Multiplex communications – Communication over free space – Message addressed to multiple destinations
Reexamination Certificate
1998-10-08
2002-04-09
Ton, Dang (Department: 2732)
Multiplex communications
Communication over free space
Message addressed to multiple destinations
Reexamination Certificate
active
06370125
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates in general to digital communication networks, and is particularly directed to a new and improved mechanism for compensating for variable transport delay encountered in a packet-based digital voice network, by dynamically updating, as necessary, the value of buffer delay injected at a receive decoder end of the network, so as to maintain the buffer delay at a value associated with the maximum value of transport delay that has been encountered in the course of a call.
BACKGROUND OF THE INVENTION
Digital voice networks are generally comprised of two principal building blocks—an encoder and a decoder—that are customarily implemented in a single piece of encode/decode hardware known as a CODEC. As diagrammatically shown in
FIG. 1
, respective CODECs
1
and
2
are installed at opposite ends of a communication channel
3
, that provides digital transport capability between equipments at respective ‘west’ and ‘east’ sites
4
and
5
of the network. The respective encoder portions
1
E and
2
E of CODECs
1
and
2
sample analog voice signals supplied thereto at a prescribed sampling frequency and convert the samples a digital representation suitable for transport over the channel
3
. The sampling rate of an analog voice signal in conjunction with the amount of data per sample determines the required bandwidth for the channel.
In a typical network, each voice channel is assigned to a dedicated time-slot of a time division multiplexed (TDM) network. Because each voice channel has a dedicated time slot, data arrives at the decoder portion of the CODEC in a very constant and predictable manner. This is diagrammatically illustrated in the timing diagram
20
of
FIG. 2
, where successive TDM data packets
21
-
25
are shown as being equally spaced apart by respective spacings a in successive ones of an arbitrary sequence of packet intervals
11
-
15
within a voice channel of interest, so as to enable the CODEC to convert the data back into an analog voice signal.
A shortcoming of such a TDM voice network is that when no voice call is being made, the dedicated bandwidth for that voice channel is wasted. To overcome this wasted bandwidth problem, statistically based packet switching networks have been developed, in which bandwidth is statistically multiplexed on an as-needed basis. In such a packet switched network, the available bandwidth may be considered to be a single large “pipe” into which packets of data are inserted.
As shown diagrammatically in
FIG. 3
, digitized data (voice or otherwise) to be transported is segmented into small segments or packets
40
, which are placed into a queue
42
and then transmitted, in the order in which they are placed into the queue, by a transceiver
44
over the digital communication channel
46
to a CODEC
50
at the far end of the channel. Each packet of data occupies the entire bandwidth for a small period of time.
Now although packet-based networks overcome the wasted bandwidth problem, they also introduce a variable amount of delay into the network. Namely, because packets do not have a dedicated time-slot, their arrival at the receive or far end of the network will depend upon on how much data has been queued up in front of them. While this variable delay may not be of particular concern for data-only applications, it can become quite noticeable in voice applications. This problem is illustrated diagrammatically in the above-referenced timing signal diagram
30
of
FIG. 2
, which shows the example of a first voice packet
31
arriving on time in packet interval
11
, while the next two voice packets
32
and
33
are delayed by a relatively large interval &bgr;, and do not arrive until interval
14
during which data packet
34
is expected.
As pointed out above, the decoder portion of the CODEC
50
requires that data be supplied at a constant interval. Since there is no data present at any time during the packet intervals
12
and
13
, and it is further delayed in interval
14
, the decoder cannot reliably reproduce the original audio during this period. This condition may or may not be detectable by the human ear, depending on the length of the delay and the mechanism employed by the decoder to handle the idle time between packets. This problem is compounded by the fact that the late packets can be expected to eventually arrive, at which time the decoder has to decide what to do with them (since their time slots (intervals
12
and
13
) have already passed). If this sequence of events occurs often enough, it can degrade voice performance to the point that the CODEC is not useable.
One way to counter this problem, diagrammatically shown in
FIG. 4
, is to delay the packets (under the control of an attendant microcontroller, shown at
55
) through a buffer
60
, such as a first-in, first-out data register (or FIFO), that is inserted into the receive or decoder end of the transport path through the CODEC
50
. The challenge is to determine the proper amount of buffering (delay), that will allow the listener to comfortably discern the voice signals that the received packets represent. One approach is to establish a fixed delay at the beginning of a call, by buffering a given number of packets that are estimated to compensate for a reasonable amount of transport variance in the network. While this approach can work reasonably well, it has the shortcoming that a fixed delay buffer delay does not take into account actual network conditions. The amount of buffering needed to ensure proper operation at worst-case may be large enough to be discernible by and unacceptable to some individuals.
SUMMARY OF THE INVENTION
The present invention solves the foregoing problem by automatically dynamically adjusting (increasing) the size of (or delay through) the buffer, based upon actual network conditions as they are encountered during a call. As will be described, the present invention operates on the premise that, if one or more packets are delayed in the network, their eventual arrival can be expected to be at a rate that is too fast to be processed in real-time by the decoder. Such eventual arrival may be exhibited as a string of immediately successive voice packets arriving a relatively large interval after arrival of a previous packet in its expected packet interval.
As will be described the delay adjustment mechanism of the invention handles this problem in two parts. The first answers the question of what to do (namely what to provide over the decoder—listener path) during the delay period (when no packets are being received). Where the delay is reasonably brief and the amount of information in each packet is small, the last received packet is replayed in place of the missing packets. However, if delays are long, replaying the same packet over and over can begin sounding very unnatural. In this case, after the expiration of a user-programmed amount of time, the decoder may stop replaying the packet and begin playing silence.
The second question is what to do with the late-arriving packets. In accordance with the present invention, late arriving packets are written into the delay buffer in the order that they arrive. They are then played out from the buffer at the nominal decoder data processing rate. Once some number of packets have been buffered in this manner, any further delay of the same duration or shorter than that currently provided by the buffer will not be heard by the user, since previously stored packets will be read out of the buffer in a periodic manner (and thereby played out to the listener) during the delay interval.
However, should a longer network transport delay be encountered, an outage may be noticed, but only for the period of time exceeding the original outage. As a consequence, in the absence of an increase in network transport delay in excess of the delay interval, the decoder will remain some number of packets behind (missing or not received during the delay interval) for the duration of the call, introducing an overall end-to-end delay into the system. Should an
Adtran Inc.
Allen, Dyer, Doppelt Mlibrath & Gilchrist, P.A.
LandOfFree
Dynamic delay compensation for packet-based voice network does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Dynamic delay compensation for packet-based voice network, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Dynamic delay compensation for packet-based voice network will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2873659