Negative Feedback in Transmitters

[KA8WTK]

  I had a couple of minutes today to do a little reading on the net about negative feedback. Some of the articles mentioned deriving the negative feedback voltage by sampling the modulated RF and rectifying it. The rectified voltage would be fed back out of phase to the first audio stage. This was mentioned as a way to modulate at a "low level" in the RF chain and still use a class C amplifier stage for output. (I have some misgivings about that.) I only saw one schematic that used this type of feedback, and it was a low power tube transmitter (Part 15) for the AM broadcast band. It used a couple of diodes off the tank circuit to get voltage directly back to the grid of the audio amp tube.
  Most of the dicussion here has been about negative feedback within the audio chain. Has anyone tried this sort of negative feedback coming directly from the sampled RF? How did it work or why wouldn't it? Seen a schematic on how to do it with a "legal limit" rig?
  Caught my curiosity...........Bill KA8WTK

[WA1GFZ]

sounds like a lot of phase shift and hard to do on a broadband transmitter. It is a bigger loop

[Steve - WB3HUZ]

Some of the old WE grid modulated rigs used this type of feedback.

I don't see how using rectified RF feedback to control the audio stage gain would be any better than the standard setup for audio negative feedback.

Pretty simple really since it includes/reduced the distortion generated in the RF final modulation process and straight audio feedback does not.

[N3DRB The Derb]

didnt collins do this in the s lline? lookit shizoid-matic.

[w3jn]

Collins did RF NFB in their big HF SSB amplifiers; so did TMC.  Both the Collins and TMC SSB handbooks detail RF negative feedback over multiple stages using capacitive dividers.  They did not, however, feed back to the audio stages.

[AB2EZ]

Bill
et. al.

The use of negative feedback from the output of a system to the input of a system, to reduce non-linear distortion, was conceived by Harold Black. Legend has it that the concept come to him one evening in the late 1920's, when he was riding the Staten Island ferry back home from his research laboratory at Western Electric in NYC (which later became famous as Bell Laboratories). He was trying to reduce (small amounts of) non-linearities in vacuum tube amplifiers, so that many amplified sections of copper-pair cable could be placed in series... to produce long distance telephone links with acceptable levels of end-to-end distortion.

http://www.invent.org/Hall_Of_Fame/16.html

http://en.wikipedia.org/wiki/Harold_Stephen_Black


Negative feedback works best when the non-linearities that are to be reduced are small to start out with (e.g., the distortion products are more than 20 dB down). It doesn't work very well if the system it is being applied to exhibits a severe limiting or clipping effect, or some other form of severe non-linear distortion. Examples of severe non-linear distortion in transmitters used in AM applications include: a) the severe distortion produced by a modulation transformer when its core saturates (i.e., when driven by a large-amplitude, low frequency audio input signal), and b) the bottoming out (turning off) of the rf output of a fixed-biased Class C amplifier* if the input rf level is reduced (modulated) below its proper value.

*If the Class C amplifier is biased by the average grid current flowing through a grid resistor to ground (which is typical), and if the time constant of this self-biasing circuit is short enough to follow the modulation (which it usually is), then the rf input-to-rf output characteristic would not be as bad (as non-linear) as it would be in the case of fixed bias. However, it would still be very non-linear! Also, on negative-going (as a result of amplitude modulation) excursions of the envelope of its input rf signal, the amplifier wouldn't be operating as a Class C amplifier.

Negative feedback will result in oscillation problems (instabilities) if there is too much phase shift in the frequency response (within the band of frequencies being linearized) from the input of the system to the output of the system, and back (through the feedback path) to the input (the phase shift "around the loop"). More specifically, the gain around the loop must be less than unity for frequencies beyond which the phase shift around the loop is 180 degrees or more**. This is usually accomplished by shaping the gain vs frequency of the feedback path, so that it rolls off (without, itself, introducing too much phase shift) before the phase shift of the system being linearized gets too large. Accomplishing this can be tricky... and the ultrasonic audio oscillation of the audio chain of some Johnson Rangers illustrates what happens if the gain around the loop doesn't drop below unity before the phase shift around the loop reaches 180 degrees.

**When the phase shift around the loop equals 180 degrees, then one has positive feedback instead of negative feedback... which results in oscillation.

