Diamond driver stage for the modular high end audio amplifier

Project presentation


This diamond driver stage is based on my research for audio amplifier output stages. The diamond driver forms a diamond buffered triple output stage together with any emitter or source follower power output stage connected to the diamond driver module. In effect, this is an improved variant of a triple emitter follower.


The schematic is pretty straightforward: This is the classic diamond driver in bootstrapped configuration.

The constant current sources are derived from my previous investigation of current sources for audio applications.

Diodes D1 and D2 are for overload management.

Q1, R1, P1, C1 and C5 present the most basic VBE or VGS multiplier. Transistor Q1 needs to be mounted on top of any output stage transistor for sensing the temperature.

Transistors Q2 and Q3 are bootstrapped to the audio output node and in conjunction with the constant current sources, they drive transistors Q4 and Q5.

Capacitor C2 may be installed for push-pull operation and is optional.

C3, C4, R6 and R7 are important for stability of the assembly together with the power output stage.

There are three output terminals. Two nodes for driving a power output stage and another one without offset that allows to use this module as low power driver stage, maybe for a headphone amplifier. I would put a test point at this node in any case and a terminal worth 25 cent makes this module even more useful.


The diamond driver module uses an own small radiator with all four driver transistors mounted to. Having the transistors on the same radiator is key to the bias current compensation scheme. In theory, all four transistors could be placed on the main radiator as well, but I believe it is better and more convenient to have an own small radiator. Many different radiators can be used dependent on availability. My preference would be the Fischer SK514 in 100mm length, but in the end I could only aquire Fischer SK481 in 50mm length and need to use two per module then.

The radiators for the current sources are not required and I plan to leave them away. However, I designed them in just in case they may be helpful. I love those small Fischer FK235 radiators because they consume almost no real estate on the PCB. The real estate consumed equals two small resistors roughly.


This module has not been built yet.

Thermal stability

The driver module tracks and compensates its own bias current so this is mostly out of the otherwise very complex equation. Only the thermal drift of the preceding output stage needs to be compensated and this is why I put provisions for a very simple VBE or VGS multiplier on the driver module PCB. The sensing transistor needs to be placed on top of any output stage transistor for sensing the temperature. The sensing transistor will track the temperature with a time delay and a deviation, which is an intrinsic error of any such temperature sensing scheme. The only way to overcome this would be on die sensing like ThermalTrak transistors.

Dependent on the use case, there are multiple options to set and track bias for the emitter or source follower output stage. It is pretty obvious that a BJT sensing transistor should be used for BJT output stages and a vertical MOSEFT for such output stages. Instead of the basic VBE or VGS multiplier, more sophisticated compensation schemes could be used. Thermal compensation is a highly complex topic and I deliberately left that open.


This driver module forms a very performant tripe emitter follower when combined with any power output stage regardless of BJT or MOSFET technology. Triples are challenging to get stable both in thermal and electrical domain. By using this diamond driver module, part of the thermal complexity is mitigated. I expect this arrangement to work very well, but this needs to be confirmed by an actual build and measurements.