Volume Control Technologies

With the introduction of Benchmark’s HDR-VC™ (High Dynamic Range Volume Control), many audio enthusiasts are taking another look at volume control methods. What are the differences between the different volume control topologies? Is one sonically superior, and why?

The important qualities to consider with respect to volume control are:

• Signal-to-noise ratio (SNR), also known as dynamic range
• Total harmonic distortion + Noise (THD+N)
• Frequency response
• Inter-channel gain matching

There are four common methods of volume control:

• Digital attenuation (via digital signal processing, or ‘DSP’)
• Analog integrated circuits (IC)
• Passive attenuators
• Active gain circuits

Overview of Qualities Affected by Volume Controls

SNR, Dynamic Range, Gain Staging, and the ‘Ceiling-to-Floor’ Analogy

‘Dynamic range’ and ‘signal-to-noise ratio’ (SNR) specifications describe the same quality. They both indicate the level of the component’s noise floor with respect to the highest possible signal level. ‘Noise floor’ describes the ‘random’ noise which is intrinsic in the component regardless of the audio material being played.

SNR measurements are conducted by measuring the noise floor without any signal present. However, this method proved to be inadequate for devices that contained auto-mute technology. These auto-mute circuits would engage when no signal was present, and the ‘noise floor’ measurements would be deceivingly low. Dynamic-range measurements were introduced as an alternative to SNR measurements. Dynamic-range measurements are conducted by measuring the noise floor with a low-level test signal that prevents auto-mute from engaging. For the purpose of this article, ‘dynamic range’ will refer to the noise floor of an audio device relative to the maximum undistorted output level.

A volume control’s effect on dynamic range can be explained via the concept of ‘gain-staging’. Gain-staging is the practice of optimizing head-room and dynamic range by properly aligning the levels of each active gain stage and/or attenuation stage (active and/or passive). A simplified description of proper gain-staging is: coordinating all devices to reach maximum input/output at the same time, operating in the upper regions of those input/output levels, and, when needed, attenuating near the end of the signal chain.

A usable analogy to dynamic range is the height of a room, floor-to-ceiling, where the ‘floor’ represents the noise floor of an electronic device, the ‘ceiling’ represents the highest signal amplitude possible, and the ‘height’ represents the dynamic range. An active audio device with analog signal paths will have an intrinsic noise floor due to thermal noise and other electronic factors. The ‘height’ between the ‘floor’ and ‘ceiling’ of a signal stage will remain consistent, but this ‘floor-to-ceiling height’ will only be utilized if the signal path has the proper gain-staging and SNR. This analogy will be used later, when discussing specific volume-control implementations.

Total Harmonic Distortion + Noise (THD+N)

All audio devices add errors to a signal. These errors are categorized as ‘noise’ and ‘distortion’. A device’s THD+N measurement quantifies the total errors added by the device.

THD+N artifacts can be symptoms of many different phenomena: harmonic distortion, inter-modulation distortion, jitter-induced distortion, non-linear frequency response, crosstalk, intrinsic noise, external noise, etc. For a given THD+N ratio, these errors may have very different thresholds of audibility.

Frequency Response

A device’s frequency response is a measurement of deviation from linear response. A perfectly linear device would reproduce all frequencies equally. The amount of gain or attenuation would be the same for all frequencies of an input signal in a device with linear frequency response.

In reality, every device has a limited bandwidth. In addition, the accuracy of the frequency response within the pass-band varies among components. An example of an intentional non-linear response is a device with a ‘bass boost’. In that case, the low frequencies will have more gain than other frequencies within the signal.

However, many components have unintentional irregularities in their frequency response. For example, some volume control circuits can change the frequency response as the volume is changed.

Inter-Channel Gain Matching

‘Inter-channel gain matching’ describes a device’s ability to achieve precise matching of separate channel output levels (i.e. between left and right in a stereo device).

Overview of Volume Control Implementations

Digital Attenuation (DSP)

DSP-based volume control of digital audio can be implemented via hardware (DSP in a chip) or software (DAW, media player). It performs multiplications on the digital audio data before the data is delivered to the D/A converter.

In most cases, a DSP-based volume control will limit the dynamic range of the playback system. This is because it doesn’t utilize the entire dynamic range of the D/A converter. Going back to the ‘height of a room’ analogy, the maximum level that a D/A converter can produce is the ‘ceiling’ (or ‘headroom’). The converter reaches the ceiling when a ‘full-scale’ (the highest possible digital amplitude) signal is present at the input.

