Bob Cordell's Power amplifier book

[jacco vermeulen]

Although some have built BJT amplifiers with such low RE (and I know some who have had to back off to a higher value)

Say Ha-lo to 0R15

 

[PMA]

Poor design of temperature sensing circuitry, should be redesigned to higher dVcorr/dT

(in the particular case you have mentioned )

 

[jacco vermeulen]

Be nice!

I thought i was, think i made a serious attempt to explain to you that going to college is not the divine solution, with a suiting example.
Mr Cordell's bookie covers a lot of stuff that benefits the casual amp builder, and i haven't even finished the brick yet.

(a lot of forum gents (can) get nasty at times, while in the world of reality they're gentle bees. An occasional diyA member is "risque" outside of the web, but tries to make an effort to be friendly/civilised in the arena of words)

 

[mnemneth]

Hi Sandy,
The answer to your question is mainly that the connection of the collector to ground prevents Miller effect feedback from the VAS output node back to the VAS input node. Such feedback is nonlinear as a result of the voltage dependence of a transistor's collector-base capacitance.

Cheers,
Bob

Hi Bob.
I exited speedreading mode and found the explanation in the book.

The mention of grounding the Collector of the primary transitor is in Para 1 on page 60, with explanation to follow.

The explanation does, indeed, follow in Para 3 on page 61 but where I lost the connection was that the primary transistor was now designated as a "buffer", which is of course correct.

My poor old brain didn't accurately connect the two paragraphs.

Sandy
P.S. workplace shut due to snow again today, so more reading for me.

 

[forr]

Hi forr,

I did struggle with the class B vs. class AB semantics issue. The best that can be said is that there is a very grey transitional region of bias between class B and class AB in any real-world design. This is largely what you pointed out.

The way that I arrived at my decision is that I consider that a stage is called class B if it is somewhat under-biased and that it is class AB if it is approximately optimally biased. When an amplifier is biased into substantial gm doubling on purpose, I think of that as what I would call class AAB.

My reasoning for the choice of terminalogy goes back to the tube days. With tubes, there is also a gradual transition in bias. Class B tube amps were generally referred to as rather under-biased designs. Most quality push-pull tube amps were referred to as class AB, so I believe that a properly biased solid state amplifier should also retain that designation. A class B tube amp generally did not have zero bias current, but had significantly higher distortion than a properly biased tube design. BTW, gm doubling generally cannot occur in a vacuum tube design (nor a MOSFET design).

As you know, even an optimally-biased BJT output stage has a class A region, albeit fairly small. This region gets bigger for amplifiers that use smaller emitter resistors and multiple output pairs. The edge of this region can roughly be defined as the point where the peak output current is twice the quiescent bias current.

Hi Bob,

Thanks for your reply.
This clarifies the topic and is certainly a good way to teach amp classification without falling in its ambiguities emanating from historical reasons.

To summarize :

As real class B of precise 180° conduction is impossible to achieve, class B in real life describes any output stage wich has conduction relying above 180° and [180°+minimum of °] to achieve optimal results.

These last results are attained for Optimally Biased class B amps wich then could be called "OBB" (not very elegant) or AB (as it is more frequently done).
However it would be better to see AB as an optimal mid position between class B and class A than as a combination of class B and class A which may not be entirely adequate in case of bipolars.

Then comes your proposal for AAB for describing output stages with real gm doubling due to overbias. This is new and should be wellcome.

Let's do not forget to mention class C with less than 180% of conduction and then having a central dead zone. They can't used in hi-fi audio without the assistance of a parallel class A stage, the famous example being the Quad Current Dumping circuit. I mention it because some persons persist to see its high power transistors as biased in class B.

I hope this will help to avoid confusion in the mind of newbies.

Regards.

Forr

 

[Bob Cordell]

Yes but my point was that higher temperature raises the current gain, which reduces the base stopper current.

This is the correct way of looking at it. Moreover, large base stopper resistors particularly pose a problem in big amps where many pairs of output transistors are in paralllel. If one transistor starts out with a little higher beta, it will dissipate more, get hotter, and its beta will rise, leading to less current flow in its base stopper resistor, higher Vbe and still higher current, while all the others are just sitting there fat dumb and happy. Local thermal runaway of that one transistor may then occur.

Cheers,
Bob

 

[richie00boy]

Sorry and I accept my maths might not be the best (actually, I don't really have any choice..) but the quiescent current is the same in EVERY one of the 18 output transistors, hence 4.68A TOTAL for the 18 off and the huge quiescent power consumption.

