fact that many designers were struggling to harness the new solid-state technology. Power transistors were slow and had poor SOA, and so required intrusive protection circuits.
It was not feedback itself that was responsible for the poor sound, but its inability to perform miracles on fundamentally poor circuit designs; that was the problem. Nevertheless, feedback got the blame for the poor sound. In a sense, a form of architectural profiling emerged, in which some treated any amplifier using negative feedback with suspicion.

这些早期设计大多通过使用负反馈实现了不错的失真测量，但声音却并不好听。从某种意义上说，负反馈允许设计师做出错误的选择或偷工减料。一个很好的例子就是在设计放大器时，设计人员错误地认为增益越高，负反馈就越多，失真就越低，至少在中频段是这样。这些设计师并没有意识到，他们正在削弱放大器提供高转换率的能力。此外，许多设计人员都在努力利用新的固态技术。功率晶体管速度慢、SOA 差，因此需要侵入式保护电路。
造成音质不佳的原因并不在于反馈本身，而在于反馈无法在基本不良的电路设计上创造奇迹，这才是问题所在。然而，声音不佳的罪魁祸首却是反馈。从某种意义上说，出现了一种建筑貌相，一些人对任何使用负反馈的放大器都抱有怀疑态度。

筑明

TIM, PIM, and IIM
During the 1970s and early 1980s several researchers sought to identify logical and measurable phenomena that were correlated with poor sound [1–9]. This effort was noble, but often the wrong conclusions were drawn. In many cases a distortion mechanism would be identified and negative feedback would be given the blame. A measurement technique for the identified distortion would then be defined [8–11]. Those very measurements ultimately disproved the assertion that negative feedback was the villain [12–14].
The distortions that were identified do indeed exist and can be measured. That is not the controversy. The point is that negative feedback itself, when properly applied, does not exacerbate these distortions. All of these distortions are in fact quite measurable in amplifiers that do not even have any global negative feedback.
20 世纪 70 年代和 80 年代初，一些研究人员试图找出与声音不佳相关的逻辑和可测量现象[1-9]。这种努力是崇高的，但往往得出错误的结论。在许多情况下，失真机制会被识别出来，而负反馈会被归咎于此。然后，针对确定的失真确定测量技术[8-11]。这些测量结果最终推翻了负反馈是罪魁祸首的说法 [12-14]。
已确定的失真确实存在，而且可以测量。这并不是争议所在。问题的关键在于，负反馈本身，如果应用得当，并不会加剧这些扭曲。事实上，所有这些失真在甚至没有任何全局负反馈的放大器中都是可以测量到的。

Negative Feedback and Open-Loop Bandwidth
负反馈和开环带宽
It may seem intuitive to many that the open-loop bandwidth of a feedback amplifier should extend to the highest audio frequencies. This allows NFB to act equally on all frequencies. It was also demonstrated by Otala that error overshoot would occur in the input stage of a feedback amplifier when the open-loop bandwidth was substantially less than the low-pass filtered bandwidth of a square wave.
The amount of NFB applied at the highest audio frequencies (e.g., 20 kHz) is necessarily limited by feedback stability considerations. The open-loop gain must usually fall at 6 dB per octave so that the NFB loop gain reaches 0 dB at a sufficiently low unitygain frequency, often on the order of 1 MHz or less. In such a case the NFB at 20 kHz will be about 34 dB. If the open-loop bandwidth is 20 kHz, the amount of NFB at low frequencies will also be about 34 dB. If the unity-gain frequency is kept the same and the open-loop bandwidth is allowed to decrease, the amount of negative feedback at lower frequencies will increase. For example, if the open-loop bandwidth is 1 kHz, the NFB at 1 kHz and below will be about 60 dB. This gives rise to the association of high feedback with bad sound, since high feedback occurs coincidentally with low openloop bandwidth when the unity-gain frequency is held constant.
对许多人来说，反馈放大器的开环带宽应延伸至最高音频频率，这似乎很直观。这使得负反馈对所有频率的作用相同。Otala 还证明，当开环带宽大大小于方波的低通滤波带宽时，反馈放大器的输入级就会出现误差过冲。
在最高音频频率（如 20 kHz）上应用的 NFB 量必然受到反馈稳定性因素的限制。开环增益通常必须以每倍频程 6 dB 的速度下降，这样 NFB 环路增益才能在足够低的单增益频率（通常为 1 MHz 或更低）上达到 0 dB。在这种情况下，20 kHz 时的 NFB 约为 34 dB。如果开环带宽为 20 kHz，低频处的 NFB 也将达到约 34 dB。如果单位增益频率保持不变，而允许开环带宽减小，低频处的负反馈量将增加。例如，如果开环带宽为 1 kHz，则 1 kHz 及以下频率的负反馈约为 60 dB。这让人联想到高反馈和坏声音，因为当单位增益频率保持不变时，高反馈会与低开环带宽同时出现。