If the digital signal is attenuated before it reaches the D/A, its peak amplitude will be further removed from full-scale. In other words, the peak output from the D/A converter will be further below the ‘ceiling’, but the ‘floor’ (inherent noise) will remain. In our ‘room analogy’, digital attenuation added a drop-ceiling to our converter, and the new ‘height of the room’ (dynamic range) has been reduced.

Also, when digital attenuation occurs, dither (noise) should be applied to avoid quantization distortion (distortion from quantization errors is beyond the scope of this article, but worth investigating). This dither can lower the SNR of the audio signal because new noise is added to a lowered signal. This is a serious noise contribution because, if the output of the digital volume control is 16 bits, the dither noise is at -96 dBFS (96 dB below full scale). However, if the output is 24 bits, the dither noise is at -144 dBFS. At 144 dB below full scale, this will not add significant noise to the system.

These dynamic-range limitations become less of a problem in D/A converters with significantly high dynamic range, since a playback system will only be as quiet as its noisiest component. In other words, if a system has a power amp with a 100-dB dynamic range, it won’t matter if the dynamic range of the D/A is reduced from 125 dB to 110 dB. The dynamic range of that system will still only be 100 dB. Generally, the higher the dynamic range of the D/A, the more digital attenuation can be applied without affecting the dynamic range of the entire playback system.

In addition to dynamic-range limitations, one should also be concerned about distortion induced by an inferior DSP algorithm. If the designer does not implement proper dithering, severe non-harmonic distortion will occur. Many computer playback systems lack dither. 16-bit systems have noticeable distortion when dither is omitted. 24-bit and 32-bit systems are much more forgiving when dither is omitted.

Other errors in DSP implementation can also result in distortion. However, properly-designed digital volume controls are becoming more common. A well-designed digital volume control will add virtually no distortion to the audio.

DSP-based volume control can work well provided that the following conditions are met:

1. Assure that the digital volume control is properly designed. This can be a difficult task for the end user, as it requires special test equipment. Benchmark has tested certain media players and other digital devices. The results are posted to a public wiki, which is updated as new test results become available. The website is http://www.BenchmarkMedia.com/wiki.
2. The D/A converter must have a spectacular dynamic range. It is not uncommon to require 20-30 dB of attenuation for normal listening levels. In those cases, the dynamic range will be reduced by 20-30 dB. If your converter only has a 110-dB dynamic range, the output will have a dynamic range of 80-90 dB – a dynamic range less then that of a 16-bit CD.
3. The peak level of the D/A converter’s output should match the maximum input level of the next device in the signal path (amplifier, pre-amplifier, etc). This is a fundamental part of proper gain-staging, as it fully utilizes the headroom of both devices. This is the reason that professional audio facilities will standardize the operating signal level between audio devices (usually at +4 dBu at -20 dBFS). With this type of configuration, the dynamic range of the amplifier will usually be the dominate noise factor.

Analog integrated circuits (IC)

Many devices implement an integrated circuit (IC) to manipulate volume in the analog domain. Analog IC’s suffer from noise and distortion due to complications of implementing many precision elements within a single chip package. Specifically, the size and proximity of these elements within the IC severely limit the quality of the elements.

The performance of analog volume-control IC’s are determined by these factors:

1. Crosstalk between electronic elements within the IC chip
   When small-form electronic elements are packed into a single wafer (chip), they are very susceptible to crosstalk. Crosstalk is a phenomenon that occurs via substrate coupling (voltage coupling from one node to another through the substrate). The amount of crosstalk is proportional to the proximity of the nodes. Therefore, small IC packages are extremely susceptible due to the components proximity to each other.
2. Power consumption and noise
   A trade-off relationship exists between power consumption and noise. Typically, an IC will be designed to minimize power consumption. This requires high-impedance circuitry. However, ‘Johnson noise’ increases with increased impedance. In other words, a high-impedance circuit causes more noise. In contrast, a low-impedance circuit will cause more power consumption, which will result in more heat. The elements are so small that significant heating and cooling of individual elements can occur at audio frequencies. This heating and cooling alters the resistance and capacitance of these elements. As a result, the response of the components becomes non-linear, which causes distortion.
3. Non-linear behavior of elements within the IC chip
   Non-linearity can be caused by a variety of issues. The issue of non-linearity due to thermal variation was discussed above. Also, IC’s have parasitic leakage paths between circuit elements. Leakage currents and DC-offsets cause non-linear behavior. Non-linear behavior results in distortion.