Remember that the output transistor pairs operate from different rails.

Sandy

Yes the same in all 18 devices, of which half are connected in series to the other half. Current sourced from the positive rail and sinked into the negative rail, a total of 2.34A. There is nowhere else for the current to flow, bar a tiny amount through the feedback resistor.

 

[mnemneth]

Yes the same in all 18 devices, of which half are connected in series to the other half. Current sourced from the positive rail and sinked into the negative rail, a total of 2.34A. There is nowhere else for the current to flow, bar a tiny amount through the feedback resistor.

Duh!!! My brain hurts.

You are, of course, absolutely correct.

I am now wearing my dunces hat, sitting in the corner and sucking my thumb.

Sandy

 

[AndrewT]

put it this way:
Where in the quiescent state operation do you find any part of the circuit passing 4.68A?

Now follow up with:
Where in the same quiescent state operation can you find some parts of the circuit passing 2.34A?

Two resistors in series pass the same current.
Let's put in two 1r0 resistors and apply a voltage of 4.68V.
The current coming out of the voltage source is 2.34A.
The current passing through the link between the two resistors is 2.34A.
The current going back into the other end of the voltage source is 2.34A.
Using 4 series parallel connected 2r0 resistors gives the same 2.34A of current.
The only difference now is that each 2r0 resistor passes 1.17A.

 

[godfrey]

Maybe you guys can agree that total dissipation = 4.68A * 55V = 2.34A * 110V = about 250W. Different ways to calculate it, but same answer.

 

[mnemneth]

put it this way:
Where in the quiescent state operation do you find any part of the circuit passing 4.68A?

Now follow up with:
Where in the same quiescent state operation can you find some parts of the circuit passing 2.34A?

Two resistors in series pass the same current.
Let's put in two 1r0 resistors and apply a voltage of 4.68V.
The current coming out of the voltage source is 2.34A.
The current passing through the link between the two resistors is 2.34A.
The current going back into the other end of the voltage source is 2.34A.
Using 4 series parallel connected 2r0 resistors gives the same 2.34A of current.
The only difference now is that each 2r0 resistor passes 1.17A.

Aye!

I am now scarfing down humble pie in copious quantity. I'v got used to the taste after all these years and find it quite palatable now.

Maybe you guys can agree that total dissipation = 4.68A * 55V = 2.34A * 110V = about 250W. Different ways to calculate it, but same answer.

Yes, I did get the correct answer, Re. dissipation, albeit by the wrong method.

So 0R33 or 0R47 values for Re appears to be the choice.

I'll also be modding the IPS for much higher value of degeneration, as per Bob's example amps.

Sandy

 

[AndrewT]

Mn,
no need for any humble pie.
We all start out as beginners.
We all make mistakes during the journey.

Some concepts are quite difficult to comprehend, particularly if one tries to learn by remote control (via the Internet).
It takes time and a lot of mind games to get your head around something you can't see.

 

[PMA]

So 0R33 or 0R47 values for Re appears to be the choice.

Poor linearity of the output stage with such values, should be no more than 0R22.

 

[mnemneth]

Mn,
no need for any humble pie.
We all start out as beginners.
We all make mistakes during the journey.

Some concepts are quite difficult to comprehend, particularly if one tries to learn by remote control (via the Internet).
It takes time and a lot of mind games to get your head around something you can't see.

BIGINNER!!, BEGINNER!!

I've been an electronics designer/prototyper for donkey's years.

I'm afraid this was a simple senile (senior) moment and me brain went out to lunch.

To be fair, it's primarily digital stuff I've done over the years with one of my faves being a huge module for a timed Snooker/Pool game (Timeframe) that unfortunately never caught on.

This is, admittedly, my initial steps into the realm of analogue and not even for my own benefit as an amp with the output capability I'm looking at would blow the windows out of my little flat.

Poor linearity of the output stage with such values, should be no more than 0R22.

Oh dear!

A poster comes in at this late stage in the thread where everyone has been singing the praises of the Audio Guru's new book and critiques his preference for higher Re values.

I, too, was an advocate of lower Re values untill I read this Auguste tome and took on board the reasons for using higher values of Re and if you go back a few posts and note my calculated quiescent dissipation horror story, you will understand the logic of using 0R33 or 0R47 for Re.

