Summary:For low impedance loads, such as those found in automotive applications or when driving  MTM speakers (where a pair of 8 ohm drivers are placed in parallel) the bridged amplifier configuration isn't really up to the task of providing a lot of power without a lot of heat.  In these situations a paralleled amplifier is in order.  Each half of the LM4780 produces half of the total current.  The IC thus thinks that it sees a much larger load.

While I  used a power supply providing +/- 28 volts (because the transformer I had on stock was 21-0-21), using a larger transformer and +/- 35 VDC will allow the parallel amplifier to deliver over 100 watts per chip.   (The maximum derived from the Overture Series Design Guide  for a 4 ohm load is about 35 volts: 

Boards for this project are available here: CHIPAMPS

The Amplifier Section:

The schematic of the amplifier section is shown below.  Nothing novel here, it follows the National Semiconductor guidelines pretty explicitly:

It is important to carefully match the resistors in the feedback and input nets.  I used Yaego metal film resistors, matched to better than 0.1%.  If you purchase a lot of 100 from Digikey you will have plenty of choices to match to an even closer tolerance.   Careful selection will allow each half of the amplifier to balance their load.  While the circuit is close to that seen on the National Semiconductor website, note that 100nF polypropylene capacitors are placed in parallel with the non-polarized 47uF electrolytics.  The use of paralleled capacitors benefits the high frequency performance of the amplifier.

Decoupling the chip from the power supply is important, as the Overture series are susceptible to high frequency oscillation.  The 100nF ceramic capacitors tackle this problem at the high end.  The 10uF/50V electrolytics, on the other hand, will reduce the likelihood of “motorboating”.  I view the 1,000uF/63 V electrolytics as a localized reservoir, limiting supply droop or I²R losses from the power supply to the operational amplifier.  The mute function can be selectively enabled, coupling pins 14 and 20 to the negative supply through the RC network as shown on the schematic, or by grounding and thus muting the amplifier.  The mute function can be bypassed by omitting C12 (10uF/50V electrolytic) and connecting pins 14 and 20 to the negative supply rail at the junction of R9 and R10. 

If you don't want to switch the MUTE function of the LM4780, just bridge pins 1 and 3 of J4.

Choice of Heat Sink:  P_DMAX  is a term which describes the power which must  be dissipated by the device into the heatsink, not into the load.   

The parallel configured amplifier heat sink calculations are pretty straightforward, each amplifier provides half of the driving current.  P_DMAX is thus half of that for a single amplifier, but remember that there are two amplifiers per chip!  Here are the equations:

A heat sink with 1.01 q_SA  is approximately 12 inches long, 5 inch base width and 10  fins with 1.3 inch height.  Coincidentally, the value which leaps off the National Semi design tool is 1.10 q_SA  Close enough for government work!  If you are going to use the LM4780 with +/- 35VDC power supply you should provide forced movement of air at a rate of at least 100 LFM.

I tapped the heat sink for a 6-32^nd screws and mount the device to the heat sink using the tensioning bar, 6-32^nd screw and a star lock-washer.  To provide electrical isolation, the LM4780’s case is connected to the negative supply rail, I used a Berq Sil-Pad cut to fit.  (Note further, if you apply too much torque to the screws you can crack the case of the LM4780.)

Tensioning Bar The tensioning bar is made from 1/8"aluminum stock.  You can mill or file the edges so that they are slightly chamfered as shown below:

The parallel configuration circuit board is similar to that of the Bridgeclone except that non-polarized electrolytics are used.  These are bypassed with 100nF polypropylene capacitors.

The Power Supply: The power supply is a straightforward, the same as that used for the BridgeClone witha  full wave bridge with 10,000 mF filtration. An A.C. input socket was “liberated” from a Dell server power supply. In case your network administrator disapproves, we have furnished the equivalent circuit below consisting of a  bifilar-wound 2 x 200 mH choke, and  “line rated” A. C.  polypropylene and ceramic capacitors.  These will serve to limit radio frequency interference into the power supply. The circuit is protected with a 5 amp slow-blow 3AG fuse.  I used a Thermometrics inrush current limiter rated for 8 amps to limit the surge current although you may find these unnecessary.  The value of the RC snubbers for the MUR860 diodes were calculated using the stated diode capacitance (100 pF at 30 VDC), the inductance of the transformer secondary windings (12 mH) and the interwinding capacitance (560 pF) of the transformer.  The supply schematic appears below:

Results: The THD+N% figures for the parallel amplifier are somewhat inferior to the bridged amplifier.  There is an obvious tradeoff between dissipation and a slightly poorer THD statistic.  In the following chart the solid lines represent the bridged amplifier, the dotted lines the parallel configuration.