[JoshK]

Of little consequence, but I think it is "theorist" not theoristician.

[MarinRider]

DVV,

Cascodes:
In my experience using listening tests, removal of casode transistors (FET or BJT) results in better sound - this applies to input stages and VAS stages (I've never tried them in output stages). The benefits (linearity and speed) seem to be outweighed by the disadvantages (getting a clean reference and the extra device in the signal path).

Implementation:
I included the example of a gross error to lead to the point that all amplifiers have many tiny errors (probably better called compromises) which lead to coupling of signals in unwanted places - electrostatic, magnetic, mechanical, rf pickup and grounding errors. The best sounding amplifiers achieve a good balance of minimising all these effects (and more, probably).

Massive decoupling:
I have not tried your suggestion of paralleling with smaller values of the same type of cap.
I will try to clarify what I meant  - my concern is with the charging currents that flow between the main resevoir caps (frequently off board) to the on board local decoupling caps. Actually, Doug Self explains this well on his website so it may be best to check there.

Whilst I'm on the subject of Doug Self and his "blameless" ampllifier, I would say that Doug goes a long way towards minimising implementation errors and it's well worth reading his website for anyone wanting to design an amp. However, I know a well trusted desginer who build a blameless amp  and the sound quality was mediocre.

BTW Doug's treatise claims that using a current mirror forces equal currents in each transistor of a differential pair. Sadly he forgot to take into account the base current of the VAS in his calculation. Here again is a case of a great technical solution to a non problem causing unecessary complication.

Dual complementary:
One good thing about this topology is that with careful matching you can do away with dc servos (here's another thing to seriously avoid) or the capacitor used to roll off the dc gain.


One thing I have learnt after many years of dabbling is that there are no rules in hi-fi. Just when you've convinced yourself that (for example) silver plated cables sound bad, someone will demo a system where they sound stunning. So I'm sure all the above can be proven wrong.

Happy Listening.


NB Here's one that might get a reaction:
Mosfets are crap as output devices but great for input devices, for power amps!

Dave

[tmd]

Is it true to say that all amplifiers built to sell are guaranteed to have more compromises that the home cooked variety?
Who cares if the heatsink in a diy job is two feet square, whereas that just won't work in something to sell. It would look ugly. I saw a pair of amps built by an enthusiast which only their maker could love. However, they were built with very few if any compromises and sounded incredible to these untrained ears. Overbuilt to the hilt they were.
So, if one were to start with a good theoretical design, build the amp with no compromises whatsoever, except maybe price to some extent, you would have to be onto a winner wouldn't you?

[DVV]

Hi Dave, all,

DVV,

Cascodes:
In my experience using listening tests, removal of casode transistors (FET or BJT) results in better sound - this applies to input stages and VAS stages (I've never tried them in output stages). The benefits (linearity and speed) seem to be outweighed by the disadvantages (getting a clean reference and the extra device in the signal path).

Disagreed on the Vas stage. I got excellent to oustanding results - however, I used a topology I never saw used before, not quite like that. I underline "not quite like that" because I have been hit a few times with electrically equivalent circuits which were much different to what they were equivalent to. In other words, you have to try it to know (as if you didn't know that already).

Implementation:
I included the example of a gross error to lead to the point that all amplifiers have many tiny errors (probably better called compromises) which lead to coupling of signals in unwanted places - electrostatic, magnetic, mechanical, rf pickup and grounding errors. The best sounding amplifiers achieve a good balance of minimising all these effects (and more, probably).

I realize that Dave, and took it as such. Just an example.

Massive decoupling:
I have not tried your suggestion of paralleling with smaller values of the same type of cap.

Do try it, Dave, it's a very educational exercise. I was sceptical initially, but after trying, I'm reconsidering my own ex-views.

I will try to clarify what I meant  - my concern is with the charging currents that flow between the main resevoir caps (frequently off board) to the on board local decoupling caps. Actually, Doug Self explains this well on his website so it may be best to check there.

I am familiar with D. Self's "faulteless amp", or whatever. However, while he did make for interesting reading, and I did pick up a few good hints, what he showed is what I built way back in the early 80-ies.

Whilst I'm on the subject of Doug Self and his "blameless" ampllifier, I would say that Doug goes a long way towards minimising implementation errors and it's well worth reading his website for anyone wanting to design an amp. However, I know a well trusted desginer who build a blameless amp  and the sound quality was mediocre.