Now, after this blatent piece of brown nosing, I'm off to the bathroom to wipe me beak.

Sandy.
P.S. PMA, please note that all comments are made with tongue firmly in cheek.

 

[AndrewT]

...........preference for higher Re values.............

Aksa promotes Re=0r47 as being an optimum for sound quality in his designs.

 

[WuYit]

Howdy,
I recently gave my view on this in the other thread:
The bipolar transistor output stage "optimal" bias theory in conjunction with Gm-doubling distortion constitutes just one of many perplexing findings made by D. Self. Since Gm-doubling distortion appears in the class A overlap region, he prescribes a low operating point. Gm-doubling distortion is subjectively mild compared to the highly objectionable switching and crossover distortion (Gm-doubling distortion is part of the crossover distortion). Rather the scanty Gm near cutoff that causes substantial distortion, particularly low level input signals being severely distorted or not amplified at all. A low operating point makes Gm vary wildly with current and temperature. Also, the load resistance limits the transconductance.
Transition characteristics are device specific, why would a certain voltage drop on emitter resistor generally signify an "optimal" quiescent current?
Output stages should be class A biased for normal listening levels.

 

[Bob Cordell]

Poor linearity of the output stage with such values, should be no more than 0R22.

Hi PMA,

Bear in mind that it is all a matter of degree, and to first order, the linearity depends on the total bias current when each output pair uses an optimal bias combination of bias current and RE. I like the value 0.22 for RE also. However, two output pairs, each with RE-0.44 and each with half the quiescent bias current, will yield about the same amount of linearity in principle.

This basically means that as you increase the number of output pairs, you are not obligated to increase the total quiescent bias current if you are willing to have the same nonlinearity as you would have with just a single pair.

Of course, if you want the best linearity, regardless of idle dissipation, than you may use RE=0.22 for each of many output pairs and accept the larger resulting idle dissipation.

Cheers,
Bob

 

[Bob Cordell]

BIGINNER!!, BEGINNER!!

I've been an electronics designer/prototyper for donkey's years.

I'm afraid this was a simple senile (senior) moment and me brain went out to lunch.

To be fair, it's primarily digital stuff I've done over the years with one of my faves being a huge module for a timed Snooker/Pool game (Timeframe) that unfortunately never caught on.

This is, admittedly, my initial steps into the realm of analogue and not even for my own benefit as an amp with the output capability I'm looking at would blow the windows out of my little flat.



Oh dear!

A poster comes in at this late stage in the thread where everyone has been singing the praises of the Audio Guru's new book and critiques his preference for higher Re values.

I, too, was an advocate of lower Re values untill I read this Auguste tome and took on board the reasons for using higher values of Re and if you go back a few posts and note my calculated quiescent dissipation horror story, you will understand the logic of using 0R33 or 0R47 for Re.

Now, after this blatent piece of brown nosing, I'm off to the bathroom to wipe me beak.

Sandy.
P.S. PMA, please note that all comments are made with tongue firmly in cheek.

Hi Sandy,

Do bear in mind that I am not an advocate of high RE values. Indeed, my favorite value for RE is 0.22, especially in amps with only one output pair. My main point was that in amplifiers with more output pairs, you do have the freedom to exchange sound quality for heat while still adhering to the class-AB optimum value of RE.

Cheers,
Bob

 

[danielwritesbac]

Hi Sandy,

Do bear in mind that I am not an advocate of high RE values. Indeed, my favorite value for RE is 0.22, especially in amps with only one output pair. My main point was that in amplifiers with more output pairs, you do have the freedom to exchange sound quality for heat while still adhering to the class-AB optimum value of RE.

Cheers,
Bob

And, thus the question that I couldn't phrase to ask, has already been answered. Thank you.

But, that made another question. So, when isn't lesser heat good news?

 

[PMA]

This basically means that as you increase the number of output pairs, you are not obligated to increase the total quiescent bias current if you are willing to have the same nonlinearity as you would have with just a single pair.

Of course, if you want the best linearity, regardless of idle dissipation, than you may use RE=0.22 for each of many output pairs and accept the larger resulting idle dissipation.

Cheers,
Bob

Hi Bob,

that's what I like

More output pairs, 0R15 - 0R22 each, each optimally biased. Then I get idle current of some 0.45A with four pairs. Of course, large heatsinks needed, but a lot of musical signal remains in class A under usual room listening level conditions and this is what counts.

Regards,