Benchmark has introduced the new HPA4 headphone/line amplifier with four 256-step relay-controlled attenuators and four 16-step relay-controlled boost amplifiers. Together these form two fully-independent, fully-balanced stereo volume controls. One volume control is dedicated to the line output and one is dedicated to the headphone output. Each has a +15 dB to -122 dB range in 0.5 dB steps. The lowest step provides a full mute. The volume controls feature high-precision metal film resistors, gold contact relays, and fully buffered inputs and outputs.
Our goal was to produce an analog volume control with the highest achievable transparency. We wanted to be able to place this volume control in front of our AHB2 power amplifier or in front of our THX-888 headphone amplifier board without diminishing the performance of either device. Our volume control would need to have lower distortion and lower noise than either of these amplifiers. Given the extraordinary performance of these THX-AAA amplifiers, this would not be an easy task!
The AHB2 power amplifier and THX-888 headphone amplifier have similar specifications; they are virtually noise free (SNR > 132 dB) and distortion free (THD < -120 dB). Our new analog-to-analog volume control would need to exceed these specifications to avoid being a weak link in the audio chain. As we looked at available solutions, it became clear that we would need to design something better.
The most basic analog-to-analog volume controls use variable resistors and these tend to fall well short of the required performance. The resistive elements introduce distortion due to changes in resistance caused by instantaneous heating and cooling at audio frequencies. This heating can be especially problematic where the wiper contacts the resistive element and where the element is terminated at each end. These thermally induced distortion effects can be reduced by decreasing the power dissipated by the resistor, but a reduction in power always reduces the SNR. However, noise and distortion are not the only problems. Stereo volume controls require a matched pair of variable resistors. Matching errors tend to introduce significant left-right balance issues at low volume settings.
There are a variety of integrated volume control chips that are appropriate for low-cost consumer audio equipment. Better chips have found their way into many hi-fi products, but they are not really appropriate for high-end applications. These chips include voltage-control amplifiers (VCA), analog multipliers, or multiplying D/A converters. These solutions offer low cost, remote control, and often have excellent left-right tracking. Unfortunately they lack transparency. None of the available chips are capable of meeting either of our noise or distortion requirements.
Digital DSP multipliers can achieve near-perfect transparency if enough of the output bits are preserved and delivered to a D/A converter with sufficient resolution. Any significant volume reduction would place difficult demands on the SNR performance of the D/A converter. In our analog-to-analog application, an A/D converter would need to be added at the front end. Any significant volume boost would place difficult demands on the SNR performance of the A/D converter. In any practical system, the converters are the performance limitation. This is why we provide analog attenuators at the output of all Benchmark DAC1, DAC2, and DAC3 converters. The attenuators match the nominal output level of the DAC to the downstream device (usually a power amplifier). This matching minimizes the use of the digital volume control. In our analog-to-analog application, the cascaded A/D-D/A would add excessive THD and noise.
In theory, switch-controlled and relay-controlled resistor dividers can provide near-perfect performance. We purchased a number of these and were surprised to find that all had design problems that were limiting their perfomance. In most cases, the impedances were too high to achieve our required noise specifications. When we replaced the resistors with lower values, other problems appeared. The lower impedances exposed crosstalk problems that were caused by circuit board layout issues. They also tended to overload upstream devices, causing distortion. All of the units we tested produced some pops and clicks in the output audio as the volume was adjusted. Clearly we would need to take a different approach if we were going to use relays or switches.
We settled on a very low impedance 16-step gain stage that feeds a very low impedance 256-step attenuator. Each channel uses 12 DPDT gold-contact relays and 64 high-precision resistors. All stages are fully buffered and fully balanced. The balanced design provides a 3 dB SNR improvement while increasing the immunity to interfering noise sources. The buffers provide high impedance inputs and low impedance outputs. This means that the Benchmark gain control places very little load on upstream devices while providing ample drive for downstream devices. The buffering also allows the use of very low resistance values and this reduces the thermal noise produced by the attenuator.
We eliminated the usual pops and clicks that are produced by relay-controlled attenuators. To do this, we chose to drive the relays with an FPGA instead of a microprocessor. The FPGA allows very precise timing of every relay closure. We use very fast make-before-break contact closures to prevent gain overshoot transients while the relays are switching. Due to this precise timing, the Benchmark attenuator and gain stages can be rapidly adjusted without producing pops, clicks, or zipper noise in the output audio.
Resistors can cause distortion when resistance values change due to instantaneous heating at audio frequencies. To eliminate this potential problem we use precision thin-film resistors that have been selected for their excellent temperature coefficients. The resistors are also physically larger than what would be necessary to handle the power dissipation. The larger package size reduces the thermal effects.