Passive Attenuator

A passive attenuator is simply a resistor network or potentiometer creating a voltage divider in the signal path. The output of a voltage divider is a scaled version of the input signal. A passive attenuator uses only ‘passive’ components, which are components that do not require a power source.

Passive attenuators have a reputation of being completely benign. However, a poorly designed passive attenuator can be detrimental to the quality of the audio. Passive attenuators can add noise and distortion, and they often change the frequency response of the system. Passive attenuators with high impedance (greater then 500 ohms) are particularly problematic.

The following characteristics dictate the performance of a passive attenuator:

1. High impedance (high-Z) will reduce common-mode rejection ratio
   Common-mode rejection is the main purpose of a properly balanced system. In a perfectly balanced system, the differential input of the device will cancel any noise that is common on both transmission lines of a balanced connection (e.g. XLR). The result of this process is a signal that is free from artifacts that were not present at the output of the previous device. A device’s capability to achieve this is described by its ‘common-mode rejection ratio’ and varies based on several factors. In general, low source impedance and high load impedance will result in a higher common-mode rejection ratio. Therefore, for the sake of common-mode rejection, it is undesirable to increase the source impedance with a hi-Z passive attenuator.
2. Variable impedance will alter frequency response
   The output impedance of passive attenuators often varies with volume setting, which causes a change in frequency response as the volume setting is changed. The output (source) impedance of a device, in conjunction with the capacitance and/or inductance of the load (the next device downstream), will create a filter. Different impedance values will cause different frequency responses. For example, a higher impedance and capacitance will result in a lower threshold frequency of a low-pass filter where attenuation occurs. This becomes a problem when the threshold frequency approaches and enters the audible range (below 20 kHz). If the cable and load capacitance approaches 500 pF, and the impedance of the passive attenuators is 15 kOhm, there will be 3 dB of attenuation at 20 kHz!
3. High-Z will increase channel gain differences
   High source impedances will have more affect on the gain of the following device’s input. Any slight differences of source impedance between the left and right channels will result in differences in gain between the two channels.
4. High-Z will increase thermal noise
   All resistors have a noise element that is referred to as ‘thermal noise’ (or ‘Johnson noise’ or ‘Johnson-Nyquist’). This noise is due to thermal excitation of electrons, causing random fluctuations in voltage. When a resistor is in series with the signal path, the noise is added to the signal. The amount of the noise is proportional to the resistance. Therefore, passive attenuators with high impedance will add significant noise to the circuit.
5. High-Z will affect inter-channel gain
   The load’s input gain is determined by the resistor network feeding the input, which forms a voltage divider. The load impedance will usually dominate that relationship because it will be several orders higher in magnitude. However, if a high-Z passive attenuator is used, it will play a bigger role in the voltage divider. Consequently, subtle variations in resistor values can affect the inter-channel gain.
6. Low impedance (low-Z) can overload the output drivers
   Using low-impedance passive attenuators is a double-sided sword. They present a very difficult load to the source. Since ‘impedance’ means resisting current flow, ‘low-impedance’ loads will present little resistance to current flow. This means that the load draws a lot of current from the output stage of the source. If the output stage of the source cannot drive high amounts of current, a low impedance load may cause severe distortion in the output stage.
7. High-quality resistors must be used
   Lower-quality resistors will add more thermal noise and variation to the electrical signal. Resistors must be precision thin film or metal film type to maintain low noise and low distortion.

Active Gain Circuits

Active gain circuits are simply line-level amplifiers. They work by sending a signal into an active amplifying component (tube/transistor/opamp/etc), whose gain is often adjustable via a potentiometer.

Certain active gain circuits may suffer from noise and distortion, if not designed and built properly. A properly designed active gain circuit will add very little noise or distortion to the audio, and can easily outperform the other volume control methods in these respects.

Active gain circuits can rescale the ‘ceiling’ and ‘floor’ of the source device to fit exactly within the ‘ceiling’ and ‘floor’ of the driven device, while adding very little additional noise or distortion. Frequency response is constant at all gain settings. Also, output impedance is very low and is constant at all settings.