With a simple cut and jump the design could be made to have essentially zero feedback. Moving the feedback connection from the output of the half H bridge to the input of the gate drive circuit would be all that is needed. This would produce the density bit stream of the delta modulator and greatly reduce or essentially remove the influence the load has on the feedback.
Another alternative is to feed a pre-recorded DSD stream directly into the gate drive. The output level would then be adjusted by changing the power supply rails for the half H bridge.
Another alternative is to feed a pre-recorded DSD stream directly into the gate drive. The output level would then be adjusted by changing the power supply rails for the half H bridge.
It is very happy to see the hot discussion here about the feedback. As I mentioned in the first paragraph of the second thread:
"No-feedback" means that neither global nor local audio signal feedback is used in the amplifier design. However, to ensure circuit stability, other functional circuits may still include feedback mechanisms, both positive and negative.
I agree @TNT, the R3 is part of the self-oscillation circuit, without it the entire circuit doesn't work. the feedback PDM pulse signal does include the modulated information of input audio (the width and period of the pulse signal), but it doesn't mean it is the negative feedback of direct audio signal as the traditional class B AMP.
To obtain the low distortion, the feedback here provides more precise PDM modulation and tens dB of PSRR, not for direct correcting the nonlinear of the amplifier circuit like Calss A/B.
That is AI generated content, it is not clear, but I think that answer is based on some public context, just lack of the human-like logic.
so called the low distortion, most of time, is the distortion at frequency point of 1kHz, according to the character of the solid-state AMP, the high frequency distortion is much higher than that of 1kHz, that is why I test the distortion up to 10kHz.
for instance, if a musician plays the violin in front of you, you will hear the rich harmonic of the tone (which usually extends into higher frequency), but if you play the same music with an AMP with higher distortion at high frequencies (whatever the nominal distortion of 1kHz), unpleasant "rich" harmonics are generated by distortion, and you won't hear the true rich harmonics. Physically, the harmonic through the AMP is much "richer" than the original music, you will sound poor rather than rich. when the AMP has low distortion in high frequency, the original true rich harmonic of music is faithfully enlarged and reproduced, allowing you hear and feel the rich harmonics and nature, transparent of the sounds as well.
Really. Why go with something like that, when you could just generate it yourself? Even Ebay these days offers AI generated descriptions for listing something and it's so blatantly obvious when used. I think I tried it once, and then cut out the thick icing, substituting my own words where it clearly went too far with a sha-na-na-na-na sales pitch shuffle, reducing the prose to facts about what I was selling...
I'm glad you pointed that out, as in retrospect it does seem a little fishy somehow. Now I know to be on the lookout anywhere for such; it's the times within which we now live.
I already pointed that out in the original thread, it is answer of ChatGPT. And I just pasted the answer as a picture there.
Indeed, there are more and more AI generated contents around us, have to lookout...
Indeed, there are more and more AI generated contents around us, have to lookout...
It doesn't matter whether it is AI generated or whether a monkey typed it. You posted it as 'your' goals.
You posted what is basically inconsistent and conflicting marketing blurbs as if it was your opinion..
So don't hide behind an AI.
Aside from your semantic wriggling to save the 'no feedback' BS.
Worse is that you can't be trusted that what you posted is really your opinion.
Ignore list.
Jan
You posted what is basically inconsistent and conflicting marketing blurbs as if it was your opinion..
So don't hide behind an AI.
Aside from your semantic wriggling to save the 'no feedback' BS.
Worse is that you can't be trusted that what you posted is really your opinion.
Ignore list.
Jan
To be clear - I have not had any ideas about what feedback might or might not be present. I tend to believe that persons like J Didden have a very deep circuit knowledge and trust their statements.
So my take is that you probably think you dint have any feedback but its quite hard sometimes to know as feedback is such a natural and good mechanism so it keeps in without one noticing it.
//
Your initial post to describe this was not so precise - the language was not clear enough. So you used AI? Bad AI. Always state f AI is used!