Odd, a few locals also built it, some of them experienced professionals, and it never made any better than a so-so sound. And they tried with different components, so that rules out poor component argument.

In my view, the problem is that his amp is sort of lazy, it somehow misses the trick in terms of dynamics. But that's just my view.

BTW Doug's treatise claims that using a current mirror forces equal currents in each transistor of a differential pair. Sadly he forgot to take into account the base current of the VAS in his calculation. Here again is a case of a great technical solution to a non problem causing unecessary complication.

True, and I could add a few comments, but I won't, I don't like discussing absent people who cannot speak for themselves. And ultimately, he did what he did, and anybody who can do better, by all means, do it, don't talk about it. Goes for all of us.

Dual complementary:
One good thing about this topology is that with careful matching you can do away with dc servos (here's another thing to seriously avoid) or the capacitor used to roll off the dc gain.

Servos can be a problem, but as ever, if done well and with some thought there will be no "pumping" or "breathing" effect. This is usually caused by using el cheapo op amps, totally unsuited for the job, such as TL071, and asking it to pump its heart out into a fairly large capacitor. Use a decent low drift op amp, such as OP97, make sure your cap is no larger than 220nF and of good quality (e.g. Wima polycarbonate) and you won't have a problem in the world.

One thing I have learnt after many years of dabbling is that there are no rules in hi-fi. Just when you've convinced yourself that (for example) silver plated cables sound bad, someone will demo a system where they sound stunning. So I'm sure all the above can be proven wrong.

Very true indeed. On the other hand, that's a problem only for those with hard wired views and habits. The rest will adjust eventually.

Happy Listening.


NB Here's one that might get a reaction:
Mosfets are crap as output devices but great for input devices, for power amps!

Dave

Agreed on crap as output devices, never used them as input devices. I sort of lean towards ultramatched, ultra low noise dual transistors, such as for example SSM 2210, etc.

Cheers,
DVV

[DVV]

Is it true to say that all amplifiers built to sell are guaranteed to have more compromises that the home cooked variety?
Who cares if the heatsink in a diy job is two feet square, whereas that just won't work in something to sell. It would look ugly. I saw a pair of amps built by an enthusiast which only their maker could love. However, they were built with very few if any compromises and sounded incredible to these untrained ears. Overbuilt to the hilt they were.
So, if one were to start with a good theoretical design, build the amp with no compromises whatsoever, except maybe price to some extent, you would have to be onto a winner wouldn't you?

Not necessarily, Tmd. To get good sound you first and foremost need a good design, one made so that it has no inherent pitfalls. Next you need quality components - I mean quality, not NASA style, price-no-object components. Following that, you need much patience to do menial jobs like buying 100 transistors only to match two pairs - but match them to almost perfection. Then you need much patience to do it the right way - for example, if you are using discrete trannies as a differential pair, to put some thermal compound around them, then squeeze them in tight - but not too tight - into a copper ring (this will give you excellent thermal matching, very important with differential pairs).

At the same time, you need a good component outlay, so that, for example, your filter capacitors sit as near to your output stage as you can make it - this will be a great help in dynamic peaks, when the caps are near.

And so forth. In other words, it takes MUCH knowledge, patience, care, and wherewithall to actually make something approaching a no compromise designs.

But all that is not worth a penny if you don't have a great design to start with.

Cheers,
DVV

[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.

http://www.zero-distortion.com . "Zero distortion" was the name I gave my radio show dedicated to audio, which ran for three years every Tuesday night for two solid hours. I grew attached to the name. Actually, in Serbian, that's a double meaning - at face value, it means "zero distortion", but it could also mean "no bull included".

Max, this thread was started to make life better for you guys who don't know as much about this particular subject as some of us others. But if you guys don't slap us, and slap us hard when we wonder off on a crazy trajectory somewhere, this thread will have missed its point.

Let's be honest, we do tend to get lost, and do need being told we're losing it (actually, my wife would tell you I lost it a long time ago ).

So all you need do is ask what you are interested in, and tell me quite straight in the face when I start to wander off.

I'll write a small piece on how the signal flows through an amp tomorrow or on Saturday and post it here, and the rest of the mad gang (Dan, Tmd and others) can correct me or add their comments, which would be really great.

Cheers,
DVV

[Ferdi]

HI DVV, this thread is great for my favorite style of learning: reading stuff way over my head and filing all the hard stuff away for future study.

