Home » Audio » Thermionic Emissions » Feedback
Feedback [message #8204] Wed, 02 June 2004 13:35 Go to next message
Wayne Parham is currently offline  Wayne Parham
Messages: 18792
Registered: January 2001
Illuminati (33rd Degree)
I see feedback brought up fairly frequently in tube amp discussions, but not as often in regards to solid state amps. Usually, when I see feedback discussed for a solid state amp, it appears to be used to sort of capture the attention of tube enthusiasts. But in either case, feedback is usually mentioned on a subjective level and rarely in the context of circuit design and the "whys and what fors." So I'd like to see a good technical discussion about feedback here.

Very simple circuits with minimal components sometimes have no feedback. The simplest circuits are limited to an active component and practically nothing else. Usually they include biasing and coupling components, but not always. Such simple circuits have fixed gain of hfe or mu and input and output impedance equals the impedance of the active device inputs and outputs.

Some biasing components will introduce negative feedback when signal current flows through them causing a voltage drop that biases the active device in a negative direction. This is actually very common. A bypass component has to be added to the circuit to reduce negative feedback. When components are placed around the active device, it chances the circuit's gain, frequency response and input and output impedance. These are usually required and their effects are usually desirable.

Negative feedback can be employed on a single stage, as is the case when an emitter or cathode resistor is used without a bypass capacitor. It can also be employed globally, by running an output signal back to the input in reduced amplitude and opposite phase. This is generally done on more complex amplifiers.

Negative feedback reduces distortion, but it does so at the expense of reduced gain. That increases the number of gain stages required. Positive feedback increases gain, but it does so at the expense of increased noise, distortion and reduced circuit stability. Excessive positive feedback will cause an amplifier to oscillate. But it was a common technique used in early radios to get enough gain; Global positive feedback was used in old radio circuits called "regenerative" receivers. Hetrodyne and Superhetrodyne circuits made them obsolete.

There are pros and cons of each kind of amplifier circuit configuration. I'd like to hear from some of the tube gurus on this one, because I think the topic of feedback is one that's not well understood.

Re: Feedback [message #8206 is a reply to message #8204] Wed, 02 June 2004 22:44 Go to previous messageGo to next message
metasonix is currently offline  metasonix
Messages: 103
Registered: May 2009
Viscount
>Such simple circuits have fixed gain of hfe or mu

Not in typical RC-coupled circuits. An active plate or collector load or else a coupling transformer or choke is required to get the full device gain as voltage gain.

>Some biasing components will introduce negative feedback when
>signal current flows through them causing a voltage drop that
>biases the active device in a negative direction

Sometimes called degenerative feedback.

>Global positive feedback was used in old radio circuits
>called "regenerative" receivers. Hetrodyne and Superhetrodyne
>circuits made them obsolete.

In most radio receivers, yes.
But simpler TRF and regen circuits are still seen today, esp. in specialized applications.

Feedback is a well-understood mechanism. If used properly, it is an extremely useful tool. But no tool is perfect. (That's what decent engineering textbooks usually say.)

Feedback, not carefully implemented, can cause phase margin errors at the frequency-response extremes. It can cause the amp to be slightly unstable with some load impedances (esp. certain costly monkey-coffin speakers with complex crossovers).

I once saw a Kenwood receiver, brand new out of the box, whose power amp oscillated at ~ 4 MHz--regardless of the load placed on it or the source material.
It made a pretty good shortwave jammer, when hooked to a speaker with a long cable.
They sold tens of thousands of that model line circa 1990.
As a special bonus, each channel oscillated at a slightly different frequency--giving some truly unpleasant IM distortion.
Most customers were perfectly happy with them....they were quite unreliable, though, for an obvious reason.
So one hopes that line, probably designed by a junior engineer at the factory (typical Japanese practice), is in the landfill. But don't think that'll be the last one!

The advantage of tube amplifiers is primarily in two areas:
1) low distortion of many triodes (and even some pentodes) allows the design of low feedback circuits;
2) tubes do not (usually) have slew-rate issues, and they never have voltage variable device capacitances, which semiconductors often have--in abundance, esp. in cheap devices intended for audio use.

And tubes still find a home in audio equipment because:
3) feedback has been (and still is) misused in solid-state amps.

The problem does not lie in the evil of feedback; the problem is inside the heads of arrogant, self-important design engineers who successfully managed to campaign in the 1960s and 70s that high-feedback transistor amps gave perfect sound. They often did not, and even today mid-fi amplifiers and professional sound reinforcement amps often have horrible feedback implementations.

