Historical look at amps

I know you aren't trying to argue. But it does get frustrating to us technical types when it seems like we speak, but no one listens. I know........maybe we don't put it in terms the layman can grasp. But all too often the layman's response comes across to us as it is us who is the one that "no capisce".

OK.......let us look at some very elementary building blocks, and you will understand what I mean. I hope.

Almost all SS amps use emitter-follower outputs. Even ones that use what is called a "complementary feedback pair", like Rowland and Threshold/Forte used in the 80s, still have an emitter follower at its core.

The emitter follower, but its very nature, has 100% local, degenerative feedback. Simple as that. Followers are all over the place in a typical amp. Even ones with ICs, use lots of them. A very handy building block. Sometimes it used to amplify current (output stage), sometimes to isolate stages or even just to shift voltage levels. So, any circuit with one in it anywhere has feedback.

Obviously.......it is local, not loop. That is the point that the guy from Ayre was trying to make.

Another example:

Take a single 1 transistor circuit......make an amplifier with it. Well, to do so, you need to stick a resistor in the emitter leg in order to bias it on. Guess what........more local feedback.

You say that you don't want feedback there.......ok......here are your options:

Take the resistor out. Connect the emitter to ground. Great, now you have something that won't bias on in a linear manner. You have what is called a Class C amp. Great for RF, useless for audio.

Bypass the resistor with a capacitor. Fine, except that it won't look like 0 ohms at all (or really any) frequency. So, there will always be a small amount of local feedback.

Come up with some bias scheme that allows the emitter to be at ground wrt AC, but not DC. You need 2 supplies......and what eventually ends up is you take the easy way out and make a differential amplifier. Ok......great.....now you have done away with the local feedback, but to make something functional, you have to apply loop feedback if want it to incorporate it into an amp design. Or apply local degeneration to make it work without overall loop feedback.

Now.....if you really want to get confused.......get a couple of amp designers with very different views and ask them if they prefer voltage feedback or current feedback. One guy will claim that current feedback is a made up term, that what the other guy calls current feedback is really voltage feedback. No, the other will claim that current feedback does exist, and that it is something that is used in conjunction with a certain topology that is characterised by bandwidth that does not change with gain......etc., blah, blah........enough squabbling to drive even me mad.

In that case, I would have to side with the layman and tell both of them to shut up. But I would not suggest that either try to run for political office. If they ran against each other, we would have to find some way to void the election, since one would have to win.

(Amps using "current feedback" have been made. Analog Devices has a paper called the "Alexander Amplifier". You can find it on their site. Rowland made an amp using that scheme.......Model 8(?), maybe. I made a CD player that had a current feedback circuit inside. The dealers hated it, and I almost lost all of them.....a story for some other day.)

But back to the original subject.......historical look at amps. You would need to include "current feedback" types to have a complete perspective. Another subject of discussion could be bandwidth........how much does an amp really need, and does it help? Some amps......Spectral........have tons of bandwidth.......and some will say that they sound bright as a result.

(BTW.....if you do build amps with overall loop feedback, the more bandwidth, the better. It just gets very hard to increase it beyond a certain point. No sure how Spectral does what they do, except the designer is a sharp dude.)

I bring this part up because I have found that Class D amps, with the same bandwidth as a typical SS amp, will sound bright to almost all listeners. You have to lower the bandwidth to make it "sound" the same. Honestly, I am not sure why. On one hand it is interesting, the other frustrating as hell for an amp designer.

The next rung??? Don't know that I would characterise it that way. I would say it is just another branch on the tree, and trust me, I have climbed all over it.

OK.....at this point someone scratches their head and goes "Doesn't this guy make digital amps now? Did he abandon "zero feedback?"

Yes, and no.

Making "digital" as a way to segue into the HT market. "Zero feedback" designs are not going to go far there. At least not for amps. Small, light, efficient, powerful, lots of "oomph"; Class D is the way to go.

As for anything else that we make.......still has "zero feedback" inside. In fact, a lot of "zero feedback" thinking in the auxillary parts of the "digital" amps.

One thing that has been missing so far is the role of transisor evolution in all of this. Most of it is way too technical, but there are a few major developments in transistor design that have led us to where are. And a few side branches, like MOSFETS. Not only do they allow for kilowatt Class D amps, but different processes led to amps like the Acoustat line (and others) which found favor with electrostat owners. The Trans Nova series comes to mind.

