Negative Feedback a deal killer?

Dob,

Please refrain from making bold statement like that. Please don't take it the wrong way but Spectron can not drive the 1 ohm Scintillas though tough load 0.8ohm from 20hz to about 2khz and then hovering around 1.5 ohm up 20khz. That is a tough load but still far cry from 0.1ohm. The Krell master reference is probably the only amplifier which can handle such load cause now we're so close to a dead short. Below is the real testimony that Spectron cannot drive 1ohm load, namely the Apogee Scintillas. Again, I meant no offense. Peace

http://www.apogeeacoustics.com/forum/viewtopic.php?f=5&t=41

H2O, welcome. A friendly note: the expected courtesy here is for those that are in the "business" regardless of how small an enterprise it may be, proclaim that each time they post. This is especially true if they're listed as a "private user".

If the amplifier has other design considerations, it will not need to have feedback to be low distortion. For example, a fully-differential amplifier will have even-order cancellation at each stage in the amp, so it will, without feedback, generate only the 3rd harmonic.

Just an example.

Now, the flip side of the coin is what happens to the amp if a feedback is used. The problem is that with any amplifier, there is a time delay for the signal to move from input to output. This propagation delay causes the feedback signal to arrive slightly behind the actual signal at the input. Now with sine waves, the amplifier can reduce its distortion after succeeding iterations of the waveform. It does not do so with non-repetitive waveforms, like real music. It has also been shown that feedback, due to the time delay, actually **increases** certain distortions, namely the 5th 7th and 9th harmonics.

Now the increase is slight, but there is a problem: the human ear uses these harmonics to determine how loud a sound is. So if you mess with these harmonics, the electronics will have an artificial loudness about them, a tonality, which audiophiles describe as bright, harsh, brittle, etc.

In a nutshell adding feedback decreases certain distortions that the ear does not care about a lot (we hear them as 'warmth', 'bloom', etc.) while **increasing** the distortions that the ear cares about a lot.

Some designers pay attention to this and others don't. The subject has been controversial for over 50 years. For more information see
http://www.atma-sphere.com/papers/paradigm_paper2.html

Finally, Chaos Theory has something to say about this, confirmed by Norman Crowhurst several decades before Chaos Theory became generally accepted: The addition of negative feedback to an amplifier results in a chaotic behaviour with both stable and unstable states. Additional harmonics and also inharmonic information is added (Chaos Theory calls distortion 'bifurcation'). This results in a noise floor quite unlike normal hiss- a noise floor that the human ear cannot hear into (whereas we can hear 20 db into hiss) that effectively masks information below the noise floor.

For those interested, Wikipedia has a nice primer on Chaos Theory: http://en.wikipedia.org/wiki/Chaos_theory

Perhaps I've brought this up before, if I have please forgive me. As I understand it great deals of negative feedback are often used in Class D amplifiers. One poster has already commented that since it is done at much greater speeds, the lag is inconsequential. I've wondered with digital designs such as the Tact, if might be possible to use some sort of, for lack of a better description, "digital fly wheel" to synchronize the feed back with the original signal. If such were the case then feedback might used to it's advantage with out the disadvantages often attributed to it, with the final signal funneled out. Any thoughts?

The faster the amplifier, the less this is a problem! However, faster is not the same as 'real time'. I would say that this is in the realm of 'progress' where we have an improvement over something that was done 15 or 30 years ago.

I think you can see though that if the input and feedback were synchronized, that the delay of the circuit is going to have to go somewhere. You might need a substantial memory for that. The faster you do the scan, the more memory you are going to need...