I think you should write in your own words - they seem better than an AI.
Now its clear.
//
I'm, for one, eager to figure out how this was achieved (but I understand if this info is a proprietary bit of a puzzle, not to be shared with us):
I did realise early in this thread that the generic block diagram was misleading - surely there must be negative feedback there, as clearly shown in a diagram... but to which signals and how it is applied, is still a mystery to me.
So, I looked for more info on class D and found the dissertation I provided the link to in post #16. So I did a bit of reading. Nevertheless, I am still no closer to figuring out what is going on with this particular design. I did learn bits and pieces here and there, so I think I understand the blurb around "other functional circuits may still include feedback mechanisms, both positive and negative."
Anyhow, I appreciate the effort put into this development work, and I got something out of it - a bit of reading about something I know very little of... apart from how class D amps sound. But... is this a typical class D amp..? It seems it is not... but why exactly..? Don't know...
I did realise early in this thread that the generic block diagram was misleading - surely there must be negative feedback there, as clearly shown in a diagram... but to which signals and how it is applied, is still a mystery to me.
So, I looked for more info on class D and found the dissertation I provided the link to in post #16. So I did a bit of reading. Nevertheless, I am still no closer to figuring out what is going on with this particular design. I did learn bits and pieces here and there, so I think I understand the blurb around "other functional circuits may still include feedback mechanisms, both positive and negative."
Anyhow, I appreciate the effort put into this development work, and I got something out of it - a bit of reading about something I know very little of... apart from how class D amps sound. But... is this a typical class D amp..? It seems it is not... but why exactly..? Don't know...
Yeah... I know. Just see my previous post... I am confused as well for the exact same reason... this statement is not helping either:
Oh well, we may get a hint or 2 from @flmhhh
Oh well, we may get a hint or 2 from @flmhhh
So, does "no global feedback" mean no feedback to the input is taken from after the output filter?
ETA: Woops, what I saw was just a preview of the PDF, thus only part of the schematic.
But I see the point of the capacitive feedback, and that's surely necessary (and likewise the hysteresis in the next stage) for this circuit to be self-oscillating.
I think the only "no feedback" circuit in this thread is the one with the input signal and triangle wave going into the comparator.
The OP mentions Mr. Pass' designs - a Class A power MOSFET without a drain resistor would have no feedback, but I seem to recall that they (generally or always) do have such resistors.
Last edited:
The sigma-delta Class D amplifier operates in a self-oscillating mode, where the feedback loop (R3) is set up to induce oscillation, essentially functioning as positive feedback. The integrator introduces a 90-degree phase shift, and together with the Schmitt comparator and subsequent MOSFET drivers, it accumulates around 360 degrees of phase shift, forming a positive feedback loop to generate oscillation. However, typical audio feedback refers to negative feedback for the audio signal, where a 180-degree phase shift (with some phase margin) is required. Can R3 in this feedback loop serve both as positive feedback for the pulse signal and negative feedback for the audio signal? Clearly, it cannot.
From another perspective, the sigma-delta Class D amplifier, after removing the LC output filter, essentially acts as an oversampling 1-bit ADC. The R3 feedback sends a 1-bit digital signal back to the input, where it performs addition and subtraction (Delta) with the input signal to generate an error signal, which is further modulated (Sigma) to reduce the quantization error of the 1-bit digital signal. This is essentially the same function as the comparator in a traditional SAR ADC (where the quantized digital signal is compared with the input signal via a DAC). The comparison with the input analog signal is aimed at reducing quantization noise and is not negative feedback of the modulated audio signal.
From another perspective, the sigma-delta Class D amplifier, after removing the LC output filter, essentially acts as an oversampling 1-bit ADC. The R3 feedback sends a 1-bit digital signal back to the input, where it performs addition and subtraction (Delta) with the input signal to generate an error signal, which is further modulated (Sigma) to reduce the quantization error of the 1-bit digital signal. This is essentially the same function as the comparator in a traditional SAR ADC (where the quantized digital signal is compared with the input signal via a DAC). The comparison with the input analog signal is aimed at reducing quantization noise and is not negative feedback of the modulated audio signal.