A few names come to mind. Names like Bob Pease....Don Lancaster....

If an amp doesn't sound good, I recommend finding the little dweeb who designed it, and kicking his ass. No, wait, better yet, find the bastard who hired him, and kick THAT GUY'S ass.

I would be so happy if more audiophiles actually removed their swollen little heads from their buttholes, and started boycotting inferior products. But since most can't seem to agree on what good sound is, I'm not gonna wait around for it.
Prefer to talk to the minority who actually listens carefully.
(sorry, that probably leaves out most of the SE-triode mafia--they often listen to amps that sound worse then cheap table radios...)


Re: Feedback [message #8207 is a reply to message #8206] Thu, 03 June 2004 01:37 Go to previous messageGo to next message
Wayne Parham is currently offline  Wayne Parham
Messages: 18792
Registered: January 2001
Illuminati (33rd Degree)
What amplifier topologies and feedback configurations do you like and which ones do you not like?

I often build amps that switch current to a load connected directly in series with the output device. That's a simple and effective configuration, one that is often used in transformer coupled tube stages. Another one I like is to use an unbypassed emitter (cathode) resistor, and sometimes use the voltage swing across it as the in-phase output. It doesn't give the full gain of the device going that way, but it is a simple configuration that is stable and works well. I can't think of any devices that don't work well in these modes, although they are only appropriate in certain situations.

What are some of your topological favorites? What are some you steer clear of? And what are those horrible feedback implementations you mentioned? I don't necessarily care to know names, but descriptions of ill-configured circuits would be interesting. Particularly if compared with others that were done right.

That may be a bit much for a messageboard post, but it would certainly make a good article. Or maybe a teaser with excerpts on the forum, followed by meat and potatoes in a printed article at a later date. I think everyone would like to know the why's and wherefor's so they'd know what to watch out for in future purchases, kits and homebrew projects.

Here's an essay on negative feedback that seemed good [message #8208 is a reply to message #8204] Thu, 03 June 2004 08:11 Go to previous messageGo to next message
akhilesh is currently offline  akhilesh
Messages: 1275
Registered: May 2009
Illuminati (3rd Degree)
http://www.hometheaterhifi.com/volume_5_4/essaynegativefeedbackoctober98.html
thanx
-akhilesh
And an even more indepth article [message #8209 is a reply to message #8208] Thu, 03 June 2004 08:19 Go to previous messageGo to next message
akhilesh is currently offline  akhilesh
Messages: 1275
Registered: May 2009
Illuminati (3rd Degree)
http://www.normankoren.com/Audio/FeedbackFidelity.html
Re: And an even more indepth article [message #8210 is a reply to message #8209] Thu, 03 June 2004 12:18 Go to previous messageGo to next message
Wayne Parham is currently offline  Wayne Parham
Messages: 18792
Registered: January 2001
Illuminati (33rd Degree)
Great articles, thanks. I had previously read Norman Koren's articles, but forgot about this one. He has done a lot of tube amp modeling with Spice, and I came across his articles while modeling Steve Bench's 6DJ8-based active crossover to include response shaping components for compression horns. Norman is very studious and I appreciate your bringing attention to his articles here.

I believe these articles address the main issues of bandwidth, gain, distortion and stability. The advantrages of NFB include potentially improved frequency response, reduced harmonic distortion, better gain control, increased input impedance, and decreased output impedance. Those are some of the things that can be gained from a good circuit that incorporates negative feedback.

The disadvantages of NFB are that it can cause peaking and other response anomalies bandwidth extremes, sometimes even enough to enter oscillation. It can also increase susceptibility to RF interference and make clipping more abrubt. These disadvantages are usually limited to global feedback implementations, and aren't as likely to occur when feedback is introduced locally, in a gain-stage. But even here, if a device is pushed to its limits, instability can result.

So I guess the thing is that each circuit should be taken on a case-by-case basis. There are likely some components that are more susceptible that others to having problems in certain configurations. Each device has its own set of strengths and weaknesses, and the challenge is to find those configurations that work best. It really isn't appropriate to view negative feedback, in and of itself, as a problem. What is best is to examine each circuit to find and address potential liabilities, and to get the most performance afforded by the components used.

