Cyril,

The Wikipedia presentation seems well organized,

and the bibliograhy is overwhelming ! 

I look forward to the unfolding of this interesting topic, 

which can serve to better understand the later OPA's.

Glen,

Thanks for the responsiveness and goodwill ... both they are very important for such a discussion where the most important circuit ideas are... and not so much our moods...

Just to clarify that the link above points to an older version of the article, which I edited on June 25, 2011. Now the current page is probably quite different:

https://en.wikipedia.org/wiki/Operational_amplifier#Internal_circuitry_of_741-type_op-amp

To be honest, I have to say that I have not read them since then... but now I would do it if there is an interest in this topic...

Though we've got much more sophisticated opamps today, the 741 is still useful - in applications and to 'get the bascis'.

If you are not aware of, there is even a discrete 'XL' variant available:

http://shop.evilmadscientist.com/productsmenu/762

Absolutely great if you want to fiddle with components, values etc., trying out the latest and greatest transistors.

The robust and durable OPA, which is suitable for any beginner; it can withstand almost anything :-)

U.Dreher,

Thank you for the URL to the "EvilMadScientist"    XL-741 . 

Cyril, 

The venerable LM-741 is not so fearsome a chip  

making no great claims for performance.

As Josef says it is "robust and durable" just what a student needs. 

I soldered  8   LM741 together , as a stack,  long ago,

just to measure the performance.

Dreher,

It is a very interesting initiative to make a discrete version of an integrated op-amp in these "simulation times"... Maybe only the Susan's desire to build a real K2W tube op-amp can rival it...

I also remember that in 80's, I had an idea about an "educational op-amp" with multiple internal points drawn out of the package.

By the way, I use 741 in labs to show to students the impact of relatively big input bias currents (I make them connect high ohmic resistors in series with op-amp inputs)...

.

Susan:

.

If I can help in any way in your quest for hollow

state perfection, please let  me know. I think the

lessons learned in designing CMOS amplifiers

may be of some value; but the big problem here

is that there is no opposite-polarity: type of tube!

.

Nonetheless, it is an interesting question as to

whether one can improve (in ANY measure)

on the K-2W performance, allowing, say, up to

a doubling of power consumption. The heaters

are the worst offenders, but there's a family of

low-power tubes designed for battery radios

which might be useful here, And,  of course

there is no solid justification for using a total

of 600V for the signal path; though one may

be obliged, in the spirit of the venture, to stay

with a +/-100V output.

.

This looks like fun!

.

Barrie

Dear Aparna,

Thanks for the exact request about the CMRR acronym. As you may have noticed, my comments are always directed to the discussion of specific circuit ideas, phenomena and definitions... so let me explain thoroughly what is the essence of this one...

Theoretically, а differential amplifier (what is the input stage of an op-amp) should amplify a differential signal produced by only one floating voltage source (as the battery VIN in the attached picture).

Practically, for some reasons (?), this differential signal is usually formed by applying two separate single-ended signals to the inputs... that are generated by two grounded voltage sources... so the floating difference between them, according to KVL in the input loop, serves as a derivative input differential signal.

So, to produce a differential input signal, we should change the single-ended input voltages in opposite directions - simultaneously increasing the one and decreasing the other, and v.v. The amplifier will amplify this difference... and we will see the amplified signal at the amp output.

Contrary, if we change the input voltages in the same direction - simultaneously increasing or decreasing both, the amplifier must not amplify this so-called common-mode input signal... and we should not see any change at the amp output.

So, in accordance to its name, an ideal differential amplifier should react only to the opposite changing input signals... and not at all to the changing in the same direction input signals. But we work with real amps that yet, although to a lesser extent, react to the latter... and it is interesting to see why...

Now, if you do not like such long explanations... and consider them as something unnecessary and annoying... you can encrypt them only with one ratio like CMRR = Ad/Acm (see as an example the Wikipedia page about it)... and, if you are a (bad) teacher, give only it to poor students to simply memorize it:)

What I wrote above I suffered personally in the late 70's when, being a student in the same Technical University, where I began to teach after my graduation, I tried unsuccessfully to find out what lies behind this acronym...

It would be interesting to see which are the basic factors determining CMRR... and discuss how they do it...

Cyril,

You have "boiled down" the difficult concept

into a clear presentation. 

There is an old proverb which Barrie reminded to me :  

We only "understand" those things

which We are able to "communicate" to others.  

Dear Aparna and Glen,

Obviously, the big specialists do not want to help us... so we, humble human beings:), ourselves should have to try to explain the matter. But perhaps it is better because their conventional explanations probably would not help us so much as we want. Here is my story that you have the freedom to support or not support...

A differential (long-tailed) pair can be thought as consisting of two main parts - a twin part (two transistors with collector resistors), and a common part - an emitter resistor or a current source. Accordingly to these two parts, we can see two main groups of factors that worsen the CMRR (make the differential amplifier react to the common-mode input signal):

The first of them consists of all kinds of differences (nonequivalences) between the parameters of the two legs - e.g., ß, VBE, temperature, etc. This errors will appear if we take a differential output (from both the collectors) as shown in the attached picture...

But usually we prefer to take a single-ended output (only from the one of the collectors). In this case there are two main errors:

• If the voltage of the power supply varies, the collector voltage(s) will vary as well... so the output voltage will be affected...
• If the emitter resistance is not high enough, when both the input voltages simultaneously vary (common mode), the collector current(s) will vary as well, so the output voltage will vary as well...

... and this was the one of the problems of the input stage in the old K2W...

Cyril,

Sub-section inspections is easy to handle. 

This is a conceptual approach ( in my terms ).  

If the functions fit together, then we will develop math to describe. 

It is clear that the ckt requires a stable Vcc/Vee,  ( CVS )

It is clear that the ckt requires a stable constant current source,  ( CCS ) . 

It is obvious that the K2W retains these weaknesses. 

Makes me wonder how the K2W ever survived the proto-type stage. 

Perhaps,

the basic functional building block logic was too good to pass by. 

I need to check the developmental history of the K2W, briefly

...  makes me wonder what type of elaborate supporting circuitry

was required.    

Question :   What is the main channel of thought in this presentation?

Is the OPA to be developed or is the KW2 to be examined ,

in the 'long run'  ?    

Students do not do well

if they are paddling their boats down two rivers at the same time. 

We need to realize early on,

while we form our organization of  your data,  

just which one is the Main topic,

and which others are important supporting topics. 

Are you accustomed to my type of 'follow along'  ? 

Aparna,

My thinking is that Common Mode signals are mixed with primary signals. 

If Common Mode Rejection Ratio is poor,

then the V(out) will contain both. 

If Common Mode Rejection Ratio is high,

then mostly primary signals are contained in the V(out). 