I am looking forward to your little story on how the signal travels through the amplifier. What would be really great is if you could refer to some other material to read up on things like different amplification topologies, implementations etc. Maybe that isn't even available anywhere though.

Thanks for starting this long story though.

Ferdi

P.S. It's quite late here already and worse for you....

[audiojerry]

don't forget video and audio streaming when you do that small piece on signal flow on Saturday, DVV.  

[MaxCast]

don't forget video and audio streaming when you do that small piece on signal flow on Saturday, DVV.  

Yeah, I'll take the 6.1 DTS version done in 4:3 please

Looking forward to the piece, DVV.

[Guan]

Hi Dejan,

May I ask for a simple explanation on what is 'damping factor' and is it true that it's the most important factor in controlling a dynamic driver?

Also why does a tube amp, even a powerful one, struggle to have damping factor in double digits but it is common for solid state amps to have damping factors in the hundreds. Does it really matter?

Cheers,

Guan

[DVV]

Re: damping factor

For any signal to be transferred properly, it is always assumed that the difference between the output impedance of the outgoing device and the input impedance of the incoming device should be 1:10 (i.e. the input impedance should be at least 10 times greater than the output impedance).

If this is so, signal transfer will be accomplished with little loss. Just as it's true for any link, so it is true for the amp-speaker line. But here more so than elsewhere, because the speaker is no ordinary load, it's far more complex. The speaker has what is known as a Q factor, where Q stands for quality. Not to get into another topic altogether, just take my word for it that it's highly desirable that the amp controls this Q factor, and to do that, it must have an output impedance at the very least 10 times smaller than the speaker impedance. This difference is called the damping factor.

If your amp has an output impedance of say 0.1 ohms, and the speaker has a nominal impedance of say 8 ohms, then your damping factor is (8:0.1) 80:1.

Now, here come the problems.

First and foremost, on the amp side, the smaller the output impedance, the easier for the amp will be to deliver larger currents (assuming the rest can cope with them, of course). In that respect, we want as large a damping factor, or as low an output impedance, as we can get. But it's not at all the same thing HOW you achieve that, nor where.

As to how, just add lots of negative feedback and your output impedance will drop, or your damping factor will rise. Of course, this will cause problems elsewhere, so it's hardly the answer we want. Add parallel output devices and your output impedance drops almost linearily; add another pair of same transistors, and you almost literally halve the output impedance, or double the damping factor.

But the damping factor (output impedance) is a composite matter, because it depends on amp topology, number of output devices, power supply output impedance and your transformer impedance. All these combine to produce the output impedance.

This should explain my strong preference for multiple output transistors, quality parallelled (which also reduces their own inherent output impedance) filter capacitors and large, seemingly overpowered toroidal transformers. All these combine to give a small output impedance even without negative feedback, let alone with it.

However, the damping factor is NOT linear. It's always much larger at the low end, say 0...1,000 Hz, than at the top end, say at 20 kHz. Several factors combine to produce this effect, such as diminishing effects of negative feedback, output inductor, rising output impedane of the power supply, etc. Typical ways to overcome this is to use better power supplies, wider open loop (without feedback) amplifiers, more output devices, avoid use of output inductor (and that implies a supremely stable amp, no easy job), etc.

On the speaker side, as we all know, their actual impedance is far removed from their nominal impedance. It's quite usual to see nominally 8 ohm speakers drop as low as 3 ohms or so, while some otherwise excellent speaker have a 4 ohm rating, and drop to 1.8 ohms. In addition to this, the speakers can be very reactive, meaning that they show large phase excusions, which tax the amp heavily, and a good way of reducing these interactions is to have a good damping factor. The better the damping factor, the less succeptible the amp to speaker funnies in very broad, general terms.

But work backwards here. A damping factor of 10:1 in case of admittedly rare but still present 1.8 ohms means an output impedance of (1.8:10) 0.18 ohms; actually, since you would want to have at least some margin of safety, that should be say 20:1, or 1.8:20=0.09 ohms. Into the full 8 ohms, this would be a damping factor of (8:0.09) 89:1, or to round it off, say 100:1.

More is better, but it's just as important how you get as it is to have it. And it can be had; for example, my favorite prototype has a damping factor of 150:1 into 4 ohms at 20 kHz. But, I use four output pairs in a wide open loop bandwidth design, large high quality capacitors and oversized, low loss toroidal transformers, because of which I can afford a low negative feedback factor and still have an excellent damping factor. At 100 Hz, it works out at better than 600:1 into 8 ohms closed loop (with negative feedback). I have to go that high if I want it to be able to do what I set out to do.