筑明

The Input Stage Error Signal
Figure 24.1 shows the input stage error signal for two amplifiers when driven with a
5-kHz square wave that is rolled off with a first-order filter at 30 kHz (as in the 10 mV
输入级误差信号
图 24.1 显示了两个放大器的输入级误差信号。
5 kHz 方波驱动时的输入级误差信号。
10 mV

0 µs        40 µs        80 µs        120 µs        160 µs        200 µs        240 µs        280 µs        320 µs        360 µs        400 µs
DIM-30 test). Both amplifiers have the same 500-kHz unity-gain frequency and the same amount of negative feedback at 20 kHz. The amplifier in Figure 24.1a has a 20-kHz open-loop bandwidth and open-loop gain of about 54 dB. The amplifier in Figure 24.1b has a 1-kHz open-loop bandwidth and open-loop gain of about 80 dB.
Both amplifiers are driven to a peak output voltage of 2 V.
The early papers on TIM made much of the concern about the overshoot in amplifiers with low open-loop bandwidth. However, it is clear that when an apples-apples comparison is done, the “overshoot” is really created by a major reduction in error as time progresses. The peak stress on the input stage is similar in both cases (7 mV and 10 mV, respectively), but the average stress is much larger in the case where open-loop bandwidth has been made large.
DIM-30 测试）。两台放大器具有相同的 500 kHz 单位增益频率和相同的 20 kHz 负反馈量。图 24.1a 中的放大器具有 20 kHz 的开环带宽和约 54 dB 的开环增益。图 24.1b 中的放大器具有 1 kHz 的开环带宽和约 80 dB 的开环增益。
两个放大器的峰值输出电压均为 2 V。
早期有关 TIM 的论文对低开环带宽放大器的过冲问题十分关注。然而，很明显，在进行苹果和苹果之间的比较时，"过冲 "实际上是由于随着时间的推移误差大幅减小而产生的。两种情况下输入级的峰值应力相似（分别为 7 mV 和 10 mV），但在开环带宽较大的情况下，平均应力要大得多。

无语密码

完美输出就可以了

筑明

Spectral Growth Distortion
频谱增长失真
If distortion products are created in the forward path of an amplifier, negative feedback will feed these new frequencies back to the input stage, where they will have another opportunity to mix with the input signal where the nonlinearities in the forward path are encountered. This reentrant distortion mechanism was described by Baxandall [15]. The process creates new spectral lines at frequencies where none may have existed in the open-loop amplifier. For this reason it is convenient to refer to this mechanism as spectral growth distortion (SGD).
如果在放大器的前向通路中产生了失真产物，负反馈会将这些新频率反馈回输入级，在那里，它们将有另一次机会与输入信号混合，在前向通路中遇到非线性问题。Baxandall [15] 描述了这种重入失真机制。这一过程会在开环放大器中可能不存在的频率上产生新的频谱线。因此，将这种机制称为频谱增长失真（SGD）是比较方便的。
Baxandall’s Findings
If a forward gain path with only second-order distortion is enclosed by negative feedback and fed a 1-kHz signal, the output will “initially” include a 2-kHz component. When this component is fed back to the input, it will mix with the input signal at the second-order nonlinearity to create a 3-kHz signal as a result of the sum and difference process that characterizes a second-order nonlinearity. On the next go-round, a 4-kHz component will be created. Baxandall showed this phenomenon by analyzing a single-stage JFET amplifier with negative feedback around it. The JFET was chosen because it creates mostly second harmonic distortion. The open-loop second harmonic distortion in his experiment was quite high, on the order of 10%. A similar circuit for illustrating SGD is shown in Figure 24.2.
It is important to note that the concept of negative feedback going around and around the loop as illustrated above is a simplistic abstraction. If that abstraction is taken too literally, it can lead to erroneous conclusions. In practice, the negative feedback traverses the loop in nanoseconds, an amount of time that is insignificant compared to the period of any frequencies in the audio band. Figure 24.3 is a plot showing the amplitudes of the different harmonic distortion components as a function of amount of NFB. This data comes from SPICE simulations of the simple JFET circuit of Figure 24.2.
As the value of negative feedback is increased, the percentage of second harmonic goes down, as expected. However, the higher-order harmonics, which start out very low, grow with increasing amounts of negative feedback as expected by the heuristic remixing argument presented above. Interestingly, after the amount of NFB exceeds about 15 dB, all harmonics decrease with increasing amounts of NFB. So, over a range of conditions, the application of negative feedback did indeed create or increase higherorder distortion products. Baxandall’s work raised legitimate concerns about the