On a side note: Usually if people quote from an article or a book or another quote they say so, and mention the author. Like ' Mr Pen described the feedback in a modulator is his article in Amazing Audio March 2022' or something like that. This makes it clear what the source is and that it is not from the poster. I wonder wheter we should not use the same 'rule' with AI generated content. @flmhhh's 1st post contained an AI generated part which clearly was contradictory and in hindsight, typical AI: all the words are correct but the text doesn't make a lot of sense.
I'll talk to Jason about that.
Comments invited (PM probably better).
/back to regular programming.
Jan
I'll talk to Jason about that.
Comments invited (PM probably better).
/back to regular programming.
Jan
There are many interesting discussions about PDM signal feedback of sigma-delta class D amplifier, the following is my understanding of that during the design of the DIY project.
A Brief Analysis of Negative Feedback in Analog Audio Amplifiers
Below is a functional block diagram of negative feedback in a typical audio power amplifier:
Figure 3, A typic 2-poles IIR filter，（ http://www.audiodevelopers.com/2-digital-processing-basics/）
Although the two systems are clearly different from a linear perspective, when viewed in terms of impulse response duration, both fall under the category of Infinite Impulse Response (IIR) systems. This means that once an input is applied to the system, it will have an "infinite" effect over time. Consequently, an IIR system has the potential to be unstable. However, with careful design of the feedback network's poles, an IIR system can operate very stably. Over time, the input signal will decay through the feedback loop and eventually approach zero.
A well-designed audio negative feedback system offers many advantages over a direct-feed system, such as a flatter amplitude-frequency response curve, lower equivalent output impedance, and a higher power supply rejection ratio. However, the phase response of a feedback system is nonlinear, which simply means that different frequencies will experience different delay times. Fortunately, the human ear is not very sensitive to the phase of sound signals.
Sigma-Delta Class D Amplifier
A typical Sigma-Delta Class D amplifier consists of several key functional modules, including an adder, integrator, comparator, D latch, and feedback loop, where the D latch can be omitted. Circuits without a D latch differ from those with one in a significant way. Without the D latch, the circuit strictly adheres to the Nyquist sampling theorem since the output digital signal can switch between high and low levels at any point in time. On the other hand, circuits with a D latch cannot directly apply the Nyquist theorem. Due to the introduction of the D latch, a higher sampling frequency is required to average out the uncertainties caused by the latch's delay.
The inclusion of the D latch allows the Sigma-Delta modulation to be implemented in the digital domain, with DSD (Direct Stream Digital) being a prime example of such a digital audio signal.
By using the above modules, a self-oscillating 1-bit ADC converter is formed. The output at the "Out" port is a 1-bit bitstream, which modulates the input audio signal.
Figure 4, from ‘Designing Audio Power Amplifiers’ By Bob Cordell
Like traditional ADCs, to reconstruct the amplified input audio signal, an additional functional module, namely a DAC (Digital-to-Analog Converter), is required to convert the digital signal back into an analog signal.
Interestingly, the reconstruction process of a 1-bit bitstream signal is relatively straightforward, achieved through a simple low-pass filter (LPF). Typically, a second-order LC low-pass filter suffices for this purpose.
One aspect that can be somewhat confusing is the feedback loop in a Sigma-Delta amplifier. Is it the same as the negative feedback loop found in traditional audio amplifiers?
Firstly, it is indeed a feedback loop, and its primary purpose is to maintain the oscillation of the Sigma-Delta amplifier. This feedback works in conjunction with the feedforward circuit to provide a 360-degree phase delay for the oscillation frequency signal. The input to this loop is the modulated bitstream, and the output can either be the bitstream itself or the bitstream that has been low-pass filtered (adjusting the delay to produce different self-oscillation frequencies).