Most people understand that active buffers create some noise and distortion. But many people falsely assume that these active components are the only source of noise and distortion in an audio circuit. The truth is that passive components are usually the primary source of noise and distortion in an audio circuit!
In a search for perfection, many audiophiles have bought into the "fully passive" myth. I suspect that this explains why most relay attenuators are fully passive. On the surface, "fully passive" sounds like a good idea.
What many audiophiles don't realize is that a single resistor can create far more noise than a good active buffer. For example, an LME49860 opamp produces about the same noise as a 440-Ohm resistor. This means that the LME49860 is 13.6 dB quieter than the thermal noise produced by a single 10 k resistor! Buffers can allow the use of much lower impedances in an attenuator's resistor ladder network. When the impedances are reduced, an active design can be far quieter than a fully-passive design. Low-noise can only be achieved when resistors have very low impedances and signal levels are kept high.
If the impedances are reduced in an fully-passive attenuator design, the noise can be reduced but distortion will increase. This distortion is caused by an overloading of the upstream device. When an audio output is overloaded, distortion increases. Contrary to popular myth, distortion can be reduced by adding high-quality active buffers.
A third problem with passive attenuators is that they may not provide enough drive for downstream devices. Cable capacitance and input capacitance can cause a high-frequency roll off. Worse yet, the frequency response may change as the volume is adjusted with a passive attenuator. Low output impedance is important for maintaining a wide bandwidth. Output buffers are essential for delivering a predictable frequency response.
In the fully-passive designs we examined, the thermal noise produced by high impedance resistors exceeded what could have been achieved with a well-designed fully-buffered design. Furthermore, the loading imposed by the passive attenuators tended to cause distortion in upstream devices. It was clear that buffers would be required in our design.
We chose the LME49860 opamps because of their outstanding THD and IMD performance, their wide voltage swing, and their ability to drive heavy loads. The 16-step gain boost ladder sits inside the input buffer's feedback loop. A second buffer drives the 256-step attenuator input, while a third buffer drives the output. Each buffer is actually a balanced pair of buffers and the entire signal path is fully balanced. The second buffer also serves as a precision fully-balanced differential amplifier. The differential amplifier delivers a voltage-balanced signal to the attenuator stage. The differential amplifier removes common-mode interference while maximizing the use of the voltage headroom in the 256-step attenuator.
It is very unusual to see a stepped attenuator with more than 64 steps. The reason for this is that the resistor precision must increase dramatically as the number of steps increase. To achieve a 256-step attenuator with consistent 0.5 dB steps, we needed to use very high precision 0.1% resistors. If we had used 1% resistors, the step sizes would have been inconsistent and there would have been variations in the left-right balance as the volume was adjusted.
The 0.1 % resistors maintain a precise left-right balance over the entire 128 dB range of the Benchmark volume control. The gain and attenuator stages have a dual mono construction and this allows independent adjustment of the left and right channels. We leverage this feature to provide a balance control that tracks over the entire gain range.
Measured from balanced inputs to balanced outputs on the HPA4 Line Amplifier:
At Benchmark, listening is the final exam that determines if a design passes from engineering to production. When all of the measurements show that a product is working flawlessly, we spend time listening for issues that may not have shown up on the test station. If we hear something, we go back and figure out how to measure what we heard. We then add this test to our arsenal of measurements.
Benchmark's listening room is equipped with a variety of signal sources, amplifiers and loudspeakers, including the selection of nearfield monitors shown in the photo. It is also equipped with ABX switch boxes that can be used to switch sources while the music is playing.
Benchmark's lab is equipped with Audio Precision test stations that include the top-of-the-line APx555 and the older AP2722 and AP2522. We don't just use these test stations for R&D - every product must pass a full set of tests on one of our Audio Precision test stations before it ships from our factory in Syracuse, NY.
Paul Seydor of The Absolute Sound interviews John Siau, VP and chief designer at Benchmark Media Systems. The interview accompanies Paul's review of the LA4 in the December, 2020 issue of TAS.
"At Benchmark, listening is the final exam that determines if a design passes from engineering to production. But since listening tests are never perfect, it’s essential we develop measurements for each artifact we identify in a listening test. An APx555 test set has far more resolution than human hearing, but it has no intelligence. We have to tell it exactly what to measure and how to measure it. When we hear something we cannot measure, we are not doing the right measurements. If we just listen, redesign, then repeat, we may arrive at a solution that just masks the artifact with another less-objectionable artifact. But if we focus on eliminating every artifact that we can measure, we can quickly converge on a solution that approaches sonic transparency. If we can measure an artifact, we don't try to determine if it’s low enough to be inaudible, we simply try to eliminate it."
- John Siau