[audiojerry]

So, as you apply less feedback, are you biasing towards class 'a' operation? Does zero feedback = class 'a', or are these two different ducks?

[DVV]

So, as you apply less feedback, are you biasing towards class 'a' operation? Does zero feedback = class 'a', or are these two different ducks?

No, class A and feedback have nothing to do with each other. You can have a pure class A amp with much feedback, and a class AB amp with very little feedback.

In real life, there is no such thing as a pure class B amp, all actually work in what is known as class AB. True, they may pass just a few miliamps of current and thus be more in class B than in class A, but that will make them more efficient and they will run cooler. The price is crossover distortion, when the imaginary sine wave has to pass through the zero signal region, when the NPN transistors stops working and the PNP transistor takes over. If the passing is not perfect, there will be  atime lag, and this we call crossover distortion.

In class A, where each device is fully open all the time, meaning it consumes as much allotted current as its ever going to, signal or no signal, there is no crossover distortion, but the amp runs less efficiently (i.e. less actual available power because the devices are very hot), it gets very hot and requires tremendous heat sinks, and you need lots of power devices for very modest power outputs.

However, this has nothing to do with how much feedback you use, although in pure class A, once you get the desired characteristics, you usually need less feedback to iron any problems out because you have less problems by default. On the other hand, with some decent design and savvy, you can make a class AB amp run just as well.

Personally, I opt for what is called enriched class AB or high bias amps. These are simply class AB amps which instead of biasing the output stage at the usual 20...50 miliamps per device, use higher bias values, like 120...150 miliamps. In my experience, there is very little to be gained sonically with more current, but there is much to be lost in terms of space and costs.

Since I typically use four pairs of transistors per channel, at 150 miliamps each my output stage is actually working in pure class A up to about 6W/8 ohms, meaning that in real world home listening terms, I am in pure class A about 98% of the time.  The amp gets reasurringly warm, as my output stage is dissipating just over 30W, but the sound is much improved over the usual design and little, if anything, behind a full class A design.

Of course, I do pull a few additional tricks too ...

When is somebody going to ask about memory distortion?

Cheers,
DVV

[Dan Banquer]

I would like to reply to memory distortion, but I can't seem to remember anything about that at the moment. Seriously; If I have a little time tomorrow morning I will address that.

[Ferdi]

DVV, can you explain about memory distortion?

(No really, I am curious...)

[DVV]

Thanks,

The last post helps explain a lot of things in this thread I didn't quite understand at first.

No problem Josh, but don't wait for it, ask the moment you aren't sure if you got it right. Let's do this the right way. I mean, we're all complaining about sales literature and its quasi-tech cliches and gabble, so let's show we can do it better. That gives us the moral right to bitch about them.

You know, the shame isn't in saying you don't know, the shame is in knowing you don't know and not asking. Hey, I wasn't born with what i know, I had to polish my butt to learn it all, and the more I learn, the more I see how much more I have yet to learn.

Let me quote an Indian philosopher here:

He who knows, and knows that he knows, he is a wise man - befriend him;
He who knows, and knows not that he knows, he is asleep - wake him;
He who knows not, and knows he knows not, he is a child - teach him;
He who knows not, and knows not he knows not, is a fool - shun him.


Hare krishna,
DVV

[DVV]

DVV, can you explain about memory distortion?

(No really, I am curious...)

Yo Fredi, all,

It's a fairly complex topic, and one I am not done studying yet, which is why it fascinates me so, but here's a much simplified explanation.

Let's take as an example a differential pair, such as are used for power amp input stages. One transistor is non-inverting (let's call it +) and the other is inverting (we call it -, minus). Now, the signal arrives at the base of the + trannie and is amplified. The feedback signal arrives at the base of the - trannie, and is also amplified. The stage is set so that the + trannie amplifies say 10 times and the - trannie amplifies say 5 times. But since the - trannie is in opposing phase of the + trannie, the overall gain of the whole stage will be the gain of the + minus the gain of the - trannie, in our example (10-5) 5.

And it will be so under constant operating conditions, which means that both trannies should be as similar (matched) as possible so there's no drift and as little asynchronicity between them. This also means we must ensure they work under as similar as possible conditions, so they stay in synch all the time. And we do do this, we join them together with thermal compound, then bind them with copper wire to ensure similar thermal operation, we select them for same gain, or buy supermatched dual transistors (two transistors in same case), etc.

Aha, but, BUT!

