Long Tail Pair Linearizer

[Electrone]

I have noticed that long tail pairs tend to be nonlinear because the current input at the emitters changes when the voltage at them changes with the input and feedback audio. The normal fix is to supply constant current to the emitters instead of using a single resistor. The constant current source is a clean solution distortion-wise.

I believe that I have discovered that there might be a remedy which is an alternative to a constant current source. Connect a suitably-valued a resistor from the audio output to the emitters of the long tail pair. Such a resistor cancels out voltage changes.

I doubt that I am the first to notice this solution. I would be interested to see who else has tried it and what the result seemed to be.

Alternately, there should be circuits where constant or near constant current is not desirable. Then connect an extra resistor and capacitor similar to a normal feedback network, but to the emitters of that LTP in addition to the one at the base of the feedback transistor. Another difference is that the capacitor and gain resistor are switched in position. Consider it as a type of bootstrapping. Like standard bootstrapping, it should increase the impedance at the base of the input transistor.

These are my observations. I will see what thoughts other people have.

[AndrewT]

was it the Patchwork that used this bootstrap in it's sims to generate that cancellation effect?

[Electrone]

Originally posted by AndrewT
was it the Patchwork that used this bootstrap in it's sims to generate that cancellation effect?

Hi Andrew. I was playing around with the idea on a circuit I have been designing in LTspice. I am working on the simplest class D amplifier that I can come up with that uses discrete components, only N-ch MOSFETs, and operates on at least +/-40vDC.

[AndrewT]

it was Abomination by MJL that used this bootstrap.

[Electrone]

Hello.

I have a new long tail pair linearizer. It offers a simple alternative to constant current sources. It looks different than other things that I have seen, but I doubt that it hasn't been tried before by someone else.

I selected the three resistor values which replaced the original 39k resistor by first setting the sum of the ones in series with the B+ equal to that 39k one. 36k + 3k = 39k. Then I used LTspice to find the third value by trial and error. That 56k resistor maintains almost constant current through the 36k one.

[darkfenriz]

Current source is still much better PSRR-wise

[homemodder]

Ive played a bit with these resitor current sources, try bootstrapping a active current source, this has benefits compared to standard one.

[Electrone]

Originally posted by darkfenriz
Current source is still much better PSRR-wise

I agree about the PSRR. If the power supply changes, this type of current source goes up or down with the voltage. However, it is in a linear manner because this circuit offers a more linear input impedance compared to a more ideal constant current source.

Originally posted by homemodder
Ive played a bit with these resitor current sources, try bootstrapping a active current source, this has benefits compared to standard one.

The latest example that I showed is close to a type of bootstrapping. If I added a capacitor in series with the 56k resistor, it would be. In order to make this project easy to assemble with p2p and very compact without SMD parts, which I always avoid, I've omitted the capacitor. it isn't really much needed since the voltage on the emitters of the long tail pair is only nominally offset by .7v above the output voltage.

I was seeking a lessening of the fluctuating of the input impedance within an audio cycle. I wanted to make operation of the circuit more like inputing the audio signal at the inverting input, but without the loss of input impedance that you get at the non-inverting input.

I have now used the simulator to confirm that a single resistor will do the job. So, I went back to using the original 39k resistor and was able to determine the optimal value of the linearizing resistor to be 910k, placed between the amplifier output and the emitters of the LTP.

Calculate it by multiplying the amplifier gain by the value of the current resistor and then go up to the next higher value. You want the normal feedback divider at the base of the feedback transistor to be the dominant actor in the circuit.

I confirmed the 910k value by temporarily adding a 1ohm resistor in series with the emitters of the LTP and using simulation to measure its current. I have not shown the 1ohm resistor in the next diagram.

I should mention that this amplifier is still being refined. If I build an amplifier this one presently is one of my favorite possibilities.

[Electrone]

However, it is in a linear manner because this circuit offers a more linear input impedance compared to a more ideal constant current source.

Oops, I mean with respect to no circuit to provide constant current, for example, a lone resistor.

[MiiB]

I think this is a great idea....If one's to use a seperate supply for the the front end and VAS stage of the amplifer the need for high PSRR i not really there... Then this could indeed be a very good way to balance the LTP....I don't fancy the CAP,...as is will ad sonic signature... If the Pair was Jfets it wouldn't be needed at all...I suspect..