The most critical point is that this feedback differs from traditional global negative feedback in that it represents a finite time response, occurring only within a single PWM cycle. Below is a simple analysis of this distinction (to simplify the analysis, we will neglect propagation delays and other factors that do not affect the fundamental operating principles).
Figure 5
Firstly, in this example, the basic principle of modulation is that the area of the negative pulse waveform is proportional to the amplitude of the input audio signal. It can be observed that the area is equal to the width of the negative pulse multiplied by the amplitude of the pulse signal.
When the audio input signal is summed with the feedback digital signal and directly acts on the integrator circuit, the magnitude of the superimposed signal directly reflects on the slope of the integral output waveform. The larger the input signal, the lower the slope of the integrator during the negative pulse, thus resulting in a wider width for the negative pulse. When the area of the negative pulse equals the input signal multiplied by a fixed gain (a * Vin), the output pulse signal flips. This process repeats in the next sampling cycle.
During this process, if the supply voltage changes, for instance, increases, then according to the principle of the area equaling a * Vin, the time width of the negative pulse signal will shorten; conversely, it will lengthen if the voltage decreases. This explains why the Sigma-Delta amplifier exhibits a high Power Supply Rejection Ratio (PSRR).
At the same time, based on the principle of area equaling a * Vin, the Sigma-Delta amplifier also serves compensatory functions, such as addressing asymmetries in the rising and falling edges and the distortion caused by gradual rising and falling edges.
It is easy to observe that the digital signal feedback to the input acts only during the current sampling cycle. Due to the 1-bit nature of the digital signal (the signal flip serves as a reset process for the previous signal, without any memory), the signal in the next cycle is unrelated to that of the previous cycle. This means that this feedback does not have an infinite response, which is one of the key differences compared to traditional audio negative feedback.
The Sigma-Delta Class D amplifier can be viewed in two parts: the first part is the 1-bit analog-to-digital conversion (ADC), and the second part is the digital-to-analog conversion (DAC).
The process of Sigma-Delta analog-to-digital conversion can be described simply as follows: within a sampling cycle, when the integral of the negative pulse signal equals a * Vin, the analog-to-digital conversion for that cycle is completed until the next sampling cycle begins. Signals between sampling cycles are independent. Therefore, this feedback signal acts on each sampling cycle, ensuring the precise generation of the negative pulse area within that cycle.
The digital-to-analog conversion process is generally conducted using an LC filter. Although both the inductor and capacitor are passive components, they also exhibit nonlinear effects, which can produce nonlinear distortion. To minimize the nonlinear distortion generated by the LC filter, traditional audio signal negative feedback can be very effective. For better results, it is advisable to increase the gain of the feedforward amplifier during the overall design. Figure 6 shows an added feedback circuit in a Class D amplifier that is identical to traditional audio negative feedback.
Figure 6， from ‘Designing Audio Power Amplifiers’ By Bob Cordell
From the perspective of Sigma-Delta analog-to-digital conversion, regardless of how it is implemented within a sampling period, the 1-bit digital bitstream theoretically represents the audio input signal without distortion. This means that not only is the amplitude-frequency characteristic completely flat, but the phase-frequency characteristic is also entirely linear. Therefore, in the absence of the negative feedback shown in Figure 6, the Sigma-Delta Class D amplifier can be considered a system without audio negative feedback, as a significant feature of traditional audio negative feedback is its non-linear phase-frequency characteristic.
Finally, I believe that the Class D amplifier has a notable advantage over traditional amplifiers: from the fundamental working principle, the distortion of a Class D amplifier is frequency-independent. My tests indicate that if the distortion at 1 kHz is 0.1%, then the distortion at 20 kHz is approximately the same as at 1 kHz. In contrast, traditional amplifiers are generally affected by the parasitic parameters of components and circuit boards, leading to distortion that is frequency-dependent, with higher distortion at high frequencies compared to low frequencies.