Using feedback to improve the linearity of a screen-modulated Class C transmitter could be done from the input of the audio chain to the detected envelope of the rf output... but it would probably be much easier (fewer "devilish details" to deal with) to use it between the point where the audio signal is being applied to the screen of the modulated tube, and the detected envelope of the rf output. This would avoid the phase shifts in the audio chain, and possibly other problems (e.g., rf getting into the audio chain) as well. Separately, feedback can be used (and typically is used) within the audio chain, to linearize various gain blocks within the audio chain.

As indicated above, feedback cannot be successfully used to turn a fixed-biased Class C amplifier into a linear amplifier, unless you use a very small modulation index (relatively small excursions of the input rf envelope level, above and below a large carrier). This is because (as mentioned above), feedback only works well to reduce small amounts of non-linearities. A fixed-biased class C amplifier has an extremely non-linear relationship between its input rf amplitude (i.e., the rectified or peak-detected envelope of the input rf signal) and its output rf amplitude (i.e., the rectified or peak-detected envelope of the output rf signal).

Also as mentioned above, a self-biased Class C amplifier might be made to work more like a linear amplifier by employing negative feedback from output to input, based on a comparison the detected envelope of the output rf signal with the detected envelope of the input rf signal... but, in the end, this might be harder to do, and the actual operation of the "linearized" amplifier less efficient (increased plate dissipation for a given amount of carrier) than using either a linear amplifier (operating in Class AB₂) or a plate-modulated Class C transmitter.

Best regards
Stu

[WA1GFZ]

Remember he said rectified and back into audio not RF feedback out of phase as used TMC.

[AB2EZ]

Frank

You said:

"Remember he said rectified and back into audio not RF feedback out of phase as used TMC"

That is what I was assuming in my comments (above)...

I clarified this by editing the my comments to emphasize that the terminology "amplitude" of an rf signal , as I used it, was intended to have the same meaning as what some might refer to as the "rectified envelope" or the "detected envelope" of that rf signal.

Stu

[Steve - WB3HUZ]

If it's negative feedback, it will SUBTRACT distortion, not add it.

[Steve - WB3HUZ]

Quite true. It was never used much in amateur circles because the adjustment of the feedback loop is very frequency sensitive. This wouldn't do for the frequency agile amateur operator. Further, distortion requirements were far less important or stringent for the amateur than for the broadcaster or commercial operator. So, good audio and RF design, with simple audio negative feedback included in rare cases was "good enough." How many commercially produced AM rigs of the past included any negative feedback? Answer: very few.

The telephone aspect of the invention of RF negative feedback would have been in relation to sideband transmission in the old sideband analog carrier systems and Stu points out in his post that it applies more to linear type amplifiers than class C amplifiers.

'HUZ says "If it's negative feedback, it will SUBTRACT distortion, not add it" but my reading in the old handbooks says it's to cancel the "distortion products created in the RF amplifier plate circuit" and shows applying the RF negative feedback to the cathode or grid of the RF amplifier itself to cancel these distortions.

I don't find any mention of using this rectified RF as negative audio feedback in any handbooks.

In very high powered class C RF amplifiers this might indeed be a worthwhile thing to pursue in the final amplifier stage only. But for the average amateur it would seem to be far too complicated to incorporate correctly.

The whole concept seems more relevant to linear amplifiers, grid/screen modulated class C RF amplifiers than to the amateur plate modulated class C RF amplifier.

Mack  

[Bacon, WA3WDR]

The problem with detected-RF negative feedback in typical AM transmitters is that it can not help distortion caused by overmodulation, and in fact it goes bonkers in that situation, producing large overdrive trying to cause negative modulation beyond 100% that can't happen.  So for such systems, overmodulation must be carefully avoided.

[KA8WTK]

These are all very educational comments.
So you all know what prompted the original post, here is an excert from Wikipedia taken from:
  http://en.wikipedia.org/wiki/Amplitude_modulation

"An approach which marries the advantages of low-level modulation with the efficiency of a Class C power amplifier chain is to arrange a feedback system to compensate for the substantial distortion of the AM envelope. A simple detector at the transmitter output (which can be little more than a loosely coupled diode) recovers the audio signal, and this is used as negative feedback to the audio modulator stage. The overall chain then acts as a linear amplifier as far as the actual modulation is concerned, though the RF amplifier itself still retains the Class C efficiency. This approach is widely used in practical medium power transmitters, such as AM radiotelephones."

