In my recent amplifier search, I've noticed that designers are declining to use as much negative feedback. Some manufacturers like BAT and Ayre Acoustics have always been against global negative feedback. Dartzeel is another famous example that tries to avoid negative feedback as much as possible. Nelson Pass penned a white paper on the subject in 2008 that is one of the best I've read on the subject:
https://passlabs.com/articles/audio-distortion-and-feedback
Martin Colloms said Dan D'agostino is down to using 5dbs of negative feedback in his latest Momentum amps- correlation with alleged better sound quality?
The last few amplifiers I've tried (Ayre VX-5 currently in my home on loan) have all sounded better without it- highs are more natural and the upper midrange has better clarity. Sure, there are plenty of other design choices that impact sound---but is the simpler circuit without negative feedback one of the bigger ones? This also goes hand in hand with some that say amps can sound better but "measure worse"- as THD is one of the goals of loop negative feedback.
Digital amps often use loads of negative feedback as well, so this could be one reason that audiophiles have moved away from this topology over the years since their highly anticipated arrival. I owned a pair of Rowland 201s for all of one week. Sure the bass was great, but the rest was awful.
Are audiophiles learning the perils of negative feedback loops?
KeithR
No. They are learning though that some designers don't know how to design a feedback loop properly. There's a difference.
Negative feedback is not completely understood for the audio. I like to use it a lot in my designs. There is a boundary when it is too much and sound subjectively starts to degrade. What is causing it I don't know. Whatever it is, we cannot measure it with methods that we are currenty using. For that reason we (designers) need to listen in combination with measurements to build best posible device.
Feedback is well-known in the rest of the industry as 'control theory'. Its also very well understood. We now have all the tools we need to measure what we need to, although I don't think that was true 35-40 years ago. IME/IMO most audiophiles are living their lives and making decisions based on measurement techniques from the 1980s and not the 21st century.
That article pre-supposes that lack of feedback is related normally to valve amplifiers and juiced second order harmonics. Every amp I've listed above is solid state and doesn't have frequency response issues. In fact, MikeL says the Dart has the most extended, natural top end he's ever heard.
This statement is false. You might want to re-read Bruno's article. It focuses much more on the idea of having sufficient Gain Bandwidth Product to allow the feedback applied to be supported at
all audio frequencies. The fact that
most amps made do not do this is why they also have rising distortion with frequency. The only way to get around that problem is to either have enough GBP
or don't use any feedback at all: that allows the distortion to be constant with frequency.
Still waiting for that to happen...
Feedback is tricky for a number of reasons!
1) if the amp lacks gain bandwidth product, the feedback will decrease at a certain frequency, causing distortion to rise with frequency. IME this contributes to brightness and harshness. GBP works sort of like this: if the GBP is 10MHz and you have a gain of 1, the bandwidth is 10MHz. If a gain of 10 in the circuit then bandwidth of 1 MHz, above which is a 6dB slope.
2) often there are frequency poles in the amplifier design that cause phase shift, such that at a certain frequency the feedback becomes positive rather than negative due to the phase shift introduced by the frequency poles (this is filter theory 101). This aspect of any amplifier is known as 'phase margin'; if you exceed the phase margin of the design with an inappropriate feedback loop it will be prone to oscillation.
3) The point in the amp where the feedback is applied is often non-linear, causing the feedback to be distorted before it can do its job. Traditionally, to get around this more feedback is needed but you can get into TIM distortion since part of the feedback node is uncontrolled by the feedback.
4) because of 3), there is a sort of bell curve where the feedback adds so much higher ordered harmonics and IM that the result is though the amp overall has lower distortion, what distortion it has is the kind to which the ear is the most sensitive because the feedback has added it through its faulty application. Since harmonics are responsible for the sound of all instruments its no surprise that adding higher orders thus generated results in brightness. If you run a little feedback this isn't so bad and if you run enough its not so bad either. In between is where you get into trouble, in particular if the feedback node isn't linear (which is true with most amps made that use feedback). This is between 12 and 30dB.
Regarding 3) above, if your feedback node is more linear then less feedback is needed to make things work and you can wind up with a more musical amplifier. I like to point to the Dynaco ST70 and ST35 as an example. The latter uses a 12AX7 section's cathode as its feedback node while the former uses a pentode. Simply due to the greater linearity of the 12AX7 section, the feedback applied to the ST35 is less distorted and so does less damage doing its job; this is one of the reasons its a more musical amplifier.
The most linear way to apply feedback is to not use any active devices!
An example of this is the feedback used around an opamp. Opamps in theory have infinite gain, which is corralled and the gain set by how much feedback is used. In practice although opamps don't have infinite gain, they do have quite a lot (+120dB) and so the gain they have is effectively set by the ratio between the input resistor and the feedback resistor, which are passive devices. If they are the same value, the opamp has 100% feedback and is said to have 'unity gain'.
If you understand all that I wrote so far then this will make sense:
A number of class D amplifiers use this technique opamps use to apply feedback. Self oscillating class D amps also use so much feedback that the phase margin is exceeded and they go into oscillation as soon as you turn them on. The feedback network is designed to allow the amp to find exactly one frequency for the oscillation, which is used as the switching frequency. In this way a very large amount of feedback can be applied and in a very linear way so it can do the most good.
Its not that far of a stretch to see that any amplifier design could be done to use feedback in this manner and thus allow the designer to get around some of the limitations otherwise imposed by limited Gain Bandwidth Product and other considerations since the feedback would do less damage on its way to reducing distortion.
The reason we avoided using feedback in our tube amps was the simple fact that if we added the amount needed, it was nearly impossible to avoid issues with phase margin and certainly there was not enough Gain Bandwidth Product to support the amount needed.
Class D came along though and presented a very easy way to develop large amounts of GBP. So with it came the possibility of larger amounts of feedback too, without the usual increasing distortion with frequency. In our class D amp, the distortion is constant with frequency just as it is in our tube amps.
