Primarily this DC AppNote is the result of my tinkering with TDA1514 and LM3886 power audio ICs and a high power transistor design. Even though Phillips (TDA1514) and National Semiconductor (LM3886) are both pretty silent on using these devices in a bridged application, I have been able to get both to work with some degree of success. I am assuming most readers are not audio engineers, so there is no detailed engineering discussion here. Instead, I will give a "lay" perspective for the audio enthusiast who has respectable soldering skills, can read a schematic and has a basic understanding of electronics principles.
I am sharing information and cannot and will not accept any responsibility for any damage a reader might cause as a result... In other words, you are experimenting on your own and must take responsibility for your own actions. Keep in mind these ICs were not really intended for BTL applications and apply a generous amount of common sense. Now, onto the good stuff...
Audio amplifiers operate either in a BTL (bridged) or single-ended ("normal") configuration. In the single-ended setup, the output lead goes to the "hot" or "+" side of the load (speaker or speaker box since we are talking audio) and the "-" or "negative" side of the load is tied to a common ground shared with the amplifier. In the BTL configuration, one amp is connected to the "+" side of the speaker (load) and a second amp is connected to the "-" side of the load. For this to work, the output signal from the second amplifier must be a "mirror image" (identical in every respect, but 180 degrees out of phase) of the output from the first amp. The BTL configuration is most often seen in low-voltage, battery-powered applications (like cell phones or "walkman" type personal tape or cd players etc) or in automotive applications over about 10 watts per channel.
In the BTL configuration, each amp drives half the load impedance. With the signals being out of phase, the voltage swing across the load appears to be doubled, and with each amp driving half the impedance the current is doubled. In theory the bridged pair will produce 4 times the power into the load that either amp acting alone could provide. In reality it seldom works that well. Usually current limiting or thermal protection circuitry will activate to protect one or both amplifier ICs, sometimes the power supply will be marginal and unable to deliver the required voltages at the required currents. Realistically it is rare for a bridged pair to produce more than twice the power each amp would deliver single-ended - especially in the case of IC amplifiers that have all sorts of protection circuitry built-in.
There are two basic approaches to bridging. One strategy is to set the first amplifier to provide all the voltage gain, then feed the output signal to both one side of the load and to the second amplifier. Either the signal has to be phase inverted between the amps or fed to the inverting input of the second amplifier. Either way the second amplifier has to operate at unity gain to provide the same signal voltage as the first. This can cause stability problems in the second amplifier unless it is compensated for operation at unity gain. This is fairly easy at low signal levels using reasonable quality op amps, but may not be so easy at high power levels. The approach I prefer is "splitting" the line level signal and using matched power amplifiers that run identical gain. I do this "splitting" with a dual op amp. The device type is not critical so long as it is reasonably low-noise and has at least a moderate slew rate. In systems starting with a nominal 1Vpp signal and going up to around 125 watts rms, noise usually isn't an issue (except from zener diodes, which should be bypassed with a 47uf to 100uF or so capacitor.) I have had excellent results using MC4558, NE5532, TLO72 and TLO82s. For production I prefer the MC4558 since it is the cheapest of these, but my own favorite is the NE5532.
It is important to eliminate as much as possible any DC offset voltage on the outputs of the op amp sections, as this dc offset will also be amplified by the power amps. Speakers really have a way of objecting to dc voltages. So will you after replacing speakers... A series cap between ground and the reference resistor on the inverting input of the first op amp section will prevent dc offsets at the outputs. Output from the first section is split. One leg is taken through a 270 ohm resistor (improves stability) and fed to the first power amplifier, the other leg is taken to the inverting input of the second op amp section. The second section (inverted input is used) is run at unity gain with the output taken through a 270 ohm resistor to the second power amp. Even though noisier that a resistor, I like to use a pot for the first section's feedback resistor. This lets me adjust the gain in the splitting circuitry, which means I can use this section as a pre-amp if I need to balance the bridged pair's output with other channels in the overall audio system. In high-powered systems, I use zener diodes with 100uF bypass capacitors to regulate voltage from the main supply rails down to the + 12 to 18V that the op amp can handle.
The TDA1514 is rated for an absolute maximum of + 30VDC power supply rails. They mean it, anything more than a 60VDC spread between V+ and V- will destroy the IC instantly. It will (only recommended for driving 8 ohms) operate on 28V rails, 24V rails are recommended. The LM3886 is rated for a maximum 84VDC spread between V+ and V- rails. For driving 8 ohm loads, +35VDC supply rails are recommended by National. For 4 ohm loads the recommendation is + 28V.