distortion-reducing ability of negative feedback, and whether some benign distortion was being exchanged for less benign distortion.
Baxandall 的发现
如果用负反馈封闭一个只有二阶失真的前向增益路径，并馈入一个 1 kHz 的信号，输出将 "最初 "包含一个 2 kHz 的分量。当这个分量反馈回输入端时，由于二阶非线性的和差过程，它将与二阶非线性的输入信号混合，产生一个 3 kHz 的信号。在下一轮中，将产生一个 4 千赫的分量。Baxandall 通过分析带有负反馈的单级 JFET 放大器，展示了这一现象。之所以选择 JFET，是因为它会产生大部分二次谐波失真。在他的实验中，开环二次谐波失真相当高，约为 10%。用于说明 SGD 的类似电路如图 24.2 所示。
需要注意的是，上述负反馈在环路中循环往复的概念只是一个简单的抽象概念。如果过于照本宣科，可能会得出错误的结论。实际上，负反馈以纳秒为单位穿过环路，这个时间与音频频段中任何频率的周期相比都是微不足道的。图 24.3 显示了不同谐波失真成分的振幅与负反馈量的函数关系。该数据来自图 24.2 中简单 JFET 电路的 SPICE 仿真。
随着负反馈值的增加，二次谐波的百分比会下降，这是意料之中的。然而，随着负反馈量的增加，高阶谐波（一开始很低）也会增加，正如上文启发式混频论证所预期的那样。有趣的是，当负反馈量超过约 15 dB 后，所有谐波都会随着负反馈量的增加而减少。因此，在一系列条件下，负反馈的应用确实产生或增加了高阶失真产物。Baxandall 的研究引起了人们对负反馈降低失真能力的合理担忧，以及对一些良性失真是否被较低的良性失真所取代的担忧。

