Transistor Line Driver

Here are some new (to me at least ...) line driver circuits that offer incredible performance for audio, using discrete components that cost pennies.

While working on a New VCA, it had occurred to me that alternative options for implementation really included doing the entire VCA in discrete transistors. A lot because I've already developed circuit designs for Voltage Adders and also a high-performance discrete op amp -- kind of the only thing missing to doing a great all-BJT VCA, using the topology of the New VCA (which originated in the Trimless VCA).

While contemplating this, an idea for line drivers came to me. This too was based on previous BJT development, based on complementary emitter follower buffers, as well as a highly improved variation. One of the minor problems I was not able to fully correct when developing the earlier BJT buffers was output voltage offset. The use of complementary NPN/PNP transistors improved the offset voltage, but it never really zero'ed it. And that was something I was interested in for DC accuracy.

Also motivated by a recent circuit design realization that the original Moog 902 VCA used a parallel transistor output circuit for differential output, that my previous buffers were in serial form: a NPN follower, then a complementary PNP follower. That got to me thinking: why not try emitter followers in parallel, then build a resistive bridge circuit between them to null out complementary offset voltages? My specific insight was that if both followers sourced their output each from 1200Ω resistor, then the combined low-impedance output from each transistor could be transformed so as to appear as from one 600Ω source. But because each transistor had it its respective ±Vbe offset, the bridge should null most of that out. I was right, that totally worked! 

The first go at this is in the schematic below, which obtained < 20mV offset without trimming at all. The bridge circuit provided offset voltage compensation that the previous serial emitter follower circuits could not attain. And the point behind a 600Ω source impedance was that it would be quite stiff if the following stages used 50-200kΩ inputs, but it would also automatically provide a voltage divider if the downstream circuit itself was terminated at 600Ω.

Then, while realizing that there could be small variances between complementary transistors, it seemed likely that the nearly equal ±Vbe could be equalized with a small trimmer in the bridge circuit. This version of the Transistor Line Buffer is shown in the second schematic below. It's capable of holding the offset voltage for the buffer at < 500µV across many ranges of output, temperature, and frequency. Just works!

Either of the circuits are very clean, and they work wideband. The -3dB bandwidth for ±10Vpp signals was measured at 13.3 MHz! The pulse waveshape is extraordinarily clean, super flat, and without abberations. Also, per FFT analysis, the buffer incurs very little distortion. 

Following the schematics are some exemplary scope fotos.

50 kHz I/O at ≈10Vpp, with 51kΩ output load.

Triangle wave I/O, ≈10Vpp, 50 kHz, 51kΩ output load.

20 kHz pulse train at 10Vpp.

The risetime of 10Vpp pulse output at 20 kHz PRR.

The falltime of 10Vpp pulse output at 20 kHz PRR.

-3db down from 10Vpp input, at 13.3 MHz, into 51kΩ load.

FFT response to 1Khz sine wave at 3.742Vpp (professional audio level of +4dBu).

Trimmed output Vos for 100mVpp output for 10 Hz sine wave.

Trimmed output Vos for 100mVpp 50 kHz sine wave.

50 kHz pulse, 100 mVpp, but now under 600Ω load, showing 2:1 voltage division.

20 kHz PRR risetime, 10Vpp input, 5Vpp output under 600Ω load. 

20 kHz PRR falltime, 10Vpp input, 5Vpp output under 600Ω load.