Negative Feedback in Audio Amplifiers and AM Modulators.
A necessary and “often included” part of published feedback circuits in good design is low level, corner compensation to eliminate low frequency motor boating and high frequency ultrasonic oscillation. All audio transformers will lose amplitude response and start shifting phase above and below some mid band frequencies. At the low frequencies this is a function of the equivalent load impedance and the shunt inductance of the transformer, and at the high frequencies it is a function of the primary winding capacitance, the leakage inductance between primary and secondary windings, and secondary winding capacitance and load. At the loss of response at the low end generally a maximum of 90 degrees of leading phase shift will occur. But at the high end the transformer leakage inductance and winding capacitances can act like a low-pass filter and can result in total phase shift of 270 degrees lagging.
Likewise any resistance-coupled stages included in a feedback amplifier will lose response at the lower frequencies due to the coupling capacitor and the grid return of the next stage. It will also produce a leading phase shift, being 45 degrees at the 3 db down frequency i.e., ( |X| = R) and 5.7degrees at 10 times this frequency, and 84.3 degrees at one tenth this frequency. The resistance-coupled stage can lose response at high frequencies due to its own output capacitance, stray wiring capacitance, and the equivalent input capacitance of the next stage, including the Miller effect of the next stage's grid –plate capacitance and produce a phase shift as above but lagging. ( Pentode resistance-coupled amplifiers can cause another amplitude and phase change if the screen is not adequately by-passed, but this is easily remedied with a larger by-pass capacitor. ) Likewise for cathode bias by-passing; a phase shift of more than 90 degrees when the loop gain in a feedback amplifier decreases to unity, can be dangerous. Phase shifts of 180 degrees or more will turn the amplifier into an oscillator near the frequency in question, and phase shifts nearing but less than 180 degrees will turn the amplifier into a “Q” multiplier having a sharp resonant rise. So stay away from 180 degrees of phase shift when the loop gain is one. That is the design goal for both low frequencies and high frequencies simultaneously.
Pros for negative feedback
1. Major compensation for phase shifts in driver, audio output, and modulation transformers.
2. When the modulator sees momentary open loads, due to carrier cutoff, the loop compensates because of the lowered output impedance, reducing off channel shot noise.
3. Push Pull cross over distortion greatly reduced or eliminated.
4. No Power Loss. If the amplifier had to be grossly over driven the output would be much less distorted.
5. Much lower Quiescent (idling) plate Current. For example 6550's AB1 total of 45 ma. instead of 100 ma.
Cons for negative feedback
1. We will lose 20 Db. Of voltage gain, but not lose one Db. Of power.
2. A few extra parts and study and design time required.
3. Extra time and patience.
Picturing how a loop works. Some of the audio output, at mid-band is fed back opposing the signal at the input to a stage included in the loop, ( not back to the low level pre-amplifier and gain control, where the distortion is low. ) If the Nyquist stability criterion is satisfied, and if there are no phase shifts , cross over modulator distortion, or distortion of any nature to correct, all that will happen is that the total gain , at mid-band will simply drop 20 Db. Next, with the feed back disabled i.e. open circuited, lets picture a perfect sine wave, audio tone inserted into the input at mid-band, 1KHz. In many cases with good design this sine wave can pass freely along unaltered, without any phase shift, to the output. Next, imagine a very low frequency of 50 cycles inserted and passing through. When a 50 cycle sine wave passes through a single tube and transformer stage it will likely look like a top or bottom clipped wave due to the elliptic load line presented to the tube by the inductance of the transformer primary, not like the original wave at all and shifted in phase by 90 degrees. Needless to say, at the output, this would look and sound terrible. A clipped wave will result again as we go up in frequency above mid-band, possibly as low as 9KHz, due to the imperfect transformers.
Although the loop is active with no time delays, Let's pretend we stop time to consider how the loop functions. In a way the loop is like some one sitting there simultaneously looking for any output shape that doesn't exactly follow the input wave shape. Say; the output wave starts to rise more quickly than the input wave. With the Negative Feedback the waves greater rise from the output is fed back to the input 180 degrees out of phase immediately forcing the input wave to be reduced to the original input shape. As long as the shape isn't altered more than 20 Db. the loop can keep the output shape exactly like the original input shape. This is how loops constantly correct (compensate) the complex signals passing through amplifiers being misshapen by transformers, etc.