筑明

Real-World Amplifiers
真实世界的放大器
Baxandall’s work was incomplete in that it only dealt with a single amplifier stage. Multistage amplifiers have a higher-order distortion characteristic due to the multiplication of individual characteristics that occur as the stages are cascaded.
Moreover, real circuits are not pure second order, but more often exponential. The BJT has an exponential characteristic that is naturally rich in high-order distortion components. Even real circuits implemented with square-law devices do not have a square-law characteristic. A JFET differential pair does not have a square-law characteristic. A class AB output stage certainly does not have a square-law characteristic. The bottom line is that real amplifiers have complex nonlinearities in their open loop to begin with. In many cases, any spectral growth that occurs may be smaller than these initial high-order products.
Baxandall was operating the JFET amplifier stage at a very high distortion level in order to demonstrate the point. Open-loop distortion was on the order of 10% in the single common-source stage he demonstrated. The effects described become much smaller at more reasonable operating levels, especially when one recognizes that the higher-order distortion products increase much faster with increases in signal level than the low-order products (e.g., fifth-order distortion products go up 5 dB for every dB increase in operating level.
Baxandall 的工作并不完整，因为它只涉及单级放大器。多级放大器具有更高阶的失真特性，这是因为在级联过程中会出现单个特性的倍增。
此外，实际电路并非纯粹的二阶电路，更常见的是指数电路。BJT 具有指数特性，自然富含高阶失真成分。即使使用方律器件实现的实际电路也不具有方律特性。JFET 差分对不具有平方律特性。AB 类输出级当然也不具有方律特性。最重要的一点是，实际放大器的开环中本来就存在复杂的非线性。在许多情况下，任何频谱增长都可能小于这些初始高阶乘积。
为了证明这一点，Baxandall 将 JFET 放大器级置于非常高的失真水平下运行。在他演示的单共源级中，开环失真约为 10%。在更合理的工作电平下，所述的影响要小得多，尤其是当我们认识到高阶失真产物随信号电平的增加而增加的速度要比低阶产物快得多时（例如，工作电平每增加 1 dB，五阶失真产物就会增加 5 dB）。
Degeneration and SGD
SGD is not limited to global negative feedback. Indeed, Baxandall’s experiment involved rather local feedback. It turns out that even emitter degeneration can be shown to exhibit the SGD Baxandall effect. A single-ended BJT amplifier stage exhibits spectral growth if its gain is reduced by applying NFB in the form of source degeneration, as shown in Figure 24.4.
Once the amount of total negative feedback exceeds about 20 dB the spectral growth stops and all orders of distortion decrease as feedback is increased. It is tempting
退化与 SGD
SGD 并不局限于全局负反馈。事实上，Baxandall 的实验涉及的是局部反馈。事实证明，即使发射极变性也能表现出 SGD 巴桑道尔效应。如图 24.4 所示，如果采用源极退化形式的 NFB 来降低增益，单端 BJT 放大器级就会出现频谱增长。
一旦总负反馈量超过约 20 dB，频谱增长就会停止，所有阶次的失真都会随着反馈量的增加而减小。因此

to generalize that this 20 dB number includes emitter degeneration or other local feedback. The existence of the SGD effect seems to make a good case for designing the amplifier for good open-loop linearity. The first 20 dB of feedback can be applied locally, typically as emitter degeneration, to get beyond the starting region of the SGD effect. The application of global negative feedback will then not be expected to cause any SGD.
一般认为，这个 20 dB 的数字包括发射极退化或其他局部反馈。SGD 效应的存在似乎为设计具有良好开环线性度的放大器提供了充分的理由。前 20 dB 的反馈可在局部应用，通常作为发射极退化，以超越 SGD 效应的起始区域。这样，全局负反馈的应用就不会导致任何 SGD。

SGD and Crossover Distortion
Crossover distortion is usually the biggest and most audible distortion in a properly designed amplifier. It starts out being rich in high-order products. Figure 24.5 shows
SGD 和分频失真
在设计合理的放大器中，分频失真通常是最大、最易听出的失真。一开始，它会产生大量高阶产品。图 24.5 显示

that the application of NFB to an amplifier reduces all orders of crossover distortion, right from the beginning.
在放大器中应用 NFB 可以从一开始就减少所有等级的分频失真

Although the Baxandall effect was an intriguing eye-opener, the audio community read too much into it, wrongly generalizing the results and asserting that feedback did not really reduce the net imperfection of the signal.
尽管巴桑道尔效应让人大开眼界，但音频界却对此过分解读，错误地将结果一概而论，断言反馈并没有真正减少信号的净不完美。

筑明

24.4 Global versus Local Feedback
Issues of open-loop bandwidth and frequency compensation largely pertain to global feedback loops that usually enclose virtually all of the amplifier stages. In contrast, local negative feedback rarely needs compensation and typically has very wide bandwidth. Emitter degeneration is a form of local negative feedback. Shunt feedback around a single stage is also local negative feedback.
Some who are opposed to the use of negative feedback are also opposed to the use of local feedback, but do not consider emitter degeneration to carry with it the supposed ills of negative feedback. Analysis shows that even emitter degeneration causes spectral growth distortion just like any other form of negative feedback.
全局反馈与局部反馈
开环带宽和频率补偿问题主要与全局反馈环路有关，全局反馈环路通常几乎包括所有放大器级。相比之下，局部负反馈很少需要补偿，通常具有很宽的带宽。发射极退化是局部负反馈的一种形式。单级周围的并联反馈也属于局部负反馈。
一些反对使用负反馈的人也反对使用局部负反馈，但并不认为发射极退化会带来负反馈的所谓弊端。分析表明，即使发射极退化也会导致频谱增长失真，就像任何其他形式的负反馈一样。

