I used the LM3886 in the inverted configuration, the reason is that I can use a trim pot to adjust the DC offset of each 3886. National's application note AN1192 suggested to use servo to set the DC offset to zero, but I prefer this simpler solution. After some adjustment, I can get the DC offset of each 3886 to about 9mV, because this is a push pull configuration, the DC offset at speaker terminal would be very close to 0mV.

The PCB is designed in a way that the center area is filled with ground. Which the +30V supply line is at the outer area of the back of the PCB. The -30V supply line is at the outer area of the front of the PCB.

The PA150 PCB - The resistors R2 and R3 are 0.1% parts, you could also buy 1% parts and hand match them to 0.1% accuracy. For the R7 0.2ohm 2W resistor, the PCB is designed to accept either SMD or thru hole component.

The unbalanced input signal is converted to balanced signal using the DRV134. The DRV134 board circuit is very simple, similar to the reference circuit in the Burr Brown DRV134 data sheet.

I choose to use SMD components because I think it is easier for board layout, and can have a shorter signal path. Using surface mount components allows me to create a smaller board. The trimmer is to adjust the DC offset of each 3886 so that the DC is at about the same level.

The 3886s are mounted to the heat sink using a aluminum bar. This mounting method can apply a more even pressure on the 3886s, and thus better heat transfer to the heat sink. I used the LM3886T, the non-isolated version for better heat dissipation. I used Solen 15uF as the DC blocking capacitor, because it not expensive and performance is quite good.
All the power input, signal input and speaker output uses screw type connector for easy assembly and testing.

I have WIMA 0.1uF bypass capacitors soldered at the 3886's supply pin directly, on the back side of the PCB. These capacitors are used to filter out the high frequency noise of the power supply. 4 Nichicon MUSE 2200uF capacitors are used in each board for supply filtering.

Power supply is a 2x25V 500VA encapsulated transformer for each channel. After rectification and at low load, the voltage is at about 34.5V. At full load, the voltage dropped to about 31V. This transformer is very quiet and vibration free. Basically I cannot hear any noise from it. These RIFA big can capacitors can deliver huge current, which is required to produce a deep and punchy bass. This is the voltage regulator board for the DRV134, I used LT317 and LT337 voltage regulator chip.

The PCB ground is connected to the chassis / AC ground via a choke and a 10ohm resistor connected in series. The choke is to block the HF noise, and the 10ohm resistor is to reduce the ground current which would cause some 50Hz hum.

I am building this as a mono-block for better heat dissipation. During testing at 300W this amplifier can become very hot. At normal music listening volume the heatsink is generally warm. Notice all the power and signal wires are twisted to minimize noise.

I used RMAA to do some basic testing.
Testing setup : PC sound card -> Amp line in -> Amp speaker out -> 4 ohm load -> Passive Preamp -> mixer preamp -> PC sound card in.

Passive preamp is to lower to amp's output voltage to line level, mixer preamp is to convert the balanced signal to non-balanced for sound card input. RMAA test result for 1W, 10W, 100W, and 294W at 4ohm load. My sound card's loop back THD is about 0.0017%, so the BPA300 THD result here may be limited by the sound card.

Listening impression: the sound characteristic of this BPA300 is similar to my other LM3886 and LM3875 gainclone. These mono blocks has deeper sound stage, better stereo imaging, much tighter and better controlled bass. Overall I can say this truly is a high end audio equipment.