Sound Quality Vs. Measurements

Status
Not open for further replies.
Best thing about the book is that it's a story. Especially for a novice, seeing how a circuit works isn't the same as knowing why it's used, which is generally a matter of history.

It's probably a very English history, and partisan in several respects, but illuminating all the same. Not up to date though, obviously.

If I remember his story, life was a struggle for transistor power output stages until reasonably matched NPN, PNP pairs became available. Can't much remember why...problem of symmetrical drive for PP I guess.

Very, very British.

There are quite a few British audio companies which used quasi-complementary output stages way after more than enough complementary devices was available. Naim, Arcam and many others.

The ideological notion behind this was (and still is on occasion) that there were inherent differences between NPN and PNP devices and that because of them, no true match could ever be made. And obviously that a true match could be made only by using NPN devices alone.

And they were not alone, quite a number of Continental audio companies did more or less the same, including some giants, namely Philips. When Philips produced its best ever series of audio devices (preamp, power amp and tuner) in their US department, their A 22AH578 200 WRMS/8 Ohms power amp used NPN devices only. A very much American topology popular in those days (op amp input, preceeded by an active buffer), however unlike US designs of the day, it used only NPN output devices (355N1, four pairs per channel).
 
Best thing about the book is that it's a story. Especially for a novice, seeing how a circuit works isn't the same as knowing why it's used, which is generally a matter of history.

It's probably a very English history, and partisan in several respects, but illuminating all the same. Not up to date though, obviously.

If I remember his story, life was a struggle for transistor power output stages until reasonably matched NPN, PNP pairs became available. Can't much remember why...problem of symmetrical drive for PP I guess.

That is your opinion from having read the book a long time ago?

Michael J.
 
Very, very British.

There are quite a few British audio companies which used quasi-complementary output stages way after more than enough complementary devices was available. Naim, Arcam and many others.

The ideological notion behind this was (and still is on occasion) that there were inherent differences between NPN and PNP devices and that because of them, no true match could ever be made. And obviously that a true match could be made only by using NPN devices alone.

Linsley-Hood published an article in Wireless World in June 1970 as a prelude to a Class AB design using complementary transistors.

In this he illustrated the assymetrical waveform of complementary pairs at HF. The article is available online at at http://sound.au.com/tcaas/jlhab1.

The assymetry arises because majority carriers are electrons in NPN transistors and these have greater mobility than the holes which are the equivalent carrier in PNP transistors.

Improved Quasi Complementary stages are worse than Complementary at LF however corrective feedback can be quite effective in that regard.

Corrective feedback is less able effective with problems that arise at HF as with complementary systems.

It is probably easier for the enthusiat to make a complementary system work properly but I would not knock the efforts of Naim, Cyrus and others who have elected to persist with the Q/C layout.

Michael J.
 
My memory says that when amps went wrong often the NPN survived . Even now when I push things to the limit the PNP will let go first . I feel that an all NPN output stage is often rejected because on paper it looks to have worse distortion . If this distortion is due to crossover or asymmetry is often overlooked , the asymmetry being less detrimental I feel . The Quad 303 has very low crossover distortion and is a Quasi-comp triple . I wonder if the Quad circuit would be better than many realize , not least with modern transistors ? I like it's primitive protection circuit . I find none work correctly so at least the Quad is easy . A CCS PSU works . It can even have a switch to disable it . For me that should have a wine detector override .

Douglas Self shows a NPN / PNP output stage where the transistors are able to fully discharge the junctions , it is also the simplest he shows . This would be justification enough for NPN/PNP . MOS FET's do not have this problem .
 
The article is available online at at http://sound.au.com/tcaas/jlhab1.
That's an incomplete URL, but I found it here:
The Class-A Amplifier Site - JLH Class-AB Amplifier

Douglas Self shows a NPN / PNP output stage where the transistors are able to fully discharge the junctions , it is also the simplest he shows .
Perhaps this is discussed in some thread or book I haven't read (I've seen Self's website, but haven't read his nor Cordell's book), but I recall something about this before, that in a push-pull BJT circuit, it's advantageous to totally turn OFF the non-conducting transistor to get lower crossover distortion. Why is this?

"Fully dicharge the junctions" sounds like the BE capacitance creates a problem, but I'm trying to see what that problem is.
 