To avoid "engineering details", make sure the supply can deliver at least 5 amps at the selected supply rails. Regulation is not critical so long as maximum voltages are not exceeded. If the power supply is marginal and heavy current demands cause supply rails to sag, maximum output power simply cannot be achieved. As a rough guide, figure100 watts output will require 150 watts from the power supply. At +24VDC rails, round to 50. For 100 watts output, divide 150 watts by 50V, approximately 3 amps will be required. "Actual results may vary", as the saying goes. When testing, I drive the circuit with a sine wave and adjust the input amplitude to drive the output (whether BTL or single-ended) to the clipping point. Then I do my measurements. This setup is extremely demanding on both amplifier(s) and power supply - Far moreso than any possible music signal. The power supply should be able to deliversufficient current to maintain the intended supply voltages continuously under this condition for at least an hour without self-destructing, likewise heatsinking should be adequate to keep the ICs cool enough they do not go into thermal shutdown under these conditions. In terms of "music", this would be equivalent to having the amp produce a steady tone at peak volume continuously. In actual practice, this may be more power supply than necessary. Music signals, even hard-core rock and roll, have quite a bit of dynamic range compared to this. Even hard rock will seldom hold a peak for more than a fraction of a second, though certainly both rock and classical music has passages that hold near peak tones for a few seconds at a time. For the human ear to detect an increase in "loudness", power levels have to approximately double. Conversely, less than peak signals require considerably less current from the power supply than a signal at or just past the point of clipping.
TDA1514Signetics used to publish an app note using a pair of TDA1514s in a BTL mode. The second device was set at unity gain, and I never got the circuit to perform with any stability. All later versions of their databooks no longer include the circuit, which tells us something. Nonetheless, a pair of TDA1514's can drive 95-100 watts rms into an 8 ohm load, if it is done CAREFULLY. "Snubbing networks" will be required on the output (1/4 watt 10 ohm resistor in series with .047uF mylar cap placed between the speaker output and ground) from each amp. Change the feedback resistor (probably 27K to 33K) down to 20K or 22K to reduce gain a bit, and make sure the load is really 8 ohms rather than 4 ohms.
LOAD IMPEDANCE IS CRITICAL. Protection circuitry built into the TDA1514 limits output current to about 3 amps. Try to exceed this and the chip simply shuts down until the input signal is small enough the output no longer tries to deliver excessive current. Keep it up long enough and the chip will self-destruct, but for usual music signals what happens is the amps shut down (silence) on the peaks and blast away the rest of the time. The effect sounds horrible, even though the sound is crisp and clean when it is there. Operating in a single-ended configuration, the TDA1514 must produce output currents right on the verge of shutdown to deliver 50 watts rms into a 4 ohm load.
To get pleasing results in BTL mode, the TDA1514 amplifiers must be "tamed down" enough to prevent peak input signals from driving the outputs into shutdown from either excessive current or thermal conditions. Big heatsinks will cure the thermal situation, the excessive current situation is a little trickier. Either a higher load impedance, limited gain (more on that in a minute) or a limited input signal can ensure the output is never driven into shutoff. The most you can do is about 100 watts into 8 ohms, and that is right on the edge of shutting down due to excess-output current protection. This protection point is not adjustable.
In and of itself, limiting the gain in the TDA1514 circuits will not prevent shutdown. What gain-limiting will accomplish is requiring a larger input signal to drive the amps into shutdown. Sometimes this is easier than limiting the amplitude of the input signal. A volume control in the input signal will certainly do the job as well, but many who like their music LOUD expect to be able to turn the volume control "all the way up" without driving the amps into shutdown. Finding the exact limit and knowing which pieces of music will go over the limit is either an art form or a matter of trial and error.... It is usually better to "tame the circuit" a bit and arrange things so there is minimal likelihood of a peak signal being strong enough to cause excessive output currents.
In essence, given adequate heatsinking and a gutsy enough power supply the TDA1514 will produce as much "rms" power as it will "peak" power. Most "consumer" audio amplifiers will produce a "peak" power that is at least twice the "rms" power level they will sustain for prolonged periods. With the TDA1514's internal protection schemes, adequate power supplies and heatsinking will allow it to sustain its "peak power" levels. This is based on using rms voltage measurements across a dummy load (4 ohms single-ended, 8 ohms BTL) on the output while driving thecircuit with a 1KHz sine wave. Increasing the amplitude of the test tone will cause the output to eventually start clipping, then further increases will result in more clipping until shutdown occurs at about 10% distortion.
