Correctly calculate DAC power consumption data

As designers of portable multimedia systems push battery life to its limits, they are spending more time than ever before poring over power consumption data from different silicon vendors. Tit-for-tat comparisons are often difficult because there are so many variables, and key differences between competing devices are often far from obvious.

Audio input and output subsystems are particularly difficult because they contain both analog and digital circuits and often require several different supply voltages. As a result, the data provided by manufacturers for these devices is often irrelevant to the actual use case and, in some cases, is downright misleading. However, a basic understanding of the relevant circuits, a solid understanding of Ohm's Law, and a refusal to take manufacturers' data at face value can help design engineers see through this confusing fog.


What exactly is included in each power figure?

It may seem obvious, but understanding what circuits are included in each power figure is key to calculating the overall power consumption of the system. However, this is often easier said than done when you have only a data sheet to work with. Now let's think about the audio output of a portable system. Figure 1 shows all the major functional blocks. The last few blocks in the chain (such as digital signal enhancement, DAC , analog mixing, and amplification) are usually integrated into a single device, loosely referred to as an "audio DAC ". However, when the data sheet for such a device lists "DAC power consumption" or "DAC supply current", it absolutely refers only to the DAC itself and does not include the amplifier and other circuitry. So what if it says "playback to headphones"? Would that include on-chip signal enhancement features such as limiting, 3D signal enhancement, or equalization? Most likely not, as silicon vendors rarely have the guts to make their devices look inferior when compared to their competitors. Some silicon vendors even specify that the DAC supply current does not include the digital audio interface. Obviously, this has no resemblance to any real use case, as the interface must be powered up to receive audio data for playback. To complicate matters further, the system architecture of these devices is also different. For example, volume control can be implemented in software on the CPU, in the digital portion of the audio chip, or with an analog gain programmable amplifier in the audio chip. A good sanity check is to determine what functions are required, check in which physical device these audio functions are implemented, and ensure that the power consumption of each function has been accounted for. The power consumption of speakers and headphones usually accounts for a large part of the overall power consumption. Since this power is not actually consumed in the IC, it is almost never included in the IC data sheet. Fortunately, it can be easily calculated from the formula P = V2RMS / Z, where VRMS is the RMS voltage across the speaker and Z is its impedance (don’t forget to multiply this by 2 for stereo speakers!). The hard part is choosing an actual VRMS. Although the maximum VRMS can be easily calculated from the swing of the amplifier output, in reality VRMS depends on the volume setting of the end user. Even at maximum volume, the VRMS on the treble and bass channels of the same piece of music will be different, so assuming a full-scale signal is almost impossible. In order to make a meaningful comparison between different audio components, a common reference is needed. For example, the Japanese JEITA CP-2905B standard states that the battery life of a system with headphone outputs should be measured while driving 0.2mW (0.1mW per channel) into a 16Ω load. What is this signal? The amplifier driving the speakers and headphones is another particularly power-hungry device. The current industry common practice is to list their quiescent power consumption, that is, when playing absolutely quiet (represented in the digital domain as a string of zeros). However, whenever there is an actual signal passing through the system, the power consumption in the amplifier (and the load) will increase. Of course, the amplifier supply current should be expressible as a non-zero signal, but what signal should it be? Some standards (such as JEITA CP-2905B) often use a 1kHz sine wave because it is easy to generate. However, it bears little resemblance to any sound or music heard by real-world users. Pink noise (as defined for loudspeakers in the IEC 60268-5 standard) might be a closer approximation to the amplifier supply current, although fundamentally no signal can map the infinite variations of music.

Another thing worth keeping in mind when comparing amplifiers is that their power efficiency depends on signal amplitude. The exact relationship depends on the amplifier (see Figure 2). For example, under quiescent conditions, a Class D amplifier may consume more power than an equivalent linear amplifier due to switching losses. Similarly, since linear amplifiers are more efficient at high volume, their efficiency at full scale can approach that of a Class D amplifier.

However, these extremes of signal amplitude are largely irrelevant because the battle for battery life is fought in the middle of the signal amplitude, where real-world amplifiers spend most of their time. This is where Class D amplifiers have won widespread acceptance in the industry, as their power conversion efficiency is much higher than that of linear amplifiers.


What about digital circuits?

