We’ve analyzed inputs, signal processing, and clocks and used the analysis to put together a workable soundbar design (see http://electronicdesign.com/author/36602/DafyddRoche). We’re now ready to get audio signals into an amplifier and out to some speakers.
AB Versus D
The strengths and weaknesses of class AB and class D amplification have been analyzed many times. For the Soundbar Reference design, size and heat generation are the primary design considerations, so we’ll limit the discussion to those points.
Unless the product’s form factor (e.g., an extremely thin TV) is significant, class AB amplifiers should be fine for anything up to about 5 W per channel. Above 5 W, heat dissipation might be a problem, unless the amplifier is in a large enclosure (such as an AV receiver) and has heatsinks. A hot-running amplifier is inherently less reliable and might result in a bad “user experience” if the user unexpectedly contacts it.
The Value Soundbar Reference Design spec calls for two channels of 15 W each. This requires class D amplification, because there’s no spare room for large transformers and heatsinks.
Chip Or Discrete?
Now that we’ve decided on class D, should the amplifier be assembled from discrete components or use a single chip? Discrete components might be less expensive, but time-to-market and having the prototype work the first time are arguably more important than parts cost. A manufacturer-tested IC significantly reduces the probability of extensive troubleshooting or product redesign.
Analog Or Digital Input?
The next step is to decide on what type of signal the amplifier will be amplifying. Class AB amplifiers require an analog input. Class D amplifiers can handle a variety of signal formats (though a given amplifier is usually designed for only one):
- Analog (including digital-to-analog converter, or DAC, output)
- pulse-width modulation (PWM)
- I2S
The single-chip TPA3110D2 amplifier selected for the soundbar has its own PWM modulator and natively handles analog inputs. You can drive it with a line-level signal, or the output of a DAC (Fig. 1).
Some class D amplifier chips omit the PWM modulator, “banishing” it to a separate chip. This minimizes interaction between high- and low-level circuitry. More importantly, it also allows the amplifier to be built with high-voltage geometry. Outputs up to 600 W from a single IC become possible.
The newest type of switching amplifier accepts an I2S input, converting the I2S data into a PWM drive signal. These amps are easy to work with, as many DSPs and analog-to-digital converters (ADCs) provide an I2S data stream at little or no cost. Some even include EQ and dynamic-range control.
Figure 2 shows such a system. The PCM1808 ADC is the clock master, driving the TAS57xx. The MSP430 loads the signal-processing coefficients into the TAS57xx amplifier and monitors the analog inputs, using its internal ADC to turn the system on when an input signal is present.
Regardless of the kinds of inputs a switching amp accepts, it allows simple two-channel home audio systems (such as those for TV sets) to have a one- or two-chip solution—just the amplifier, or a low-cost ADC directly driving the amplifier.
Design Caveats
Most class D amplifiers have fixed gain. (The TPA3110D2 offers four preset gains.) If your DAC can’t supply sufficient voltage to drive the amplifier to full output, it’s not unlike driving a sports car with the parking brake on! Users then have three options.
First, they can use a DAC with a larger output voltage swing. Second, they can use an additional operational amplifier to boost the signal before the power amplifier. Third, they can use a class D amplifier that has external gain control, allowing the customer to “dial in” the exact amount of gain to take their maximum DAC voltage to maximum power amp output.
Also, most DACs are based on sigma-delta architectures, which generate significant out-of-band (above 22 kHz) energy. This energy is removed with a low-pass filter, so the class D amplifier doesn’t “reflect” it (alias it) into the audio band—not a nice sound at all! Figure 1 shows the connection between the audio codec (PCM3070) and the class D stereo audio power amplifier (TPA3110D2) in the Value Soundbar Reference Design. The LPF required is a simple RC filter.
Finally, a PWM data stream generated directly from an analog signal is fully analog. (The pulse widths vary continuously. They aren’t quantized.) Yet pulse waveforms represent the data. Rise and fall times and jitter become important. Layout is critical to get the best performance.
For a theater-in-a-box system, a minor reduction of sound quality caused by not paying attention to these issues probably isn’t important. Still, you should be aware of the potential problems.
Preventing ESD Damage
Real-world products require immunity to electrostatic discharge (ESD). When people wear rubber-soled shoes and walk across carpets, they generate static electricity that can damage low-voltage devices (discrete and integrated).
Where possible, add ESD protection diodes at inputs and outputs. These devices are designed specifically for ESD, with multiple diodes that conduct high voltages to the power supply rails, or through a small resistor to the ground rail.
In the Value Soundbar Reference Design, ESD diodes protect the ADC inputs of the audio codec, the coaxial input of the S/PDIF receiver, and the external GPIOs to the switches (Fig. 3). These are critical points where ESD can attack low-voltage ICs. The analog inputs are additionally protected by a shorting jack that grounds them when the signal cable is disconnected.
Power Supply Solutions
Switching amplifiers usually require three dc voltages. Two are a ± pair for the output rails, usually ±35 V or ±50 V. The third powers the digital (and low-level analog, if any) circuitry at 3.3 V or 5 V. The rail voltages are usually obtained from a power supply connected to the ac power line. The digital voltage is commonly derived from the positive rail voltage, using a switched-mode power supply (SMPS) or linear regulator (Fig. 4).
The rail supplies can be linear or switching. Linear supplies typically require large, heavy transformers, but their rectifiers and capacitors are inexpensive. Switched-mode supplies use small, light transformers, but more, and more expensive, electronic components are required.
Switching supplies are compact and cool-running—necessities if the supplies are mounted within the product. Voltage regulation is an inherent part of their architecture and doesn’t require additional power-wasting components. They can be designed to operate on line voltages from 100 to 250 V, without having to rewire the transformer. This universality makes it easy for a single product to serve domestic and foreign markets.