Imagine a voice signal, and 60 Hz hum , on a two-wire cable , 

as the cable travels from the microphone to the amplifiler.

The voice ( primary ) signal is driven on the (+) wire. 

The 60 Hz hum is driven on both equally,

being induced onto the cable 

as the cable travels from the microphone to the amplifiler. 

(1) The 60 Hz hum is Common to Both wires,

equally in the same direction.  

(2) The voice signal is driven on only one wire , in one direction, 

with the other wire as a passive ground for return currents.  

The Differential stages will provide a method

to attenuate the Common Mode Signal,

and ( at the same time ) amplify the primary signal. 

With that in mind, lets move forward, one stage / function at a time 

and we will see how the 741 does this. 

HTH, Glen

'

Ali:

.

This discussion about the innate brilliance of the 741

op amp overlooks the fact that it was the outcome of

a large number of previous steps toward this "good"

solution - so good, in fact, that the design has lasted

right down to the present day.  Bob Widlar's principal 

contribution was seeing how to merge elements into

what I referred to at the time as superintegration, We

both felt that this was the way forward for integration

and area/cost reduction. But we were both wrong. It

is scaling which has brought about the revolution in

the cost-per-function.

.

The Lone Arranger

Barrie,  Thank you. 

"""super-integration""" is a term I have never used,

and you place it in perspective with """scaling""". 

Perspective in the use our language is always useful,

as it can translate into a larger picture of our current study. 

Perhaps, it is seldom that the students that Cyril addresses

have a long historical perspective  of their subjects 

(  they miss so much  by being so very young ).  

You make me wonder just how long it takes to traverse 

the silicon die of a CPU ?   

Did the reduced clock-cycle-count of the R-6502 cpu 

make up for the slower clock speed ? 

Is it true that many OpAmps have transistor counts similar to the 741 ? 

If so, then what might be the reason for this ? 

just a curious apprentice, Glen

Dear colleagues,

Today I had an accident with one of my relative; I will try to join the discussion tomorrow.

Cyril

Dear Cyril and colleagues,

Hope every thing is well with you.

The analog integrated circuits are based on operational amplifiers where the integrated digital circuits are based on the so called dominant logic gate either NAND or NOR gates. In the sense that they are building units of the analog systems and digital systems, respectively. The merits of the integrated operational amplifiers is that, it is built from many transistors and very few resistors in the opposite of hybrid or discrete op amp which is built from few transistors and many resistors.

In the integrated circuit one is limited by the device area. It is so that the  resistors are area consuming especially when their resistance is too high.

The intelligent solution is to substitute the biasing and the load resistors of the amplifiers with current sources and current mirrors. This is the great structural difference to the predecessors.This solution is called active biasing and active loads.

The idea of the active current mirror is a consequence of the differential amplifier  configuration where it is assumed that  the two collector resistors of the differential pair must be equal to achieve the output balance.

So, the current sources, the current mirrors and constant voltage source in addition to the basic amplifier circuits became basic building blocks of the analog circuits.

Best wishes

Dear Abdelhalim, how have you managed to collect as much wisdom in your comment? Thanks for the generalizing findings and conclusions about the implementation of integrated operational amplifiers...

Dear Barrie Gilbert, thanks for the shared experience in creation of these pioneer circuits. Your participation is a real happiness for us...

Dear Aparna and Glen, thanks for the specific questions and posts expressed in a human way...

Each of you has a unique personality... but the combination of all your contributions forms a unique atmosphere in this forum. Thanks you again!

Your Cyril

General discussions are useful... but even more useful would be more concrete discussions about the fundamental ideas behind specific circuit implementations.

So maybe it's time to begin discussing the op-amp internal structure in details?

In my opinion, the ultimate perfect idea can be understood best if we follow, step by step, the trail that led to this idea .... the evolution of the idea... I will try to do it figuratively, in a free and liberated way, to make it understandable to everyone...

Dear Aparna,

When talking about great principles, the kind of devices does not really matter...

Step 0. Imagine that we are, I suppose, somewhere in the early 30's ... and so far we had only the classic common-cathode (lately, common-emitter and common-source) amplifier stage...

This means that, to control the device (tube, BJT or FET), we should apply the input voltage across the device's 2-terminal input. To do it in these configurations, we have grounded the cathode (emitter, source), and applied the input voltage (of the grounded input source) to the grid (base, gate).

As a result of such a connection, when "moving" the grid (base, gate), the cathode (emitter, source) is "immovable"... the entire input voltage is applied to the device input... and the maximum output voltage... and accordingly the maximum gain, is obtained...

(If you want to explain it in a more obscure way, say that there is no negative feedback in this stage:)

I hope you will agree with me that there is nothing interesting in this configuration... it is just obvious...

Step 1. But here the first big idea arises - to put something in the cathode (emitter, source)... to connect it not directly (by a wire) to the ground... but via some additional element...

Now, if the voltage across this element varies when varying the input voltage, it will modify the behavior (gain) of this amplifying stage. For example, in the simplest "resistor case", the device output current will change the voltage drop across the resistor. Thus the device will "move" itself its cathode (emitter, source) when "wiggling" the grid (base, gate) in such a direction so that to decrease the effective input signal.

(Again, if you want to explain it in a obscure way, say that a negative feedback... or "emiitter degeneration"... is applied in this stage:)

Then, since we are curious enough, we begin experimenting with the value of the cathode (emitter, source) resistance. Meantime, we have already realized that the device output voltage VE (across the resistor) tries to follow the input voltage VIN... as though the input voltage VIN = VE is applied across the resistor RE.

So this configuration acts as an active voltage-to-current converter... or simply, as a current source... whose current is converted back to voltage by the passive resistive current-to-voltage converter RC... and the whole combination acts as a voltage-to-voltage converter.

Thus, the more we increase the resistance RE of the element, the more the voltage gain RC/RE decreases... and finally, when RE becomes infinite, the gain should be zero...

Aparna,

I see that you have academic experience in India and Canada, 

that is a good thing. 

-

You appear to be using the same description of Common Mode Signals 

as before, and also using the term CMRR

( Common Mode Rejection Ratio )

in the same sentence, as if they are the same thing. 

Check the WIKI page on these terms.  

You will have a clearer picture

of what great things the OpAmp Differential Amp must do  

to attenuate these trouble-some C.M. signals.   

I found it helpful to review the standard terminology from Wiki

and I Hope that is helpful to you also. 

. still the apprentice, Glen

We can describe this kind of input signal in two ways:

https://en.wikibooks.org/wiki/Circuit_Idea/Walking_along_the_Resistive_Film#Varying_both_V1_and_V2:_a_resistive_summer

Now let me continue with my story...