Of course, at normal to moderately loud listening levels in a car (or even fairly large room in the home) power levels should be well below the "critical stage" where shutdown might occur. 100 watts is a lot of power. Most people cannot stand the pain of being in a 20'x20' room with a 1KHz test tone playing at 100watts unless they are wearing GOOD ear protection. At 100 watts a 30Hz tone will not be deafeningly loud, but anything in the room that can possibly move will be rattling and vibrating - including loose window glass, acoustic ceiling tiles that are loose, items laying on a table (and for that matter probably the table itself) pictures hanging on the wall and so forth. For most people, a comfortable listening level in an average sized room at home is between 3 and 5 watts average output with decent speakers. Peaks, expecially deep bass notes, may hit 40 to 50 watts or more while the normal level is remarkably low.
LM3886The LM3886 features a unique self-protection scheme that is in many ways more desireable than the Signetics products. Rather than simply reaching a critical point and shutting down, National's "SPIKE" protection has a "transition area" of sorts. Going slightly "over the line" results in a proportional "notching out" of the backside of each waveform on the output until the transgression is severe enough to cause a total shutdown like the TDA1514. Conveniently enough, this "notching effect" becomes noticeable to the ear at about the same point as total shutdown. Like the TDA1514, the LM3886 has internal protection for thermal overload and against excessive output currents. Like the TDA1514, the shutoff points are not "user adjustable".
As far as I know, National has published only one app note (BPA-200) on using LM3886 in either bridged or parallel configurations. It seems to be sometimes available and sometimes not, written more to show what can be achieved with LM3886 than as a recommended application. Like the TDA1514, the LM3886 can be used sucessfully in a BTL configuration as long as its output parameters are not violated. The same schemes used with the TDA1514 apply to the LM3886 in terms of signal inversion for the second channel and taking steps to ensure the input signal never tries to drive the outputs into cutoff. UNLIKE the TDA1514, the LM3886 CANNOT sustain its maximum peak output capability.
National claims 68 watts "average" output for the LM3886 and 135 watt instantaneous peaks. Since I am accustomed to making measurements and comparisons in "rms", I rather doubt the 68 watt claim. I found closer to 50 watts rms was the maximum I could get into 4 ohms before thermal protection began to engage. Improved heatsinking did not improve results, my conclusion is the limitation is in how fast heat can be transferred from the die to the exterior of the IC package. The 135 watt "peak" claim is conservative, for short bursts I was able to get almost 140 watt peaks before the thermal protection began to engage. Like my other measurements this is using rms voltage, or .3535 times the peak-to-peak voltage (1/2 the Vpp times .707). Fairly accurate measurements can be made using a calibrated oscilloscope and calculating down from the peak-to-peak signal.
I've also found that although the LM3886 has higher output current capabilities than the TDA1514, the LM3886 requires higher voltage supply rails to produce the same power as the TDA1514. On the other hand, the LM3886 performs nicely at voltages that will destroy the TDA1514. "Sound quality" between the two is virtually indistinguishable although the specs make the LM3886 look better on paper. Connected to good speaker boxes, most listeners would not be able to tell which was which, the very few who could distinguish they were "different" would not be able to reliably identify which was which...
My own preferences are for the TDA1514 in automotive applications where power comes from a regulated switching power supply and the LM3886 for line-powered applications where a transformer can be used to provide an unregulated supply. Neither chip is useful in BTL mode for driving 4 ohm loads. As popular as 4 ohm speakers are in car audio it isn't practical for DC to offer a kit that uses either chip in BTL mode, and price considerations would make such a kit rather expensive. For 50 watts, either IC makes sense economically. For 100 watts and higher, discrete transistor designs make more economic sense when production quantities are involved. For the do-it-yourself builder, an extra three or four dollars is insignificant in a $50.00 to $100.00 project. For production quantities that $ 3.00 is quite signficant.
Our LM3886 Experimenters Package circuit board has a bridging circuit on-board in addition to provisions for low and high pass active filters and a summing network. For those who wish to make their own boards, the bridging circuit will take an area slightly over 1 square inch.