Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

UTILIZATION OF UNIFORM DELAY ACROSS N CHANNELS FOR TRANSPARENT PERFORMANCE

IP.com Disclosure Number: IPCOM000004640D
Original Publication Date: 2001-Mar-09
Included in the Prior Art Database: 2001-Mar-09
Document File: 2 page(s) / 8K

Publishing Venue

Motorola

Related People

Michael Kloos: AUTHOR [+2]

Abstract

UTILIZATION OF UNIFORM DELAY ACROSS N CHANNELS FOR TRANSPARENT PERFORMANCE

This text was extracted from a RTF document.
This is the abbreviated version, containing approximately 60% of the total text.

UTILIZATION OF UNIFORM DELAY ACROSS N CHANNELS FOR TRANSPARENT PERFORMANCE

by Michael Kloos and Esteban Yepez, III

THE PROBLEM

As Digital Signal Processors (DSPs) get faster, voice encoding/decoding (or vocoding) software developers will be able to support more voice channels per DSP. For example, consider a task that executes in N seconds on a 1x DSP. The same task will execute in N/3 seconds on a 3x DSP. The 3x DSP can be redesigned to support up to three channels simultaneously. Although the DSP can process certain instructions in parallel, the bulk of instructions must still be executed in a serial fashion. This means that each process must be completed before a new process can be invoked.

The migration to more channels per DSP allows infrastructure equipment to support the same capacity with fewer pieces of hardware. The infrastructure manufacturers who are able to reduce inventories, while adding a price premium for the enhanced capacity also share these cost reductions.

Ordinarily, one would expect that this enhanced speed would reduce audio delay. This may not be the case. In the iDEN system, there is a Just in Time Audio algorithm (JITA) which is a feedback loop from the base radio (BR) to the XCDR that adjusts the packet timing of the outbound packets to mini mize audio delay.

This process runs independently for each audio channel. When multiple channels are placed on the same DSP, this timing adjustment can no longer be made optimally for each channel, because processing for the other channels have their own timing constraints.

Further, since every adjustment in timing can cause an audio discontinuity or "click" in audio, proper design dictates that timing adjustments on one channel should not affect the timing of any other channel.

Figure 1 illustrates that running each channel at optimal timing will not meet this goal. Prior to the JITA adjustment, Channel 1 is encoded before Channel 2. Channel 2 must wait for Channel 1 to complete before it can encode and transmit packet 2. The BR is satisfied with the transmit time for Channel 2.

When Channel 1 makes a JITA adjustment, however, it allows Channel 2 to begin encoding as soon as it is time to start encoding packet 2; there is no more waiting for Channel 1. The result is that Channel 2 will transmit packet 2 earlier for all subsequent packets. At this point, the BR detects a change in Channel 2's transmit time; it will make a JITA request to force Channel 2 to revert to the original transmit time.

In the meantime, Channel 1 has delayed its encode to start D ms later. Since it must now wait for Channel 2 to complete, its transmit time is delayed. Again, the BR will detect the change in transmit time and may request a new JITA adjustment.

The timing will change again when Channel 2 makes a JITA adjustment. The result is an endless cycle of JITA adjustments for both channels. As menti...