Step 2. The next great idea is to replace the high-ohmic emitter resistor with a dynamic (nonlinear) current-stable resistor (as they usually say, "current source"). As a result of applying this idea of genius the stage does not amplify at all:( Let's see why...

The "current source" in the emitter will (try to) keep up a constant emitter (collector) current IE = VE/RE = VIN/RE in such a clever manner:

• If we increase the input voltage VIN, the "current source" will increase, in the same extent, its present resistance RE, and v.v.,
• if we decrease the input voltage VIN, the "current source" will decrease, in the same extent, its present resistance RE.

Thus, both the numerator and denominator do not change... the ratio VIN/RE, the emitter and collector current... and the output collector voltage stay constant...

Then why the hell we did it?

The answer is that, in this way, we can control the gain of the stage:

• If we want the stage not to amplify, we insert the "current source" in the emitter, and v.v.,
• if we want the stage to amplify, we remove the "current source" from the emitter and connect the latter to ground

Thus, to make the emitter voltage "soft", we have to insert a constant current source (actully, a current-stable element) and v.v., to make the emitter voltage "hard", we have to insert a constant voltage source (a voltage-stable element).

In other words, we have to switch (commutate) two elements in the emitter - a current-stable (current source) and voltage-stable (voltage source)...

Here the next powerful idea appears - instead to replace the current source with a voltage source, we may just connect the voltage source in parallel to the permanently connected current source. The voltage source will dominate in this combination and will determine the high gain...

https://en.wikipedia.org/wiki/Talk:Emitter-coupled_logic#...from_the_side_of_the_emitter

Step 3. This idea is realized in a brilliant way in the differential (long-tailed) pair where two stages work at a common emitter resistor (or a "current source").

• At the common-mode input signal, each of the parts (including the emitter element) can be considered as a common-emitter stage with emitter degeneration... and both they are connected in parallel. So, when the input voltages simultaneously change... the current does not change... and the gain is almost zero.
• At the differential input signal, the emitter element is shunted ("shorted") by the other part... the current vigorously changes... and the gain is maximum.

It is not so important to keep a constant current in the configurations above having a differential output since when the collector voltages vary their difference will not vary.

But it is crucial in configurations with a single-ended output (when taking the output only from the one collector). In this case, it is absolutely neccessary to insert a "current source" (instead the humble ohmic resistor) in the emitter.

Now we can realize how imperfect the input stage of the legendary K2-W was... Its common-mode gain is significant - 220/220 = 1... i.e., the anode voltage follows the cathode and grid voltages...

 ... even when made by FET:)

Now let's see how the initial idea has evolved into the present 741 circuit solution... to show the connection between the classic differential pair and 741 input stage.

As until now, I suggest to put ourselves in the place of its inventors to understand the meaning of this development.

(Indeed, it is not so easy to do that without the approval and encouragement of the wise and able ... but I am used to this... and will still try to do it... as I have been doing it all my life...)

Step 4. In the name of this noble mission - to improve the classic long-tailed pair, we bravely decide... to turn it upside down!

For this purpose, we join the bases of PNP transistors Q3 and Q4 (instead their emitters) to control them from the side of the emitters... i.e., we build something like a "common-base long-tailed pair".

There are two benefits of this:

• The undesired Miller effect is eliminated
• The voltage levels are shifted downwards
• The reverse Vbe rating is increased (the base-emitter junctions of NPN transistors break down at around 7 V)

But the common-base stages possess low input resistances since the input sources have to ensure the big emitter currents.

Step 5. We remember the "good old emitter follower"... and decide to enormously increase the op-amp input resistance by connecting the NPN emitter followers Q1 and Q2 before the common-base pair Q3 and Q4... 

But how do we realize the clever trick with "moving the emitters" (emitter degeneration) introduced in Step 1?

The classical emitter-coupled pair is biased from the side of the emitters by the constant current source connected there. The series negative feedback (emitter degeneration) forces the transistors to adjust their VBE voltages so that to pass the desired current through their collector-emitter junctions. As a result, the quiescent current is β-independent.

Here, the Q3/Q4 emitters are separated and already used as inputs... so we cannot use them to bias the transistors. Maybe we can use the collectors for this purpose? No, this is a bad idea since they are also separated and cannot be used as a common input for the quiescent current (we cannot control by current the transistor from the side of the collector since it will behave as an opposing "current source").

Step 6. So, we can set the quiescent current only from the side of the bases by connecting the constant current source to their common point... i.e., we have to make a kind of a "base-coupled differential pair"... where the bases should be "movable" (at the common mode) and "immovable" (at the differential mode). To do it, we should introduce some negative feedback again... and it is the next brilliant idea in this circuit solution...

The total quiescent current is mirrored by a current mirror implemented by Q8 and Q9, and the negative feedback is taken from the Q9 collector. Thus it makes the transistors Q1-Q4 adjust their VBE voltages so that to pass the desired quiescent current...

Step 7. Next, to obtain an extremely high gain, we decide to use another brilliant circuit idea named "dynamic load".

According to this idea, the output part Q9 of the current mirror acting as a current source is loaded with another current source (sink) - the output part of the current mirror Q10-Q11. The two transistors Q9 and Q10 are connected with their collectors thus forming a kind of a nonlinear voltage divider with vigorously changing output voltage in the common collector point (extremely high gain)... or, more professionally speaking, Q9 and Q10 form an amplifying stage with dynamic load.

Passing we notice a little clever trick - one resistor of 39 kom sets the current through the input parts of two current mirrors - Q11 (of the lower mirror) and Q12 (of the upper mirror)...

Step 8. We can see the powerful dynamic load idea also in the input differential pair where another but improved current mirror (Q5–Q7) serves as a collector load of Q3 and Q4. The input of the current mirror (Q5 collector) is connected to the left output (Q3 collector) of the differential amplifier and the output of the current mirror (Q6 collector) is connected to the right output of the differential amplifier (Q4 collector). The emitter follower Q7 increases the accuracy of the current mirror by decreasing the amount of signal current required from Q3 to drive the bases of Q5 and Q6.

This current mirror converts the differential current input signal to a single ended voltage signal and enomously increases the voltage gain. Let's see how...

The input current is "copied" from the left to the right leg where the magnitudes of the two input signals are added. Figuratively speaking, the left current is reversed and moved into the right leg where it opposes the right current (Widlar used the same trick in μA702 and μA709).

Step 9. Thus imperceptibly we came to the most interesting part of our trip where we will examine the behavior of the circuit at the two input modes:

At common mode, the input voltages change in the same direction. The negative feedback makes Q3/Q4 base voltage follow (with 2VBE below) the input voltage variations. Now the output part (Q10) of Q10-Q11 current mirror keeps up the common current through Q9/Q8 constant in spite of varying voltage. Q3/Q4 collector currents and accordingly, the output voltage in the middle point between Q4 and Q6, remain unchanged.