The trouble starts in the fact that we can NEVER have identical devices, they will always be at least a little different. Then, they will NOT work under the same conditions - when one has a large signal (a crescendo, for example) at its base, the other will still be relaxed, as it takes time for the signal to propagate throughout the amp and be taken back to the inverting trannie, by which time the signal might have decreased. And lastly, as there is no perfect conductor, so there is no perfect semiconductor, so a part of the signal is always stored as a small charge inside the trannie, which takes time to discharge.

When this goes on for some time, all transistors/tubes/FETs will heat up, but because they are not the same, each will heat up differently. This will further move them apart in terms of synchronicity, and in the end, we will end up with an amp which is quite a bit removed from what it was initially. Now remember, in this context, one milisecond is a long time, here we are talking about nanoseconds and picoseconds - but with enough repetition, this starts to become significant.

And remember, while working, our entire amp heats up, so by deafult everything changes anyway.

As auditioners, we hear this as blurred sound, we sense lack of definition, no inciseveness where there should be, it starts to sound slack and not quite in control.

So, the idea is to bleed off those stored charges as quickly as possible. There are several techniques to do this, from applying more local feedback, i.e. feedback across each stage individually (a good idea anyway, as it removes the need for much overall feedback), from adding discharge resistors in places, to adding transistors in now somewhat more complex differential stages.

But the odd thing about it is that once we reduce magnetic distortion, the amount of overall feedback starts to matter much LESS. In fact, it would appear that more can be better, a total reversal of what we have done so far. On the other hand, the whole concept is still very new, and we lack experience to evaluate it fully.

Very interesting, even inspired work was done on this by an engineer from Lavardine, a French company, but the man was unfortunately killed in a car crash, and by Lundahl, a Swedish company best know for their magnetic amplifier, which received rather good reviews in England.

I have the papers for the Lavardine patent, the whole documentation, and i think I'll post it on my site this coming weekend, but be warned, its completely technical, much maths, much diagrams, but nothing a hobbyist could use as is. It is, after all, a patent application.

Cheers,
DVV

[audiojerry]

And when do you go to bed DVV? What time is it now in Belgrade?

[ehider]

Here's some more information about the supposed "best sounding" of all capacitors; Jensen "four terminals":

I've been told that the special variant Jensen terminal capacitors are a copy of the twenty year old Sprague capacitor developed for missile military applications (among other unique applications). These capacitors were specifically designed to maintain ultra low impedance all the way up to ultra high frequencies (beyond 1 Megahertz). Jensen basically duplicated the Sprague design but with modern day materials to create it's four terminal variant that is supposedly the "best of the best" sounding of any capacitor currently available.

If anyone on reading this thread is familiar with Jensen four terminal capacitors and how they compare sonically against other capacitors, please let me know! There are about a dozen hard core audiophiles that would be interested in obtaining another capacitor other than Jensens. Apparently everyone is sick and tired of the hassle involved with acquiring the Jensens ( even though everyone swears by them once they get they've installed them in their amp, pre-amp or cd player's power supplies).

[DVV]

And when do you go to bed DVV? What time is it now in Belgrade?

Why, when I get tired, Jerry.

It was gone 1 AM when I posted that, I'm 7 hours in front of you and 1 hour in front of Ferdi, who's in the Netherlands.

But the last few days were local holidays, and today, the ballad stops, it's a working day today. And before you ask, it's 8:25 AM just now locally. Outside, it's -6 centigrade, there's some snow from yesterday's snowstorm, but it's sunny anyway.

Cheers,
DVV

[tmd]

This is really great stuff. I can't wait to read more. I have roughly one million questions myself but the main ones I would like to ask at the moment are;
1. What the hell is a digital amp? The eAR two for example.
2. You guys are all talking solid state. Are the issues different for tubes? I own a Decware Zen amp and can't help thinking that the only limitation is that very sensitive speakers are needed. If you can find speakers which are 100dB sensitive, the search is over. However, I still haven't found those speakers so maybe this limitation is the kicker.
Keep up the good work, Neil.

