Why Zero Feedback

[TomicTime]

We do suffer….. Stop by anytime RV 😉 Watch planes and tip a glass 🙂 Best to you

Jim

[atmasphere]

There are several reasons feedback has a bad rap in high end audio.

The main reason has to do with where the feedback is applied. In most tube amps its the cathode of the input tube. In many solid state amps is the input of a transistor that is part of the differential input of the amp.

In both cases these devices aren't linear. As a result, the feedback signal gets distorted prior to being able to do its job! So of course its going to add distortion as a result.

The distortion added is IMD and higher ordered harmonics. A lot of the innate distortion of the circuit does get suppressed.

If the resulting distortion profile lacks the 2nd and 3rd harmonics in sufficient amount, the higher orders will be audible as they are not masked by the lower orders. This will cause harshness and brightness. We've all heard it!

There are two solutions. One approach is to make the feedback node as linear as possible. Thus the feedback isn't as distorted as it would be otherwise. Here are two examples of that- the feedback used in an opamp, where the feedback is entirely external to the opamp itself, instead being mixed with the incoming audio thru a restive divider network. The other way is to use a really linear tube like a 12AX7 section to receive the feedback. In this way you might not have to use a lot of feedback and the circuit might still sound quite good. The Dynaco ST35 is a good example of that in practice. The ST70 is not.

The other solution is to use a lot of feedback. Like over 35dB; this can overcome the non-linearity of the feedback node. The problem here is that you have several things that can stop you in the circuit. One is that the circuit might not have the gain or bandwidth to support the feedback. If this happens distortion will rise above a certain frequency. Distortion rising with frequency isn't a good thing- its a clue that the amp might sound bright and harsh although the bass might be just fine (since the feedback is supported at the lower frequency).

The second problem you can have with this approach is circuit stability. Because of frequency poles in the circuit, feedback at audio frequencies might be negative, while at some ultrasonic frequency it can become positive due to phase shift associated with the frequency pole(s). Sources of a frequency pole might be a coupling capacitor, a coupling transformer, Miller effects in the active devices (tube or transistor), stray capacitance and the like. If the feedback loop isn't properly designed, the amp could go into oscillation if a signal of high enough frequency can get through the feedback loop and phase shifted enough to be positive feedback.

Class D amplifiers offer a way around both of these problems. They are particularly easy to develop the gain/bandwidth product that is needed to support a constant +35dB or more  of feedback at audio frequencies. If the amp is of the self-oscillating variety, you can add so much feedback that the amp goes into oscillation, but then the oscillation is used as the switching frequency. The feedback network can be designed so that the amp will always find the same switching frequency, killing two birds with one stone.

A large amount of feedback can compensate for phase shift that might otherwise be present if no feedback is used and bandwidth is limited. It also allows the amplifier to reject AC noise problems. So there's a lot of advantage to this if you can pull it off.

Mind you, we've been building zero feedback amplifiers for nearly 50 years so I'm not saying this idly. A zero feedback amplifier will not have the dreaded rising distortion with frequency and if designed properly, will not exhibit harshness and brightness since the higher ordered harmonics are masked. So this can be a very musical amplifier, but also can be load sensitive, so the choice of speaker used can be critical to really getting this kind of amp to show off its stuff.

A very high feedback amp gets you to the same place if designed properly and has the advantage of being usable on a wider range of speaker loads.

[GdnrBob]

Ahh, thanks Ralph.

You explain it so simple and eloquently.

Bob

[Ipspam]

Cool thread. I have been wondering as well,  since I follow the time/phase philosophy for speakers, treatments, dac. But use a Nad M22 with lots of open loop feedback. And also vanishingly low amounts of distortion. They do a lot of things well,  and I especially love the real time impedence correction. Can really tell.it has a good hold on the woofers.

SO I guess the question is, at what point does the THD get so low that the timing errors it's adding are no longer relevant.

OR am I missing something.  Can THD be functionally 0, and there still exist timing errors that are very audible?

I understand it's easy to "hide" timing errors in distortion as a fraction of a ms Timing difference is going to be like 0.0000x% THD. Rather than distortion in the FR range,  which is not relevant much because speakers and room, the distortion is in the timing, which seems to be relevant.