Figuratively speaking, when we "move" the Q3/Q4 emitters, the negative feedback "moves" the bases in the same direction... we have the feeling that the bases are "floating"... and practically we do not do anything. Remember this operation of the classic differential pair where we "moved" the bases... but the negative feedback "moved" the emitters in the same direction... and we did nothing.

Another advantage is that this following negative feedback (bootstrapping) increases virtually the effective op-amp common-mode input impedance.

At differential mode, the input voltage sources are connected through two "diode" strings, each of them consisting of two connected in series base-emitter junctions (Q1-Q3 and Q2-Q4), to the common point of Q3/Q4 bases.

So, if the input voltages change slightly in opposite directions, Q3/Q4 bases stay at relatively constant voltage... they are virtually grounded. The common base current does not change as well; it only vigorously steers between Q3/Q4 bases and makes the common quiescent current distribute between Q3/Q4 collectors in the same proportion.

The current mirror inverts Q3 collector current and tries to pass it through Q4. In the middle point between Q4 and Q6, the signal currents (current changes) of Q3 and Q4 are subtracted. In this case (differential input signal), they are equal and opposite. Thus, the difference is twice the individual signal currents (ΔI - (-ΔI) = 2ΔI) and the differential to single ended conversion is completed without gain losses. Since the collectors of Q4 and Q6 appear as high differential resistances to the signal current (as we seen above, Q4 and Q6 behave as current "sources"), the open circuit voltage gain of this stage is very high.

More intuitively, the transistor Q6 can be considered as a duplicate of Q3 and the combination of Q4 and Q6 can be thought as of a varying voltage divider composed of two voltage-controlled nonlinear resistors. For differential input signals, they vigorously change their instant resistances in opposite directions but the total resistance stays constant (like a potentiometer with quickly moving slider). As a result, the current stays constant as well but the voltage at the middle point changes vigorously. As the two resistance changes are equal and opposite, the effective voltage change is twice the individual change.

...........................................................

This was my story about the input stage of the legendary 741 op-amp...

Cyril,

In your STEP 2,  you wrote :

"""

If we increase the input voltage VIN, the "current source" will increase,

in the same extent,

its present resistance RE,      and v.v.,

if we decrease the input voltage VIN, the "current source" will decrease,

in the same extent,

its present resistance RE.

"""

....... Which seems to be a repeat.  

And, 

"""and v.v."""  requires a tag. 

I know you mean "visa versa" , but to 'what?' exactly ?

Edit required ?   

"""

This was my story about the input stage

of the legendary 741 op-amp..."""

Will return tonight to pick up from page 4. 

Glen 

Dear Glen,

Your remark is interesting from a methodological and didactic point of view. Maybe you are right that, in some sense, the second part of my explanation is redundant.

Indeed, in most cases it is sufficient to consider the operation of electronic circuits only when increasing (decreasing) the input signal since the other case is predictable. Exceptions are maybe circuits with hysteresis and reactive elements.

So, in Step 2, I wanted to consider the reaction of the nonlinear resistor connected in the emitter when we try to change the voltage across it. For completeness, I decided to examine both possible cases - when increasing and when decreasing the voltage. Since the latter is opposite to the former, I have used the phrase "vice versa" (v.v.) But if I used it correctly, only you - a native speaker, can judge.

Cyril,

OK.  It is just language 101, to me. 

I need to return this evening, to peruse the last page of posts again,

and with due respect for the design theory you are presenting. 

I am increasingly aware that you have been

around these trees in the electronic forest many times.  

still the apprentice, Glen 

.

It should be Analog 101 to all in this blog ally.

.

The Lone Arranger

.

Barrie, 

You could remind us of your   Dr. Leif article  , beautifully composed, 

about the "The Four D's " . 

It  describes several major characteristics about Analog Design. 

Certainly the first "Durable" applies to the 741 design. 

That first one led to the "Diverse" family of OpAmps. 

All of these OpAmps work with "Dimensional" physical realities. 

And the last "D" is the most important quality of designer

who you would want on your team,

a person who loves "Discovery".

Design in Analog is about solving the mysteries all around us. 

* I find it useful to look at the "Analog 101" level of the 741, 

as a reminder of the Durable place this basic OpAmp has

in our engineering developmental history. 

* Understanding the 741 will help understand

the Diverse family of OpAmps that followed,

each with special qualities. 

( especially this morning  

as the sun did not rise very early for me ).  

Cheers to all ,  Glen

Dear Cyril,

Thank you for your effort to re synthesize the input stage of the operational amplifier 741 demonstrating the new circuit ideas contained in it compared to the hybrid operational amplifiers. Your scenario with its steps is very useful for the people interested to understand the operational amplifiers. I think you fully succeeded to play the role of the inventors as if you were with them. 

However i want to add that  the start point in this circuit may be the biasing current  circuit composed of Q12 , Q11 and the bias resistance 39 KOhm. This current is mirrored by Q10 by a Widlar current mirror to bias the differential input amplifiers through the simple current mirror Q8 and Q9. This type of biasing is different from the classical biasing using resistors.

Thak you again Cyril.

Dear Abdelhalim,

You always join the discussion at the right time:) Thanks for your positive attitude to this kind of historical trip back in time...

Maybe you are right that it is good to start with the biasing since, from a didactic point of view, it is better to introduce the input signal after biasing...

I have only one objection. In my opinion, the differential pair is biased by sinking the biasing current from Q3/Q4 bases.The Q8/Q9 current mirror serves only as a "current sensor" that senses the common Q1/Q2 collector current (Q3/Q4 emitter current)... then reverts its direction... and finally passes it against the Q10 biasing current... thus introducing a negative feedback...

So, if we compare the classical differential stage with this bisare stage in regard to biasing, we can say that:

• The classical differential pair is biased from the side of the joined emitters; it is a "long-emitter-tailed pair"
• The 741 differential pair is biased from the side of the joined bases; so it is a kind of a "long-base-tailed pair":)

If we compare them in regard to the common-mode operation, we can say that:

• As in the classical differential pair the joined emitters follow the common input signals, this pair acts as a "common-mode emitter follower"
• As in the 741 differential pair the joined bases follow the common input signals, this pair acts as a sort of a "common-mode base follower":)

Dear Cyril,

Great! Sticking on the definition of biasing, it  is to set the DC operating point of the device. For a bipolar transistor  strictly speaking IC, VCE , IB.and VBE. The bias arrangement makes the sum of the collector currents of the npn transistors and the sum of the base currents of the pnp transistors is equal to the Widlar current source. Here the base current is the difference between two current sources, the Widlar current source and the mirror current of the the sum of the collector current. The base current itself is not a direct current source.This arrangement may keep both sum of currents constants as the transistors are active and the collector currents are proportional to the base current.