My memory says that when amps went wrong often the NPN survived . Even now when I push things to the limit the PNP will let go first . I feel that an all NPN output stage is often rejected because on paper it looks to have worse distortion . If this distortion is due to crossover or asymmetry is often overlooked , the asymmetry being less detrimental I feel . The Quad 303 has very low crossover distortion and is a Quasi-comp triple . I wonder if the Quad circuit would be better than many realize , not least with modern transistors ? I like it's primitive protection circuit . I find none work correctly so at least the Quad is easy . A CCS PSU works . It can even have a switch to disable it . For me that should have a wine detector override .

Douglas Self shows a NPN / PNP output stage where the transistors are able to fully discharge the junctions , it is also the simplest he shows . This would be justification enough for NPN/PNP . MOS FET's do not have this problem .

Apart from my incorrect spelling of asymmetry the link I provided in my last post was corrupt.

It should have read The Class-A Amplifier Site - JLH Class-AB Amplifier. There is a brief mention of the Quad 303 in that reference. The 303 receives a mention in the reference above as well as in Linsley-Hood's book.

In regard to the assertions that quasi comp is very,very British I would like to add that amplifier applications based on Baxandall's version of these output stages made it into RCA Power Devices handbooks possibly as early as 1970.

My edition of the Handbook 1978 includes a complementary output design with cross coupling of the driver emitters. I built a few amplifiers using this idea soon after acquiring the Handbook.

I once had a Quad 33/303 on loan from a friend - it was pretty bullet proof. Sometimes you learn something from restoring such equipment but I had to give this one back.

Michael J
 
That is your opinion from having read the book a long time ago?

Michael J.

Yes. I was into valves at the time. The stuff on transistors is all I've ever read about solid state amps, so I guess if I have an idea it must have come from there.

Just fetched it from the attic...cough, splutter...and yes, the chapter on transistor power amps appears as a problematic search for symmetry until he gets to complementary pairs, which appears to be end of story. He does note the problems of perfect matching and PNP frailty but doesn't pursue it. He writes that the availability of PNP power transistors

"...provided an incentive to the circuit designers to provide amplifier systems which took advantage of this new technology, and offered the possibility of reducing low signal level crossover distortion to a level where it would no longer be audibly detectable."

Moving to the chapter on "contemporary" design, I see a symmetrical complementary output stage.

I think symmetry is a holy grail rather like zero distortion. I can see that the perfect amplifier would be symmetrical, and have zero distortion, but this doesn't exclude the possibility that a more assymetrical, more distorting amplifier could be a better bet in an imperfect world. As long as imperfections exist, their quality is just as important as their quantity.

Fortunately the world has since become perfect, more or less.
 
That's an incomplete URL, but I found it here:
The Class-A Amplifier Site - JLH Class-AB Amplifier


Perhaps this is discussed in some thread or book I haven't read (I've seen Self's website, but haven't read his nor Cordell's book), but I recall something about this before, that in a push-pull BJT circuit, it's advantageous to totally turn OFF the non-conducting transistor to get lower crossover distortion. Why is this?

"Fully dicharge the junctions" sounds like the BE capacitance creates a problem, but I'm trying to see what that problem is.

Hello benb,

Nigel refers to a complementary output stage - inwhich the emitters of the driver transistors are cross coupled by a resistor (which may have a small capacitor in parallel with it).

To turn on a transistor you need to charge up the base emitter junction and for audio purposes the time this takes is significant.

If you remove the charging source, the charges already occupying the junction area linger and the transistor is not fully switched off.

Cross coupling driver emitters enables stored charges to be removed through the output transistor base regions on the cut off cycle so these are clear before the relevant output need to be turned on again.

With regard to this arrangement - Marshall Leach believed that the drivers operate in Class A as the output devices traverse the crossover region and this eliminates notch distortion.

You can read more on his claims if you look up his Low TIM amplifier pages. These describe a fully complementary amplifier from input to output - I built one 25 years ago.

Linsley-Hood discusses stored charges in the context of his Class AB amplifier - refer to the link you corrected for me.

I suggest you check both of these references out for more information.

