Philosophy    The reproduction system    The recording system    AVS Research    Technical specifications    
Press reports


Andrea von Salis Hi-End Audio Software for MAC

Technical specifications

Basic Design Notes

The AVS patented audio processing relies on hi-precision DSP software algorithms which were developed matching numerous and diverse knowledges in these technology fields:

● Computer fast number crunching
● Real-time digital audio processing
● High performance DSP filters and delay lines design
● Operating systems and user interfaces expertises

The technology chosen to implement the AVS process is Personal Computer. 
Apple Mac Os X is a proven and appreciated platform for multimedia processing, so AVS has been implemented as a user program running on an Apple Mac computer, exploiting the extreme computational power of modern multi-core CPUs.

Basic process architecture

From the user-selected  input device the  audio signals are delivered to the AVS 3D●VR application at the sample length and rate supported by the  converter streams. 

After that, processing takes place which changes the sampling speed from 32kHz up to 192 kHz (controlled by the user) and the quantization to 32 bit floating point.

This guarantees that a very high dynamic range and a large bandwidth are allocated where they really are needed, that is to say into the DSP process (filters and delay lines).

After processing the PCM stream is restored to the sample length and rate supported by the  output  audio device chosen in order to be compatible with its required PCM stream characteristics.

The main challenges which this architecture bears are linked to the necessity to maintain a real-time stream of information between all the audio chain parts. Drop-outs and other artifacts are heard when the real-time condition is not satisfied, in this case rarely.

On the other side, the main advantages are flexibility and the possibility to exploit a DSP power that cannot be challenged easily. Modern x86 and PPC CPUs are provided with DSP extensions and clock speeds which render them most apt for perceptual signal processing.

Advanced Design notes

The main components of this processing system are digital filters and accessories (delay lines, attenuators, switchboards).

Digital Filters

The type of digital filter used throughout the AVS 3D●VR software is a second-order IIR class stabilized filter. While its novelty could appear dubious, it is in fact a not conventional architecture, based on a stabilized structure which makes use of supplementary signal process taps, in order to improve the stability of the filter cell while obtaining the desired response diagram. This way a filter structure has been obtained which is very little prone to ringing and instability. This yields in turn a “better sounding” filter.

The coefficient generators are calibrated to match theoretical amplitude filter figures and are all-round 32 bit units. The resulting filter is a 32 bit variables – 32 bit coefficient, high-quality, high-stability, full parametric unit, meant to compare favorably with whatever filter is today on the market, regardless of implementation and price.


Delay lines which are employed in the AVS 3D●VR software are all 32 bit units, consisting in allocated memory cells which are used to store the PCM signal for the needed sampling time multiples. Clock precision and fine resolution guarantee a transparent “time-delayed wire” behavior.

“Wiring” (signal lines), switchboards (signal commutators) and attenuators are all 32 bit floating point units, in order to preserve the signal quality throughout the DSP processing software. Fixed point PCM is restored only in the end, together with sampling rate and word length required by the conversion device.

Basic data (with ECHO Audiofire 4 Firewire I/O device)

Sampling: 32 - 192 kHz/32bit float

Total dynamic range (including I/O): >110 dB (ECHO)

DSP Program internal dynamic range: >200 dB

Frequency response: +/- 1 dB 20 Hz – 20 kHz  or better (ECHO)

Filters: 	40

Filter characteristics:
0.05     <   Q  <  3
-15 dB  <   A  <  15dB
20Hz    <   F  <  20kHz

Delay lines:	6

Step Delay lines: 
Room Correction: 1 cm (33 microsec.)
Correlation: 10 microsec.