As I have not found detailed analyses of the Sigma-Delta feedback loop, many references describe this feedback loop differently. This article is merely a collection of my personal insights gained during the DIY process, primarily comparing the functionality and effects with traditional audio negative feedback. If there are any inaccuracies, please feel free to correct me!
A Brief Analysis of Negative Feedback in Analog Audio Amplifiers
Below is a functional block diagram of negative feedback in a typical audio power amplifier:
Figure 1， from Figure 2.9 of ‘Audio Power Amp Design Handbook 4th Edition’ By Douglas Self
By simplifying the diagram above, the following model can be summarized: Fw(s) represents the feedforward network, which describes the open-loop response of the amplifier. Fb(s) is a feedback network, whose impulse response (especially the phase response) must ensure that the network operates in a negative feedback mode.
Figure 3, A typic 2-poles IIR filter，（ http://www.audiodevelopers.com/2-digital-processing-basics/）
Although the two systems are clearly different from a linear perspective, when viewed in terms of impulse response duration, both fall under the category of Infinite Impulse Response (IIR) systems. This means that once an input is applied to the system, it will have an "infinite" effect over time. Consequently, an IIR system has the potential to be unstable. However, with careful design of the feedback network's poles, an IIR system can operate very stably. Over time, the input signal will decay through the feedback loop and eventually approach zero.
A well-designed audio negative feedback system offers many advantages over a direct-feed system, such as a flatter amplitude-frequency response curve, lower equivalent output impedance, and a higher power supply rejection ratio. However, the phase response of a feedback system is nonlinear, which simply means that different frequencies will experience different delay times. Fortunately, the human ear is not very sensitive to the phase of sound signals.
Sigma-Delta Class D Amplifier
A typical Sigma-Delta Class D amplifier consists of several key functional modules, including an adder, integrator, comparator, D latch, and feedback loop, where the D latch can be omitted. Circuits without a D latch differ from those with one in a significant way. Without the D latch, the circuit strictly adheres to the Nyquist sampling theorem since the output digital signal can switch between high and low levels at any point in time. On the other hand, circuits with a D latch cannot directly apply the Nyquist theorem. Due to the introduction of the D latch, a higher sampling frequency is required to average out the uncertainties caused by the latch's delay.
The inclusion of the D latch allows the Sigma-Delta modulation to be implemented in the digital domain, with DSD (Direct Stream Digital) being a prime example of such a digital audio signal.
By using the above modules, a self-oscillating 1-bit ADC converter is formed. The output at the "Out" port is a 1-bit bitstream, which modulates the input audio signal.
Figure 4, from ‘Designing Audio Power Amplifiers’ By Bob Cordell
Like traditional ADCs, to reconstruct the amplified input audio signal, an additional functional module, namely a DAC (Digital-to-Analog Converter), is required to convert the digital signal back into an analog signal.
Interestingly, the reconstruction process of a 1-bit bitstream signal is relatively straightforward, achieved through a simple low-pass filter (LPF). Typically, a second-order LC low-pass filter suffices for this purpose.
One aspect that can be somewhat confusing is the feedback loop in a Sigma-Delta amplifier. Is it the same as the negative feedback loop found in traditional audio amplifiers?
Firstly, it is indeed a feedback loop, and its primary purpose is to maintain the oscillation of the Sigma-Delta amplifier. This feedback works in conjunction with the feedforward circuit to provide a 360-degree phase delay for the oscillation frequency signal. The input to this loop is the modulated bitstream, and the output can either be the bitstream itself or the bitstream that has been low-pass filtered (adjusting the delay to produce different self-oscillation frequencies).
The most critical point is that this feedback differs from traditional global negative feedback in that it represents a finite time response, occurring only within a single PWM cycle. Below is a simple analysis of this distinction (to simplify the analysis, we will neglect propagation delays and other factors that do not affect the fundamental operating principles).
