The PLL in your D/A Can Taint A/B Listening Tests!

A/B and A/B/X listening tests are important methods of comparing two audio sources or two audio components. In the studio an engineer may want to switch an effect or EQ setting on and off to decide if it contributes positively to a mix. Hi-Fi enthusiasts may wish to compare audio components, signal sources and interconnects. The DAC1 and DAC2 converters have input selector switches that allow fast and easy switching between signal sources. Before attempting to conduct these tests, it is important to understand how these converters and their switches work. A/B tests using the DAC1 input selector can be very misleading. In contrast, A/B testing using the input selector on the DAC2 will produce reliable results. This application note provides guidance for conducting reliable A/B or A/B/X listening tests with your D/A converter.

A/B Tests With the DAC1 - Caution!

The DAC1 converters allow instantaneous switching between digital sources but the audio unmutes before the low-jitter UltraLock PLL (phase locked loop) has fully settled. When we created the DAC1, some of our professional customers needed a D/A converter with a very short lock time. To this day, the DAC1 has one of the fastest lock times in the industry. Audio begins playing only a few milliseconds after a digital signal is applied, and this happens before the clock system has fully settled. This can cause a very slight pitch shift in the first 250 ms of playback. This pitch shift is imperceptible when switching between different musical selections, but it can become perceptible when switching between two identical (or nearly identical) signal sources. While it is convenient to run A/B tests with the input selectors on the DAC1, we do not recommend this! We strongly recommend switching the source within a computer, an upstream digital console, or DAW. The upstream switching will produce a continuous data stream to the DAC1 and the PLL will not be transitioning between unlocked and locked states.

False Results!

There have been a number of anecdotal accounts of optical and coaxial cables producing differing results when a DAC1 input selector switch is used to switch between two cables that are receiving identical data. Usually this is done by simultaneously feeding the DAC1 from two outputs on a disk player. Conducting an A/B test in this manner will give erroneous results with the DAC1 but not with the DAC2. The DAC1 unmutes just before the end of the pitch shift. The DAC2 unmutes just after the end of the pitch shift.

Pitch-Shift Artifact - The Cause of False Results

If you use the DAC1 input selector to conduct an A/B comparison, the very first thing you hear after the switch is a pitch shift that persists for about 250 milliseconds. It is a very short event, but your ears will detect the momentary change in pitch and this will color your opinion of what you hear after the pitch shift. It is a really interesting psychoacoustical effect and I don't know that anyone has really done any research on this. Nevertheless, the effect is real and it is easily reproducible. With two identical inputs to the DAC1, one will be perceived as being brighter than the other. For this reason, we do not recommend using the DAC1 input selector for A/B comparisons (use the console or DAW instead). Other brands of D/A converters have similar pitch-shift artifacts, but the magnitude and duration of these artifacts will vary. It is probably safe to say that most D/A converters will have pitch-shift artifacts that can produce false results in A/B tests (if D/A is switched between two different digital inputs).

A/B Tests With the DAC2 - Slightly Longer Mute Time

A slightly longer mute delay on the DAC2 prevents the erroneous results of A/B tests that are tainted by the pitch-shift transient of the PLL. The trade-off is that there is a momentary silence when switching. This muting can be avoided by switching within the DAW or digital console, because these devices normally provide a continuous digital data stream while switching between signal sources.

A/B/X Testing

Remote Control for QSC ABX Test SetIn an A/B test, signal source A is simply compared to signal source B. In an A/B/X test, the listener has access to A and B and an unknown X which is either A or B. On each trial, the listener must determine if X is A or B. The listener is free to switch between A, B, and X as many times as necessary before making a decision about X. X is randomly selected for each trial and the system keeps score to determine if the listener actually heard a difference. Rigorous A/B/X tests can scientifically confirm that a listener can hear a difference between two sources. We use these types of tests in our listening room here at Benchmark. Our listening room is equipped with a QSC ABX test set, DAW ABX testing, and software-based ABX testing. We also have an A/B system that uses multiple DAC2 converters. The photo shows the remote control for the QSC ABX system in Benchmark's listening room.

Rigorous A/B/X tests can also be conducted with specialized software running on a Windows, Mac, or Linux computer. Lactinato ABX/Shootout-er is a free software-based ABX system that operates across each of these three major operating systems. Other similar software solutions can be found with this search.

Avoid PLL Lock and Unlock Events when Conducting Listening Tests

The PLL lock times on the DAC1 and DAC2 are among the fastest in the industry. Unfortunately, low-jitter phase locked loops cannot lock instantaneously. A/B tests should be conducted in a way that prevents the locking and unlocking of a PLL. In general, D/A converters will produce some pitch-shift anomalies when switching between two different digital sources. The DAC2 mutes for about 500 milliseconds to eliminate the pitch-shift transient.  All of these effects can be eliminated by switching the audio source within a computer program or a digital mixer (instead of using the input switch on the D/A converter).

John Siau
John Siau


John Siau is VP and Director of Engineering at Benchmark Media Systems, Inc.