[Dan Banquer]

Hello All;
         Before I reply to "memory distortion" I would like to address some things ehider brought up on circuit boards and caps. If someone could put up a web link to the jensen caps you have talked about I would like to investigate this further.
    I have found some similar things on circuit boards as ehider as. Coming from a strong RF background this has been thoroughly reinforced. At low level and line level, having an exposed ground plane on the top of the board (no solder mask) can help absorb stray signals and assorted interference. I have found using exposed ground plane to be helpful in this application, weather it be mixed signal such as DAC or line level. At Power Amp level, however, this has little or no effect. The signals are to high, and so is the amount of current.
    Having a solder tinned ground plane does however have some drawbcks and is not presently compatible with mass production assembly techniques, it can also be prone to contamination. However, being that the audio eqipment we use is in a rather controlled envronment generally lessens this problem. In addition, cleaning boards with a exposed surface ground plane is very easy. A vacuum cleaner for the dust, and Isopropyl Alcohol (91%) from you local drug store does the trick. Vacuuming out your equipment every 3 to 5 years is not a bad idea.

[JoshK]

If you can find speakers which are 100dB sensitive, the search is over. However, I still haven't found those speakers so maybe this limitation is the kicker.

I think Von Schweikert makes a pair of speakers called the 100db which are 100db sensitive.  Also check out the Coincidence Total Victory.  I know a guy who owns a pair and runs them on a WAVAC 300b amp.

[DVV]

Here's some more information about the supposed "best sounding" of all capacitors; Jensen "four terminals":

I've been told that the special variant Jensen terminal capacitors are a copy of the twenty year old Sprague capacitor developed for missile military applications (among other unique applications). These capacitors were specifically designed to maintain ultra low impedance all the way up to ultra high frequencies (beyond 1 Megahertz). Jensen basically duplicated the Sprague design but with modern day materials to create it's four terminal variant that is supposedly the "best of the best" sounding of any capacitor currently available.

If anyone on reading this thread is familiar with Jensen four terminal capacitors and how they compare sonically against other capacitors, please let me know! There are about a dozen hard core audiophiles that would be interested in obtaining another capacitor other than Jensens. Apparently everyone is sick and tired of the hassle involved with acquiring the Jensens ( even though everyone swears by them once they get they've installed them in their amp, pre-amp or cd player's power supplies).

Eheider, why don't you join the DeZorel tryout list? My point is that before rushing out to buy $2.4K worth of capacitors, you might want to investigate a really good line filter costing less than $400 first. And instead of improving your amp, it hits ALL of your system, cleaning up not only the amp(s), but the rest of it too.

That should put super caps in better perspective for you, as it would clearly demonstrate why they may not be necessary at all. Why pay for functions you can get better done elsewhere, and not only locally, but on an overall basis, and for less money?

Cheers,
DVV

[DVV]

This is really great stuff. I can't wait to read more. I have roughly one million questions myself but the main ones I would like to ask at the moment are;
1. What the hell is a digital amp? The eAR two for example.

Basically, that's an amp which uses a digitized input signal and does its entire work in the digital domain. In other words, it uses its output stages as basiclly switches, 0 for off, 1 for on. This avoids quite  afew problems with analog amps, and makes your output stage as efficien as it will ever be, around 99% efficient, allowing for less power devices, BUT forcing you to MOSFETs, because only they have the required short rise times. It's a switching amp, in other words.

But while ridding you of problems of analog designs, it introduces problems inherent to the digital domain. This is to say it's a very new field just now, and we have only just started to tread it. Also, while digital can be good, it's not perfect, as witnessed by problems such amps typically run into in the high range, or the 10...20 kHz region.

But they are the future.

2. You guys are all talking solid state. Are the issues different for tubes? I own a Decware Zen amp and can't help thinking that the only limitation is that very sensitive speakers are needed. If you can find speakers which are 100dB sensitive, the search is over. However, I still haven't found those speakers so maybe this limitation is the kicker.
Keep up the good work, Neil.

Somewhat different for tubes, but then tubes have problems inherent to them - such as the most obvious one, which is that after about 100 hours of operation, tubes begin to change their characteristics - permanently.

In MOST simplified terms, you could say tubes and transistors are opposites - tubes love voltage and hate current, transistors love current but hate voltage.

This makes tubes viable for preamps, but unsuited to power amps, and transistors far better suited to power amps than preamps. Bear in mind this is generalization only, and as such, just skims over the surface. The hard science is far more complex.

As for speakers, well, I don't want to be FORCED to go for ultra high efficiency speakers because I own an 8W per side tube job. High efficiency is most desirable, but it can bring about its own problems as well. I like 'em balanced, effcient but without sacrificies elsewhere (such as in distortion, bass extension, etc). Not least because you'll find extra efficient speakers cost extra money.

Cheers,
DVV

[MarinRider]

I believe the way a design is implemented is what causes the most differentiation in sound quality between amplifiers. After all, most amps have similar topologies and (except maybe at the high end) similar components, yet no two sound alike.

