Hi everyone,
I’m working on a Amplifier 100W and looking to implement a system where I can inject a negative noise signal directly into the signal path after the output buffer, specifically between the output buffer and the speaker output terminal. The goal is to cancel out unwanted noise via destructive interference, without affecting the main audio signal's quality.
Here’s what I’d like to understand:
Thanks in advance for your help!
I’m working on a Amplifier 100W and looking to implement a system where I can inject a negative noise signal directly into the signal path after the output buffer, specifically between the output buffer and the speaker output terminal. The goal is to cancel out unwanted noise via destructive interference, without affecting the main audio signal's quality.
Here’s what I’d like to understand:
- Can I use capacitive coupling to inject this inverted noise signal into the high-power signal path?
- If yes, what would be the appropriate capacitor value and voltage rating for this setup?
- Are there other methods (e.g., inductive coupling or resistive networks) that could work better for this purpose?
- What precautions should I take to ensure that this injection method doesn’t degrade the audio quality or interfere with the main signal's integrity?
- The amplifier runs at (100 W into 8Ω), so the components need to handle significant currents and voltages.
- I’m trying to avoid transformers, op-amps, or any components that could introduce distortion or phase shifts or are not suited in terms of Voltage rating.
Thanks in advance for your help!
Hi,
Any signal processing, even a simple sum as you need, should be done at low signal level, so at the amplifier input.
You can use an opamp - at unitary gain, it will not add any relevant phase shift or distortion.
At low signal level, you can even sum signals with 2 resistors only if you want to do so.
Do you have special needs to sum signals at power level? (by "output buffer" I understand the final power stage of the am)
Any signal processing, even a simple sum as you need, should be done at low signal level, so at the amplifier input.
You can use an opamp - at unitary gain, it will not add any relevant phase shift or distortion.
At low signal level, you can even sum signals with 2 resistors only if you want to do so.
Do you have special needs to sum signals at power level? (by "output buffer" I understand the final power stage of the am)
Hi,
Thanks for your reply!
I understand your point about summing signals at low levels, but in this case, it’s not a viable option for me. The issue with a feedback system, especially at the input stage, is that it always introduces delay and, as a result, phase shift. This is unavoidable due to the time it takes for the signal to propagate through the amplifier stages and for the feedback loop to correct errors.
That’s why I’m aiming for a direct injection of the inverted noise signal into the output signal path, immediately after the output buffer and before the speaker terminals. This method should avoid the latency and phase issues of feedback and allow for a more precise cancellation.
Do you see any challenges with using capacitive coupling for this injection at the power stage? Or is there another method you’d recommend for summing signals at this point?
Looking forward to your insights!
Thanks for your reply!
I understand your point about summing signals at low levels, but in this case, it’s not a viable option for me. The issue with a feedback system, especially at the input stage, is that it always introduces delay and, as a result, phase shift. This is unavoidable due to the time it takes for the signal to propagate through the amplifier stages and for the feedback loop to correct errors.
That’s why I’m aiming for a direct injection of the inverted noise signal into the output signal path, immediately after the output buffer and before the speaker terminals. This method should avoid the latency and phase issues of feedback and allow for a more precise cancellation.
Do you see any challenges with using capacitive coupling for this injection at the power stage? Or is there another method you’d recommend for summing signals at this point?
Looking forward to your insights!
The scheme you are proposing will only work with noise that originates from things like power lines, ripple in power supplies etc. In other words, periodic signals. Thermal noise is RANDOM -- there is no such thing as a "negative noise generator" for it.
Unfortunately, the term "noise" covers a lot of territory regarding the mechanisms and sources that generate it.
Unfortunately, the term "noise" covers a lot of territory regarding the mechanisms and sources that generate it.
You are trying to reduce the amplifier's noise level? This has been considered a solved problem in recent years.
And where would you get the noise signal from? Most of the amplifier noise is from the early stages with gain,
but you could not separate the noise from the signal or distortion.
And where would you get the noise signal from? Most of the amplifier noise is from the early stages with gain,
but you could not separate the noise from the signal or distortion.
Are you talking about actual stochastic/thermal noise or linear/non-linear distortion?