And a link to the low power AM transmitter information:  http://www.dogstar.dantimax.dk/tubestuf/amtx-2.htm
This is where diodes are used to rectify the RF and send the resultant voltage back to the audio inout.

Bill KA8WTK

[Bacon, WA3WDR]

Terman noted that the linearity of the high-efficiency AM linear and high-efficiency grid-modulated systems left something to be desired, and he recommended RF-demodulated negative feedback to improve their linearity.

But class C as such does not result in high efficiency.  Proper loading, resulting in low voltage during the high current part of the RF cycle, does that.  And that requires a proper impedance load.  And with a fixed B+, that impedance is different for different output signal levels.  Hence high efficiency AM linear or grid modulated amplifiers have multiple level sections - one for carrier level and below, and another for carrier level and above.  These either have different B+, or different loading, or possibly both.  The Doherty amplifier uses the impedance-inverting property of a 1/4 wave line to allow the carrier tube to assist the peak tube with the positive peak, which is a plus.   This trick actually increases the loading on the carrier tube on positive modulation peaks.  If I recall correctly, the Terman-Woodyard high-efficiency grid modulated amplifier also uses a 1/4 wave line, and also has this advantage.  The simpler Taylor system cuts off the carrier tube on positive peaks, and the peak tube has to carry the entire positive peak.

With fixed B+ and a fixed load, even a class C amplifier is only efficient near peak output.  At 1/2 output voltage, the plate only pulls down to about 1/2 of B+, and dissipation is relatively high for the amount of current that is drawn.  But at full output, the heaviest current is drawn at relatively low plate voltage, so the dissipation is relatively low for the amount of current drawn.  Dissipation is a little lower with class C than with class B, but the main source of dissipation is the high plate voltage at the instant of maximum plate current.

Of course there is less current drawn at the lower output level, so the dissipation is not as great at 1/2 output voltage as it might have been.  So in fact, dissipation peaks somewhere between 1/2 and full output voltage.  But efficiency is approximately proportional to the output level.  This is why the high-efficiency linear and grid modulated systems supply the carrier from a section that is running at full output into its load.

[KI4YAN]

Dogstar is simply hosting Fred's sites. Fred has been dead for some time, and those who knew him took a big hit when it happened. his transmitters were set up to provide Hi-Fi quality for entertainment, not for communications, so the goals of the LPAM transmitter circuits were different from ours, but that does not mean that we can't apply his ideas. Besides, The 4W exciter circuit works pretty damn good, i listen to it daily at work. makes for great Foxhunt transmitters, good quality signal, but not a lot of power, allowing for lower band foxhunts!

[WU2D]

The article in this months Electric Radio on the National Modulator includes a very interesting schematic. This modulator uses negative feed back but it is only used to rectify the negative going waveform of the audio. Therefore it can be considered as a negative peak compressor.

Mike WU2D

[KA8WTK]

Ah, finally found a reference to this feedback scheme in an Editors and Engineers "West Coast" handbook.
In the 11th edition on page 130 "rectified carrier inverse feedback" is discussed.
"Degenerative feedback may be used satisfactorily from the rectified carrier back to the audio system in transmitters using Heising plate modulation, and supressor or control-grid modulatio." It also notes that this is especially suited to grid bias modulation.

The reason this got me so curious is that it might be useful today in order to build high power AM transmitters, using tubes, without the expense of a huge modulation transformer (and associated scarcity, weight, size, etc.). When this idea came about, modulation iron was easy to come by and so the technique may not have gained favor. It does not surprise me that ARRL handbooks do not mention this idea.

The up-shot is, if this can be implemented properly, you might be able to build (for example) a low level AM transmitter that would drive an un-modulated Class C amplifier and have an undistorted signal. Or at least be able to build a high power grid-bias or control-grid transmitter with cleaner modulation.

 

[Bacon, WA3WDR]

Anything that improves linearity is good.  Negative feedback is relatively simple to implement, and today it is simple to make a highly linear detector.