Switching supplies generate high-frequency noise. Close attention to board layout and bypassing is needed to keep noise out of analog circuitry.
Texas Instruments offers reference power-supply designs for AV use. One is a 720-W class G supply that can power many of TI’s class D amplifiers. A class G power supply uses power rail switching to minimize idle power losses in the power stage.
In this specific case, a logic input pin on the power supply can be used to tell the power supply that either a 50-V or 25-V supply is required. Halving the power supply voltage from 50 V to 25 V will divide the idle power consumption by 4. (P=V2/R)
“If we electrocute the user, he won’t buy any more of our stuff.”
The ac line (100 to 250 V) often directly powers consumer electronics. Line voltage is potentially lethal, so there are stringent Conformité Européenne (CE) and Underwriters Laboratory (UL) safety standards for line-powered products.
On the other hand, if an amplifier is powered by an outboard supply (“wall wart”), only the supply has to meet tight safety standards. As 30 V (ac or dc) is considered “safe” (it isn’t high enough to produce a strong shock sensation, and it definitely isn’t lethal), an amplifier powered by a 30-V external supply has safety standards that are more relaxed and easier to meet. You can start selling the product sooner, and there’s less trouble getting certification for foreign markets.
If you use an external power supply, you’ll probably want a commodity model from a reputable manufacturer. Keep an eye on the following:
- Output: If the amplifier has a continuous average output of 50 W (100 W peak), a 24-V power supply has to deliver at least 4.5 A.
- Voltage regulation: Voltage regulation can be important in systems that use open-loop class D amplifiers, which typically switch the power rail directly to the speaker load, so any PSU noise will couple to the speaker load. However, all of our analog input power amplifiers are internally closed-loop amplifiers. They will have good PSRR.
- Tolerance for abuse: How well does the supply tolerate shorts (one or many, brief or sustained)? “Accidents happen.” Good supplies bounce back from abuse. Bad supplies die.
For the Value Soundbar Reference Design, we used an off-the-shelf 24 V supply. Surprisingly, we had to go through three vendors to find a reliable product. My office looked like the elephants’ graveyard of power supplies.
Selecting A Regulator For The Low-Voltage Supply
Once you’ve found a reliable supply of the right capacity, you need to decide how to generate the low voltages. (Depending on the design, these might power analog as well as digital stages.) You can use inexpensive linear regulators (such as 78xx devices) or an SMPS. A linear regulator will usually be the least expensive solution, but pay attention to the regulator’s thermal and current limits.
If the circuit needs 5 V, and the supply is 24 V, the regulator drops the difference, 19 V. The current through the regulator is the current through the load, so a 1-A load would require a linear regulator to dissipate a blistering 19 W. For heavy loads, an SMPS is therefore the best choice.
If the load draws only 100 mA, only 1.9 W need be disposed of. A low-dropout (LDO) linear regulator might be the best tradeoff between cost and wasted power.
If the low-voltage circuits include analog processing, there’s a good reason for not using an SMPS. Switching noise can get into the analog circuitry. In such cases, an SMPS can drop the voltage to within a few volts of the desired value, with an LDO linear regulator finishing the job.
Each low-voltage stage can be powered with its own linear regulator. Though this minimizes the amount of heat each regulator has to dissipate, it doesn’t reduce the total power wasted.
There is no single choice of regulator quantity and type that’s appropriate for every application. The best choice will vary with the design and its budget.
Analyzing Power Consumption
Given the need to hold down cost and reduce energy waste, we analyzed the 3.3-V rail’s power consumption. Moving the PCM2705 USB DAC to host-powering from the PC provided a meaningful reduction.
There was no obvious way to reduce the LEDs’ power use (5 to 10 mA each), until we realized they didn’t have to be on all the time. They could be illuminated during changes, then switch off after a few seconds. This didn’t reduce the peak current required, but it significantly reduced long-term heat dissipation.
Our frugality reduced current consumption to less than 100 mA (when the LEDs weren’t on). This made it possible to use an inexpensive UA78M33 regulator to drop 24 V to 3.3 V. Only about 2 W had to be dissipated, so the PCB could perform heatsink duties. With a 30-W class D amplifier running at 90% efficiency, the amp generated more heat than the regulator.
A single linear regulator was the least expensive solution. In retrospect, a better solution would have been to use an SMPS to drop the voltage to around 5 V, followed with an LDO linear regulator to get 3.3 V. This would have given better efficiency, at a slight increase in cost.
Some customers asked why we didn’t use an SMPS to directly produce 3.3 V. As explained before, we didn’t want switching noise riding along the rails and references into the analog circuits. A final LDO regulator suppresses most of that noise.
Reducing Standby Power
There’s increasing pressure (especially in the European Union) for reduced standby power consumption. It’s now required to be no more than 0.5 W.
The soundbar’s external power supplies consume around 300 mW when idling, so the soundbar itself can waste as much as 200 mW. This is less than 9 mA from a 24 V power supply, so there’s almost no “reserve” after the 4.5 mA bias current of the regulator. Fortunately, the MSP430 host controller draws just 0.1 µA during standby, and the audio codec pulls a similar amount. The amplifier (when shut down) draws just 200 µA.
In Part 6 of this series, I’ll discuss the layout process and decisions made for optimum performance while maintaining as much flexibility for layout reuse for other SKUs.
References
- For more information about audio, visit : www.ti.com/soundbar-ca.
- Access parts 1-4 of this series here: Part 1, Part 2, Part 3, Part 4.