Re: And an even more indepth article [message #8211 is a reply to message #8210] Thu, 03 June 2004 18:05 Go to previous messageGo to next message
Manualblock is currently offline  Manualblock
Messages: 4973
Registered: May 2009
Illuminati (13th Degree)
As a untutored amatuer may I offer this Quote for comment," when conventional transistor amplifiers enter undefined operating regions the amp clips and feedback loses it's grip on the amp. Conventional designs use feedback to compensate for gain transitions and absorb back EMF from the drivers. Speaker drivers are always creating back-EMF as a result of numerous resonances, many with very high Qs and long time durations. When a conventional amp clips, the source impedance seen by the driver suddenly changes, and the back-EMF is suddenly reflected back to the driver instead of being dynamically absorbed by feedback correction. So although a conventional feedback amp may look benign clipping into a resistive load, things get a lot uglier with real world speakers, and no speaker driver is going to be happy with a power source that has sudden transitions in the driving impedance." Any help with this one?

Re: And an even more indepth article [message #8212 is a reply to message #8211] Fri, 04 June 2004 00:31 Go to previous messageGo to next message
Wayne Parham is currently offline  Wayne Parham
Messages: 18792
Registered: January 2001
Illuminati (33rd Degree)

That's a pretty big deal. The issue of back EMF and load impedance is significant.

I think the best way to get a handle on this is to start from a simple perspective and visualize the situation, progressing to more complex attributes in steps.

1. Take a look at a simple voltage divider using two resistors. Use a DC voltage as the source. You'll notice that there is a fixed division of voltage between the resistors.

2. Now use the same voltage divider, but use an AC signal to drive them. You'll see it also has a fixed division of the signal between resistors.

3. Substitute the load resistor with a purely reactive component. Try a capacitor first, then an inductor. You'll find that the capacitor has the highest amount of voltage across it at low frequencies and the lowest at high frequencies. Current through the circuit is highest at high frequencies and lowest at low frequencies. The inductor acts just the opposite.

4. Now make the load a complex impedance having resistance, inductance and capacitance. Try several configurations for the load network. Make some having mostly series inductance and resistance, like a voice coil (sans diaphragm w/mechanical resonance). Make some circuits that are tuned circuits like a radio tuner, notch filter or loudspeaker port. What you'll find now is that the voltage across the load changes in a predictable fashion depending on the frequency.

5. In each of the preceding steps, the source voltage and resistance were held constant. The source voltage frequency was changed, but the amplitude was fixed to a reference value. The source impedance - the series resistance of the voltage divider - was also held constant. But what if we no make this variable too? That's what happens when amplifier output characteristics change.

A good amplifier acts as a constant voltage source with very low output impedance. But the output impedance is not zero, so it becomes the source impedance. The output circuit, including the speaker wires, is the source impedance and forms a voltage divider with the loudspeaker as the load. So the higher the output impedance, the more fluctuation in signal results from a changing load impedance. And when the amplifier itself fluctuates impedance because of instability or whatever, that adds another level of complexity to the voltage division between these two things, the output source and the load.
Re: And an even more indepth article [message #8217 is a reply to message #8212] Fri, 04 June 2004 14:07 Go to previous messageGo to next message
Manualblock is currently offline  Manualblock
Messages: 4973
Registered: May 2009
Illuminati (13th Degree)
Thanks again for the in-depth response; I hope you know these posts are very helpful. In response to the feedback query, if I understand you correctly instability is the factor that most affects the amps ability to reproduce an accurate tracing of the waveform. What would be the requirement for the most stable amp not using global feedback?

Re: And an even more indepth article [message #8220 is a reply to message #8217] Sat, 05 June 2004 09:23 Go to previous messageGo to previous message
abajaj11 is currently offline  abajaj11
Messages: 14
Registered: May 2009
Chancellor
HI Manualblock,
My solution is: ensitive speakers that don;t take the amp over 3/4 of the way up...i.e., no clipping. Apprently extends tube life as well.
-akhilesh

Previous Topic: Ламповый Завод "Светлана"
Next Topic: Ultralinear
Goto Forum:
  


Current Time: Tue Dec 03 23:43:30 CST 2024

Sponsoring Organizations

DIY Audio Projects
DIY Audio Projects
OddWatt Audio
OddWatt Audio
Pi Speakers
Pi Speakers
Prosound Shootout
Prosound Shootout
Miller Audio
Miller Audio
Tubes For Amps
TubesForAmps.com

Lone Star Audiofest