It it’s the former, you can’t subtract noise from the output, but you’d have to resort to the “noise summing” (might not be the correct terminology) that played a prominent role in the hilarious Tom Evans/MendItMark feud.
If you want to do feed-forward of a distortion correction vector, well, it’s complicated. My dad actually made a career of that for RF amplifiers and held a few patents in the field. I doubt you’d ever have to resort to that in audio, other than for audiophile bragging rights.
It it’s the former, you can’t subtract noise from the output, but you’d have to resort to the “noise summing” (might not be the correct terminology) that played a prominent role in the hilarious Tom Evans/MendItMark feud.
If you want to do feed-forward of a distortion correction vector, well, it’s complicated. My dad actually made a career of that for RF amplifiers and held a few patents in the field. I doubt you’d ever have to resort to that in audio, other than for audiophile bragging rights.
To nullify output noise just connect the return path of your speaker with the output of a -100W amp buld from dark matter.
This is called the anisotropic noise dumpster.
This is called the anisotropic noise dumpster.
@Damnati I'm intrigued by this. What are you trying to do exactly?
1. Where do you take the noise from?
2. How do you extract the noise from the signal?
3. What are you hoping to gain (no pun intended)?
4. What kind of noise is it? Thermal? Phase noise? Distortion?
The best place to mitigate noise is usually BEFORE amplification. What's the reason you're trying to remove it after amplification?
1. Where do you take the noise from?
2. How do you extract the noise from the signal?
3. What are you hoping to gain (no pun intended)?
4. What kind of noise is it? Thermal? Phase noise? Distortion?
The best place to mitigate noise is usually BEFORE amplification. What's the reason you're trying to remove it after amplification?
Where do you get this "inverted noise signal?" At least this makes more sense than the "negative noise signal" as you called it in your first post, but it still doesn't completely make sense.
You could try a dual op amp and make both halves a nab tape amp. Then invert one and see if it cancels noise.
I think you might see 6db. Same as parallel input transistors on a MC preamp.
I think you might see 6db. Same as parallel input transistors on a MC preamp.
Hi,
Sorry, but my question was specifically about how to couple a low-level signal from an op-amp to the output of a power amplifier. It is completely irrelevant what the signal represents—whether it's noise, distortion, a fart, or the sound of a falling teacup. Please stay on topic.
I have a negative noise signal that is a perfect inverse of the noise in the main signal. It is not phase-shifted and runs perfectly parallel to the original signal. I just need an RC network to couple it to the main output. I only need to know how to do this.
I'm also not interested in your family stories.
Thanks in advance.
Sorry, but my question was specifically about how to couple a low-level signal from an op-amp to the output of a power amplifier. It is completely irrelevant what the signal represents—whether it's noise, distortion, a fart, or the sound of a falling teacup. Please stay on topic.
I have a negative noise signal that is a perfect inverse of the noise in the main signal. It is not phase-shifted and runs perfectly parallel to the original signal. I just need an RC network to couple it to the main output. I only need to know how to do this.
I'm also not interested in your family stories.
Thanks in advance.
- There is no such thing as negative noise? Really?
That’s simply not true:- Every signal has an inverse – no matter what you call it. If you take any signal S(t)), its inverse is simply −S(t). The same applies to noise. If you take any noise signal and invert it, you get the negative of the original noise. That’s literally how active noise cancellation works.
- Noise isn’t truly random – at least not in the way people think. Sure, thermal noise and some types of noise are modeled as random processes, but in reality, noise has predictable characteristics. It has a spectrum, it can be correlated, and in many cases, it has deterministic components.
- Any noise can be turned into a common-mode signal and canceled – This is basic signal processing. If you can measure noise at two points and it affects both equally, it can be treated as a common-mode signal and rejected. That’s exactly how differential amplifiers, balanced audio (XLR), and noise-canceling circuits work. You take the unwanted signal, invert it, and cancel it out.
- Every signal has an inverse – no matter what you call it. If you take any signal S(t)), its inverse is simply −S(t). The same applies to noise. If you take any noise signal and invert it, you get the negative of the original noise. That’s literally how active noise cancellation works.
How do I couple the output of an op-amp to the high-power output of an output buffer?
