Discrete Buffer Design

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Ok, here's the first real working schematic! I'm calling it the Barefoot Buffer! :grin:

BarefootBuffer-rev4-Schematic-rev4.gif


The main changes from previous versions are as follows:

1. Input stage transistors are now 2N3904 and 2N3906 types, only because I readily found spice models for these.

2. All diodes are 1N4001 type. Initially I tried fast switching types like the 1N4148, but the distortion dropped by nearly half using the 1N4001. I didn't check into it thoroughly but most likely the 1N4001 has a higher voltage drop which increased the bias current.

3. Moved C1 and C2 from bridging the feedback resistors R11 and R12. They now act as Miller caps for Q5 and Q6 and this greatly enhanced the circuit stability.

4. Q5 through Q12 are now MPSA06 and MPSA56. These come in TO-92 packages with 500mA Ic. Using all the same device types in the midsection and output stage reduced odd order distortion significantly. I assume it has to do with lower crossover distortion. Despite the TO-92 cases the output devices can easily drive a 600 Ohm load to the full 22V p-p output.

This buffer has a gain of 2, so the previous stage never needs to swing near the rails. The circuit is also fairly efficient drawing only 40mA of quiescent current (2W dissipation).

Frequency Response

BarefootBuffer-rev4-FreqResponse-600Ohm.gif


The frequency response looks fast and stable with a smooth, low Q roll off out past 10MHz and the -3dB at about 60MHz. Even baking at 100˚C it never peaks. Increasing C2 and C3 generates more of a knee just above 10MHz. If you don't completely trust the simulator, using 33pF caps will give a little more stability insurance. But I wouldn't go above 47pF. The phase response also looks nice and smooth under normal temperature conditions.

Distortion

BarefootBuffer-rev4-Distortion-600Ohm.gif


The resistor values are optimized to deliver the lowest distortion for large signals into a 600Ohm load within the dissipation limitations of the output devices. And according to the simulator the distortion looks outstanding. 2nd and 3rd harmonic distortion are both less than 0.025% at full 22V p-p output. Just a little higher and odd order distortion shoots up drastically as the waveform clips. So, like the Bryston amps this circuit was inspired by, it delivers very clean output right up to the clipping point. Even if the simulator is off by and order of magnitude, these numbers still look great because this buffer is going inside an op amp feedback loop!

Unfortunately the 50 component limit in my demo software doesn't allow me to see how the BB performs in the feedback loop of an op amp.

Next step is to build it and see if it lives up to its promise!
 
I imagine I will.

I've already learned a ton so far. Several hours with this sim software and I've developed a much deeper understanding of how discrete BJT circuits work. If the simulator works even half as well as LEAP (spice based loudspeaker and circuit design) I'll be happy. The op amp circuits I design using LEAP are right on target with the real world.
 
Barefoot, looks nice but compare your circuit with for instance TPA6120 which has much better specs. One single IC made of 5 GHz transistors, if I'm not mistaken to a cost of 2-3 dollars and small and cute and capable of 700 mA! Extremely fast, extremely low distortion, 120 dB S/N and sounds alright in my ears too.
 
Back
Top