Electronics 101: Use of Negative Feedback; a Q

I've taken an interest in the past year or so of trying to understand the exact origins of this idea . . . that is, that Global Negative Feedback is a Bad Thing. I think that there are three opening points here that can be accepted as fact:
- The ubiquitous negative view of global NFB in high-end audio is not found in other analog electronic engineering disciplines
- There were many decades between the first patents of the electronic feedback amplifier, and when it started to achieve a pariah status in high-end audio
- It is extremely rare to find a discussion of global NFB in an audio forum or publication where it isn't linked with the promotion of some accompianying viewpoint regarding components or topology . . . i.e. a proxy for a tubes-vs-transistors discussion.

From what I have found, this idea of Global NFB being Bad really came into being in the very early 1970s, in Audio Engineering Society papers written by mainly by Matti Otala, sometimes in association with John Curl. This is the origin of the TIM acronym, meaning Transient Intermodulation Distortion. These days, TIM is more accurately described as slew-limiting distortion, and has been very, very throughly investigated and discussed within AES, including standardized measurement methods.

Matti Otala and others did put forth the viewpoint that global NFB was at least part of the root cause of slew-induced distortion, and in his papers proposed some topologies with greatly reduced open-loop gain (hence less global NFB) as a solution. But while the association of TIM and global NFB somehow remained, Otala's suggested circuit topologies remain obscure. The reason for this is that while slew-induced distortion is indeed affected by circuit choices that also affect the amount of global NFB . . . slew-induced distortion isn't CAUSED by feedback.

The other source of this Bad Feedback idea seems to be a article written by Peter Baxandall for Wireless World, from December 1978. This article contains a graph where Baxandall plots the amplitude of the second through the sixth distortion harmonics (Y-axis) as a function of the feedback ratio (X-axis). Here, at a glance that increasing the feedback up to about 20dB causes an increase in the fourth, fifth, and sixth harmonics while it reduces the amplitude of the second. Continuing to increase the amount of feedback further then reduces all the harmonics in a linear fashion.

This is the ONLY well-documented scholarly paper I have found from which I can imagine comes this idea of global NFB decreasing the amount of lower harmonics, at the cost of increasing the energy in the upper harmonics. But this article is just one in a series of six on power-amplifier design, and if you actually READ them and not just glance at the pretty graphs, it's obvious that this is NOT Baxandall's conclusion from this experiment.

Rather, Peter Baxandall unambiguously states his conclusion from this experiment is that global NFB needs to be INCREASED to reduce all the harmonics to a negligible level, rather than avoided so that the second harmonic can completely dominate the distortion signature. He also specifically emphasizes that "even FETs used without feedback generate high-order harmonics - and therefore . . . high-order intermodulation products".

It's rather unfortunate that in discussions of negative feedback on the internet, the seminal papers from which most of these ideas stem remain largely ignored. This is of course because they're copyrighted, and one simply cannot pop up a link to them. But if anybody's truly interested, the Otala papers (and responses to them) can be obtained from the Audio Engineering Society http://www.aes.org, and the Baxandall series has been re-printed by Jan Didden at http://www.linearaudio.net

it would appear that electrical matching concerns between an amp and speakers having fluctuating impedance stats as a function of frequency may be mitigated in whole or part by using NF.

Here's what I think is the most precise response this question:
- Due to a varying impedance-vs-frequency characteristic, the response of a loudspeaker will be affected by the output impedance of the amplifier, thus . . .
- If the amplifier's output impedance differs significantly from that which the speaker designer used for evaluation, then the response of the speaker will be different from what the designer intended, and . . .
- Although there are no specific standards, *most* loudspeakers are designed with low-output-impedance amplifiers in mind, and . . .
- The use of global or local Negative Feedback is the (almost universally applied) method of acheiving a low output impedance in an amplifier.

Thank you Kirkus for your cogent and thoughtful response. Hopefully, we will read more posts that address the very narrow issue about using NF to compensate for "varying impedance-vs-frequency characteristic[s]" of speakers.

You also said that "[i]f the amplifier's output impedance differs significantly from that which the speaker designer used for evaluation, then the response of the speaker will be different from what the designer intended."