Figure 5
Firstly, in this example, the basic principle of modulation is that the area of the negative pulse waveform is proportional to the amplitude of the input audio signal. It can be observed that the area is equal to the width of the negative pulse multiplied by the amplitude of the pulse signal.
When the audio input signal is summed with the feedback digital signal and directly acts on the integrator circuit, the magnitude of the superimposed signal directly reflects on the slope of the integral output waveform. The larger the input signal, the lower the slope of the integrator during the negative pulse, thus resulting in a wider width for the negative pulse. When the area of the negative pulse equals the input signal multiplied by a fixed gain (a * Vin), the output pulse signal flips. This process repeats in the next sampling cycle.
During this process, if the supply voltage changes, for instance, increases, then according to the principle of the area equaling a * Vin, the time width of the negative pulse signal will shorten; conversely, it will lengthen if the voltage decreases. This explains why the Sigma-Delta amplifier exhibits a high Power Supply Rejection Ratio (PSRR).
At the same time, based on the principle of area equaling a * Vin, the Sigma-Delta amplifier also serves compensatory functions, such as addressing asymmetries in the rising and falling edges and the distortion caused by gradual rising and falling edges.
It is easy to observe that the digital signal feedback to the input acts only during the current sampling cycle. Due to the 1-bit nature of the digital signal (the signal flip serves as a reset process for the previous signal, without any memory), the signal in the next cycle is unrelated to that of the previous cycle. This means that this feedback does not have an infinite response, which is one of the key differences compared to traditional audio negative feedback.
The Sigma-Delta Class D amplifier can be viewed in two parts: the first part is the 1-bit analog-to-digital conversion (ADC), and the second part is the digital-to-analog conversion (DAC).
The process of Sigma-Delta analog-to-digital conversion can be described simply as follows: within a sampling cycle, when the integral of the negative pulse signal equals a * Vin, the analog-to-digital conversion for that cycle is completed until the next sampling cycle begins. Signals between sampling cycles are independent. Therefore, this feedback signal acts on each sampling cycle, ensuring the precise generation of the negative pulse area within that cycle.
The digital-to-analog conversion process is generally conducted using an LC filter. Although both the inductor and capacitor are passive components, they also exhibit nonlinear effects, which can produce nonlinear distortion. To minimize the nonlinear distortion generated by the LC filter, traditional audio signal negative feedback can be very effective. For better results, it is advisable to increase the gain of the feedforward amplifier during the overall design. Figure 6 shows an added feedback circuit in a Class D amplifier that is identical to traditional audio negative feedback.
Figure 6， from ‘Designing Audio Power Amplifiers’ By Bob Cordell
From the perspective of Sigma-Delta analog-to-digital conversion, regardless of how it is implemented within a sampling period, the 1-bit digital bitstream theoretically represents the audio input signal without distortion. This means that not only is the amplitude-frequency characteristic completely flat, but the phase-frequency characteristic is also entirely linear. Therefore, in the absence of the negative feedback shown in Figure 6, the Sigma-Delta Class D amplifier can be considered a system without audio negative feedback, as a significant feature of traditional audio negative feedback is its non-linear phase-frequency characteristic.
Finally, I believe that the Class D amplifier has a notable advantage over traditional amplifiers: from the fundamental working principle, the distortion of a Class D amplifier is frequency-independent. My tests indicate that if the distortion at 1 kHz is 0.1%, then the distortion at 20 kHz is approximately the same as at 1 kHz. In contrast, traditional amplifiers are generally affected by the parasitic parameters of components and circuit boards, leading to distortion that is frequency-dependent, with higher distortion at high frequencies compared to low frequencies.
As I have not found detailed analyses of the Sigma-Delta feedback loop, many references describe this feedback loop differently. This article is merely a collection of my personal insights gained during the DIY process, primarily comparing the functionality and effects with traditional audio negative feedback. If there are any inaccuracies, please feel free to correct me!