With lots of feedback, when transformers and other factors shift the feedback loop too much we will have an oscillator. If the loop gain is falling , with a change in frequency, but is still greater than one and the loop phase shift is 180 degrees, then the loop no longer provides negative feedback, but becomes regenerative and the amplifier can certainly oscillate. Therefore we purposely roll off the gain within the amplifier loop, at the low and high frequency ends, without allowing the loop phase shift to become 180 degrees before the loop gain falls below unity. Frequency compensating networks must be designed and added before closing the loop, for loop stability. These networks must lie within the loop and are most easily placed at the lowest signal level area.
When adding negative feedback loop we loose 20 Db. of gain. Therefore if we barely have enough overall gain to start with, new gain will have to be added. Sometimes this new gain is simply an added Mike Preamp or added gain by increasing the value of plate resistors in existing RC coupled stages. Mic or channel gain external adjustments should not fall within the loop. The loop should encompass all the distortion, phase shift areas. Such as transformers, modulated screen grids, etc. Sometimes if the leakage inductance of a output or modulation transformer is low enough the feedback take off point can be on either side of the transformer.
A note pertaining only to amateur radio AM modulators;
A necessary but often missing part of published amateur transmitter negative feedback circuits is low level, corner compensation to eliminate low frequency motor boating and high frequency ultrasonic oscillation. All audio transformers are broad band tuned circuits and will start shifting phase above and below some mid band frequencies. More phase shift than 90 degrees is very dangerous, as In oscillator design, on purpose, we use negative feedback with 180 degrees of phase shift . Unfortunately many times someone placing a newly modified transmitter on the air will be unknowingly, transmitting squirrelly, unstable sidebands 25 to 55 KHz. above and below their main carrier frequency. This inaudible on channel ultrasonic audio oscillation will raise their modulator idling ( quiescent ) current causing distorted over bias compensation and can overheat and damage the modulator tubes and transformer, besides QRM'ing adjacent QSO's while being transparent to the operator and their friends on their QSO channel. Often after the ultrasonic is discovered the workaround cure is to reduce the amount of feedback and or place a shunt capacitor anywhere to stop it. After implementing negative feedback, transmit into a dummy load, and listen with a receiver which isn't being overloaded and tune above and below your main carrier listening for any squirrelly sidebands up to 100 KHz. above and below your carrier frequency. Low frequency motor boating is always apparent because its on channel and audible to the person you are transmitting to. Proper initial design makes all of this unnecessary.
Here is a page from the Radiotron Designers Handbook editor; F. Langford-Smith, fourth edition 1952 Chapter 7 page 66, 7.59 A+ B <<< here >>> 116 Kb showing construction of simple networks to reduce the within loop gain to make a stable audio amplifier without motorboat or ultrasonic oscillations. In the 50's many of us wannabee's called each other "Tron's" as we aspired to know everything in this "Audio Bible".
Here is a Quick Reference Chart for Cornering, Reactance and Resistance in Parallel <<< here >>> . Print as landscape, 153 Kb. We are often determining the Minimum size required for a capacitor or the minimum primary inductance of our driver or modulation transformer or audio decoupling choke. As we are using 20 Db. of mid band feedback, which will overcome many phase shifts at the extreme lower and higher end, we simply locate twice the frequency ( 140 cycles ) on the chart, of the lowest audio frequency ( I use 70 Cycles ) we wish to pass for our frequency bandwidth. This 140 Cycles would be our 6 Db. low corner design frequency. For my voice I use 6 Db. corners of 140 cycles Low and 4.5 KHz. High, which after closing the negative feedback loop affords more than a total flat bandwidth of 40 through 8 KHz. Essay by W1ECO & K1DEU
Measuring Audio Phase Shift in Audio Amps or Modulators with a Oscilloscope and audio tone generator by W1ECO
I strongly recommend adding The ( so called ) 3 diode negative peak limiter by Steve, WA1QIX for Plate modulated Transmitters.
An example of Negative Feedback cleaning up the distortion in a Transmitter without transformers but feeding the ever varying complex load presented by an Screen Modulated RF final. <<< Heath Kit DX-60 Screen Modulation >>>
An Astatic D-104 CrystalFET Mike Pre-Amp with Bass and Treble boost by W1ECO & K1DEU drawn by K1DEU <<< here >>> . This pre amp uses a very unusual bass boost circuit. This circuit also works also well with condenser, electroFET and dynamic elements. I mostly use Panasonic ElectroFET Condenser elements. Prices vary between $3 to $18 for the larger diaphragm elements. The 9.7 mm close talk Panasonic WM-55D103 element is pretty good
Back to Ham Index Page Back to Home page
07/30/2009 by K1DEU Back to Home page