[Holmz]

Like Anton Segur in the movie says, “I wouldn’t worry about it”.

it looks like it is built on the Hypex N-Core, and the Class—D need feedback to work.
If it sounds good, then I would not fret.

But I am not sure how one compares the feedback in Class-D to Class-A or AB.

There are some videos with the designer where he discusses feedback in Class-D.
And that same group is also doing the low distortion motor drivers.

 

11 minutes ago, Ipspam said:

...

SO I guess the question is, at what point does the THD get so low that the timing errors it's adding are no longer relevant.

…

I would think that the multitone test should catch it.
It looks about the same as that from both of my preamps.

https://www.stereophile.com/content/nad-masters-series-m22-power-amplifier-measurements

[Holmz]

On 11/22/2022 at 8:36 AM, Ipspam said:

Cool thread. I have been wondering as well,  since I follow the time/phase philosophy for speakers, treatments, dac. But use a Nad M22 with lots of open loop feedback. And also vanishingly low amounts of distortion. They do a lot of things well,  and I especially love the real time impedence correction. Can really tell.it has a good hold on the woofers.

SO I guess the question is, at what point does the THD get so low that the timing errors it's adding are no longer relevant.

OR am I missing something.  Can THD be functionally 0, and there still exist timing errors that are very audible?

I understand it's easy to "hide" timing errors in distortion as a fraction of a ms Timing difference is going to be like 0.0000x% THD. Rather than distortion in the FR range,  which is not relevant much because speakers and room, the distortion is in the timing, which seems to be relevant.

@Ipspam can you describe how it sounds compared to anything else you’ve run?
I keep thinking about Class-D, for a few reasons including a trigger to turn it on, XLRs, and reducing rack space, etc.
So more inputs are helpful to me to understand what others seem to “feel” about these amps in practice.

And hearing from those that have them, or chose them, pique my interest more. 

[BKDad]

Might be of interest:  https://www.temporalcoherence.nl/cms/images/docs/FeedbackHvM.pdf

Note - this comes from a Dutch (!) company that actually sells products like loudspeakers and electronics, so there's the usual potential (emphasis on potential) conflicts to consider.

And, another:  https://www.temporalcoherence.nl/cms/images/docs/AmplifierIssues.pdf

Edited November 28, 2022 by BKDad

[Steve Edwards]

On 6/22/2022 at 10:54 PM, TomicTime said:

Recently I had a nice 2 year run with a variable NFB tube amp the Music Reference RM-9. It has three is NFB ( aka gain ) settings that w careful level matching allowed for an excellent learning experience and a LOT of discernment about sonic differences between the 3 levels of NFB. Image depth was greatly impacted

Jim, as the now proud and grateful owner of that Music Reference RM-9, I'll have to do some more experimenting with that switch!  In the manual, I think Roger recommends using the HIGH setting for MC cartridges, which is what I have with the Hana MH.  

Play on

Steve

[TomicTime]

Steve ; You are so welcome…. Roger understood feedback and THD ( and very importantly that the components of the # ( order ) are WAY more important than the total. And he understood the tradeoff between THD and TIM. Of course both Vandersteen amplifiers are based on that AND better approach…. Understanding the ear brain and using a reference back to original acoustic events are imo essential…

In your case…experiment away ! 
 

Jim

[atmasphere]

On 11/21/2022 at 5:06 PM, Ipspam said:

SO I guess the question is, at what point does the THD get so low that the timing errors it's adding are no longer relevant.

What is meant by 'timing errors' in this post?

[Bruskie]

@atmasphere

Welcome to the forum Ralph, I am fairly new here too. We talked a few times on the Gon and in PM. I have always welcomed your posts for their humility and hugely informative nature. And I know I speak for others! We are very fortunate to have you over here. Welcome.

Bruce

[GdnrBob]

On 6/21/2022 at 9:19 AM, Richard Vandersteen said:

Feedback is time distortion.

 

On 11/28/2022 at 5:25 PM, atmasphere said:

What is meant by 'timing errors' in this post?

@atmasphere, this is what Mr. V. posted.

