3 transistor feedback but not quite a differential amplifier

[iainpunk]

hello,

ive been experimenting with this design. its usable a differential amplifier, but with a bit of a twist regarding the bias. it needs a bit of an offset.

here's the obligatory 20dB gain stage, i used BC327-40 and BC337-40 transistors. it has a max gain of about 36dB when a 1mVrms signal is put in, but this is somewhat dependent on the bias voltage.


i came up with this when i was trying to minimise the boss style opamp, this topology saves me 2 resistors! the bias current in the input transistors is about 40uA when the input is biased at 4.8v, so input bias currents can be very low. ive breadboarded an overdrive and a boost with success using this ''''opamp'''', the maximum gain i seem to get from this circuit is about 36dB, so on its own a bit lack luster for a fuzz or high gain distortion, but i found it to be very suitable for overdrive. ive put in a negative feedback loop with the typical soft clipping diodes, and it sounds quite good, nothing special, but good. it also clips quite nicely in open loop configuration.

cheers

[PRR]

[ElectricDruid]

I'd find you a lot more educational if you weren't *quite* so oblique, Paul!

What exactly do those two schematics have to do with the one that Iain posted, and why?

While I don't doubt you're making some interesting point about those circuits, I feel like I need to know a ton of stuff to be able to get it, and I don't know that stuff. So I don't get it, and I don't learn anything. That doesn't help me. And I'd venture I'm not the most newb person on here, although I'm pretty green on transistors.

Please, help a guy out!

[Rob Strand]

i came up with this when i was trying to minimise the boss style opamp, this topology saves me 2 resistors! the bias current in the input transistors is about 40uA when the input is biased at 4.8v, so input bias currents can be very low. ive breadboarded an overdrive and a boost with success using this ''''opamp'''', the maximum gain i seem to get from this circuit is about 36dB, so on its own a bit lack luster for a fuzz or high gain distortion, but i found it to be very suitable for overdrive. ive put in a negative feedback loop with the typical soft clipping diodes, and it sounds quite good, nothing special, but good. it also clips quite nicely in open loop configuration.

Interesting twist on the circuit.  You sometimes see this pattern in power-supplies where the transistors carry more, and varying, current but not so much for audio.

Can't do much about the bias shift but it's not a problem in your case (base bias resistors).   In the Boss circuit the two bias resistors are reduced to a single base bias resistor because you always have the access to the "global" Vcc/2 rail.

With this topology you can use much larger feedback resistors than the single input transistor version.  Be interesting to see if you can have a variable gain set-up.  Boss use JFETs so you don't get scratchy pots when you vary the feedback resistors in variable gain amp.

Which way to go often comes down to very specific requirements.

[PRR]

I feel like I need to know a ton of stuff to be able to get it...

Note the 2 inputs (each case).

Note how the two Emitters are polarized and come together.

It gives a current semi-proportional to the difference of the two inputs.

[Rob Strand]

Note how the two Emitters are polarized and come together.

It gives a current semi-proportional to the difference of the two inputs.

You can even do a two-input differential amp with one transistor by feeding signal into the emitter.
Your second circuit is the buffered version of that.

We see the differencing on feedback amps all the time but with a slight switch of perspective
you have a two input circuit.

[PRR]

> feeding signal into the emitter. Your second circuit is the buffered version of that.

I don't see that.

[Rob Strand]

> feeding signal into the emitter. Your second circuit is the buffered version of that.

I don't see that.

It's hard to find examples one the web because they aren't a common pattern.  However, if you start with this one,

From,
https://electriciantraining.tpub.com/14180/css/The-Two-Input-Single-Output-Difference-Amplifier-92.htm

R2 is just for biasing (it also lowers the input impedance of input 2).

Input 1 is high impedance and input 2 is low impedance.    So the next step is to add a buffer to feed into port 2.   At first think of the buffer as AC coupled - no problems there.

The next step is to remove the AC coupling and make the buffer DC coupled.    If we choose a PNP transistor for the buffer we get your circuit but with the resistor between the emitter set to zero.   If we choose an NPN transistor for the buffer we get the common differential amplifier.

There's one more detail.  When we DC couple the input impedance into the top transistor is the same as the output impedance of the buffer stage.   That means we loss some gain.   However we don't lose the differencing.  The difference voltage appears across the two 're' transistor emitter impedances.     When we add a resistor between the emitters we further reduce the gain.   Adding the resistance between the emitters might seem weird but in fact we do it all the time when we add RE resistors to a diff-pair to reduce the gain, as done in many amplifiers.

The important difference between these variants and the common differential amplifier is the common differential amplifier is balanced for DC and will do differencing on both DC and AC signal.     The variants do differencing on AC, sort of DC differencing as well but there's DC offsets that we need to factor in.   You can see an example of that here, the reference voltage is the zener voltage less the two Vbe drops,



In full your circuit is like an unfolded differential pair.   It's actually got a name but I can't remember it right now.   The folding and unfolding ideas are common in oscilloscope and wide-band output amplifiers for signal generators.   The books by Feucht might have the details.