Just one comment re. Ar_t's prev post:
"You say that you don't want feedback there.......ok......here are your options:

Take the resistor out. Connect the emitter to ground. Great, now you have something that won't bias on in a linear manner. You have what is called a Class C amp. Great for RF, useless for audio."

even when that resistor in the emitter leg is removed there IS local feedback in that device. It's the intrinsic emitter resistance called "little re" + any package lead resistance. The issue is that it's really very small & of no practical consequence i.e. not large enough to bias the transistor in its linear region.
========================================================

IMHO, there are several reasons that amps s.s. amps (or sand amps as I like to call them) sound better:
* firstly, the transistor devices themselves have gotten radically better than they were 5, 10, 20 years ago. Ar_t did cite this as well. The distortions from these newer devices is much lower than what it used to be so amps using them can play louder for longer. Tighter manuf tolerances also makes it easier to match them. Many manuf like Pass, Rowland, Symphonic Line & a whole host of others use several of them in parallel for high current outputs.
* 2ndly, the s.s. amp designers themselves have grown in their skill set to design these amps. I bet that Ar_t can testify to this! :-) It takes a while & several generations of products to understand how the semiconductor devices *really* works & how to coax the best from it.
You can see this in the realm of CD music too. When was the CD 1st introduced? If anyone of you has a CD, say, from 1985 or so & you compare it to a recent re-issue of the same music, you can tell that the people re-mastering the CD have a vastly better understanding of the whole process.
The growth of the individual amp designers to skillfully use the semiconducting device to its inherent strengths is what I consider a more important reason for better sounding amps. You can have the best BJT or MOS but if the implementation is poor, you'll still get bad sound.
* 3rdly, *more* s.s. amp designers understand & believe within themselves that there is no or very little correlation between THD, TIM measurements & sonic character of the amp. Thus, applying large amounts of negative feedback to ensure that the amp measured superbly in some reviewer's lab is not a top priority any more. There have always been s.s. amp designers thru the yester years that believed in less global feedback & we consumers have voted w/ our money by owning these products.
* 4thly, there is a lot of admission from the s.s. amp design camp that the vacuum tube, tho old & to some unreliable, was & remains a really fantastic amplification device for audio where lack of harmonic & inter-mod distortion is king. You'll often find people seeking a "tubey sounding s.s. amp" - the forums are littered w/ such posts. Why are these people seeking such an amp?
Also, note that some of the best s.s. amps sound like a good tube amp. Anytime I have ever tried to find an adjective to describe a good sounding s.s. amp I've mostly come up with "sounds like a tube amp!". Fancy that!!! I wrote this in my orig post (which Unsound picked up on). The vacuum tube might be a real old fart but it remains the most linear amplification device that we can work w/ practically. Any world class system will have it somewhere in the chain. Real good s.s. amps only approach that quality of sound.
FWIW. IMHO. YMMV.

Thanks Ar_t, I didn't think you meant any disrespect by it. I was really just making sure you weren't thumbing your nose at us with your comment "do you have any idea...I didn't think so." I taught electronic theory at a junior college for 10 years including transistor biasing and circuits so I do understand what you are saying. I also appreciate the effort that went into your last response and getting some insight from someone trying to design a practical, marketable amplifier. It is obvious you have a passion for what you are doing and I wish you luck in your endeavors.

I use tubes in my amps but the theory is pretty much the same. The amp I'm using now has 3 stages; all are common cathode with a bypassed cathode resistor which gives it a lower cutoff frequency of below 10 Hz. To me that is zero feedback, but as you point out there is some small amount and more as the frequency decreases

That brings us back to Hansen's defense of using the term "zero feedback," an idea which was an integral part of the original post. He is correct that there is no textbook definition so he defines it to suit his marketing needs. I don't need my lower cutoff to be any lower and the feedback in the audio band in my opinion is negligible so I describe my amp as having zero feedback. While I think he is taking liberties with the term to the point of being deceptive, a point born out by Keis' belief that Ayre wasn't using any feedback whatsoever, a belief based on Ayre's advertising claims, Hansen could also point his finger at me and accuse me of the same so it is an argument that can't be won. However, I have nothing to gain by using the term and I truly believe that for all practical purposes my amp is zero feedback. On the other hand, it is just as obvious to me that Ayre is twisting the term to capitalize on the current "feedback is bad" frenzy gripping the audiophile community, and relying on the fact that most audiophiles are non-technical and will therefore believe they aren't using any feedback at all.

This is also born out by his reference to the Maxim 4200 data sheets. By selectively quoting from the sheet it gives the impression that his use of the term is an accepted industry practice. If you actually read the sheet and take the phrase "without negative feedback" in context, it is obvious they are only talking about a global feedback loop from output to input. Further reading from the same data sheet:

The MAX4200–MAX4205 include local
feedback around the buffer’s class-AB output stage to
ensure low output impedance and reduce gain sensitivity
to load variations.

