[DVV]

I forgot. I mentioned switching noise, so here's probably where Dan will kick in to expound the virtues of Motorola power devices. Hugh and I will call him a rogue and challange him for tomorrow at dawn, humming in a memorable duet of the virtues of Toshiba devices and their brave deeds.

But truth be told, Dan is right in stating that Motorola power transistors have an unusually low switching noise floor, typically far better than most, and better than Hugh's and my Toshibas. On the other hand, Hugh's and my Toshibas have shorter switching times (Ton and Toff), a wider bandwidth (not too important as output devices because they work in pure current gain mode, i.e. their voltage gain unity), but an incomparable gain "curve", which is much more like a straight line than a curve (highly desirable).

I quote this to illustrate what one faces when designing. Many compromises, and tubes are no exception, just a different set of compromises.

Cheers,
DVV

[AKSA]

Dejan,

Very interesting.  You chaps are very erudite on amplifier technology;  I doffs me cap!  I was always very drawn to the extreme linearity of the Toshibas, and virtually ignored the switching noise issues which I manage to eliminate with some circuit trickery.

I agree with the overall judgement;  tubes add an organic, flowing, human sound to music.  Transistors add slam, pace, dynamics and resolution.  But I would exhort people to think about hybrids;  with proper design, they combine all these things, and the best option is tube pre and SS power......

Cheers,

Hugh

[blizzard]

Hi Dejan,
  Now thst's an answer!  The fog is almost lifted.  One more clarification, and I think we're all set.
  You stated:

Very simple, Steve. In class A, the devices are always in their ON state, they conduct their full current all of the time irrespective of whether there's any signal at all or not. Since they are operating on a constantly open basis, and do not switch

  Let me know if these statements I am going to make are correct.  Or, if I need further reprogramming.
  In class A the device is conducting full at all times.  At no signal this current is all DC and through some feedback circuitry it gets cancelled out.  It also produces heat.  When it gets an audio signal, some to most of the DC current becomes AC current which is directed to the speakers.  The residual DC gets cancelled out in the feedback.
  In class A/B it's pretty much the same thing except for the fact that at idle it does not conduct full current.  So, when a signal is given to produce an AC current (to the speakers), which is greater than the DC idle current, the transitor has to allow more current to flow than was flowing at idle.  This is what people refer to as turning on.  Even though it was already flowing idle current, which I would call being on.
  Here is the crux of the whole thing.  I think my question is and always has been largely that of semantics.  When it is said that a transistor that has a bias current is not conducting, it really is conducting.  It's conducting its bias current.  It's just not conducting the full current that is required when it receives the signal.  When it gets the signal it has to increase the current flow from idle to whatever is needed for output.  
  So, that delay (in class A/B) which is causing the crossover distortion, is caused by the transition from idle conduction current to full signal coduction current.  In class A, the transition is still there, but being that idle current is never less than ful signal current, there is never a lag while the transistor is calling for more current.
  Fewwwwwwwwwwww!  Do I have it now?  Even if my explanation was as convoluted as all hell.
          Thank a million Dejan -- that was exhausting,
              Steve

[DVV]

Hi Dejan,
  Now thst's an answer!  The fog is almost lifted.  One more clarification, and I think we're all set.
  You stated:

Very simple, Steve. In class A, the devices are always in their ON state, they conduct their full current all of the time irrespective of whether there's any signal at all or not. Since they are operating on a constantly open basis, and do not switch