Amplifiers are not the only circuits that consume less quiescent power than active power; other analog circuits (such as mixers and gain-programmable amplifiers) and digital CMOS circuits do the same. For CMOS circuits, power consumption is largely a function of the frequency of 1- and 0-state bit transitions, so a signal consisting of only 0-state bits (i.e., quiescent) requires only very low supply current. To get meaningful data, all devices should process a true nonzero signal.

Another factor to consider is the sampling rate of the digital audio signal. Most digital and mixed-signal circuits convert once per sample, so their average power consumption is directly proportional to the number of samples per second. When comparing audio DACs or ADCs , you should pay attention to whether the power supply current specified in the data sheet is based on the same sampling rate.

Further up the signal chain, the quality of the encoding of the source audio file (e.g., the bit rate of an MP3 file) can affect the power consumption of the decoder. The bit rate and buffer size determine how often data must be retrieved from the storage medium. This is particularly important in hard disk-based systems, as each disk read results in a large spike in battery current.

Many audio ICs (e.g., DACs or ADCs) can be configured as either masters or slaves. In master mode, the audio IC drives the digital audio interface and therefore draws more current than in slave mode, so, not surprisingly, their power consumption in the datasheet is usually measured in slave mode. So, does this mean that slave mode is always preferred? Of course not. After all, if the audio IC does not drive the interface, the counterpart on the other end must do that work, so power consumption is simply moved from one end of the system to the other, not eliminated.

When the datasheet states power consumption in master mode, it is important to pay attention to the load capacitance, as this determines how much additional current is required. If the datasheet numbers are assumed to be measured with a "worst case" large load capacitance, the reality may be much better than the datasheet specifications. The opposite is true, however, and IC vendors may artificially manipulate these power figures by using unrealistically low load capacitance.

Some audio devices have special clock modes that eliminate the need for a very power hungry low jitter PLL, but this mode can only be used in master mode. For example, many Wolfson
audio DACs and CODECs have a "USB mode" in which the audio clock is generated directly from a 12MHz USB clock. In this case, the power saved by integrating the clock often far exceeds the power consumed in the audio IC.

Power Supplies

All but the simplest audio ICs use more than one power rail. A typical circuit includes at least one analog supply, a digital I/O supply for the audio and control interfaces, and a separate digital core supply. The total power consumption of the IC is the sum of the power consumed on each supply rail. One of the most obvious ways

to save power is to use the lowest possible voltage for each supply. For the digital I/O voltage, the lower limit may be dictated by the other system components to which the audio IC must interface. On the other hand, the digital core voltage can use a lower voltage that is usually stated under "recommended operating conditions" in the data sheet.

Some data sheets contain a graph of supply current versus voltage for a given mode of operation. If the data sheet does not have such a graph, you can make some reasonable logical assumptions. For digital logic such as CMOS ICs, the current is proportional to the applied voltage. This means that reducing the voltage can get you two times the benefit, that is, reducing the supply voltage by half can actually reduce the power consumption of this supply rail by three-quarters. For

analog circuits, things are more complicated, because analog circuits often contain constant current sources. After halving the analog supply voltage, the power consumed by the analog portion of the IC (excluding any load) is usually between one-quarter and one-half of the original power consumption. A clearer

understanding of the power consumption data in the data sheet

In order to make a true and meaningful comparison of the power consumed by different audio ICs, the test conditions between different audio ICs must be realistic and consistent, including the power supplied to the load, the nature of the signal (such as pink noise), the sampling rate, and the supply voltage.

In addition, the functionality must reflect the expected actual application situation, all the required functions on the IC must be fully enabled, and any unneeded functions must be turned off as much as possible. The digital interfaces of the audio ICs to be compared should all operate in master mode, or all operate in slave mode, and the load capacitance should be the same in each case. The master clock of each IC should also be the same, and if a PLL needs to be clocked from the audio clock, its power consumption should also be calculated.

Of course, in real life, different vendors tend to use different test conditions for their audio ICs. However, if they know which factors have the greatest impact on power consumption, system designers can quickly identify some key indicators and extrapolate some important data from the vendor's test conditions based on their own actual application. This allows them to have an in-depth view of IC power consumption, which is much more meaningful than the "headline" specifications that can often be found on the first few pages of the data sheet.