24.5 Timeliness of Correction
纠正的及时性
Some critics of negative feedback argue that NFB represents an electronic attempt to correct an error after it has happened and that the finite time delay and sequence of events make the correction faulty. The electronic time-of-flight delay and phase delay due to feedback compensation do indeed exist, but these delays must be small in order for the circuit to be stable. In an amplifier with a 1-MHz unity-gain frequency  the delay must certainly be less than 0.5 ms. This is 100 times smaller than the period of a 20-kHz sinusoid. This delay is merely a different way of recognizing that the distortion-reducing properties of negative feedback are less effective at very high frequencies. It does not suggest that negative feedback is failing to correct an error at 20 kHz.
一些批评负反馈的人认为，负反馈电路是在错误发生后试图纠正错误的一种电子尝试，有限的时间延迟和事件顺序使纠正错误成为泡影。由于反馈补偿造成的电子飞行时间延迟和相位延迟确实存在，但这些延迟必须很小，电路才能稳定。在单增益频率为 1 兆赫的放大器中，延迟肯定要小于 0.5 毫秒。这比 20 kHz 正弦波的周期小 100 倍。这种延迟只是一种不同的认识方式，即负反馈的失真抑制特性在极高频率下效果较差。这并不表明负反馈无法纠正 20 千赫的误差。

24.6 EMI from the Speaker Cable
来自扬声器电缆的 EMI
The speaker cable is a big antenna. The concern about negative feedback here is that EMI from the loudspeaker cable will get back to the input via the feedback path [16]. This concern is not completely unfounded. In fact, the use of a phase lead capacitor across the feedback resistor can make the input stage unnecessarily vulnerable to EMI that makes its way into the amplifier via the speaker cable. Such EMI will be attenuated by the shunting impedance of the output stage and by the feedback network before it arrives at the input stage. Nevertheless, this is a good reason to employ an input stage that has good signal-handling capability to high frequencies, such as a JFET stage or a well-degenerated BJT stage operated at a healthy bias current. Such an input stage is more resistant to EMI effects from the input port as well.
扬声器电缆就是一个大天线。人们对负反馈的担心是，扬声器电缆的 EMI 会通过反馈路径传回输入端 [16]。这种担心并非毫无根据。事实上，在反馈电阻器上使用相位引线电容器会使输入级不必要地受到通过扬声器电缆进入放大器的 EMI 的影响。这种 EMI 在到达输入级之前，会被输出级的分流阻抗和反馈网络所削弱。不过，这也是采用对高频具有良好信号处理能力的输入级（如 JFET 级或在健康偏置电流下工作的良好衰减 BJT 级）的充分理由。这样的输入级还能更好地抵御来自输入端口的电磁干扰效应。


24.7 Stability and Burst Oscillations
稳定性和突发振荡
An amplifier that does not employ global negative feedback does not need to be properly compensated (it does not need to be compensated at all) for global loop stability. Such amplifiers will tend to be less prone to burst oscillations due to global feedback loop instability. It is true that there are more possibilities to make a bad design with negative feedback. Negative feedback is a powerful tool that can be abused. However, abandoning NFB does not ensure that a power amplifier will be free from burst oscillations. This is especially true of oscillations that can originate locally in the output stage.
不采用全局负反馈的放大器不需要为全局环路稳定性进行适当补偿（根本不需要补偿）。这样的放大器往往不容易因全局反馈环路不稳定而出现突发振荡。使用负反馈的确有更多可能做出糟糕的设计。负反馈是一种可以被滥用的强大工具。然而，放弃 NFB 并不能确保功率放大器不会出现突发振荡。输出级局部产生的振荡尤其如此。


