Somebody explaining this design?

Heyy...this schematic is by Phil Allison. I think he's on this group.. wait for answer..

It's current feedback. Q2 is the current feedback transistor. Basically...the more Q2 turns on the more Q1 turns off.. D from BC

Looks like crap to me ;-)

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

Got any idea what a "differential pair " is ?

The topology does look a bit strange when first encountered - but is just a basic differential pair operated in fully differential mode with adjustable gain supplied by varying the emitter to emitter resistance.

Forget about the BC549s, and imagine only one of the two inputs is driven.

See what it breaks down to then.

....... Phil

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

I saw one a little while ago that used 8 or 16 transistors in parallel as an input stage. Hard to believe this is better.

That was to lower the noise, this is to lower the high-level signal distortion, without excessive feedback. Of course it's better.

As Eeyore said, "Q1/Q2 and Q3/Q4 are 'compound pairs' that have very high linearity compared to a single transistor."

Actually, this, and a few other tricks, are widely used in discrete-circuit designs with good results. For example, early oscilloscope stages usually did not have any negative feedback, and relied on highly-linear circuits to begin with. I've also seen this in some IC designs, although it could be that the commonly-available IC PNP process capabilities don't well justify this scheme. JT will have to fill us in on that.

"Winfield Hill"

" The compound pairs of 2N4403 and BC549s are far more linear than any single transistor. "

Forget the read the article, did you Win ?

....... Phil

That was my lame attempt at a super basic explanation. I guess I should have mentioned that added feedback is from more transistors for more linearity. D from BC

Sorry, Phil, I didn't see it. Rereading all 16 articles in this thread, I still don't see it, where was your article?

Anyway, Jim's comment, "Looks like crap to me ;-)", implies an incorrect analysis, as several people correctly saw. In a differential amplifier the addition of two transistors and resistors in this way makes a dramatic improvement. If the amplifier doesn't have feedback, as in this case, some trick like this is essential. Graham says he designed this scheme into Studiomaster's 8-4 rack mixer in 1979, but it's an old trick, going back to the 60's at least, SFAIK.

Oh, wait, sorry, now I understand: you wrote the article in question in this thread,

The scheme correctly "stiffens" the emitters of Q1,Q3, but blows it by taking the outputs from the emitters of Q2,Q4.

The proper way is left as an exercise for the student ;-)

...Jim Thompson

Perhaps you can spell it out for us.

There are other configurations, but this looks right to me, for this one. Q2 serves to keep Q1's current constant, more or less, as does Q4 for Q3. This means the changing signal current, Vin/R9 (where R9 includes the variable part), goes through Q2 and Q4, rather than the input transistor pair. The input pair's nearly constant current means a nearly constant Vbe voltage, hence the low distortion. I'd like to see current sources in place of R2 R6 to improve CMRR, and a maybe mirror for R4 R8, but this is a good start.

There's another scheme I like by Barrie Gilbert, but it requires six transistors. I'm not sure it performs any better at low audio frequencies.

Barrie's scheme is an entirely different manifestation.

...Jim Thompson

"Winfield Hill"

But did not read it or see who it was by.

What a posturing ass you are.

....... Phil

"Winfield Hill"

You missed that too.

....... Phil

"Jim Thompson"

F*ck off.

........ Phil

OK. Back from lunch and purchasing a new power supply for my PSpice machine. Fan on the old one developed a bad case of the "grindies".

Now I would like to see a computation of the gain of this "marvy" amplifier.

Win?

...Jim Thompson

Wow! Amazing how-slowly increasing noise goes unnoticed until it becomes a grind! Changed out the PS and the silence is deafening ;-)

...Jim Thompson

Well? Anyone? Mathematical solutions only... no simulations please ;-)

...Jim Thompson

what I see is a differential pair of PNP Sziklai connected transistors (AKA "complementary Darlington" ), with a Pi emitter tail to the positive rail. Also, there is the standard feature of the resistor parallel to the "BE" of the second transistor of the darlington pair.

The differential pair is out of the feedback loop, introducing offsets, etc. what is the linear range for this pair ? => dynamic input range ?

would it be better to close the loop to the bases and have a fully differential system. Bartlett is my friend...(grin)

Jure Z.