Making balanced output from single end

You can connect the input of this circuit to a balanced or an unbalanced source and the output to a balanced or an unbalanced load

This one also can be used (of course with source/emitter resistors added or even being cascoded).

Phase splitter with common mode control.PNG
 
Are the collector and base of Q6 supposed to be shorted?

This circuit doesn't have the main feature of the circuit I posted: compensation of the distortion of the input pair by a matching pair in the feedback path. It works completely open loop except for the common mode feedback.
 
Are the collector and base of Q6 supposed to be shorted?
Yes, you're clear eyed, my mistake in drawing by using F6 (copy) in LTSpice.

This circuit doesn't have the main feature of the circuit I posted: compensation of the distortion of the input pair by a matching pair in the feedback path. It works completely open loop except for the common mode feedback.
Not clearly so because this is just a barebone for fully-differential (and because of this inverting) amplifier. And, yes, it can be DC coupled from input and output.

The other approach with two dedicated LTPs can be like this:
Phase splitter with common mode control _2.PNG
 
Last edited:
I always toy with the idea of a such bal line driver mated with a matching transformer so that its once again single ended gain stage. It seems like a needless design but hear me out.

The 1:1 output trafo impedance matches the cable. It ground isolates the next stage. Offers a very natural sounding lpf. And the line driver/phase splitter lends a very high input impedance. The line driver's distortion is halved.

Thats enough positives to make it an idea worth entertaining, imho
 
To remain in the spirit of the concertina phase-splitter, it is possible to use a Rush variant, which eliminates the drawback of the load on one output influencing the other, and adds more flexibility if required:

RushConcertina.png


Normally, R1, 2 and 3 would be equal, but you can change the ratio to tweak the symmetry or the gain.
 
  • Like
Reactions: xrk971
The resistors are chosen according to the normal biasing rules, the same as for a common emitter amplifying stage.
With a bipolar supply+GND, some could be dispensed with.
I will work out an application example or two to illustrate the way it can be done practically.
It is also possible to include some temperature compensation, and maybe a linearity correction too.
Buffering the outputs is also an option, but in this case there would be practically no differences between the Rush and classical concertina.
Note that the circuit doesn't have particular merits: it is just an option amongst other, one more tool in your toolkit in case you need it
 
Here is a first example, based on a single supply and including bias temperature compensation:

RushConc1.png


The gain is ~1, but it can be altered by changing the value of R1 (and R5 if the change is large). If the temperature compensation is not required, you can omit D1 and D2, and increase slightly the value of R5.
At a gain of 1, the maximum output p-p swing is Vsupply/3 (minus inevitable parasitic voltages).

At higher gains, the swing will increase, up to Vsupply/2 (theoretically)
 

Attachments

It's a common collector stage, resistor and common base stage, just like a differential pair with emitter degeneration and with single-ended drive, but with opposite polarity devices in the "pair".

Compared to a normal differential pair, differences are that you don't need a tail current source and that you don't get any even-order distortion compensation.
 
Last edited:
Indeed, which is why I mentioned the fact that the circuit has no particular merit: it just replicates the operation of a single device concertina, without the effect of one output loading affecting the other.

The type of expansive distortion (opposed to compressive distortion in LTP) generated in such circuits seems to be favoured by "subjective" listeners, but I am not going to venture on such grounds: my preference goes to an inexistent harmonic profile, no THD at all
 
In this example, the supply is bipolar, the gain is 2, and error correction diodes have also been added. The result is a thd down by almost 20dB. The compensation is first-order, very crude, but using more elaborate correction schemes (like two complementary tringlotron's partially merged), another 20dB would be reachable.
The circuit is somewhat lopsided, because the input of the common-base part is connected to the GND, but it doesn't matter unless you need the maximum output swing. In case it is needed, you can revert to a symetrical configuration.
Many other variants are possible, but I think you get the idea: you can have unbalanced gain outputs, use CFP's or darlingtons instead of simple transistors, etc

RushConc2.png
 

Attachments

If the diodes were ideal, you would need four of them on the positive and four on the negative side for optimal distortion compensation at a gain of four (two per output), like you have now. With non-ideal diodes, you need whatever integer is closest to 4 divided by the emission coefficient. To make it more predictable, you could use four diode-connected transistors and reduce the 2 kohm resistors to 1.82 kohm each. You can also use the same trick with a normal differential pair.
 
If the diodes were ideal, you would need four of them on the positive and four on the negative side for optimal distortion compensation at a gain of four (two per output),
Well spotted: I missed the 2 factor of the two junctions. And compensating a transistor junction with a gold-doped diode is extremely crude indeed, but it does work in some way, as the sim confirms.
My aim is to show a number of options, certainly not to present a finished, refined design
 
It is 1.752, which explains the relatively good compensation
If you can accept a bit less gain, you could reduce the 2 kohm resistors to 1.752 times 910 ohm = 1594.32 ohm, nearest E24 value 1.6 kohm. If there are no other second-order effects (such as bulk resistances or high injection effects) messing it up, that should get it in the distortion optimum.