I think so, but Q3 is current driven, not voltage driven. Trying to use resistors to set its current introduces heat inflicted DC drift. If I am interpreting you correctly, you would obstruct the DC feedback. Your centered output point is connected back to a centered differential pair. The centering of the output stage is done by adjusting R1 and R2. Schematic post #29 needs a capacitor between R7 and ground to force the DC levels to be correct while allowing you to dial in any gain you want.

 

thank you, so I'm getting the Aol Acl idea, actually makes my design easier, as i can just aim for good gains at each stage.
why will i have/how can i deal with these saturation issues?

Yes, exactly this makes designer life much easier.

why will i have/how can i deal with these saturation issues?

The max positive saturation is Vcc - (VceQ3 + VbeQ4)
But the negative one will depend on R6 value also.
Iload_max for negative half is equal to:
-ILmax = (0.5Vcc)/( R6/HfeQ5 + Rload) so for Vcc = 20V; R6 = 1K (IcQ3 ≈ 10mA); HfeQ5 = 100 we have
-ILmax = 10V/(10Ω + 4Ω) = 0.714Apeak and this give as VLmax = 4Ω*0.714A = -2.8Vpeak

Amplifier that should never be built in real life




We can fix this by using a bootstrap capacitor. Also this amplifier has many more secrets than you think. Have you consider for example a thermal stability?

 

Jony130 seems to be covering the other end of the amp, and he's right about, "more secrets than you think". This design has been around for decades and is well polished. The differential input stage is a crude form of operational amplifier, which I'm pretty good at. Push-pull outputs...not so much.

 

ok, slowly getting there i think guys, ill have a bit more of a read into bootstrapping and get back to you, its a term i've come across but not dealt with a lot. I'm aware that my thermal stability will not be great, i've opted to look at something without any form of active loading as I'm only just getting my head around a simple circuit like this. hopefully by the time i've got my head around this a bit i can think about revising things and building something a bit more stable.

all i can say is thank you guys ever so much! you have really helped a lot, no doubt i will be back with more stupid questions....
will

 

Do you know anything abut stability of amplifiers?

The amp Jony said should not be built, will oscillate madly at some elevated audio frequency ( perhaps even enough to destroy the output transistors) and needs stabilising by a capacitor between Q3 collector and base or better between Q3 collector and Q2 base to control this.

Perhaps Jony will run these for you, next time he runs his circuit through the simulator?

 

This cap would be separate to that in series with r7? Could it just be placed before r8 in the feedback loop? Im assuming it would have to be a rather large cap...

 

See how I place C1 in this amp
http://forum.allaboutcircuits.com/attachments/6-png.77467/

 

This was as i understood it, this should sort the oscilations at the output pair, but your saying this amp would still be unstable, is this mostly down to thermal variation in q points or am i missing some other dc aspect?

 

This was as i understood it, this should sort the oscilations at the output pair,

C1 do not prevent oscillations , we need C1 because we don't want to amplify any DC voltage present at the input.
To prevent oscillation we need a capacitor between Q3 collector and Q3 base.

but your saying this amp would still be unstable,
is this mostly down to thermal variation in q points or am i missing some other dc aspect?

Make sure that the output stage is thermally stable.
Please read this hole article
http://sound.westhost.com/amp_design.htm

 

To prevent oscillation we need a capacitor between Q3 collector and Q3 base.

But you did not show that in your drawing (post #47) and that is confusing.

Personally, I am about 2 beers short of a six pack on frequency control in feedback amplifiers and I would love to see how this is done.

 

We can go through the details of why each section does what is does if you like, and why I still think you should have started at the output and worked back to the input.

Do you know why you are using a long tail pair at the input?

As regards the stability, no it has nothing to do with temperature, it has to do with phase shift.

A plot of phase v frequency will show that the phase shift through the amp becomes large and more negative as frequency rises.
At some point in the several hundred kHz to 1/2 MHz range the phase whift will be greater than -180 .
At this point there the gain will still be grater than 1

Since the feedback is input to the inverting input of the long tail pair the minus tims minus will make a plus and the feedback will become positive.

The amp will oscillate vigourously at this elevated frequency.

