&%PAGE&

-  &%page&  -
Annex 3 to Doc. AVC-22R
Annex 3 to Doc. AVC-22R


Source: Experts Group for ATM Video Coding in SGXV
Title : Liaison statement to SGXVIII on requirements for B-ISDN network
        model as it impacts on video coding 


The Experts Group for ATM Video Coding as established at the July 1990 
meeting of Sudy Group XV has initiated its activities by holding the first 
meeting in The Hague (13 - 16 November 1990).

Since this Group is intending to make Recommendation(s) for video coding in 
B-ISDN environments, we need correct understanding of the ATM netowrk 
characteristics.  For this purpose, we have formulated a list of questions at 
the first meeting as attached which are expected to be answered by SGXVIII 
at the earliest occasion.

The Experts Group would welcome the participation of a liaison 
representative from SGXVIII to assist in clarification of the B-ISDN 
definition and performance.  This would be particularly valuable at the 
second meeting (tentatively scheduled for 12 - 15 March 1991), since 
multiplexing of variable rate sources under certain cell loss conditions, 
and refinement of our network model, is expected to be discussed in detail.

Liaison to SGXVIII: Requirements for B-ISDN network model as it impacts on 
video coding

Abstract

The SGXV Experts Group on Video Coding for ATM held its first meeting in 
The Hague, 13-16 November 1990.  An important outcome was the recognised 
need for a realistic model of the target B-ISDN to permit progress in the 
definition of suitable video coding methods.  This liaison statement outlines 
the impact of various network parameters on video coding issues, and 
requests guidance from SGXVIII concerning parameter values.  To permit 
advancement of video coding work, an initial network model is offered for 
comment by SGXVIII.


Introduction

The Experts Group for ATM Video Coding was established by SGXV to 
investigate new possibilities for video coding offered by service support on 
the B-ISDN, and to develop appropriate coding algorithms.  There is a 
significant impact on video coding as a result of ATM transport, and on 
multimedia system design as a result of virtual connections.  The impact of 
certain network parameters on approaches to video coding and video service
provision are outlined below.  The intention is to both obtain guidance from 
SGXVIII regarding suitable parameter values to allow video coding work to 
progress, and to provide SGXVIII with some input that may influence aspects 
of network design.  The first section briefly itemises the main issues and 
the addenda will give a more comprehensive explanation.


Issues

1. Cell loss ratio: 

     -	What will be the  expected values, for both priority levels?

2. What is the cell loss burst behaviour?

3. How is the CLP bit used?

     -	Will there be seperate negotiations for the two priority levels?
     -	When will the service provider set this bit? 
     -	What are the restrictions for the use of this bit?
     -	Is the quality of service selectable?

4. Usage parameters:
 
     -	What parameters will be used for policing and admission control?
     -	What policing mechanism will be used?
     -	What averaging intervals can be used to measure mean, peak, etc.?

5. Multimedia connections:

     -	How will the admission and monitoring of an ensemble of VC's be 
handled?
     -	What is the limit on differential delays between different VC's?
 
6. Bit Error Rates:
     -	What is the expected rate?
     -	What is the impact on the AAL?

7. Cell delay jitter:

     -	What values are expected?

8. Network model for hardware experiments:

     -	What will be maturity of the B-ISDN network by mid '93?

9. AAL

The Experts Group for ATM Vide oCoding is willing to collaborate with 
SGXVIII in defining the AAL for video services, and will forward appropriate 
input to SGXVIII.


Addenda

Ad.1 Cell los ratio

Various cell-based video coding systems have been developed or simulated, 
collectively capable of satisfactory performance in the face of a variety of 
cell loss ratios.  However, the actual figure to be expected from the 
network for a particular video service application and bit rate will 
determine both the need for cell loss protection or recovery and the 
method to be used.  The Experts Group expects to identify appropriate cell 
loss ratios for video services on the B-ISDN, and will input this information 
to SGXVIII when available.

The cell loss ratio has fundamental implications for the video coding 
strategy and its efficiency.  If, for example, layered video coding systems 
are to be used, exploiting the availability of the cell loss priority 
indicator, an indication of the cell loss rate for each priority level is 
necessary.  Figures for expected cell insertion rates are also required.

Ad.2 Cell loss burst characteristics

The question of whether cells are lost in isolation or in bursts is 
fundamental for the video coding approach.  Some coding schemes are 
proposed which provide a means of protecting against bursts of cell loss, 
but they may not be necessary if cells are lost in isolation (i.e. if cell 
losses are uncorrelated).

     -	Will cell loss be dominated by network congestion?
     -	Will bursts of cell loss result from network congestion?
     -	Will the cell loss burst length be service rate dependent?
     -	Will high priority cells be affected by network congestion?