I assume from a lay person's perspective, in plain English, that means if a speaker was voiced to be driven by a "Voltage Paradigm" amplifer (as described in Ralph Karsten's White Paper, a SS amp), then using a "Power Paradigm" amp (usually a tube amp) to drive the speakers may affect the sonic presentation. This assumes of course, that the tube amp has a relatively high putput impedance if little or no NF is used.

Therefore, the amount of acoustic deviation between the actual versus intended presentation presumably will be affected by how flat (or "tube friendly") the speaker's impedance curve is, and how much NF is being using to reduce the output impedance of the amp, be it SS or tube. I surmise that if a tube amp uses NF to reduce its output impedance, it will behave somewhat like a "Voltage Paradigm" amp. In other words, the acoustic deviation between what the designer intended and actual performance may be mitigated (i.e., reduced). Is this correct?

So Kirkus, the follow up to my Q was how do so called zero NF amps manage to compensated for changing speaker impedances?? Is the bottom line objective simply to reduce the amp's output impedance by whatever means is possible, i.e., through NF or otherwise?

I believe that Ayre SS amps do not use NF in the circuit design, yet have relatively high DFs, thus suggested low output impedances. I don't know how Ayre achieves these results without NF, but they do. Is that all that is needed, low output impedances and the amp will be a champ??

Thanks again.

BIF

Quick follow-up to my last post. As I recall, some time back, one or more of our EE audio tech members commented that a DF of more than 20 DF doesn't add much acoustic benefit. I think a DF of 20 correlates to an output impedance of .4 ohms.

In late 2011 or sometime in 2012, ARC released a new Referance tube amp, the Ref 150. This baby has a 1000 joule power supply, which I assume is a lot. In addition, the ARCDB web site says the Ref 150 uses 14 db of NF and has a DF of 17, the later stat getting close to the DF factor of 20 mentioned above.

Synthesizing all of this at a high level, what I understand is that lowering output impedance much beyond .4 ohms may yield little benefit. However, at .4 ohms or lower, the amp (be it tube or SS) will behave like a Voltage Paradigm amp and be better able to compensate for the changing impedances values of a speaker in a way the designer intended. Hence, using the Ref 150 as an example, can one expect that it might do a fair job handling a speaker that was designed to be driven by a tube amp, even if the speaker does not have tube friendly impedance curves and negative phase angles??

I suspect I'm probably mixing and matching concepts here, but hopefully you can untangle what I just wrote in this post.

Thanks again.

I believe that Ayre SS amps do not use NF in the circuit design, (...)

I find that hard to believe, Ayre is a very serious constructor... Maybe what they are saying is there is no global FB, only local?

Is that all that is needed, low output impedances and the amp will be a champ??

Usually, a lowish impedance does yield good results if/ when the speakers are designed with that in mind. Many speakers are...
If not (take the Lowther based designs for example), then output impedance in the ohm range (rather than milli-ohm) would be better!

I believe that Ayre SS amps do not use NF in the circuit design, yet have relatively high DFs, thus suggested low output impedances. I don't know how Ayre achieves these results without NF, but they do.

No. They use feedback, even if they say they don't, or come up with a reason why their feedback is "different". This is akin to all the "LED Televisions" on the market - virtually all of them are in fact LCD televisions with LEDs used as a backlight. But they needed a new acronym to differentiate them from the previous models, and marketing departments aren't exactly known for their terminological precision. How many times have you seen "solid-state" on the outside of a television who's principal component was a picture tube? "Zero-feedback" is a similar label.

In reality, virtually every audio power amplifier without an output transformer (tube or transistor) uses some sort of unity-gain voltage follower circuit as the final stage to lower its output impedance. This type of circuit has 100% local negative feedback, hence the unity gain.

I assume from a lay person's perspective, in plain English, that means if a speaker was voiced to be driven by a "Voltage Paradigm" amplifer (as described in Ralph Karsten's White Paper, a SS amp), then using a "Power Paradigm" amp (usually a tube amp) to drive the speakers may affect the sonic presentation. This assumes of course, that the tube amp has a relatively high putput impedance if little or no NF is used.