  Let me know if these statements I am going to make are correct.  Or, if I need further reprogramming.
  In class A the device is conducting full at all times.  At no signal this current is all DC and through some feedback circuitry it gets cancelled out.  It also produces heat.  When it gets an audio signal, some to most of the DC current becomes AC current which is directed to the speakers.  The residual DC gets cancelled out in the feedback.
  In class A/B it's pretty much the same thing except for the fact that at idle it does not conduct full current.  So, when a signal is given to produce an AC current (to the speakers), which is greater than the DC idle current, the transitor has to allow more current to flow than was flowing at idle.  This is what people refer to as turning on.  Even though it was already flowing idle current, which I would call being on.
  Here is the crux of the whole thing.  I think my question is and always has been largely that of semantics.  When it is said that a transistor that has a bias current is not conducting, it really is conducting.  It's conducting its bias current.  It's just not conducting the full current that is required when it receives the signal.  When it gets the signal it has to increase the current flow from idle to whatever is needed for output.  
  So, that delay (in class A/B) which is causing the crossover distortion, is caused by the transition from idle conduction current to full signal coduction current.  In class A, the transition is still there, but being that idle current is never less than ful signal current, there is never a lag while the transistor is calling for more current.
  Fewwwwwwwwwwww!  Do I have it now?  Even if my explanation was as convoluted as all hell.
          Thank a million Dejan -- that was exhausting,
              Steve

That's it. THERE IS NO SUCH THING AS A PURE CLASS B! A "classB" amp is actually working in class AB, but its quiescent current is very small, in commercial units typically 20...50 mA. I have seen a top line Yamaha power amp with multiple output trannies conducting just 50 mA, whereas if I had made it, it would have been conducting say 600 mA.

In both cases, it's working in class AB, in my case it would simply have been pushed more towards class A, and as a consequence, heated up and less prone to temperature variations if things got tough, and would have produced much less crossover distortion.

Another small logical paradox. Pure class A dissipates much heat at zero signal, but actually reduces dissipated heat as its power output rises, because more current is being sent to the speakers and thus less is left over to turn into useless heat.

Cheers,
DVV

[JohnR]

Tie up any two transistors or tubes, and you will have an amplification factor of at least (20x20) 400:1, even up to several million to one.

On the tubes, a bit of an over-generalization, mate, I guess you're not too familiar with the tube databook, eh? Not to mention that output transformers don't exactly have a 1:1 turns ratio.

We'll let you off, not your field is it

JohnR

[blizzard]

Well Dejan,
  I guess that's it.  
  Next question:  Quantum mechanics and string theory.  Should amp designers account for this in their designs?

  I'm kidding of course.  Everyone knows string theory can only be applied to speaker designs!
    Thanks for all your help Dejan,
            Steve

[DVV]

Tie up any two transistors or tubes, and you will have an amplification factor of at least (20x20) 400:1, even up to several million to one.

On the tubes, a bit of an over-generalization, mate, I guess you're not too familiar with the tube databook, eh? Not to mention that output transformers don't exactly have a 1:1 turns ratio.

Never was, John. Ever since I got into electronics, it's been transistors for me, never tubes. In all honestly, I know a bit about how they work, but regarding actual data sheets and performance, I don't have the foggiest idea.

We'll let you off, not your field is it

JohnR[/quote]

Whew, that's a load off my chest! I thought I was looking at say 10 years down the river.

But seriously, although I know very little of tubes, I have never discounted for them, much less elimintaed them from my field of interest. I have heard some Audio Research and VTL stuff which quite frankly impressed me with their sound quality.

Cheers,
DVV

[DVV]

Well Dejan,
  I guess that's it.  
  Next question:  Quantum mechanics and string theory.  Should amp designers account for this in their designs?

Nah, it's time for introducing the shishkebab into the whole deal.

  I'm kidding of course.  Everyone knows string theory can only be applied to speaker designs!
    Thanks for all your help Dejan,
            Steve

Not at all, Steve, that's what this forum is for.

Cheers,
DVV

[Raj]

Hi,

Are there any side effects to having a wide frequency response (upto 200khz) and beyond?


