Tubes Asylum Questions about tubes and gear that glows. FAQ |
|
In Reply to: NFB Question posted by Will C on September 18, 2004 at 11:07:49:
If your amp has a relatively high output impedance or the circuit is inherently non-linear (i.e. distorting) then you MUST USE NFB to correct these faults.
The concept of Hafler et al in the 50's was that we can use the crappiest components but build a lot of gain into the circuit, then add a hefty amount of NFB and fix all these problems with no sweat.
The problem with these assumptions is the simplistic feedback theory they rely on. Over the years we gained some knowledge regarding the bahavior of feedback under a large loop delay. While some readers may feel lost at this point, the idea is quite easy to understand. If you return signal from the output and subtract it from the input, you essentialy negate the added products, i.e you reduce distortion. If the delay of the returned signal is a significant portion of the input waveform (like 0.1 mSec for a 1 KHz signal), the subtraction process is not that "clean". We can show mathematically that some NEW PRODUCTS appear at the output, not directly related to the input!
That scenario is your sonic nightmare. The ear can relate 2 KHz and 3 KHz harmonics to the fundamental signal at 1 KHz, but it would have no "explanation" to an ARBITRARY PRODUCT like 1.1 KHz!
From this little explanation one can derive a few scenarios here feedback might work well. For example, in an amp with a very small delay from input to output.
It would not be a good thing to have global feedback in a complex amp with more than say 3 stages made of inherently slow devices.
Let's look now on the arguments from a purist point of view with some analogy. If you want to build a wheel for your race car, you can take a standard rim from a production car and examine it. If it's not balanced, you add weights in strategic points. Have you seen serious race car rims balanced that way? Probably not.
This method is good enough for a road car. The racing mechanic may opt to machine the rim for a significantly better balance tolerances.
This solution is the IDEAL SOLUTION for a street car as well, but the practical and COST EFFECTIVE solution is adding weights!
If you are in the business of selling mass production amps, you'll go with a cost-effective method that Hafler et al have chosen and apply feedback to a reasonably cheap implementation.
If you want a BETTER RESULT you will start from a more naturaly "cleaner" design and tweak it AS LITTLE AS POSSIBLE. Just common sense.
A good amplifier would use circuits and components that are inherently not distorting. There is no perfection in the real world, so you might add SOME LEVEL OF FEEDBACK to that design. You use feedback as the last resort, not as a magic fix (which it's not!)
That's the theory.
When it comes to an actual amplifier, you should first determine how good it is without feedback. If it's all triode (inherently linear) and inherently maintains a stable operating point, then you don't need much feedback. In fact, you would prefer not to add feedback, so you don't introduce delay-induced distortion products!
If your amp has a high-gain pentode gain stage and tetrodes in the output stage, you don't really have a choice but to use LOTS of feedback... OUCH...
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- It all depends on your starting point - serus 22:25:31 09/18/04 (2)
- Very good post ... Heard the Hurricanes? - KIS 08:06:08 09/19/04 (1)
- Excellent observations! - serus 08:29:01 09/19/04 (0)