Ralph really seems to favor a binary viewpoint of the subject with these two clearly-defined camps of amplifiers and speakers . . . but it's indeed correct that certain some loudspeakers produce far more variation in response with a high output impedance than others. This is an important subject to him because his circuit design preferences place practical limits on how low his amplifiers' output impedances can be.

As for the subject of how low an amplifier's output impedance *needs* to be, I actually think John Atkinson approaches this subject in a rational and practical manner when he measures an amplifier's response into an IHF speaker load, and the corresponding impedance/phase plots for the loudspeakers he measures. It's also possible that one may prefer a loudspeaker's response to be somewhat different that its designer, and that a higher-output-impedance amplifier may yield a pleasing result.

Personally, I can't say I've found such a modification to a speaker's frequency response to be a pleasing one, except for a few very rare occasions. But your mileage may vary . . .

Thanks again Kirkus. Admittedly, I am not familiar with Atkinson's test involving an IHF speaker load. Care to explain what that test is all about?

So, going back to the use of NF to reduce an amp's output impedance, in the case of the ARC Ref 150 which uses 14 db of NF and probably has an output impedance between .4 and 1 ohm (probably closer to .4 ohm because its "rated" and reported DF is 17), can one assume that the Ref 150 will behave "SS-like" if presented with a speaker load that was voiced to be driven by a high current SS amp?

Isn't that the bottom line consideration here? Short of shlepping a 75 pound tube amp to a dealer to properly audition speakers, it's helpful to know (or at least be able to reasonably predict) if a tube amp can "switch-hit" and function in a "SS-like" way if presented with a speaker load which was voiced to be driven by a low-impedance, high DF, high-current SS amp. If the foregoing proposition is technically accurate, is it also fair to consider a tube amp's rated output impedance and correlatively its DF, when matching it to a speaker load? Any back of the hand guidelines?

Also, is output impedance less critical in the higher frequencies? Many 3-way speakers have an impedance bump at the midrange-tweeter cross-over point. I believe Ralph's White predicts that a "pure" Power Paradigm amp will deliver more power (watts) than a "pure" Voltage Paradigm amp at higher impedance levels. If so, it may matter.

Thanks again.

P.S. check the Ayre web site re the zero NF point. And I do have a very high-level understanding of the difference between local versus global feedback.

Admittedly, I am not familiar with Atkinson's test involving an IHF speaker load. Care to explain what that test is all about?

Start with http://www.stereophile.com/content/audio-research-reference-150-power-amplifier-measurements and then read the loudspeaker measurement section of some loudspeakers. Please don't take this as my endorsement of Stereophile's reviewers in general . . . but specifically Atkinson's measurement sidebars tend to be very consistent an well-reasoned, and he explains himself well.

in the case of the ARC Ref 150 which uses 14 db of NF and probably has an output impedance between .4 and 1 ohm (probably closer to .4 ohm because its "rated" and reported DF is 17), can one assume that the Ref 150 will behave "SS-like" if presented with a speaker load that was voiced to be driven by a high current SS amp?

In the above link, you'll see that Atkinson actually measured the output impedance as 1 - 1.4 ohms from the 8-ohm output tap, corresponding to a Damping Factor of 8 - 5.7 respectively. This amp actually has a somewhat higher output impedance than most of the high-quality, correctly-operating vintage tube amps that I've measured . . . most of which actually have lower output impedances than many modern "low-feedback" solid-state designs. I think with the ARC Ref 150, the interaction between the amplifier's output impedance and most loudspeakers' impedance curves will be a significant part of its sonic signature.

Short of shlepping a 75 pound tube amp to a dealer to properly audition speakers, it's helpful to know (or at least be able to reasonably predict) if a tube amp can "switch-hit" and function in a "SS-like" way if presented with a speaker load which was voiced to be driven by a low-impedance, high DF, high-current SS amp.

The loudspeaker/amplifier equation is complex, and schlepping an amp to the dealer (or demoing the speakers in your home) is absolutely the best way to be sure you'll be happy with it. After all, can you really study automotive specifications and get an idea whether or not you'll like a car without at least putting forth the effort to take a serious test drive?

If you really don't want to pair up your amplifier with speakers you're considering, and you want to be happy with the way they work together . . . then I feel the most logical solution is to purchase the amplifier, loudspeakers, and loudspeaker cables all together, or purchase active speakers.