So, to conclude - a good damping factor makes an amp less susceptible to speaker load variations and ensures a good signal transfer, with little dropouts at frequency extremes and good speaker cone control. But it is very important how it was achieved, whether it has been forced by excessive feedback, or whether it is designed in, inherent to the design as such.

In my experience, amps which have been fed back to kingdom come to achieve impressive results will tend to sound less powerful than their specifications and measurements show, they will tend to sound weedy and aenemic, as if they lack real power. Conversly, amps which use lower feedback factors, larger power supplies, etc, will tend to sound full bodied, confident, with true grit and would seem to have more power than their specifications and measurements suggest.

Some months ago, I took my DeZorel filter to a friend's place, where it was connected to his Naim amp. That amp is rated at 30W/8 ohms, but as is it sounded like a typical 70-80W job. We added the filter, which of course made the amp's power supplies far more efficient, and it sounded like my Yamaha AX592, which is rated at 100W/8 ohms (though in all fairness, that Naim cost about twice what my Yamaha cost), but it's an obvious case of a job well done. The late Julian Vereker really did know his stuff.

Cheers,
DVV

[DVV]

don't forget video and audio streaming when you do that small piece on signal flow on Saturday, DVV.  

You want a detailed descripition of what and how to do with the showgirls, as well?

Jerry, video signal is treated more or less the same way as any audio signal, though using different components, of course. The video signal bandwidth is well into the 3+ MHz range, which is why it's usually shielded. Harman/Kardon, for example, has a working bandwidth of 9 MHz even on their lowliest of receivers, they like to play it safe.

Ciao,
DVV

[tmd]

Dejan,
It seems to be very important to match the output transistors as best possible. Is this because they are in push-pull mode and you want the zero transition to be smooth or is it way more complicated than that?
Also, you say in an op amp article I read that one of the advantages of op amps over discrete devices is that they are well matched because they are on the same substrate. Can you get matched transistor pairs in a single package? I assume so as my LeAmps have a chip to output instead of discrete devices although I am also guessing that the chip has more to it which may mean that the design of the circuit is pretty set by the chip maker allowing less choice in design of the overall amp it goes into.

[Guan]

Thanks DVV for a masterful explanation.  

Guan

[MarinRider]

I keep trying to find something to argue with in this thread but there is a lot of sense talked here!

I will have a go at damping factor:
1/   Some valve amps sound good and have very poor damping factor.
2/   Except for poor power transfer even a damping factor of 1 could sound fine if the load is mainly resistive and consistent.
3/   John Lindsley Hood added 0.22 Ohm resistors on the output of his 80W Mosfet amps (not a bad sounding design by the way) on the basis that it damps reflected energy from the loudspeaker. This gives a best case DF of 18 into a 4 ohm load.

And Naim amps;
IMO the Julian Vereker designed amps tended towards harshness and lacked tonality. The new stuff keeps many of the strengths of the old but with excellent musicality as well. (No disrepect intended to Julian, he had his views, and may he rest in peace), It's interesting to note that the Naim amps are an ancient design with truly excellent implementation and good but sensible components. That's the way to do it!

Dave

[DVV]

Dejan,
It seems to be very important to match the output transistors as best possible. Is this because they are in push-pull mode and you want the zero transition to be smooth or is it way more complicated than that?

No, that's basically it. Well, they do sound better that way, too.

Also, you say in an op amp article I read that one of the advantages of op amps over discrete devices is that they are well matched because they are on the same substrate. Can you get matched transistor pairs in a single package? I assume so as my LeAmps have a chip to output instead of discrete devices although I am also guessing that the chip has more to it which may mean that the design of the circuit is pretty set by the chip maker allowing less choice in design of the overall amp it goes into.

Of course you can. For example, SSM 2210 (NPN) and SSM 2220 (PNP). Ultramatched to better than 0.1% (try that with discrete!), DIL 8pin plastic package, noise figure better than 1 dB COMBINED (sic!), etc. The only drawback is their voltage, just 40V, recommended to be kept at 35V or less.