[antonis]

Sorry guys but I stll can't see any "differencial" configuration..
(at least, at a first glance..)

What I see is a complementary feedback pair buffer (with gain) with inperpolation of a "weird" common Base BJT..

edit: Now I've seen the light..

IMHO, unless it's a current matter, same (or better) results could be obtained without lower p-n-p ..
(Do I miss something consequential..??)

[amptramp]

It looks like you are approaching something like the LM2900/LM3900/LM3301 Norton amplifiers National Semiconductor put out.  The input stages were not symmetrical but they were used as op amps even though there were some differences from ordinary circuits.

https://www.ti.com/lit/ds/symlink/lm2900.pdf?ts=1656501518136&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FLM2900

They were designed as single-supply amps and their circuitry had some differences from normal circuitry since the non-inverting input needed some current drive.

[mac]

Looks like a Heathkit TA28



mac

[antonis]

Looks like a Heathkit TA28

CFP configured to give voltage gain (as output & feedback point aren't inherently the same) is an ancient topology with countless applications..

[PRR]

...(Do I miss something consequential..??)

You have lost the voltage gain. Actually you make gain at an internal node (a collector) and then throw it away before the output.

Yes you can shove NFB in at the Q1 emitter, as Rob shows. But that wants a lo-Z NFB network. Adding iain's 2nd transistor lets the NFB Z be like 100 X higher, which may be more suitable.

But here's one without the emitter buffer:

(Flip it top/bottom to look like iain's plan; transistors won't know which way is up.)

hIE of Q1 is maybe 600r. The NFB terminates in 330r, so the emitter loading is light. But you also have to figure the leverage of the NFB net on the output. The output is 6.8k. It is loaded by 220k subsonic, 12k midrange, but dropping below 6k well before the top of the audio band. The heavy loading is in practice compensated by increased NFB due to RIAA and practical music spectra (dropping through the treble). This is not a stellar phono preamp, but it was excellent performance per buck in its day.

[antonis]

You have lost the voltage gain. Actually you make gain at an internal node (a collector) and then throw it away before the output.

I knew that something seemed odd..

OK.. Let me move output directly on p-n-p Collector..

[PRR]

Here's another serendipitous discovery:



This would maybe work better upside down. As a single stage under battery it may not matter, but in a large system you'd like more bias filtering and isolation.

[iainpunk]

Sorry guys but I stll can't see any "differencial" configuration..
(at least, at a first glance..)

What I see is a complementary feedback pair buffer (with gain) with inperpolation of a "weird" common Base BJT..

edit: Now I've seen the light..

IMHO, unless it's a current matter, same (or better) results could be obtained without lower p-n-p ..
(Do I miss something consequential..??)

i found that leaving out that transistor, having a gain control and / or clipping diodes in the feedback loop isn't quie ideal. the clipping diodes will have one of them forward biased if the gain is set high enough, making it cut out when playing really dynamically, and having a gain pot there makes it crackle and change the bias voltage.

been working on a bass overdrive for my dad's birthday using this basic topology, but with a directly coupled gain stage/filter thing boost bass and cut a bit of treble.

this is what i have on my breadboard now, and it sounds great. the gain control needs to be log, instead of the lin i have on it now, but the tones really work in our rehearsal space with the rest of our band. with gain above 1/4th and the tone control high or maxed, it has tonnes of sustain and a fat burpy character on the attack.

cheers

[antonis]

the gain control needs to be log,

Actually, super-log..

[Rob Strand]

i found that leaving out that transistor, having a gain control and / or clipping diodes in the feedback loop isn't quie ideal. the clipping diodes will have one of them forward biased if the gain is set high enough, making it cut out when playing really dynamically, and having a gain pot there makes it crackle and change the bias voltage.

That added transistor has to help. 
The good thing is it works and sounds good with the gain pot.

I suppose if you use high gain transistors it would help lower the crackle.

Another way might be to raise the 1k collector resistor.   That will drop the final PNP's collector current then that will drop the collector currents on the input NPN and PNP.   The whole idea is to scale back the base current on the input/feedback PNP relative to the gain pot value.   Of course you can't raise the 1k resistor indefinitely, it will be a delicate balancing act.

[iainpunk]

for the transistors in the bass overdrive ive put together im using BC327 and BC337 with the ''-40'' suffix to indicate a higher gain range of 250 or more with a Ic of 100mA. my DMM indicates that the two ''differential'' transistors have a bias current of about 16uA so im not really afraid of crackle in the version with both transistors.

cheers

[Rob Strand]

for the transistors in the bass overdrive ive put together im using BC327 and BC337 with the ''-40'' suffix to indicate a higher gain range of 250 or more with a Ic of 100mA. my DMM indicates that the two ''differential'' transistors have a bias current of about 16uA so im not really afraid of crackle in the version with both transistors.

FWIW, I like your transistor circuits.   You always try to push things a bit beyond the normal textbook circuits.