Here they doing it with a Transformer:
In the Benchmark AHB2, they do it via an RC network, but the patent doesn’t explain it—of course not.
https://patentimages.storage.googleapis.com/87/3b/ca/b85fd90cbe2d2d/US8421531.pdf
Thanks in advanced!
Last edited:
Looks like this is feedforward to reduce harmonic distortion, and has little to do with noise cancellation. Noise cancellation, since noise is a stochastic process, will gain you a whopping 3dB improvement in S/N ratio.
A approach to give noise reduction is to parallel shunt feedback opamps, with output summing resistors in the range 10-47 ohms. Two opamps, 3dB. 4 opamps 6dB. 8 opamps 9dB. So every doubling of the number of opamps adds another 3dB in S/N ratio. So 16 opamps gets you 12dB etc etc.
A patent, by proceeding to grant, has to give enough information so that someone "skilled in the art" can build one.
A approach to give noise reduction is to parallel shunt feedback opamps, with output summing resistors in the range 10-47 ohms. Two opamps, 3dB. 4 opamps 6dB. 8 opamps 9dB. So every doubling of the number of opamps adds another 3dB in S/N ratio. So 16 opamps gets you 12dB etc etc.
A patent, by proceeding to grant, has to give enough information so that someone "skilled in the art" can build one.
Alright, let’s cut the noise (pun intended) and actually solve the problem instead of getting lost in definitions.
Restating the question for those who got sidetracked:
I need to couple a low-level signal from an op-amp to the high-power output of a power amplifier. The signal is a pre-inverted version thats made by the subtraction of the amplified signal and the reference signal. It will have some amount of the following components:1️⃣Amplifier-Induced Distortions
These distortions occur because real amplifiers are not perfectly linear:- Harmonic Distortion (Additional harmonics introduced by nonlinear amplification)
- Intermodulation Distortion (IMD) (Nonlinear mixing of different frequencies)
- Phase Distortion (Phase shifts affecting the audio signal)
- Transient Intermodulation Distortion (TIM) (Distortion caused by fast signal transitions)
- Slew Rate Limiting (Limitation of the maximum signal rise rate)
- Crossover Distortion (Zero-crossing distortion in Class-AB amplifiers)
- Memory Distortion (Long-term changes due to thermal or capacitive effects in the amplifier)
2️⃣ Internal Amplifier Noise
This type of noise originates inside the amplifier and remains after subtraction:- Thermal Noise (Johnson-Nyquist noise in resistors)
- Shot Noise (Semiconductor noise in transistors)
- Flicker Noise (1/f Noise) (Low-frequency noise in active components)
- Current Noise (Current fluctuations from transistors and op-amps)
- Power Supply Noise (Residual ripple and interference from the power supply)
3️⃣ Electromagnetic and External Interference (EMI/RFI)
If the reference signal does not contain the exact same interference, these components remain after differentiation:- Electromagnetic Interference (EMI) (Coupled interference from external sources)
- Radio Frequency Interference (RFI) (Interference from radio signals, mobile networks, etc.)
4️⃣ Frequency and Phase-Dependent Errors
Since real amplifiers behave differently across frequencies, the following issues remain:- Group Delay Distortion (Different delay times for different frequencies)
- Envelope Delay Distortion (Frequency-selective signal delay)
I am NOT asking for:
- A lecture on whether "negative noise" is a thing. I already explained what I mean.
- Discussions on whether noise is stochastic, deterministic, or generated by fairies in a vacuum.
- Explanations on why feedback-based cancellation exists—yes, I know, and I don’t want to use it.
What I need:
- Can I use capacitive coupling to inject this inverted signal into the high-power output?
- If yes, what capacitor values and voltage ratings should I consider for a system running at 100W into 8Ω?
- If capacitive coupling isn’t ideal, what alternative methods (inductive coupling, resistive networks, transformers) are best suited for this kind of injection?
If you just want to argue terminology or talk about patents without explaining how they achieve it, kindly start a new thread instead of derailing this one.
Thanks in advance to anyone who provides actual technical guidance.
For excellent articles on feedforward error correction, Giovanni Stochino in Linear Audio V8 and V9 (September 2014 and April 2015) first discusses ideal topologies, and then reports results from a 200W power amp that has 5ppm harmonic distortion and 1ppm intermodulation distortion at full power from 20Hz to 20kHz.