This is my point of view. i agree with you but i wished to add some explanations concerning the role of the current mirror at the collectors of the differential pairs.

I am glad to see that the discussion begins accelerating...

Obviously, the biasing is a key point in this circuit solution; that is why I would like to examine it in more detail. It seems the main question is, "How does the current source Q10 (current mirror Q10/Q11) set the (common) bias current?"

In my opinion, the mechanism of this setting is the same as in the input part of any current mirror - a voltage type (parallel) negative feedback. There it is implemented in the possibly simplest way - by connecting the collector to base (as in the Q10/Q11 current mirror in the attached picture below). As a result of this humble connection (sometimes named "active diode"), when we try to inject/draw current into/from the collector, the transistor adjusts its collector current so that to pass the desired current. Thus the parallel negative feedback does the impossible - it allows controlling the transistor from the side of the collector.

Now imagine that the biasing current is set by connecting a biasing current source in the collector... and we insert an input voltage source in the emitter... and begin to vary its voltage. What will happen? The base will follow the input voltage variations of the emitter... but the collector current will stay unchanged...

https://www.researchgate.net/post/Can_we_reverse_a_BJT_by_injecting_the_input_current_into_the_collector_and_taking_the_base-emitter_voltage_as_an_output#share

Let's now see this mechanism in the 741 input stage. Here Q3 and Q4 transistors as though have two collectors - own (connected to Q5/Q6) and "foreign" (Q1/Q2's collectors). The latter is mirrored by Q8/Q9 and connected to Q3/Q4 bases. As a result, the whole combination (Q3/Q4, Q1/Q2 and Q8/Q9) can be considered as the Q11 arrangement in the picture above.

So when we begin simultaneously changing the input voltages at Q3/Q4 emitters (through Q1/Q2 emitter followers), the Q3/Q4 common base voltage will follow the input voltage variations...  the common collector current and accordingly, the partial Q3/Q4 collector currents, will stay unchanged. As a result, the output voltage of the input stage at the Q4 collector will not change at these common-mode input variations.

For comparison, in the classic long-tailed pair, the biasing current was set from the side of the emmiters by means of a current type (series) negative feedback.

Dear Cyril, thank you for your efforts to explain the current mirror biasing of the differential amplifier of 741. Yes the common mode rejection is accomplished by connecting the common base to the common collector point of  Q9 and Q10. These two transistors act as  current sources which adapts the voltage across the current sources to keep the current constant. It is the same concept for the conventional common emitter differential biased with a constant current source. I think you hinted this previously. In this way the common mode voltage will be absorbed without changing the biasing currents.

Dear Abdelhalim,

You are right when saying "these two transistors act as current sources which adapt the voltage across the current sources to keep the current constant"... but this sentence sounds quite general for me. When I asked this question, my idea was to unravel to the smallest details the mechanism of negative feedback, which makes this circuit solution insensitive to common-mode variations. So let me thoroughly explain it in another way...

We new insight is that we can think of the pair Q2 and Q4 (Q1 and Q3) as of an emitter-couipled differential pair whith two heterogenius transistors in series. The bases of these transistors serve as inputs of this kind of a "differential amplifier". The joined upper Q1+Q2 collectors serve as a common current output... that is reversed by the current mirror Q8+Q9... and then converted by the amplifying stage Q9 with dynamic load Q10 into an output voltage. Thus the common point between Q9 and Q10 collectors serves as an output of this odd "differential amplifier".

Now note that the Q2/Q1 base is the non-inverting input... and the Q4/Q3 base - the inverting input of the "differential amplifier". As the input voltage at terminal 2 (1)  is applied to the non-inverting input (the upper base)... and the output is connected to the inverting input (the lower base), this configuration is nothing else than the well-known voltage follower...

So, when the input voltages at the non-inverting input(s) simultaneously change, the voltage of the inverting input will follow them. For example, if the input voltage increases, the Q1+Q2's and, accordingly, Q8's collector current will try to increase... respectively, the mirror copy - the current through Q9 will also try to increase (Q9 will try to open more). As a result, the output and accordingly the base voltage, will increase to follow the input one.

We can think of this voltage follower as of a kind of a "voltage servo" ...

Now let's consider thoroughly how the common collector bias current is kept equal to the setting Q10 current...

I suggest another fresh viewpoint - to think of the Q9+Q10 stage as of a high gain comparator with two current inputs - Q11 and Q8 input currents, and a voltage output - the common point between Q9 and Q10 collectors. The Q9 input current is voltage controlled... and the Q10 input current is constant. The comparator output is connected to the voltage input controlling the first current input.

When the common collector (biasing) current through Q8 tries to change, the Q9 current will also try to change... and will be not already equal to the setting Q10 current. The output voltage will change to restore the previous biasing current. For example, if the input voltage increases, the Q1+Q2's, Q8's and Q9 collector current will also try to increase. As a result, the comparator output will increase until restores the previous current.

We can think of this arrangement as of a kind of a "current servo"...

Cyril,

I was reading in Wiki,

and found someone who claims to have invented

the label "Long Tailed Pair".  

"""IMO "long-tailed pair" is more colorful and figuaritive;

as though it shows implicitly the circuit structure.""", 

from C...Fantasist.  

This engineer appears, to me, to be from another language,

and is very much at ease with transistor theory,

and very open and responsive to group discussion.   

I followed the logic for many pages,

but the level of understanding that these engineers have

with transistor theory was very much beyond my experience. 

And, that is about where I stand in our RGN discussion.  

Following the logic, not able to contribute,  

but finding it more interesting than I had imagined it would be. 

I have been re-introduced to the differential amp 

in much greater detail, and current mirrors,

and the "long tailed pair" .   This is really good. 

I would not have imagined that controlling CM signals

would be so very complex.  

This analysis is surely a topic for gifted 'thinking' engineers. 

IMHO,

The OpAmp is a system in constant feedback ...  

always seeking equilibrium and always changing. 

That has always been my analysis of the OpAmp. 

A True Null Node at the Inverting-Input is a dream ...

it is the equilibrium that is the system's target. 

Being an old apprentice may be a good place to be in life, 

Glen

To finish this story about the 741 input biasing, we should also consider in detail how, when the setting current changes, this servo system makes the common collector current equal to the input setting current...

Imagine that, for example, we want to increase the biasing current. For this purpose, we decrease the 39 k resistance. As a result, the current through Q11 increases... its VBE and accordingly, Q10 base voltage increases... and Q10 opens more to increase the current through itself... Q3/Q4 base voltage decreases... the difference VBE increases... the common collector current through Q8 increases... its VBE and accordingly, Q9's VBE increases... and Q9 opens more to increase the current through itself... until it becomes (almost) equal to the input setting current...