Thanks
Raj

[Dan Banquer]

About 12 years ago when I started to get into this Motorola had the best reputation in the States for discreet bipolar. What I have found is consistency in my production runs, which is a big help. What I have also found is that most of  spec's are on the conservative side. To tell you the truth, Hugh, Dejan, I wouldn't want to do the LNPA 150 without Motorola, and to take the point even further I wouldn't want to do any discreet design without Motorola. My old Linestage pre amp was all Motorola also.
Their Voltage regulators, even though they maybe specified for 2% are typically below 1%. I have also observed other brands of regulators with much more noise than the Motorolas.
Quite frankly; I think that quality goes unrecognized and under appreciated in this business.

[DVV]

Hi,

Are there any side effects to having a wide frequency response (upto 200khz) and beyond?


Thanks
Raj

There are many, Raj, some wanted, some unwanted.

Wanted:
1. Frequency linearity 20 Hz - 20 kHz;
2. Lower phase shift;
3. faster response (shorter rise time).

Unwanted:
1. Possible instability, even oscillation if not well done;
2. Ringing (longer settling time);
3. Possibility of high frequency overload;
4. Danger of RF breakthrough.

This is just an off-hand list, there's more. It depends greatly on HOW was that frequency range achieved, with lower gain and hence more local feedback requiring less overall feedback, or by violating the amp with much global feedback, what are your power supplies like (more or less linear, smaller or greater, lower or higher quality, etc) and which semiconductors you use in your circuits.

People like Goldmund os Switzerland normally sell power amps with a response out to 3 MHz, whereas the British school likes to cut it at below 80 kHz (but is now forced to over 100 kHz by 24/96 sources).

Cheers,
DVV

[DVV]

About 12 years ago when I started to get into this Motorola had the best reputation in the States for discreet bipolar. What I have found is consistency in my production runs, which is a big help. What I have also found is that most of  spec's are on the conservative side. To tell you the truth, Hugh, Dejan, I wouldn't want to do the LNPA 150 without Motorola, and to take the point even further I wouldn't want to do any discreet design without Motorola. My old Linestage pre amp was all Motorola also.
Their Voltage regulators, even though they maybe specified for 2% are typically below 1%. I have also observed other brands of regulators with much more noise than the Motorolas.
Quite frankly; I think that quality goes unrecognized and under appreciated in this business.

I agree with that, I have also found them to be very consistent. However, the same could be said of Toshiba, which has also never let me down in consistency, their data sheets usually reflect the worst case which I have yet to see in real life.

But Dan, it's also a fact that I find equivalent transistors made by others to be better than Motorola's. Admittedly, this applies to European BC and BD types. For example, BC 546B as made by Philips is somewhat better performing than Motorola's version, and BC 414/416 (NPN/PNP very low noise types) by Siemens lead the pack, though none falls below the nominal ratings.

And SGS-Thomson, the Italo-French conglomerate, makes better Texas Instruments licenced transistors than Texas Instruments itself, and by no small margin. I refer to TIP 35/36C, where TI rates them at 40 amps impulse, while SGS rates them at 50 amps - and they will survive surges of 50 amps, I tried.

Cheers,
DVV

[DVV]

I forgot to mention Dan's comment on Motorola's unrecognized quality.

Like it or not, the audio business is highly succeptible to fashon. Right about now, Sanken power devices are "in", Motorola power devices are not "in". People blindly read data sheets most can't interpret and certainly don't have enough background knowledge to understand, and look for better figures.

A wonderful example of this is National Semocinductor's LM61xx family; people saw the 3,000V/uS slew rate, and presto, it was the king of the road overnight. The fact that it doesn't really sound like much is of no importance, nor have most ever tried anything but that op amp. It usually replaces the very mundane op amps, and people fail to see that just about anything is better than what they originally got (as with NJR 4558, for example).

Cheers,
DVV

[DVV]

This is just to let interested parties know that on my site, http://www.zero-distortion.com, in the download section, you can find quite a few schematics of well and little known audio products, several very interesting technical texts, and a number of data sheets for better known transistors and op amps.

Free for all.

Cheers,
DVV

[AKSA]

Dejan,

Thank you for your services to our community, and your continuing informative posts on the technology.  I learn a lot!  