Many topologies can be made to sound good even with basic components, as long as the implementation is good. The best topology (if there is such a thing) and a bag full of Jensens  and bulk foils etc will sound bad if there are mistakes in the implementation.

Here is a hypothetical example:- take the best sounding amp you know of and add 2ft of the best twisted pair wonderwire in the known universe between input terminals and the inputs on the pcb, untwist the wire and with the biggest loop area possible route it around the power transformer and in parallel with supply rails and output wires as much as possible. This will result in poor sound quality.

This is clearly an example of a gross error in implementation, however all power amps have some compromises in their implementation. Below I list a few common circuit techniques and topologies that I now avoid because in my opinion they don't sound good. I think this is either because there is no intrinsic benefit or because they are difficult to implement really well.

a)    Cascodes
b)    Dual complementary circuits
c)    Active current sources
d)    Massive de-coupling

And the reasons why:
a)    Cascodes:- Another device in the signal path, also the base (gate) of the cascode tr cannot be perfectly decoupled - remember this is another input to the amplifier.
b)    Dual complementary - Why use nearly 2X the number of parts required (with 2X the pcb area and 2X the soldered joints)? Why do many people love SET amps? The parts can never be perfectly matched leading to the poles and zeros being slightly different for each half of the circuit.
c)    Current sources - I'm not saying these are bad per se, but you have to very careful how any ground related nodes are physically connected.
d)    Big decoupling caps - these have to recharge after a transient. The bigger (lower Z) they are the more the charging current, with perfect star grounding this should not be a problem.

Big, complicated high power amps are much harder to get right than small, simple low power amps.

In conclusion (IMHO):
Get the basics as ideal as possible - grounding, pcb layout, physical layout.
Keep it simple otherwise implantation becomes difficult.

Even the dreaded "green solder resist smearing" will be reduced with a simple, small pcb!

BTW, I would love to know what is in ehiders $1600 amps. Because there certainly are a few people around who think on a totally different level to most of us.

Cheers,

Dave

[DVV]

I believe the way a design is implemented is what causes the most differentiation in sound quality between amplifiers. After all, most amps have similar topologies and (except maybe at the high end) similar components, yet no two sound alike.

Many topologies can be made to sound good even with basic components, as long as the implementation is good. The best topology (if there is such a thing) and a bag full of Jensens  and bulk foils etc will sound bad if there are mistakes in the implementation.

I agree completely with the above. Too much is assigned to sheer topology, it's come to that if you're not using fully complementary, it doesn't really work at all. Yet it's hardly ever mentioned that there are intrinistic differences between NPN and PNP transistors, it's as if they don't exist at all.

Here is a hypothetical example:- take the best sounding amp you know of and add 2ft of the best twisted pair wonderwire in the known universe between input terminals and the inputs on the pcb, untwist the wire and with the biggest loop area possible route it around the power transformer and in parallel with supply rails and output wires as much as possible. This will result in poor sound quality.

This is clearly an example of a gross error in implementation, however all power amps have some compromises in their implementation. Below I list a few common circuit techniques and topologies that I now avoid because in my opinion they don't sound good. I think this is either because there is no intrinsic benefit or because they are difficult to implement really well.

That is indeed a gross error, to say the least. He who doth it should be smitten by the Lord, for no such stupidity should be allowed to multiply.

a)    Cascodes
b)    Dual complementary circuits
c)    Active current sources
d)    Massive de-coupling

And the reasons why:
a)    Cascodes:- Another device in the signal path, also the base (gate) of the cascode tr cannot be perfectly decoupled - remember this is another input to the amplifier.

Hang on - are you talking about cascodes in the input stage, usually a differential pair, or cascodes in general? Not nearly the same thing.

b)    Dual complementary - Why use nearly 2X the number of parts required (with 2X the pcb area and 2X the soldered joints)? Why do many people love SET amps? The parts can never be perfectly matched leading to the poles and zeros being slightly different for each half of the circuit.

Ah, this is the greatest disagreement between Dan and me. Dan likes fully complementary, and I don't for exactly the reasons you outline above. I find fully complementary circuits extremely hard to make work well and generally avoid them.

c)    Current sources - I'm not saying these are bad per se, but you have to very careful how any ground related nodes are physically connected.

I would never give them up, but I do agree with your words of caution. Fail to do it right and you not only have no benefits, but you gain problems.

d)    Big decoupling caps - these have to recharge after a transient. The bigger (lower Z) they are the more the charging current, with perfect star grounding this should not be a problem.