[AB2EZ]

Bill

You referenced and quoted from:

"Ah, finally found a reference to this feedback scheme in an Editors and Engineers "West Coast" handbook.
In the 11th edition on page 130 "rectified carrier inverse feedback" is discussed.
"Degenerative feedback may be used satisfactorily from the rectified carrier back to the audio system in transmitters using Heising plate modulation, and supressor or control-grid modulation." It also notes that this is especially suited to grid bias modulation."

There are several things about this quote that don't make sense to me:

a) the reference to "suppressor grid modulation" ()

[If he meant to say "screen grid modulation", then I agree that the proposed feedback scheme could be used to reduce residual non-linearities in a screen modulated transmitter]

b) the reference to "grid bias modulation" ()

[Is he saying that you would attempt to apply a feedback signal (derived from the difference between the audio modulating signal and the detected envelope of the rf output) to the grid bias of the output rf tube... presumably to compensate for residual non-linearities of a linear amplifier whose input rf signal is already modulated?]

c) the reference to "control grid modulation"

[I'm guessing that this refers to a linear amplifier... in which the amplitude of the rf signal applied to the grid is modulated. If so, I agree that the proposed scheme could be used to reduce residual non-linearities associated with a practical linear amplifier]

Again, I believe that the reference should be viewed in the context of reducing residual non-linearities in a modulated transmitter... not in the context of making an extremely non-linear transmitter behave in a linear fashion.

Stu

[KA8WTK]

Just to clarify, the book reads "supressor OR grid modulation".

Here are the schematics in the book. Notice that the rectfied output is going to the audio chain (speech amp, mic transformer, mic preamplifier, etc.)

[Ian VK3KRI]

In increasing order of Obscurity...
 Terman devotes 2 pages to this in Radio Engineering 3rd Ed. He states the primary use is reducing noise , distortion and hum, allowing the use of  less filtering for B+ .  He mentions the issues of phase shift due to sideband triminng of Hi Q circuits reducting the amout of feedback befor the Bode requirements fail to be met.

In 'Radio Engineering Vol 1' by Sandeman (BBC Engineering 1947) There is an example of the design of a 'Grid Modulated Doherty Amplifier' (Sect 6.22)   This explains the use of an Envelope detector on the output feeding back to the grid of the first audio stage to reduce distortion

Radio-1  (Australian Postmaster General's Training Document 1952) It also has a small section on RF feedback and shows a block diagram of the 10KW Standard Metropolitan Transmitter used by the ABC (local version of BBC) , Using 2 x 5KW linear amps in parrallel, providing envelope detected feedback to the cathode of the first triode in the audio cct.

I had assumed that envelope detected feedback was  standard on Broadcast transmitters. Or was it just a BBC/ABC  thing where Govt. engineering departments were able to dictate specs to manafacturing compaines with  price  no object, as opposed to the US where private companies would have that had to pay their own way would have chased the cheapest and easiest to operate transmitter?

                                                 
                                                                                     Ian VK3KRI
                                                       

[Steve - WB3HUZ]

as opposed to the US where private companies would have that had to pay their own way would have chased the cheapest and easiest to operate transmitter?

Not true. All broadcast transmitter (no matter where they were produced) had to meet FCC (gov't) specs.

[Ian VK3KRI]

as opposed to the US where private companies would have that had to pay their own way would have chased the cheapest and easiest to operate transmitter?

Not true. All broadcast transmitter (no matter where they were produced) had to meet FCC (gov't) specs.

No, thats not quite what I meant. In the state run radio / telecommunications area, equipment was quite often overspeced because the organisation was to a large degree run by engineering sections, and specs or designs for equipment  would be produced that were quite often a wish list of nice to have, or someones pet design ideas. 

On the other hand , running a commercial operation, you're more likely to factor costs into the decision of what  to buy.
                                                                                            Ian VK3KRI

[K1DEU]

About 20 years ago I detected my Viking Valiant's Rf out put at the antenna connector(rf choke to ground) and fed it back to a audio stage following the mike gain pot. It worked extremely well until one over modulated as Bacon, WA3WDR pointed out.
  Its much better to sample either the modulation primary or secondary. When we look into an open or distort going positively quite a bit of correction is available with a closed loop 20 Db system. As the modulation xfmr usually has low leakage reactance I sample the primary.

One of my articles  http://hamelectronics.com/k1deu/pages/ham/transmitters/am/pages/negative_feedback_design.htm

  73, John,K1DEU