Clarification of these points is sought from SGXVIII, and additional guidance 
to an appropriate statistical model to characterise bursts of cell loss 
would be welcome.

Ad.3 Use of CLP indicator

The use of CLP is useful for some coding schemes to provide tolerance to 
cell loss.

     -	Under what circumstances would the service provider set the CLP 
indicator?
     -	Could the CLP be changed by the service provider after a user has set 
it?
     -	Will the usage monitoring structure encourage the use of both high 
and low priority cells?
     -	Will the rate of high and low priority cells be negotiated 
independently with the network?

Other information concerning call admission control and usage monitoring 
that would impact on the user's choice of a combination of high and low 
priority cells would be welcome.

Ad.4 Usage parameters


Ad.5 Multimedia connections

The ability of the B-ISDN to perform the multiplexing task provided by the 
terminal on circuit-switched networks makes it attractive to consider cell-
by-cell multiplexing (by use of different Virtual Channels, or possibly on a 
single VPI/multiple VCI's) for the provision of multimedia connections.  
However two issues arise here, as discussed below:

     -	Will the network be capable of providing connection admission and 
monitoring based on the group of VC's constituting a multimedia 
connection?  If not, would users see a penalty in the use of multiple 
VC's, and be encouraged to perform multimedia service multiplexing at 
a higher layer?  SGXVIII should be aware of this possibility and 
consider whether this capability can be accomodated.  Previous 
experience with multimedia services suggest that a group of at least 
seven VC's may be necessary, but we would like to know if there is an 
upper limit.

     -	Differential VC delay.  If multimedia connections (video and associated 
audio in particular) are supported over multiple VCs, there exists the 
possibility of differential delay.  If excessive, this may require end-
to-end signalling overheads to add time-stamps and permit 
resynchronisation.  What is the expected limit on differential delay 
between VCs?

Ad. 6 Bit error rates

We assume that cell payloads are subject to a small probability of 
transmission bit errors.  The statistics of such errors will determine the 
need for, and type of, error correction mechanism and the overhead 
necessary to achieve this.  It could also influence approaches to, and 
efficiency of, video coding and choice of code word assignment schemes.  

What is the expected probability of transmission bit errors, and are these 
errors likely to be uncorrelated of bursty?  Draft Rec. 

I.363 notes (section 2.3 and 3.3) that one of the functions of type 1 and 2 
AALs is the '...monitoring of user information field or bit errors and 
possible corrective action...'.  The Experts Group wishes to work with SGXVIII 
to further clarify the functionality of the AAL in this respect.

Ad.7 Cell delay and jitter

The fixed component of end-to-end delay is an important factor for 
conversational video services.  It will impact on the choice of coding 
method and allowable buffering within the encoder and decoder.

What is the expected maximum B-ISDN delay, including processing and 
queueing within the B-ISDN switching equipment?

The variation in delay, or jitter, determines the size of receive buffers 
necessary for its removal, and therefore again influences the total end-to-
end delay.  What are the expected statistics of cell delay jitter?  Is a 
hypothetical reference connection available or planned, that would assist in 
these matters?

Ad.8 Network model for hardware experiments

It is the initial intention of this Experts Group to target hardware trial of 
ATM video codecs for the second half of '93.  Success of such trials will 
depend on the availability of network equipment or simulators.  Would these 
be expected in this timeframe?

Ad.9 AAL

Draft Rec. I.363 describes AAL type 1 & 2 structures which could be used for 
real-time video services.  To make progress, the experts group intends to 
distinguish between:

     -	an AAL suitable for existing video services (e.g. H.261), that could be 
standardised in the relatively short term,
     -	for future ATM video coding standards, an AAL matched to the specific 
coding algorithms will be necessary.  It is premature to define an AAL 
for these applications at this time. 


Conclusion

Close liaison between the Experts Group for ATM Video Coding and SGXVIII 
will be necessary to harmonize and optimize B-ISDN network design and video 
sevice provision.  It is the Experts Group's intention to provide input to 
SGXVIII on requirements for network performance and to assist SGXVIII in the 
definition of the AAL for type 1 & 2 service catagories.

In the first instance, however, the Experts Group requires guidance to 
provide bounds for certain network parameters crucial to the development 
of appropriate video coding methods.  The main issues have been highlighted 
in this document.  A model of the network is necessary to permit 
commencement of video coding studies.  The parameters of such a model, 
along with some estimates of possible parameter values, is provided in the
Appendix.  This is a first attempt, to initiate studies.  Refinements will be 
made at future meetings.  SGXVIII is invited to comment on, or correct this 
model.


END

Attachment: Annex 4 (netowrk model) of this meeting report