This shows that they do employ feedback and that they, unlike others, are not trying to hide the fact.

I don't want to blow this out of proportion. Like Bombaywalla, I have no real axe to grind with Ayre. I admire their products and even went to the unprecedented length (at least for me) to purchase one of their CD players new from a dealer because I could not find a used one. Even paying retail I thought it was a comparative bargain. It is just that Hansen’s defense of his marketing campaign reminds me of Bill Clinton asking for the definition of the word “is” when defending himself in the Monica Lewinsky scandal. I suppose I shouldn't worry about this advertising claim any more than I worry about Miller's claim that their lite beer tastes great.

Yes, the intrinsic emitter resistance does constitute local feedback, but I was trying not to get too technical.

Ok, speaking of emitters and such, the development of "ring emitter" transistors lead to a radical change in transistor design. Back when I started getting serious about amp design, you had 2 choices: RCA and Motorola. RCA (for whatever reason) did not make high-power PNP devices. Motorola did. With those, we had amps like the SWTP "Tiger" series. A bit unstable, but probably the first step towards modern amp design. Even then, the transistors were made with diffused processes, and were not the most rugged in the world. Eventually, they learned how to make epitaxial processes, and things started to take off. Some firms, Bedini as example, stuck to using only NPN devices in the outputs.(The RCA approach.) But most everyone else went to complementary devices. However, out of that grew the 0.000001% THD wars, and the resultant bad sound.

(Looking back......in hindsight......there may not be a convincing reason to use complementary devices in closed loop amps. Remember, the little 20 watt Bedini did sound good.)

I suspect that the guys who came up with the ring-emitter concept were probably used to designing RF transistors. Sanken, Toshiba, and Fujitsu all had strong contenders. Linear, fairly rugged, and perhaps most important: low capacitance. This allowed designers to push the bandwidth higher, as we were all concerned with TIM, SID, and a host of other "new" mechanisms that we were becoming convinced explained why our amps all sounded like doo-doo. Somewhere, things had gone horribly wrong.

To me, the thing that really got my attention was not only the linearity and low capacitance, but the new packing concept. WOW! You can bolt the transistor to the heat sink, on the inside, bend the leads 90 degrees, and hook it right to the PCB! No more drilling hole through the heat sinks, using nasty sockets, steel cases with screws going through them to make electrical contact, etc.

But let me pause and give praise to the guy who may have been the first to "think outside of the box", when it came to using ring-emitters, and in an entirely different manner.

John Iverson.

Not only did he incorporate the new transistors, mounted in a different manner, but he came up with an usual input stage, followed by an even more unusual gain stage. I had not seen anything like it before. He refined the gain stage somewhat in the later versions of the Eagle amps.

OK......what was so great about it?

Some will argue, but transistors are basically current controlled devices. (Yes, you have to create a voltage to have current......not the point here.) If you think as the input/control signal as a current, and design with current linearity, not necessarily voltage linearity, as the parameter to optimise, you come up with ideas that have not been used before.

At least not in audio power amplifiers. I suspect John may have worked on some military/government electronics somewhere in his career. Regardless, guys who thought like him gave us things like folded cascodes, and other techniques that increased both linearity and bandwidth.

The more linear it is to start with, the better it will sound if you use feedback to lower it. Likewise with bandwidth: the higher you can get it, the more stable an amp should be.

So, a lot of factors came along that made it easier to build amps that were inherently more linear than the junk we designed in the 70s. Some of us decided that designing by specs was even more meaningless than the rest of the crowd, and we got rid of all the loop feedback. But none of it would have possible 25-30 years ago. The semiconductors did not exist, we had our head(s) screwed on backwards, and it took some cock-eyed ideas (which may have been invalid!) to get them oriented back in the right direction.

Actually......now that I think about it......Audio Research was on the leading edge in SS design with the notoriously unreliable D-100. It used a "zero-feedback" output stage....what was it...mid 70s?......long before anyone else thought of that concept. (The problem was mostly a heat sink issue. The amp could have been reliable with about 2, maybe 4, times the heat sink surface area.) The input stage may have been bad.........I don't know, the modules were potted, but the output stage concept was a good one. I know..........I have used it the last 10+ years. With decent transistors on much larger heat sinks.

OK.....that ought to be enough to digest for a while. I appreciate the encouraging e-mails. Thanks.