So, at common mode, the simultaneously changing input voltages disturb the biasing "servo system"... it reacts to this "intervention" and compensates it.

Cyril, 

Just like a really good story teller  

"""until it becomes (almost) equal to the input setting current"""  

you posted a 'hook' to make the reader come back for more !  

 A True Null Node at the Inverting-Input is a dream ...

it is the Equilibrium that is the System's Target.

IMHO,

The OpAmp is a System in constant feedback ...  

always seeking Equilibrium and always Changing. 

That has always been my analysis of the OpAmp. 

Yes Glen, really I (Circuit-fantasist) wrote it in 2009, in connection with the emitter-coupled logic (ECL)...

I fully share your observations about op-amp circuits with negative feedback...

Regarding your surprise that things can be (presented) so complicated ("I would not have imagined that controlling CM signals would be so very complex"), I would say:

The naked truth is that as a rule, they do not explain to us the essence of things... and give us only superficial explanations... The reasons for this are diverse... and I have tried to reveal some of them in a few stories in Wikibooks.

When someone "dives" deeper into the essence of things and bothers explaining them in detail, the explanations become bulky and verbose. This, for some reason irritates people... and they react negatively... or do not react at all...

https://en.wikipedia.org/wiki/Talk%3AEmitter-coupled_logic#Talking_points_in_Explanation_section

https://en.wikibooks.org/wiki/Circuit_Idea/Why_Circuit_Ideas_are_Hidden

In this respect it would be interesting to answer the question, "Which is the 'long tail' of the 741 input differential pair?"

Cyril,

So introduce the "Long-Tail" of the Differential Amp ? 

How is it a productive model of bias ? 

Returning again to the booth of Cyril,

Thank you for your effort to explain the effect of the the common mode voltage on the differential amplifier of the famous 741 op amp. There may be an alternative way to explain this effect. It is the way of simple circuit analysis.

It is so that the common base of the pnp transistors differential pair is connected to the common point of collectors of the current sources Q9 and Q10. Assuming that they are ideal current sources, their source resistance will be infinte.

The common mode voltage applied to the base of the differential npn transistors will be dropped on the forward biased emitter junctions of the npn and pnp transistor differential pairs and the infinite resistance of the theoretically infinite resistance of the ideal current sources of Q9 and Q10. According to the potential division rule, then the common mode voltage will be absorbed  by the current sources which means that the potential of the common base node will follow the common mode voltage. 

So, if the tail is a resistance it must be very long theoretically infinite to have the effect as the current source. Since the tail must be limited to pass the biasing current , the best solution is to substitute the resistive tail by  a current source which satisfy the  required bias current and the very high resistance. 

Best wishes

Dear Glen,

It is surprising that such an interesting discussion can be held even only by two participants. Something more, if there is a good will, such a discussion can be conducted even by only one participant:) I remember I have done this before time by asking questions and answering them myself:)

(Here I have to break my statement because it turns out that though we are not only two... we are not alone:)

----------------------------------------------------------------------

Well, I have read the useful Abdelhalim's remarks and will use them in the next statement about the fancy name...

Now about the figurative name "long-tailed pair"... Imagine we are good teachers who want not only to say but also to illustrate what they teach...

Electrical quantities - voltage, current and resistance, are inaccessible to our senses... they are invisible... So it is a good idea to visualize them... and then to operate with them as geometric objects - moving, stretching, rotating them...

First, we can represent voltages by voltage bars with corresponding length, and currents - by current loops with corresponding thickness...

Similarly, if we use resistors with linear form and linearly distributed resistance, we can visualize their resistance by a bar with length that is proportional to the value of the resistance. I have widely used this technique in the Wikibooks story about the Ohm's law...

So in the emitter-coupled differential pair, the common emitter resistor with high resistance can be figuratively thought as of a long "tail" (see the H&H picture below).

https://en.wikibooks.org/wiki/Circuit_Idea/Philosophy#Visualisation

https://en.wikibooks.org/wiki/Circuit_Idea/Walking_along_the_Resistive_Film

In my opinion, in the 741 input differential stage, such a "tail" is connected to the Q3+Q4 bases... and this "lower tail" is composed by the two current "sources" Q9+Q10 in parallel. It has extremely high resistance... so it is a "long tail"...

But while the classic long-tailed pair can be thought as of a "Y suspender"... this 2-tailed pair can be represented by an "X suspender":)))

But we can see another "upper tail" - the transistor Q8, which has extremely low resistance... so it is a "short tail"...

So, it seems the 741 input differential stage can be thought as of a differential pair with two "tails" - lower (long) and upper (short)... but I cannot find such a kind of suspender to represent it:)

Cyril,

Now, I am beginning to see diagrams that resemble what I visualize.

(1) Once I can grasp the visualization diagrams, real world, 

(2) then I can merge the transistor biasing, differential world,

into my mind. 

(1)

The CM is ONE signal, applied to the two conductors,

then applied to the (+) and (-) differential inputs 

thereby effectively canceling.  

(2) 

The Real Signal (RS) is ONE signal which is referenced to ground.

The RS was originally generated with reference to Ground.  

The RS and the Ground are the two required poles. 

The RS and Ground are applied to the differential inputs

and there IS a measurable Difference Signal to be used thereafter.

These are the two classes of signals

that I would measure with the Oscope

in the real world,

as I deal with my real world project.

The OpAmp system will see these two classes of signals  

from the Differential Amp Perspective.

( I realize that I am anthropomorphizing the OpAmp system ).  

... On The Other Hand  

My target project ( with all its inherent real world problems ) 

can be handled from an investigative and logical and math approach

and make good use of the tool-of-choice "OpAmp"

which is designed to be used in a logical and math approach. 

The OpAmp System is a contained unit ... a "black box".  

This thread is largely about designing the "black box", 

and that requires the real talent of special electrical engineers,

which is why we are following your lectures

and watching your story method unfold 

and morph into something good for your students. 

just the apprentice, Glen

Oh Glenn,

You make me think deeper and deeper on things that I thought I knew... but it turned out that it is not exactly like:)

So we finally went back where we started - "What is a common-mode signal? Is it only one or two?"

You said: "The CM is ONE signal, applied to the two conductors, then applied to the (+) and (-) differential inputs thereby effectively canceling."

Exactly! This is one of the most important manifestations of the common-mode signal. The twisted pair arrangement is an excellent example of this passive compensation technique...

Really, there one disturbing signal is simultaneously "copied" in two wires... and then the two "copies" are subtracted by such a differential amplifier...

