July 13, 2018

MQA Music production workflow for professional [ Technical seminar report ]

by Bob Stuart and Mick Sawaguchi

This is a seminar report in which held at InterBEE in 2017 November. On the consumer side, MQA (Master Quality Authenticated) technology has attracted attention for several years but since there was not much opportunity to introduce it systematically to the domestic production side, we had a session at the professional exhibition called InterBEE this time has been held.
The session was held on the afternoon of November 15th, when Bob Stuart, MQA developer and C. E. O, came to Japan on MQA's basic technology and MQA-CD production applying it. I also focused on MQA's technology, continuously released albums in MQA format from UNAMAS-Label, and released the world's first MQA-CD [A. Piazzolla a Strings and Oboe] from OTTAVA-Records So I will introduce the overview as I participated in the lecture.

1 Purpose of MQA coding and trend of music

Bob: Thank you for coming all the MQA sessions. Today, I am very pleased that we have the opportunity to systematically introduce MQA to the professional people on the production side and what kind of application it is.

First, let's see the relationship between quality and convenience in music software for the past 50 years.

Analog ·open reel tape can say that quality is equivalent to master, but it is not easy to handle. The quality of the cassette tape is not good, but it can be handled easily. Although DVD-A and SA-CD are high quality, this also requires a special player. Internet stream delivery is extremely convenient, but the quality is not good.

So what is the goal that MQA aims for? As the goal is to have high quality and convenience as shown on the right side of the figure MQA has developed three goals for this purpose. I will introduce MQA approaches in the following order.
2 Studio Master for quality:
MQA-De-Blur technology that emphasizes time axis information rather than frequency domain

Music production by high bit / high sampling exceeding CD quality has been done and sound source has been spreading in the form of downloading. The figure shows the result of subjective evaluation of the relationship between various formats and the quality obtained by our top mastering engineers in Europe and America.

From this result, we can see that no significant improvement in quality has been seen for the amount of data at 96 KHz sampling or later. In general, I thought that recording high-bit and high sampling improves frequency characteristics and improves quality. What we focused on is the result of a recent brain hearing study. That is, "accurate reproduction of time axis characteristics is an important factor to quality" beyond the expansion of the frequency domain. It is said that high bit / high sampling is being performed because it records this time axis information accurately as it makes high bit / high sampling.

The time axis is accurate means that the precision with which the sound source can be perceived correctly improves, and in the case of music it means that the sound quality, localization, depth, distance feeling can be accurately reproduced. Human auditory detection limit is said to be 5 μsec - 8 μsec. This means that you can detect sound sources up to distances of 1 m - 2 m, centered on yourself.

This figure shows the accuracy of the time axis characteristics at Fs: 48 KHz and 192 KHz.

MQA developed with the goal of reducing the fluctuation of the time axis fluctuated in the workflow from recording to playback to the extreme limit and delivering it to the end user.
We call this "De-Blur" [De-Smear].

That is, it is a technology to reproduce the time axis of the digital signal A / D converted in the recording to the same time axis information at the D / A conversion at home.

The accuracy of time axis reproduction of MQA is currently up to 10 μsec resolution, and in terms of distance it can detect distances up to 2.5 m.

3 Coding technologies to realize convenience Audio Origami

In the era when LP and CD were mainstream of music software, I was able to enjoy music simply with just one format. When it comes to the age of digital audio, a vast variety of formats ranging from compressed sound sources to Hi-Res sound sources are flooding, and listeners are confused in the playback equipment. We will provide music in a standardized format as much as possible and consider how to do so in order to be of high quality and the result will be the method called Audio-Origami, which will be described later.

The research, which became the basis, gained a hint from the analysis result of many Hi-Res music files. The average frequency of the music signal analyzed so far has a distribution of triangles we call Golden-Triangle.

In case
The high-frequency region in the high-bit-high-sampling large data volume means that there are many margins. The principle of Audio-Origami is that by dividing this music signal area into three areas ABC, it folds the information which each has, while enciphering while reducing the total amount of data, and at the time of decoding it opens the folded data It is the idea of ​​handling master quality with light data.

Sampling frequency Vs In the graph of sound quality, normal high sampling ·
Although the sound quality is gradually improved from 96 KHz if it becomes high bit, it becomes almost constant from 96 KHz in MQA encode ·decode data. The graph of red is a graph when playing with existing playback equipment without decoding MQA data, but also in this case the sound quality is also improved as the blur of the time axis information is eliminated.

The graph on the upper right shows how far the blur on the time axis (Temporal Blur) is distributed. For normal high bit / high sampling etc. SF of about 1536 KHz is required to realize 10 μsec, but in MQA it is almost 10 μsec after the master of 192 KHZ.
The graph on the lower left shows the data transfer rate. In normal digital format, the high sampling rate and high bit will result in a huge transfer rate, which requires time for downloading, but the MQA file is stored at almost 1.4 Mbps. If it is the current network infrastructure, it can be handled as very light data.

How do you embed B and C area signals to a signal level of -140 db or less? And how does De-Blur work? Although I will ask you a question like this, I will spend a lot of time when explaining this principle, so I will skip this time but since I can download MQA's WEB commentary and AES paper freely, I refer to it please. I think that an article translating my paper by Mr. Bike Suzuki to the JAS journal will be helpful.

REF:Stuart, J. R. and Craven, P.G., A Hierarchical Approach to Archiving and Distribution, 9178, 137th AES Convention, (2014). http://bit.ly/JRS_AES

Oppenheim, J. M., et al., Minimal Bounds on Nonlinearity in Auditory Processingq-bio.NC (Jan 2013). arXiv:1301.0513 

Lewicki, M.S. Efficient Coding of natural sounds, Nature Neurosci. 5, 356363 (2002). 
Jackson, H. M., Capp, M. D. and Stuart, J. R., The audibility of typical digital audio filters in a high-fidelity playback system, 9174, 137th AES Convention, (2014).

Stuart, J.R., Sound Board: High-Resolution Audio, JAES 63 No 10 pp. 831-832; (October 2015) http://bit.ly/AES_Soundboard
4 Compatibility with existing media MQA-CD

MQA data is characterized by compatibility with existing formats. Here we introduce the workflow and various tools for encoding to MQA file. MICK SAWAGUCHI will introduce the workflow of MQA-CD production after this.

5 Practice of MQA-CD production workflow

From here, Mr. Mick Sawaguchi of UNAMAS-Label will introduce MQA-CD production practice and points of note. In UNAMAS-Label, the sound source (192-24 WAV format) of distribution downloading is also released at MQA, but in June 2016, the world's first MQA-CD was released from OTTAVA RECORDS, after that the CD of OTTAVA RECORDS, Since it is released as MQA-CD, here we introduce the workflow of MQA-CD.

So, we prepared the actual sound source, so we added the same sound source for you.

1 192-24 WAV Master
2 MQA-CD-MQA decode reproduction (decoded and played from 44.1 - 16 to 176.4 - 24)
3 Normal playbacks without MQA-CD-MQA decoding (played as normal 44.1 - 16)
4 Sound sources produced in normal CD format

Please listen with the type of. Even if you do not decode MQA-CD, since blurring of sound is improved by De-Blur, I think that you can hear that it is superior in localization and sound field reproduction.

Bob: Everyone, thank you very much. I also thank Bike Suzuki of MQA-JAPAN who interpreted my academic story.