My response to this is to use same type, same line, same manufacturer larger cap, say 10,000uF, followed by a 4,700uF in parallel. Larger caps filter better but are slower, smaller don't filter as well, but are inherently faster.

Big, complicated high power amps are much harder to get right than small, simple low power amps.

Hear, hear. I would only add that a well executed small amp can sound subjectively just as powerful as a poorly executed objectively much more powerful amp. Sometimes even more powerful, though that's rare.

In conclusion (IMHO):
Get the basics as ideal as possible - grounding, pcb layout, physical layout.
Keep it simple otherwise implantation becomes difficult. ...

Exactly, though I would add that there is another trend at work these days. Just as it's wrong to overcomplicate a circuit, so there is a Scandinavian gig to take out everything you can take out so long as it still works. I think this is just as dangerous as what it sets out to rectify. In other words, there is such a thing as too simple.

What is often overlooked is that to use extremely simple designs one is soon relegated to the sky-high costing components, and thus simplicty soon becomes dreadfully expensive because of the need for very expensive parts.

As ever, I believe one should stay sane and do it rationally, without trying to prove any particular point.

Cheers,
DVV

[DVV]

That's it? We've exhausted the subject? Everybody understands everything, like damping factors, problems with feedback, power supplies, energy reserves and requirements, etc?

Nobody is interested in tuning your own seto of output devices?

Or is its CES time?

Cheers,
DVV

[tmd]

Well that certainly isn't it for me. I have devoured some of the info on your website DVV and I replaced a crappy op amp in my CD player based on one piece you wrote. I am only getting started on a long, frustrating and ultimately very rewarding road to audio nirvana. I would dearly love to know enough about both tubes and solid state to design and build my own pre amps and amps.
The only thing I even think I know enough about to comment on is the power supply. If I were to make my own amp, I would try to overbuild the power supply to a rediculous extent. It seems that you can never have too big a transformer or too much capacitance. However, as I am cheap, I would try to get as much gear as possible from old equipment. I have ripped apart old printers and the like recently and not a switched supply in sight. I found a huge transformer with two very meaty bridge rectifiers in one printer. Might be worth playing with.
I believe that for me the best way to learn is to build and tinker with various pieces. I may start at the source though and work from there as that is my current focus for tinkering. How much better might my CD player sound with separate power supplies for each section? Can't wait to try. Sorry, a bit off topic but all very interesting.
Here is an amp question.
Can you explain why the type of speaker load is such a critical thing and why amps with no feedback still work if it is so crucial?
Thanks, Neil.

[MaxCast]

That's it? We've exhausted the subject? Everybody understands everything, like damping factors, problems with feedback, power supplies, energy reserves and requirements, etc?

Nobody is interested in tuning your own seto of output devices?

Or is its CES time?

Cheers,
DVV

I think there are a limited number of people here that can actually follow what you guys are saying...and I am one of them  

I do enjoy reading the thread though.  Maybe for us beginners you can describe the journey a signal takes through an amplifier and then we can ask questions or see how/where the above fits in along the signals journey.

DVV, could you list your web site (or list it in your sig) I lost it when my puter crashed.

[Dan Banquer]

Hello All;
          No I'm not at CES. Since I do direct mail and CES is predominantly for retailers it really doesn't make sense for me to show up. Some of my equipment has been there, as well as the Stereophile Show, over the years.
    Complimentary topologies are tough. This is not the kind of stuff you can put out as a kit. The offset, current, voltage and bias sets in the LNPA 150 would be over many hobbyists heads. The complimentary topology does have some real advantages when properly implemented. It's called divide and conquer. When you need to swing large voltages having two transistors instead of one really helps on a number of issues. Transistors that are listed as complimentary are not truly complimentary, as Dejan points out. However with a little work ( a little circuit tweaking) things can be made to work quite well. As I pointed out in earlier post the "secret" here is to make the complimentary amp as complimentary as possible on an AC and DC basis. Most implementations forget that on some level.
   Let me explain this through a wonderful joke.
   An Engineer and a Theoretician are placed equidistant from a very beautiful and very naked woman. They are told that they can go to this woman but only 1/2 the distance at a time. The Theoretician yells out , "I'll never get there". The Engineer says,"Start;I'll get close enough!"
    Dejan; if and when you are stateside and come to Boston, we'll go through the schematics and I will beak it down for you.