24.8 Clipping Behavior
剪切行为
The use of global negative feedback does tend to alter the clipping behavior of an amplifier. It sharpens up clipping edges and makes the onset of clipping more abrupt. The use of Baker clamps in the amplifier design will make the abrupt clipping cleaner, but will usually not soften it. If you are going to clip your amplifier often, you may not want to use negative feedback. Guitar amplifier designers learned this many years ago. Soft clipping circuits can eliminate this problem by gradually clipping the input to the amplifier before the amplifier itself clips. Unfortunately, soft clip circuits are rare because they increase circuit complexity and they increase measured amplifier distortion at levels below clipping. It is notable that some amplifiers that do not employ negative feedback clip rather sharply as well.
全局负反馈的使用往往会改变放大器的削波行为。它会使削波边缘更加尖锐，使削波开始时更加突然。在放大器设计中使用贝克夹钳会使突然的削波更加干净，但通常不会使其变得柔和。如果您要经常对放大器进行削波，您可能不想使用负反馈。吉他放大器设计人员多年前就知道了这一点。软削波电路可以在放大器本身发生削波之前逐渐削波放大器的输入，从而消除这一问题。遗憾的是，软削波电路并不多见，因为它们会增加电路的复杂性，而且在低于削波电平时会增加测量到的放大器失真。值得注意的是，一些未采用负反馈的放大器也会出现相当严重的削波。


References
参考资料

1.        Otala, M, “Transient Distortion in Transistorized Audio Power Amplifiers,” IEEE Transactions on Audio and Electro-acoustics, vol. AU-18, pp. 234–239, September 1970.
2.        M. Otala, and Leinonen, L., “The Theory of Transient Intermodulation Distortion,” IEEE Transactions on Acoustics, Speech and Signal Processing, vol. ASSP-25, no. 1, pp. 2–8, February 1977.
3.        Leach, W. M., “Transient IM Distortion in Power Amplifiers,” Audio, pp. 34–41, February 1975.
4.        Leach, W. M., “Suppression of Slew-rate and Transient Intermodulation Distortions in Audio Power Amplifiers,” J. Audio Eng. Soc., vol. 25, no. 7–8, pp. 466–473, July–August 1977.
5.        Greiner, R. A., “Amp Design and Overload,” Audio, pp. 50–62, November 1977.
6.        Otala, M., “Feedback-generated Phase Modulation in Audio Amplifiers,” 65th Convention of the Audio Engineering Society, preprint No. 1576, London, 1980.
7.        Otala, M., “Conversion of Amplitude Nonlinearities to Phase Nonlinearities in Feedback Audio Amplifiers,” Proc. Of IEEE International Conference on Acoustics, Speek and Signal Processing, pp. 498–499. Denver, CO, 1980.
8.        Otala, M., “Phase Modulation and Intermodulation in Feedback Audio Amplifiers,” 69th Convention of the Audio Engineering Society, preprint No. 1751, Hamburg, 1981.
9.        Otala, M., and Lammasniemi, J., “Intermodulation Distortion in the Amplifier Loudspeaker Interface,” 59th Convention of the Audio Engineering Society, preprint No. 1336, February 1978.
10.        Leinonen, E., Otala, M., and Curl, J., “A Method for Measuring Transient Intermodulation Distortion (TIM),” J. Audio Eng. Soc., vol. 25, no. 4, pp. 170–177, April 1977.
11.        Leinon, E., and Otala, M., “Correlation Audio Distortion Measurements,” J. Audio Eng. Soc., vol. 26, no. 1–2, pp. 12–19, January–February 1978.
12.        Cordell, R. R., “Another View of TIM,” Audio, February–March, 1980; available at www.cordellaudio.com.
13.        Cordell, R. R., “Phase Intermodulation Distortion–Instrumentation and Measurements,” Journal of the Audio Engineering Society, vol. 31, March 1983; available at www.cordellaudio.com.
14.        Cordell, R. R., “Open-loop Output Impedance and Interface Intermodulation Distortion in Audio Power Amplifiers,” preprint No. 1537, 64th Convention of the AES, 1982; available at www.cordellaudio.com.
15.        Baxandall, P. J., “Audio Power Amplifier Design–5,” Wireless World, December 1978.
16.        Thiele, A. N., “Load Stabilizing Networks for Audio Amplifiers,” J. Audio Eng. Soc., vol. 24, no. 1, pp. 20–23, January–February 1976.

