ABX Testing and Hi-Res Audio/Music

First things first, I continue to read headlines and other references to “high-res audio” when in fact, the author means to say “high-res music”. Remember that the DEG and major labels decided it would be less confusing to have TWO names and TWO logos for “high-res” experiences. So yesterdays discussion about how Amir has without doubt shown that “high-res audio is legit” besides being untrue also used the wrong terminology. High-resolution audio has already been proven. The term and associated logo refer to hardware that is capable of capturing and reproducing frequencies up to 40 kHz and dynamics of 130+ dB. Although, it is difficult to find equipment that can handle the very stringent requirements originally proposed by Sony…it can be done. We equipped a room that met these requirements at the AXPONA show back in April. Benchmark, Revel, DH Labs, and Dolby Labs collaborated on a setup that one attendee called, “clearly the best sound of the show”.

You may have noticed that I routinely use “high-res audio/music” to be safe. It’s unfortunate that I have to fold both terms into a single group, especially since they don’t share the same requirements, but what can you do? I haven’t seen the “hi-res music” logo used much yet, but I’m sure it’s coming.

All of the buzz following Michael Lavorgna’s article at Audiostream was quite a surprise to me. I haven’t followed any of the recent discussion going on at WBF, but I knew that Amir was claiming victory in the battle of “hi-res” perceptibility…by listening to a high quality recording of a set of keys! A recording that had been subjected to at least two sample rate conversions. I’m OK with a serious ABX test…I resist the “emo” type research that talks about “low level” details and other subjective non-measures.

So what would a real test be like. What method would I suggest…what procedures and results would I have to see to be convinced that everything I’ve been doing for the past 15 years means something. I can’t and won’t worry about the statistical methodologies and any of the other things that advocates on both sides of the issue quibble about. Here’s my approach:

1. Use only music recording that have been recorded using the highest quality equipment from microphones to a PCM digital recorder running at 96 kHz/24-bits.

2. Make sure that the studio or recording venue has an extremely low noise floor.

3. The performances music include instruments that have large amount of ultrasonic partials and the compositions/tunes must have extreme dynamics…both transient and gradual.

4. Passively split the outputs of the microphone preamps and send one to an ADC converter running at 96 kHz and the other to an identical ADC at 44.1 kHz. The two converters should be of the highest quality and calibrated.

5. Playback the two recorded streams through identical DACs into the ABX switcher, which should be balanced and completely noiseless when switched. Allow the person being tested to listen to each stream as long as they like and allow them to switch back and forth at will. They can choose the volume of the playback and modify it, if they prefer.

6. Make sure that the entire signal path is capable of achieving the hi-res audio requirements. This means very high-end speakers, extremely quiet and dynamic power amps, and high quality DACs. The power should be monitored and regulated.

The testing procedure would be explained to the participants (one at a time) by an uninterested third party. Participants would be able select A or B or “I don’t know”.

To be continued…