Dec 2000

The Goldmund/Job/Stellavox Fundamental R&D

Some preliminary precautions...

Fundamental Research has nothing to do with product development. We have, since Goldmund came to Switzerland in 1980, clearly separated the product development departments, different for each company because they develop products for different fields, from the Fundamental Research which is common to all our brands.

Fundamental Research has nothing to do with technology development. We are experimenting on several advanced technologies all the time like Firewire, Wireless, Voice Activation, UPNP components (see our technical page) or fast Internet and broadband links between sites for Audio and Video. We may be among the very few who currently do it, but it is not Fundamental Research either.

We also master the Computer Technologies that will be used a lot in the future of Audio and we are certainly the most advanced company in that field among hi-end Audio manufacturers, but it is not Fundamental Research.

Fundamental Research addresses the long term development fields and work for that reason on the most advanced technologies. There is usually no immediate product application of the results even if, for most of them we hope, they will be used one day in products developped by more than one of the brands.

Sound is Everywhere

in the Hifi playing music.
in the Radio playing music or speeches.
in the TV programs.
At the Movies.
in the Telephone.
and more recently in the Computer field as :
- Background music (using the computer as a CD or DVD player or downloading from internet when programming or working on business).
- Games sound.
- Speaking computer (very soon everywhere).
- Voice Recognition.
- Storage and distribution (MP3, AAC or other compression systems) through the Internet.

Sound is Vital

It is half of the human communication. In most of the applications seen above the sound is at least half the message, the other half being the image.
The human perception can be split into two separate domains :
- The visual perception (the image)
- The audible perception (the sound).

These two domain are both necessary to rebuild the full perception. However, for the accuracy of the human perception, sound is even more important than the image. Even a movie is more understandable without the image than without the sound. Video conferencing is nothing without the sound, but telephone is often sufficient for human communication.
It is the emotional half. The sound is the part of the message carrying the most emotion. Try a Movie without sound, or a computer game without the sound ? On the contrary, Radio and telephone prove that emotion can easily be carried by the sound only.

Sound Quality is mandatory

It guarantees the emotional content. A poor quality sound is not impressive. Top quality sound induces greater emotional impact.

Sound Quality is needed for helping recognition of the original source. Intelligibility is mandatory in movie sound reproduction because otherwise dialogs would not be understood. Accuracy is mandatory for voice recognition.

Sound quality is a guarantee for health and safety. Noise and severe distortion (especially in the digital domain and time domain) have detrimental effect on the human nervous system. As well has the today's computer screens have to avoid radiation of detrimental rays or magnetic fields, best sound systems have to avoid creating irritating noise and distortion.

At Goldmund, our work is to Master Sound Quality

Within few years, most sound (music, movie sound or video conference sound) will be coming from a computer. Here is a simple basic possible configuration (finally quite accurate despite the fact we introduced it already 3 years ago !).

Our fundamental research, for the time being (we may extend some of our research in Audio to Video too...), covers the following 4 domains :

The Digital Sound Signal Processing inside the computer or in DSP's.
The A/D and D/A Audio converters (converting the analogue signals to the digital domain and backwards).
The Power Amplifiers (powering the speakers with maximum accuracy).
The Speaker Systems (to make the sound signal audible to the human ear).

These 4 domains are fully explored in the 3 following principle directions :

Space Recreation.
Stereo, multichannel (AC3, etc..), are due to reproduce space in sound imaging. We do extensive research on all of these ways to properly recreate the Sound Imaging.
The most significant of the technologies we use or develop today are the Single Stereo Speaker (the JOB Speaker), the Goldmund 100% software multichannel decoder (AC3, Prologic, etc..), as well as some experimental 2/3/4 channels with 3/4 speakers formats to recreate a more accurate spatial imaging in very large rooms (where the single stereo speaker is not sufficient in image width). The same program covers the holophonic recreation needed in virtual reality to precisely locate a sound in space.
All these technologies involve a lot of digital signal processing.

Intelligibility.
Everything (or nearly everything) in sound is based on signal transients. Speech as well as music. The capability to reproduce credible sounds (which induces intelligibility) for sound equipment is directly related to its capability to rebuild transient with extreme time accuracy. It means the signal has to be treated by circuits both extremely fast and time-accurate. Today, the time accuracy we reach in the lab (and in the "Alize" D/A converters) is already better than 100 picoseconds and the circuit bandwidth necessary is around 10 MHz. These circuits provide already a far better intelligibility. But speed and time accuracy have also to be improved in speakers and digital domain signal processing.
The largest domain of application, and one of our main concern, is voice recognition, for which the inclusion of one of our circuits may immediately increase dramatically the performance of the existing software in recognition ratio.

Naturalness.
The sound naturalness is what produces recognition by resemblance to reality. It is brought by a combination of several factors :
- Time Accuracy in the reproduction. Time Distortion doesn't exist in mother nature and is literally impossible to avoid in electronic circuitry. The Human Brain is impossible to fool on this matter. It is probably the most important in the Recognition Factor.
- Low sound Coloration. Mechanical vibrations produced by speakers, or even by power amplifiers, coloration induced by non-linear electronic components or cabling, natural mechanically-related component (capacitors, coils but also semi-conductors) coloration, all these phenomenons are hiding the original timbre of the sound by modifying its harmonic/temporal distribution.
- High sonic Dynamics (difference between high level and low level sound). Dynamics is needed to reproduce accurately the very strong impact of the change in sound level in music, noises or even human speech. Most of today's circuits are limited inherently in dynamics and so are the usual speakers.
- Low Distortion (especially those not found in mother nature like time distortion and intermodulation distortion). Distortion affects the tonal balance and the credibility of sound by adding spurious components not existing in mother nature, which confuses brain.
- Absence of sonic artifacts. It is quite easy to make most music and speech sound "nicer" by elaborate digital signal manipulation. But naturalness (as we want it) implies that no sound artifact of any kind is applied, in order for the original sound to carry all its emotional content. And the technology to avoid all artifacts is even more complex than the technology to produce them...