Bob

[Ipspam]

On 11/22/2022 at 11:42 PM, Holmz said:

@Ipspam can you describe how it sounds compared to anything else you’ve run?
I keep thinking about Class-D, for a few reasons including a trigger to turn it on, XLRs, and reducing rack space, etc.
So more inputs are helpful to me to understand what others seem to “feel” about these amps in practice.

And hearing from those that have them, or chose them, pique my interest more. 

I don't have a whole lot of experience on quality amps, in a treated room. The majority of my 5 or so Sansuis and 2 threshold amps before this were "underpowered" in the 60-90 watt range,  and before I had seriously worked on my room.

I will try my best.

Class A is sweet and sings? (Threshold S-150 and S-200) Yes. The second order distortion registers to the ear as a reinforcement,  similar to how a room mode builds on itself to stand out. It's a fun sound.. It can also be extremely grating, depends (but usually grating from the room, not the amplifier!) I remember listening to Bob Dylan's Blowin' in the Wind, and near the end there is a note that just reaches up and rings. I thought for sure that Harmonica sounded like that in the studio, because it was such an experience to behold, how could it not have? Some time months later, with room treatments and probably some speaker placement changes I listened again, and I was expecting this sweet note to connect and dazzle my eardrums.... Except it was just a regular Harmonica note. It was a total letdown. The doubling, the shimmering glare, it was gone. That note wasn't doubled up in the studio. It was doubled up in my room.

The moral of the story is that people can like their distortions, and Class D can be totally unlikable for its lack of distortions and personality.

With that said,  there is great potential to connect with the artist and music rather than the sound. Many people love the sound of their "slammin bass" or "sweet highs" or "lush mids". It can be more about the sounds and how they touch ear. Like a drug for the ear that gets it buzzed.

With class D, that experience is not going to come from the amplifier. That leaves the  speakers,  and room to tweak the sound you want. Room sound is generally always horrible. Nobody ever says "this small untreated room is actually pretty nice sounding. I never knew that's what the lyrics say". That leaves the Speaker's to color the sound if one wants a "sweet" or "fun" or "pounding" or whatever sound.

So a funny thing happens with class D and Vandersteens,  I feel anyway. It leaves only the room to color the music, and most all rooms color music poorly.

So do you want to hear your ugly room, or a singing amp? For many, the appropriate answer is a colored amplifier that really sings and sounds beautiful.

... ....or do you want to go with ugly room sound and then work on prettying it up to let true musical beauty shine through? 

That's how I feel about class D. Its pretty boring, but that boringness is offset by being able to truly see and hear into the music and into the mind of the composer and what they intended for the instruments to play, and when they intended it to play. The whole rhythm from bass to treble is accessible in all parts of the soundstage. Subtle delays stand out. Little chorus change cues usually not noticed smash you in the face with their out-of-placeness. But the sound isn't fun. The music has to be the fun.

I guess I would say Class D is a nice tool. But many people are looking for a nice toy. 

I also have a 2016ish designed Class D Nad M22. There is now the NAD M22 v2 released in 2019 and the NAD M23 released in 2022. So I'm interested in those as 2016 class D was still largely in its infancy re Hifi.

On 11/28/2022 at 3:25 PM, atmasphere said:

What is meant by 'timing errors' in this post?

The way I visualize (which is to say I'm guessing!) error correcting feedback is that it is always behind in its correction.  If there is 0.1% distortion and zero jitter, the feedback can correct that 0.1%, but because it's not instantaneous, maybe that timing error measures as 0.0004% distortion, even if the 0.1% is "perfectly"  corrected,  it's now the time lag that is distortion, and should measure as distortion. It's not jitter because it's always behind rather than potentially ahead or behind in timing.

Edited December 9, 2022 by Ipspam
Minor touch up

[BKDad]

I came back to this thread while looking for something else entirely.  Google will do that for/to you.

Anyway, earlier I recommended a book about power amplifiers.  I see that the link I posted (https://linearaudio.net/books/2220) is now in 404-ville.

Here is a current link of where you can purchase Dr. Kolinummi's book:  Audio Power Amplifiers

Edited August 27, 2023 by BKDad