Fascinating!

I've copied an excerpt from Mr. Atkinson's review of the Ref 150:

"As expected, the Ref 150's output impedance varied according to the transformer tap selected. The 16 ohm tap measured 1.4 ohms at low and middle frequencies, rising to 1.9 ohms at the top of the audioband. The figures for the 8 ohm tap were 1 and 1.4 ohms; for the 4 ohm tap, they were 0.55 and 0.87 ohm. All three taps offer quite a low source impedance for a transformer-coupled design; as a result, the modulation of the amplifier's frequency response, due to the Ohm's Law action between that impedance and that of our standard simulated loudspeaker, was relatively mild. From the 8 ohm tap (fig.1, gray trace), it was ±0.8dB; the 4 ohm tap offered ±0.4dB, the 16 ohm tap ±1dB. Fig.1 indicates that the Ref150 has a wide bandwidth, particularly into loads higher than the nominal tap value, which correlates with a well-defined 10kHz squarewave (fig.2)."

I assume based on the results of these tests that the Ref 150 performed like a Voltage Paradigm amp to a large degree. That is, even though the simulated speaker's source impedance varied with frequency, the amp was able to maintain relatively level amounts of gain notwithstanding. I further assume that gain correlates to power (watts).

In other words, if I have this right, the Ref 150's circuit topology, using NF, was able to appropriately compensate for the changing impedances of the simulated speaker, which change as a function of FR.

Bottom line: the use of judicious amounts of NF in a tube amp enable the amp to perform, to some extent like a tube amp. Hence, the proposition that some speakers are "tube friendly" versus "not tube friendly" should be accepted with circumspection. It may very well be that the tube amp's use of negative feedback is a compensating factor, to some degree, when driving speakers that may otherwise be considered tube unfriendly.

Do I have it??

If I do, I think folks need to take Electronics 101 when mathing speakers and amps. It ain't so simple. ;>')

WOW fokls It's much more simplier than you're describing:

--No need to get into one or another's review details
--No need to know EE101
--You can be just a trivial consumer

and the rule is

NEVER JUDGE OR VALUE THE AMP BY PRESENSE, ABSENSE, LARGE OR SMALL VALUE OF NEGATIVE FEEDBACK.

For consumer these are only marketting term values.
For engineers it's simply a value to keep an amplification stage stable to prevent self-oscillations that can damage your speakers.

Marakanetz, based on what littleI do get, I think Kirkus is spot on. I doubt the outcome would be diffent though. I think both would advise that in order to make the best decision, you gotta listen to the gear.

However, as others have also written, auditioning gear can be challenging because there aren't many hi-end B&M retail stores around. And even if you're close to a good hi-end outlet, it may not have all the gear you're intersting in.

So, what are you left with?? Go with your gut, research as much as you can, read the Forum to see if others have been where you are going??? And, as I mentioned above, sometimes it's just trial and error. I've done a lot of that. So far I've been lucky because I haven't lost too much on the flips.

Marakanetz, I think that negative feedback can make amplifier unstable since feedback can easily become positive at higher frequencies, while "no feedback" should be always stable.

I agree with Kirkus that there is always some form of feedback. It should be used in moderation. By definition it improves all specifications making it tempting to use as a fix for design shortcomings. For instance, TIM can be completely eliminated by limiting amplifier's bandwidth at the input to one that amp had before feedback was applied, but it requires decent bandwidth to start with and that calls for good design and components.
Even with this bandwidth limitation negative feedback lowers distortions - both IMD and THD while also reducing output impedance. Again, I wouldn't go crazy with amount of feedback since distortions, especially THD, might not be that audible while output impedance is not that important. We should take into consideration that inductor in series with the woofer is most likely >0.1 ohm (limiting DF to 80 no matter what) not to mention that speaker/transducer impedance itself is mostly resistive. Some speakers work well with higher output impedance amps while others (like B&W 802) expect amplifier to have low output impedance - to be voltage source doubling current to half impedance. Driving complex load is very complex (as name suggest)and I would follow Kirkus advice to get speaker, amp and try them together. Judging by spects or descriptions can yield poor results IMHO.