As I said before to prevent this you need to control (roll of or reduce) the upper frequency response so that the gain falls below 1 before the phase shift becomes 180.

I suggest capacitors C4 (about 100p to 220pf) to control the open loop gain and C5 (15pf to 33pf) to control the closed loop gain.

 

C4- is a dominant pole compensation capacitor or miller compensation capacitor. This capacitor has influence on unit gain frequency as well as on slew rate. SR ≈ IcQ1/C4 ≈ 1mA/100pF≈ 10V/μs
But in general we select the size of this capacitor by testing it on the lab.
C5 - together with R8 set -3dB corner frequency , Fc = 0.16/RC = 0.16/(51kΩ * 100pF) = 31Khz

 

This capacitor has influence on unit gain frequency as well as on slew rate. SR ≈ IcQ1/C4 ≈ 1mA/100pF≈ 10V/μs

Which is why I offered an alternative connection in post#45

You know this and I know this.

However the OP want to use building this circuit as a learning exercise, something I have done in the past and greatly approve of.

So my objective here is to help this learning process and stimulate the OP to ask 'Why'

 

Thankyou for your continued input, i'm going to start researching paired transistors for thermal stabilization as soon as possible, i have a couple of assignments i need to hammer out first before i can really get stuck in on the amp design.

As for starting at the input, it was simply a case of knowing my understanding of the long tail pair was poor, so i figured i would try and get my head round this first. So far its just been analysing individual sections of the amp in order to greater understand their purpose and how they should be designed. When i start selecting components and calculating values properly i will attempt work from the output, i still have a long road to go here as this is all new to me. Am i right in thinking that i should design my output, then design the q3 to give collector current equal to the current i want for my desired q point for the output transistors. If this is the case then i at least get why i'm working backwards ^^

I chose to use the long tail pair to as i knew i could create a differential input with feedback this way, although obviously my understanding wasn't great. im sure theres other reasons for it?

your explanation as to phase shift was brilliant, i hadn't considered the implications of frequencies above and beyond the audio spectrum, i had pretty much just written them of thinking "well i'm going to get the bandwidth i need for the audio signal so thats fine." but it makes sense that the transistors will still be doing the work for these frequency ranges and thus i need to consider there effects in the design. The solution seems simple enough for even me to get whats going on first time, a refreshing change!^^

 

Yes, design it backwards.
Why a differential pair? You could do the input stage with a single transistor and apply the negative feedback to the emitter, but the differential pair has less distortion.

 

OK this is a feedback amp.
You have probably seen the simple amp diagram showing feedback add to the input.
The feedback signal has the same phase relationship to the input as the amplifier output, since it is tapped off from the output.
But we want negative feedback so if the signal output from the amp is opposite in phase to the input, all well and good we can add it straight to the amp input.
But if the output is in phase we must invert it or ad it to an inverting input, we cannot just add it into the signal.

The current thinking for this amp is :-

Input the signal to the base of an input transistor (Q1) and take the output of this from Q1 collector after boosting the voltage gain.

Input this signal to Q3 base and gain take the signal from the collector.

Feed this to a pair of emitter followers.

So there are two phase inversions at Q1 and Q3, but none at the ouput emitter follower pair.

So the output signal is in phase (in the pass band) with the input.

As #12 says we would therefore have to feed it back to the emitter of Q1.

The problem with this is that the emitter of Q1 presents a moderate to low impedance to the feedback signal.

Alternatively we can add an extra stage (Q2) and invert the signal again making the output of the opposite phase to the input so we could then feedback to the higher input impedance of the base.

But if we add an extra transistor stage, it would be better used as a long tail pair with the base of Q2 ofering high impedance to the feedback so we can choose relatively high value feedback resistors (we want voltage feedback not current and the output is low impedance developing significant current) we do not then have to change the phase of the output signal since Q2 is of the opposite of phase from Q1. That is an oft forgotten meaning of 'differential pair'.

There are also biasing advantages for Q1 if it is part of a long tail pair, rather than both input and feedback input transistor.

Sorry if this is a bit garbled, but I thought you'd rather have some explanation than none.