lin889889

机器翻译的"分频失真"应该是“交越失真”。

wopp

谢谢分享

丹麦王子

争啥呀，基本上都是带负反馈的，极少数例外，比如我这个牌子

无语密码

这种资料仅仅适合停产的晶体管

docmd

好文收藏了，此文与我猜想居然相同。果然英雄所见相同。

capa

Douglas Self的中译版都读不下去，Bob Cordell的英文版更令人头大

minor

听过无负反馈功放后，不再想听有负反馈的了。
自然，野性，从容，无拘无束。。。

swsw4321

差分动态过小，高频相移反馈延迟造成，提升电路转换速率，限制放大带宽

king5555

折衷是负反饋不从输出的喇叭端，而提前由前面的电压放大級进行负反饋。等于输出級是0dB的。

筑明

king5555 发表于 2023-7-21 08:08
折衷是负反饋不从输出的喇叭端，而提前由前面的电压放大級进行负反饋。等于输出級是0dB的。

后面一章有关于电流放大级不参与反馈的说法，那将需要更多的晶体管来并联，且电流会需要很大，这意味着更高的成本：
输出级是放大器中造成失真的最主要因素之一，因此任何不包含全局负反馈的放大器都面临着巨大的挑战。在没有反馈的情况下，将交越失真降至最低尤为重要。最直接的方法是使用更多对输出晶体管，并使用相当小的发射极电阻器进行最佳偏置。放大器的运行温度会更高，但它将受益于更大的 A 类区域，而不会出现 gm 倍增。
输出级的贝塔下降会导致 VAS 负载与信号有关。这将导致 VAS 的中等高阻抗输出节点失真。因此，输出级的电流增益必须非常高。输出三极管可以提供足够的电流增益，但四级输出发射极跟随器（输出四极管）可能更能减少这种失真。输出晶体管中的厄尔效应也会导致失真。
由于没有 NFB 来降低输出阻抗，某些 NNFB 放大器的阻尼系数可能很低。幸运的是，如果将更多的输出对与输出 Quad 结合使用，阻尼系数就会很好。
直流偏移可能是 NNFB 放大器的一个问题，因为放大器增益通常会保持在直流以下的满值。在没有 NFB 的情况下，没有反馈网络并联电容器（通常是电解电容）来将增益降至直流时的统一值。因此，输入级直流偏移乘以 NNFB 放大器的全增益。此外，由于在某些设计中输入级增益可能低至 1，因此 VAS 中的直流偏移会增加偏移问题。严格遵守无反馈理念就排除了使用直流伺服的可能性。

king5555

筑明 发表于 2023-7-21 22:19
后面一章有关于电流放大级不参与反馈的说法，那将需要更多的晶体管来并联，且电流会需要很大，这意味着 ...

这样是提前负反饋，有时候在本坛看的到，尤其是那位专搞MOS输出管的网友，如图。
输出阻抗绝大分额是由输出管及其输入侧之阻抗決定，有了负反饋只能微微再降低输出阻抗。而引入负反饋的目地不为阻抗，而是为了稳定增益和直流电平。
如图，提前负反饋，缺点是输出端的直流电平可能会偏差。红虚线相接了就是一般的负反饋。
以上一些了解，恐有偏失。

筑明

king5555 发表于 2023-7-21 23:35
这样是提前负反饋，有时候在本坛看的到，尤其是那位专搞MOS输出管的网友，如图。
输出阻抗绝大分额是由 ...

20年前的国产功放基本上都是你说的这种反馈，认为这样能隔离音箱的反电动势，各种杂志也在吹嘘这种新的无负反馈技术，但是这种结构需要更多的电流管并联来降低输出阻抗，然后加大静态电流来改善交越失真，所以取消末级反馈没有那么简单。

cgpad

                         并联大功放管。





1 Gryphon Tabu Century+-50V.rar (76.64 KB, 下载次数: 104)

2023-7-22 20:56 上传

点击文件名下载附件

2 Gryphon Tabu Century +-50V.rar (96.79 KB, 下载次数: 86)

2023-7-22 20:54 上传

点击文件名下载附件









https://audiocircuit.dk/download ... gnature-pwr-sch.pdf

http://bbs.hifidiy.net/thread-755116-1-1.html
、