(Interesting ... I long ago I was not peeking in this page... and now I see that it is visited by almost 60,000 people... so obviously someone benefits from our "naive" discussions...)

https://www.researchgate.net/post/What_is_the_basic_idea_behind_the_twisted_pair_Why_are_the_two_wires_twisted_How_does_this_arrangement_compensate_undesirable_disturbances

But let's continue to "philosophize" what is this thing called "common-mode signal"... while clearing all ambiguities in this term...

I think that all these problems, requiring a distinction between two types of signals (common and differential) had occurred after they decided to introduce the concept of "ground". If we worked only with a floating source and load - either the one, the other or both, we would not have these problems...

For example, the ordinary battery-supplied voltmeter has only one type of input voltage... it is neither common or differential... it is just a voltage measured between two input terminals... Or, for another example, the humble battery produces neither common or differential voltage... it produces simply a voltage between its two output terminals...

Once entered the requirement of ground, we are forced to use more sophisticated arrangement where the needed (differential) voltage is obtained by subtracting two (single-ended) voltages. So these (common) voltages are auxiliary... they serve only to produce the true (differential) voltage.

The Wheatstone bridge is such a typical system where the true (differential) voltage VOUT is obtained by a serial subtraction, according to KVL, of two grounded (single-ended) voltages VR2 and VR4. If we supply this bridge by a floating source, then we can speak only about one voltage - VOUT...

Now let's return to our favorite device...

I think the problem is that the differential (long-tailed) pair is not a true differential amplifier. If it was, we would be able to connect a floating voltage source (e.g. a battery) between its input terminals.... but we cannot do it so easy...

The differential pair (classic, 741's, etc...) is composed of two single-ended amplifying stages that need grounded input sources... Then the true (differential) input signal is extracted from these partial voltages.

Each of these stages imposes certain restrictions on the input sources (for example, they cannot be too high resistive... or to produce too high voltage...)

Welcome colleagues!

I see that Cyril is talented in demonstrating and qualitatively describing the  operating concepts of the electronic circuits. 

For sake of completeness, i want to analyse the effect of the common mode voltage in op amp 741 from other side. It is so that the total collector current of the npn differential pair is constant and equal to the current source of  Q9 and at the same time the total base current of the pnp differential pair is the difference of the current source of Q10 and Q9. This is irrespective of the common mode voltage. Which means that the bias currents of the transistors are constant at any common mode voltage. It follows that the base to emitter voltage of the two transistors, the npn and the pnp transistor is constant.  This can only valid if  VBnpn - VBpnp = constant . This means that the pnp base voltage must follow the npn base voltage, which is the common mode voltage. 

This insight has the advantage of demonstrating the usefulness of  using the current mirror biasing fo the bases of the pnp transistors and the collectors of the npn tranistor.

Best wishes

A great conclusion, Abdelhalim!!!

Honestly, I was prepared another perspective on the operation of the biasing system, which fully coincides with yours. If you let me, I would expose it here with the risk of repeating?

OK, let's do it... in any case it would be useful...

So far we have explained the output of the biasing system in terms of voltage. But, with the same success, we can do it in terms of currents...

According to this viewpoint, we can think of Q3/Q4 transistors not as of voltage-controlled devices but instead as of current-controlled devices. So the transistors are controlled by the common current sank by their bases...

This current is a small difference between two relatively big currents - the one of them is injected into (by Q9), and the other is sank (by Q10) from the common point between their collectors... and this erroneous current is kept constant... like the erroneous voltage between the op-amp inputs.

So the biasing system keeps up a constant very small difference (Q3/Q4 common base current) between the desired (Q11's) and real (Q8's) current... that is a ß fraction of the common collector current.

I think this topic is closely related to the question about current feedback amplifier (CFA).

https://www.researchgate.net/post/What_is_the_truth_about_the_exotic_current_feedback_amplifier_Is_it_something_new_or_just_a_well_known_old_Is_it_really_a_current_feedback_device

 Cyril, I think your description of the CFA is somewhat "simplified".

Josef

Josef,

Long ago I have not been back to it... but I have nothing against to revive and this discussion:)

Glen,

I am thinking about the nature of the signals applied at the DA inputs... and again going back to my point above...

I will repeat in other words that actually the so-callel "differential amplifier" does not accept the floating voltage across its inputs. It measures two voltages relative to ground... then "calculates" their difference... makes it relative to ground... and sends it to the output...

So, in this arrangement there are two "real signals" relative to ground (single ended common-mode signals). The third (floating differential signal) is derivative of them...

Perhaps you meant to say something like below?

"The Real Signal (RS) is ONE signal which is referenced to ground.

The RS was originally generated with reference to Ground.

The RS and the Ground are the two required poles.

The RS and Ground are applied to the differential inputs

and there IS a measurable Difference Signal to be used thereafter."

If you still want to apply the input voltage directly between the two DA input (e.g., to connect a floating battery as in the attached picture), you have to connect resistors between each of the inputs and ground to close the input loop. Then, the floating input voltage source will pass a current through these resistors... and will create two opposite voltabe drops across them...

But still the differential amplifier will continue accepting not the battery voltage but again the two single voltages in respect to ground...

Cyril, 

"""It measures two voltages relative to ground... then "calculates" their difference... makes it relative to ground... and sends it to the output...""" 

in other words could be like this : 

"""The RS and Ground are applied to the differential inputs

and there IS a measurable Difference Signal to be used thereafter."""

Good correction.  

You are a most flexible thinker, and very exacting. 

I have always operated in a ground-referenced system. 

Did not mean to imply there was some magic to this ! 

The more often I go over these diagrams, 

the more the visual language is bringing my own language into a line. 

Barrie mentioned having developed a "v-mirror".  

Glen,

Saying "The RS and Ground are applied to the differential inputs and there IS a measurable Difference Signal to be used thereafter", you probably mean tha the one terminal of each input source is connected to ground, and the other - to the corresponding op-amp input?

I was expecting some reaction and even disagreement with my suggestion above about where currents flow when a floating voltage source is connected between the op-amp inputs. Let's see why...

Our human intuition leads us to think that, in this arrangement, the input source passees its current through the amp inputs ("blows" it into the inverting input... and "sucks" it from the non-inverting input). So there is something strange to claim that the input current bypasses the amp input (inner loop) and "prefers" to pass through the outer loop (resistors RB1 and RB2).

My explanation is simple. The input current cannot pass through the amp differential input since there are two diodes (base-emitter junctions) that are opposite connected in series... so that always one of them is off... 

For me this paradox is extremely interesting and important for understanding the circuit. That is why, three years ago, I dedicated a special question to it.

https://www.researchgate.net/post/Can_we_connect_a_floating_voltage_source_between_the_two_inputs_of_a_differential_amplifier_If_so_at_which_conditions

Cyril,

I wrote "I have always operated in a ground-referenced system. " 

Meaning that All my electronic project signals

are developed and applied

in a ground-referenced system

on my simple work-bench.   

