My Final Audio Amplifier

 
 

After 35 years in high-end audio, I decided to sell off my big, heavy, expensive, power-hungry equipment and simplify my system. Since I’ve also spent these last decades doing electronic engineering (starting with audio while I was in high school in 1975!), a design of my own was in order. Here’s what I wanted:


  1. An integrated amplifier that includes a phono input.

  2. At least 50 W/chan into 8 ohms, but fully rated into 2 ohms.

  3. Reasonably efficient class AB outputs.

  4. Reliable, conservative, robust design that will last “forever.”

  5. Very simple controls: source select, volume, balance, and power.

  6. Headphone output.

  7. Compact, fairly light, and pretty packaging.


Click on the thumbnails below to download the overall system block diagram and interconnect diagram.

Design Objectives

Power Amplifier



Integrated_Amp_2.html

Next

I settled on a fairly simple power amplifier topology, using lateral power MOSFETS (Excicon 20N20/20P20) because they are rugged, thermally forgiving, support very high current, and are fun to design with. Modest power supply voltage (±38 V) achieves the desired output power without stressing any components.


For those familiar with such designs, the most interesting point is the frequency compensation scheme. It’s derived from Peter Baxandall’s Inclusive Miller Compensation. This scheme encloses the output stage in the feedback loop over a greater bandwidth than is normally achievable, resulting in lower distortion at high frequencies. In this case, a factor of five better at 20 kHz when compared with conventional Miller compensation of the voltage amplifier stage. More information on this and related compensation techniques is found in Bob Cordell’s excellent book, Designing Audio Power Amplifiers.


Compensation was optimized with the help of my N2PK Vector Network Analyzer. The goal is to achieve maximum loop gain while maintaining sufficient gain and phase margin for stability. Most folks working in audio don’t use such an instrument since it’s nominally for RF design. That’s unfortunate, because it gives the designer an instant and unambiguous view of an amplifier's high-frequency gain and phase. Results can be compared to simulations done in LTSpice or other circuit analysis packages. Simulation, while not terribly accurate in some areas such as these old-school power MOSFETS, is very useful when deciding which component to adjust to achieve a certain result.


When completed, the amplifier specs were as follows:

  1. 60 W into 8 ohms, and 125 W into 2 ohms.

  2. THD at 20 kHz: 8 ohms, 0.006%,  4 ohms, 0.025%, 2 ohms, 0.18%.

  3. 18 A peak into 1 ohm.

  4. Slew rate 200 V/microsecond

  5. Continuous operation at 1/3 power into 2 ohms without overheating.

  6. Unconditionally stable into any reactive load with the addition of a 750nH inductor (not present on the amplifier board).

(Click to down- load large gain-phase plot)

(Click to down- load graphs)

Here’s the “ugly style” prototype and final product.

For DIYers out there, here is the PC board design. It was done in ExpressPCB. All boards for the entire system are on a 1-foot square layout, of which I had two fabricated. You can copy and paste parts out of it.   Download:  Preamp and power amp AS-BUILT.pcb