Obviously, they are meant for extremely low noise input stages, in which your metal film resistors will contribute more noise than the two trannies combined. But because of the low voltage, I am generally forced to make a virtual battery voltage source for the pair, which is bad becose it makes me work and spend money, and is good because you get an input stage second to none - literally. No Levinson, no Krell, no nobody can do any better.

I'll tell you the rest of the trick too, so you're not shortchanged on crucial data - use Motorola's BC 650 NPN transistor for an NPN differential input stage using SSM 2210. Ultra low noise, hyper linear, very hard to find, but not expensive at all transistor. Only Motorola makes it, as far as I know.

Say hello to Connie Dover for me if you should meet her on the street. I'm a fan of hers of sorts.

Cheers,
DVV

[DVV]

Thanks DVV for a masterful explanation.  

Guan

As a sifu, I am obliged to share my meager knowledge with others, is that not so, Guan? You said "sifu", I didn't.

As I understand it, "sifu" is a Chinese word for an honorable and learned master - not sure that describes me well at all, but there you are. At my age, I honorably and learnedly chase girls downhill only. After much meditation, I realized downhill was less exerting than uphill.

Seriously though, no problem at all Guan, all, I always cheered for active sites, places people could aks be told in plain terms (well, I try!) what they wanted to know. Just reading is like buying a paper mag, totally passive.

Cheers,
DVV

[DVV]

I keep trying to find something to argue with in this thread but there is a lot of sense talked here!

Of course you try, Dave, you're a pig headed Irishman, and if you don't argue, there's no point. That's bad. I would strongly advise you never to come to Serbia, because here, it's worse - FAR worse. Compared to Serbs, pig headed Irish come out smelling of roses.

Which is probably why Serbs have a band called "Orthodox Celts", specilizing in Irish music, which packs any hall they play in, or why every self-respecting CD pirate has a selection of Irish music to choose from. Which makes my life MUCH easier, as I also happen to be an Irish music freak. Birds of feather ...

I will have a go at damping factor:
1/   Some valve amps sound good and have very poor damping factor.
2/   Except for poor power transfer even a damping factor of 1 could sound fine if the load is mainly resistive and consistent.
3/   John Lindsley Hood added 0.22 Ohm resistors on the output of his 80W Mosfet amps (not a bad sounding design by the way) on the basis that it damps reflected energy from the loudspeaker. This gives a best case DF of 18 into a 4 ohm load.

Yes, don't pass up this chance to be argumentative, Dave, you are after all Irish. But seriously, what you said above is all true, and if your intent was to put some additional perspective on the damping factor issue, you did it, and you are quite correct. As I said, it's a very complex question, but in essence, it revolves around the speaker characteristics, both impedance and the phase problem.

As for J. L. Hood's resistor, I have yet to try that, and I do have it down in my log as a to do item. Until I do, I honestly cannot comment.

And Naim amps;
IMO the Julian Vereker designed amps tended towards harshness and lacked tonality. The new stuff keeps many of the strengths of the old but with excellent musicality as well. (No disrepect intended to Julian, he had his views, and may he rest in peace), It's interesting to note that the Naim amps are an ancient design with truly excellent implementation and good but sensible components. That's the way to do it!

Dave

My friend, the owner of the integrated amp, commented he had never heard Naim deliver such wonderfull bass as when the filter preceded it.

Naim is notoriuos for being very choosy what it works with, I am told.

Cheers,
DVV

[audiojerry]

DVV. I was trying to pull your leg, and now I think maybe you are pulling mine. In case you were not, my tease meant that I wanted your discourse on signal travel to be presented with audio/video, THX 7 channel format.  

[hairofthedawg]

You mention that the amp and speaker are very much intertwined as far as the damping factor and speaker load.  I realize that it depends largely on a designer who has the ability to design both, but would you recommend purchasing both from the same designer?  I wonder if most designers build things as systems or as the best they can do with each individual item?  It would be nice if they looked at them as systems, but I have no idea.  It's worked for me in my case, Symphonic Line, but I just go by the sound, not the specs.  It was a big leap of faith but I haven't been disappointed.  From what I've read it sounds to me like the most useful aspect of specs is determining synergy, but damned if I know enough to do that.  Has anyone ever put together a chart of the relationship of various specs between components and described what each value does in relations to the others.  That would be a hell of a project but I think it would be pretty useful, provided you could get the useful, and accurate, specs from the manufacturers.  Sorry if I don't make sense, like others have said, this is mostly over my head, but it does make me think, and that's good...thanks all!

cheers,

Dick