The concept of a 'floating voltage source'

applied to a Diff OpAmp is interesting.  

Thinking on the 'inside' of the OPA 

requires more thinking,  new thinking. 

Your "link" is a very different 'read' for me.   Thank you. 

At the university, 1980,  

we collected myographic signals from patients,

where the 'ground' would have been the patient, per se.   

That would explain why I had designed/built 10 instrument amps.  

That is where I first encountered Common Mode Signals

in the form of audio from nearby Amplitude Modulated Radio Stations.   

Of course, today, we would make good use

of some of Barrie's ADI instrument amps.  

Glen

Glen,

Thanks for the very informative "letter" about galvanically isolated power systems; it was very interesting for me. It seems it is widely used nowadays in all kinds of switching power supplies for phones, laptops, etc.

BTW, I have heard about a 2-wire power system where the supplied device somehow determines what is the neutral wire and connects the body of the appliance to it. Is this true... or it is just another joke? Have you heard about such a clever technique? If so, how is it implemented?

Cyril,

Have not heard of any such news of a device that automatically detects the type of supplied power.  Not in those words. 

OTOH,

if I wanted one, I would feed the power supply through a full-wave bridge, and then chop at a high frequency, then back down,  in Switcher-Supply style.  Polarity should not be a problem.    

I have several CCTV cameras that accept 12V DC or 15V AC.   The internal operating voltage for the vidicon tubes should be 70V DC or more.  They might follow the method I described.  

Just guessing about the technique to change the unknown incoming power into a circuit standard power.    

But speaking of electrical utility power, many of our hand-electric-drills have been manufactured with "double-insulation" , thus dropping the requirement for a third Green-Ground conductor ( to carry hot-to-ground faults ).  These are supplied with a two-wire cord.  

Then maybe manufacturers can benefit from this idea:)? Just imagine what a large gain by saving the third wire this will be...

By the way, I have seen 2-wire powered hair dryers in some hotel bathrooms... and then I was afraid of electric shock:) Now I realize this was your isolated power supply...

But maybe we should return to our op-amp topic?

I have another insight about the length of the tail in the long-tailed pair...

As we know, the idea of this metaphor is to represent the resistance by a length. So, in the classic differential pair (e.g., in K2-W), where the emitter resistance is high, the tail is long... but with a constant length.

In the differential pair with a current source in the emitter, the emitter resistance is dynamic... and it varies in the common mode. So, the length of the tail should vary as well...

So imagine such an attractive animated experiment where we vary the common input signals... and, at the same time, the tail changes its length... Then we vary them in a differential manner... and the tail does not change at all...

Cyril,

The "double-insulation" is simply another physical layer of plastic.

"Isolated power supply" is a different animal. 

Two different things.  A simple issue to be handled elsewhere. 

So, yes, it is obvious that the Long-Tailed-Pair is dynamic. 

* Logically, under your examination,

it must be dynamic in order to handle the Common Mode signal. 

* Logically, the Differential Signal is handled in the relations of the transistor super-structure (-) and (+) transistors.

(1) A single transistor, handling one signal, has a static tail.

(2) An OpAmp, a Differential stack, pulls power from a dynamic tail.

Back to the Monkey Analogy : 

* The monkey with a tail that coils shorter or un-coils longer 

to handle Common Mode signals.  

* The monkey's tail is static to support the Differential Signal

and allow the monkey to work the Differential with both hands. . 

* summary:

(1) Two hands dynamically working together on Differential Signal, 

(2) a Dynamic Tail for the Common Mode Signal. 

That would be a 'memorable example' for the students. 

So, we can introduce another name for the long-tailed pair with emitter current source - "dynamic-tailed pair":)

Cyril,

Since we are in the midst of fanciful daydreaming, 

("wool-gathering" in Irish slang) :

The Monkey's ""dynamic-tailed pair""  is a more descriptive label.  

I think these are Q5 and Q6. 

The Monkey's Dynamic Hands might be Q1 + Q2  and Q3 + Q4 +Q7. 

and might be newly labeled ""dynamic-headed pair""

there being a "left Q1 hand" and a "right Q2 hand" . 

I would say the "right hand" is the "Major function", 

and the "left hand" is the "Minor function". 

These are some labels more memorable

than anything I ever read in a textbook. 

Already, in my old mind, they are taking shape 

as a "living" differential-structure, called the "Monkey". 

I might remember the Monkey analogy ,

with its   "dynamic-headed pair"   and    "dynamic-tailed pair"  ,

for a long time to come.   

I would hope that a rising student

would see this analogy of 741 Differential Amp

as something having a basic structure

that can be a reference point for all other OpAmps in his future. 

just an old apprentice busy "gathering-wool" ,  

Glen 

Dear Glen,

Thanks for the nice figurative explanations where otherwise dull and dead circuit structures are personalized in such a funny "monkey way"...

But would you mind telling me what is this paradox that the older we become, the more we indulge in our own imagination... and allow ourselves to have fun with such serious things in life that others worship? Perhaps in this way we try to provoke "more serious professionals" to intervene and make to pieces our "monkey approach" in electronics:) ... and so to get a real heated discussion as 741 deserves? Maybe...

Above you have put for consideration the next legendary circuit technique in the 741 internal structure - the more sophisticated dynamic load Q6 included in the Q4 collector. So we can already expect interesting explanations of this clever trick...

Cyril, 

I do hope that a good discussion of 741 will be forthcoming ,

even though it may quickly go above my head.   

Yes, Q6 is next.

Glen,

There is such a wise Bulgarian proverb that, I think, is very suitable for our situation - "Докато умните се наумуват, лудите се налудуват". I have not found it in English... and it is hard to translate... but it means something like, "While clever ponder over what to do, mad manage to do it":)

Then let's do it:) ... to reveal the essence of the sophisticated 741 current mirror dynamic load. I only keep wondering though how this is possible - not to be a circuit designer... not to have access to modern research laboratories and manufacturer's departments... but still manage to reveal the underlying ideas of such intricate circuit solutions... relying only on the natural human common sense, intuition and imagination... Perhaps the role of some innate motivation is crucial here...

Unfortunately, it turned out I have already presented my point of view about this famous technique on Page 5 (steps 8 and 9 of the "741 building scenario")... so I will have to fabricate some other story here:)

Well, let's try to find some extremely simple, clear and striking explanation of what is happening in this part of the circuit. For this purpose, I suggest to model it by the humble voltage divider...

The role of the output part of the differential pair, like each output stage, is to produce voltage. So, if we decide to play this role, we can do it by the ubiquitous voltage divider. In this arrangement, we can change the one resistance and keep constant the other as in the ordinary common-emitter stage. To implement this idea, we connect a varying resistor (rheostat) and a constant resistor in series.