 

OK this is a feedback amp.
You have probably seen the simple amp diagram showing feedback add to the input.
The feedback signal has the same phase relationship to the input as the amplifier output, since it is tapped off from the output.
But we want negative feedback so if the signal output from the amp is opposite in phase to the input, all well and good we can add it straight to the amp input.
But if the output is in phase we must invert it or ad it to an inverting input, we cannot just add it into the signal.

The current thinking for this amp is :-

Input the signal to the base of an input transistor (Q1) and take the output of this from Q1 collector after boosting the voltage gain.

Input this signal to Q3 base and gain take the signal from the collector.

Feed this to a pair of emitter followers.

So there are two phase inversions at Q1 and Q3, but none at the ouput emitter follower pair.

So the output signal is in phase (in the pass band) with the input.

As #12 says we would therefore have to feed it back to the emitter of Q1.

The problem with this is that the emitter of Q1 presents a moderate to low impedance to the feedback signal.

Alternatively we can add an extra stage (Q2) and invert the signal again making the output of the opposite phase to the input so we could then feedback to the higher input impedance of the base.

But if we add an extra transistor stage, it would be better used as a long tail pair with the base of Q2 ofering high impedance to the feedback so we can choose relatively high value feedback resistors (we want voltage feedback not current and the output is low impedance developing significant current) we do not then have to change the phase of the output signal since Q2 is of the opposite of phase from Q1. That is an oft forgotten meaning of 'differential pair'.

There are also biasing advantages for Q1 if it is part of a long tail pair, rather than both input and feedback input transistor.

Sorry if this is a bit garbled, but I thought you'd rather have some explanation than none.

OP, I like your "Thread" you've help me find this......... http://www.allaboutcircuits.com/vol_3/chpt_4/11.html

It's going to be my next experiments in my search to understand audio circuits and hopefully on a good path in building "Audio Circuits" by myself.

Since, I'm such noobie; I don't know if it will help you out or if your beyond this page?

Good luck.

kv

Edit: Oh, and to help me understanding my new Christmas gift. "Function Generator"

 

Jony I have some serious doubts about the output pair in your circuit.

At 10 watts into 4 ohms the peak current is nearly 2.25 amps

The guaranteed gain of a BD441/442 at 2 amps is only 15.

This makes the base current 133mA.

The standing current in R6 is 10/1000 or 10mA.
Swinging the colelctor of Q3 down to zero and up to about 18 volts will not generate enough current swing in R6 to drive the output pair and some form of distortion will result in their strangulation at the peaks.

You really need to add a transistor to each to form darlingtons that raise the gain at least an order of magnitude.

 

Jony I have some serious doubts about the output pair in your circuit.

At 10 watts into 4 ohms the peak current is nearly 2.25 amps

The guaranteed gain of a BD441/442 at 2 amps is only 15.

This makes the base current 133mA.

The standing current in R6 is 10/1000 or 10mA.
Swinging the colelctor of Q3 down to zero and up to about 18 volts will not generate enough current swing in R6 to drive the output pair and some form of distortion will result in their strangulation at the peaks.

You really need to add a transistor to each to form darlingtons that raise the gain at least an order of magnitude.

Just read your signature. lol

Nice,

kv

 

Jony I have some serious doubts about the output pair in your circuit.

At 10 watts into 4 ohms the peak current is nearly 2.25 amps

The guaranteed gain of a BD441/442 at 2 amps is only 15.

This makes the base current 133mA.

The standing current in R6 is 10/1000 or 10mA.
Swinging the colelctor of Q3 down to zero and up to about 18 volts will not generate enough current swing in R6 to drive the output pair and some form of distortion will result in their strangulation at the peaks.

Yes and this is why I give this note "Amplifier that should never be built in real life". Also notice the lack of emitter degeneration resistors at the output stage. Also the negative max current is -ILmax = (0.5Vcc)/( R6/HfeQ5 + Rload) ≈ 0.7A or less.
And the power dissipation in the output transistor is P ≈ 0.1 * 10V^2/4Ω = 2.5W at last no problem with this .

You really need to add a transistor to each to form darlingtons that raise the gain at least an order of magnitude.

Well I prefer Sziklai pair as a output stage.