Michael J
 
Last edited:
dSWa99w.jpg


This is the type of output I was taking about . Self says it is often stated that taking a resistor to the opposite rail is better ( two resistors ) as it is class A . As he says this single resistor option will reverse bias the junctions which should ensure what can be discharged is , so no great advantage . He also says some claim better bass if class A . It sort of makes sense if using fast transistors to get the last beans of performance from them . As said it is also the simplest arrangement . The traditional Dalington resistors add feed-forward whilst dealing with the same problem . Perhaps add a little ? The Blomley amplifier attempted to avoid these problems .

High Performance Audio Power Amplifiers - Ben Duncan - Google Books
 
As far as I can see, Nige, you are showing us a very classic driver with a speedup capacitor, something that has been used for the last 30 years at least as the default driver stage.

The usual variation is adding a predriver stage to raise the overall current gain and to better isolate the VAS from the output. This also can be done in several ways.

At the time JLH wrote his book, this was no standard at all, since PNP power devices were rather rare, but that did change by say 1975, when a fair selection of complementary devices was available from Motorola, NEC, Toshiba, etc. Not nearly as wide as today, of course, but still fair.

If I'm missing something, do tell.
 
...

Moving to the chapter on "contemporary" design, I see a symmetrical complementary output stage.

I think symmetry is a holy grail rather like zero distortion. I can see that the perfect amplifier would be symmetrical, and have zero distortion, but this doesn't exclude the possibility that a more assymetrical, more distorting amplifier could be a better bet in an imperfect world. As long as imperfections exist, their quality is just as important as their quantity.

Fortunately the world has since become perfect, more or less.

This brings us back to the usual difference between theory and practice.

In which we can have one fully complementary, fully symmetrical amp which measures well but nevertheless does not sound as well as it measures, and another Single Ended input amp, which does not measure so well but just to spite us manages to sound really good, well above what one would expect from its measurements.

"Theory is when everyone knows how it should work, but it doesn't.
Practice is when no-one has a clue how and why, but it works like a charm."
(Technicher Rundschau, Zurich, Switzerland, circa 1969)
 
My point was the simple circuit is the one that works best . This reason alone might make PNP a better choice . As I pointed out the usual feed forward resistor are now not required . I suspect most people never would see these resistors as being for dispersing the stored charge . It is a difficult question if the resistors help distortion in feed forward ( conventional Darlington ) , I am not sure . My instinct is to say they do . In the Complementary feedback pair the resistor has even more to do . I read and it makes sense that the CfbP can have one transistor doing all the work ( hogging ) . It will measure better than a Darlington , if working correctly better still .
 
My point was the simple circuit is the one that works best ... .

A highly questionable statement. What is "simple"? I would imagine "simple" as relatively uncluttered FOR A GIVEN PURPOSE.

You'll agree, I hope, that it's hardly the same thing whether you want 10W or 150W from a circuit, whether it's supposed to be fine into 8 Ohms only or if it should keep its cool when the load drops to 2 Ohms or less, etc, etc.

It also depends on how you distribute your feedback - if you go for high local degeneration, you will very likey need to use more components.

And so forth, the point being that "simple" is a VERY relative term and idea. Just as I am sceptical about American behemoth amps delivering 500+ WRMS, I cannot turn a blind eye to the logic of some of their arguments.

You don't need to read a ton of books to verify this - just compare head to head a high quality low power amp with a high quality amp which is say 6 dB more powerful nominally. Listen for high power demanding sounds like the tympani, or large scale orchestral works and the difference become obvious soon enough.

Remember, the typical road of evolution for companies you love, such as Armstrong, Quad, etc, was to start off with simple circuit, low power models and then progress to not so simple higher power models. Look at Quad 33 and what came after it, look at Armstrong's line of wonderful receivers yielding 30 and 50 WRMS, and their last effort, a power amp yielding 200 WRMS.
 
Even so . The get it to switch is the most important thing . The Cray super computer was if I understand correctly worked in class A so as take the switching speeds to a new dimension . It went through the origin rather than switching a transistor on or off as binary . Forgive me if I have been misinformed . It seems to me this type of principle is of prime importance . The JLH amp being class A didn't need it to be so . Also the speed of the transistors of less importance . 3055's are more than good enough . Doubtless the compliment would do no harm . The typical class A was a CCSink following the amp below . Turning from CCSink to CCS might be of some importance .
 
Status
Not open for further replies.