Cheers,

Hugh

[DVV]

Dejan,

Thank you for your services to our community, and your continuing informative posts on the technology.  I learn a lot!  

Cheers,

Hugh

I am old school - I have been taught that knowledge becomes truly meaningful only when shared.

Perhaps altruistically, but I feel that if my finding a small gig somewhere is all that differentiates me from others is a poor differentiation indeed. So I share it, and go on to the next small gig. My last was a DC/overheat circuit of wonderful simplicity, yet full functionality.

Cheers,
DVV

[MaxCast]

I second the thanks to Dan B and DVV for their many posts to this thread and the whole board.

I don't know if it was covered before (this is a long thread ), but could someone explain the Stasis design?  What makes it so expensive?  Is it any good?  I've seen it in Threshold, Nakamichi, Sony gear.  Is it still used today?
Thanks,

[DVV]

I second the thanks to Dan B and DVV for their many posts to this thread and the whole board.

I don't know if it was covered before (this is a long thread ), but could someone explain the Stasis design?  What makes it so expensive?  Is it any good?  I've seen it in Threshold, Nakamichi, Sony gear.  Is it still used today?
Thanks,

Boys, boys! Stop it. We're all here because we want to, and we will never have any more than we put in. I just happen to have a little more to put in, that's all, and if you want me to say it, OK, I'll say it - I enjoyed every minute of it.

Right, with the formalities out of the way, we can rock now.

To answer your question, Max - Statis is a word denoting a constant state of affairs, an unchanging state (presumably in the positive meaning of the word). Essentially, this is a circuit topology developed by Nelson Pass which keeps the amplifier in something very near its optimum state given the variations of the signal. In other words, it is kept in a changing mode, but in such a way that it anticipates the signal, and by the time the signal gets to a part of the circuit, it is wide open and ready to pass it on with a minimum of distortion and almost zero delay.

This is superficially a form of sliding bias, variable bias, "Sustained Plateau Biasing" (Krell's version), etc. Please note the word "superficially", because while similar, it is not quite the same, just as Nelson Pass is not just another amp designer (you can't expect straightforward variations on a theme from people like Nelson, he is too good for just that).

In cicruit terms, this type of biasing uses an ultra fast op amp, like for example LM6171 (3,000V/uS) to open up the bias spreading transistor before the signal gets to the output stage. It is often forgotten that a signal must pass through some circuitry, and this is called propagation delay. It is an infinitesimal time interval, say 3-4 microseconds, but that op amp will act in less than 1 microsecond, so when the signal reaches the bias transistor, it will have already been open wide, kicking the class AB output stage into pure class A. As such, it will kick the signal out to the speakers. But it will revert to class AB as soon as the signal level goes down, so you are spared the usual penalities of pure class A, such as excessive heat and the consequent enormous heat sinks and incredible power supplies, and hence much of the pure class A cost as well.

Neslon is a bit different insofar that he expands on the theme by adding some local feedback for the output stage as well, all in one go, which then allows him to do away with gloabl feedback altogether and still have reasonably low distortion (inaudible, actually).

In my opinion, Nelson's system is the best around bar none. It's the most logical one around.

Cheers,
DVV

[MaxCast]

Thanks for the info, DVV.  Have a selection from the Canarvel Wagon on me.

[Dan Banquer]

Actually sliding bias does not have to work quite the way DVV describes. The topology that I am presently using in the LNPA 150 does not use an op amp but has discreet circuitry that "feeds" the bias control transistor. The bias will change as function of signal voltage. At idle and up to approximately one watt the DC bias is pretty much constant. As it starts to move above the one watt the DC bias will increase. Propagation delay for the amp is a constant 1 microsecond to the best I have been able to measure.
Mr. Audioengr raised a concern earlier in this thread about parasitic capacitance on P.C. Boards and power amps. I have not found this to be at all applicable to date. The capacitance that I have encountered is due to the "Miller Effect: on the class a drivers which represent the second stage of the amplifier.