Saturday, March 24, 2012

Evolve Random Access Channels for IoT: IV Power Control p.k. Access Timing

Access Channel Enhancements for 1x Rel. F
One Eighth Rate Access Probes for Smart Terminals
Evolve Random Access Channels for IoT: I Introduction
Evolve Random Access Channels for IoT: III Slotted ALOHA Models

During the discussion of evolving access channels for machine-to-machine (M2M) communication liked services, two major requirements are defined. As stated by 3GPP2 system requirement document, one is "[a]n M2M service shall not impact any ... existing applications and services that are currently being supported". The other one is "[t]he scalability of M2M Device deployment shall be supported allow flexible M2M Device and service deployment scenarios for large numbers of M2M Devices."  More details of these two requirements can be found in 3GPP2 Machine-to-Machine Communication System Requirements, S.R0146-0 and 3GPP M2M Service Requirements, ETSI TS 102-689. Due to the nature of M2M traffic pattern and QoS requirement, recent studies find out that uplinks, especially existing access channels, will become the bottleneck when M2M services are widely deployed in the future. Evolving access channel once again becomes a hot topic since it was optimized for short message service (SMS) in 1990s.

Minimize Collision and Interference: Power Control and Access Timing

Of the two requirements, let's start from how to minimize the impact of new on legacy services at first. Doing so is not only required by service providers but also because these two requirements don't always conflict with each other. From random access channel fundamentals, it is not hard to understand two key factors determining the collision and interference between probes, access timing and power control. Access timing information is about when access probes will start using access channel. Access timing is not only the key assumption for many ALOHACSMA MAC protocols, but also important for new access probes to decide the best time using the shared access channel with the minimum overlap to legacy probes. The second factor is perfect power control. This becomes critical is because when access channel is loaded in the future,  it is impossible to completely avoid the overlapping between access probes even though the perfect timing information of the channel might be available. At this time, perfect power control on new access probes becomes necessary in reducing interference to legacy services. Based on these two factors, two optimal access channel designs will be presented and discussed in the following.

Optimal Access Probes: Dumb Probes p.k. Smart Probes.

If the perfect access timing information of legacy access probes is available, one optimal access channel design is to use very short and smart access probes, which know the activities of the other probes and are able to try its best to arrive inside some access gap. In order to have little overlap with other access probes, it is nature for a smart access probe to be short. While it is necessary to maintain a minimum payload during a short period, a smart and short access probe usually demands high transmit power and data rate. How smart probes work with legacy probes can be illustrated in Figure 1.

Figure 1. A traditional Thinking of Improving Access Probe Design
Though smart probes can have minimum impact on legacy probes, it is very hard to be implemented. In reality, it is extremely challenging for an access probes to know the instant access timing information of other probes, even though they may be co-located next to each. In a typical scenario for mobile communication, each mobile has no timing information of any other mobile without the help of base stations.  When the perfect access timing information is unavailable, one optimal access channel scheme is to send very long and dumb access probes instead. Here “dumb” means each of these access probes has no any access timing information of other probes. However, in order to minimize any impact to other probes, including current and future probes, each dumb probe still wants to take advantages of access gaps of access channel. To achieve this, each probe has to be transmitted at a very low data rate and span a very long duration with perfect power control.  A low transmit data rate spanning a long duration usually means a low transmit power. Though a dumb probe does look "slow" and cautious, it also means potentially more diversity opportunities are achievable.
Figure 2. A New Thinking of Improving Access Probe Design
Now as you can see, there are two different starting points for designing optimal access probes for minimizing the impact to legacy services. For smart probes, timing information is the challenge and key.For dumb probes, perfect power control is the key. A comparison of these two optimal designs can be found in the following table.

Table 1. A Comparison of Dumb Probes and Smart Probes

Monday, March 5, 2012

RTP Packetization for H.264 NAL Units

H.264 Network Abstract Layer Header
How to Broadcast Multimedia Contents?

RFC3984 (RTP Payload Format for H.264 Video) defines 3 modes: 0 (single NAL unit mode), 1 (non-interleaved mode), and 2 (interleaved mode).  Mode 0 requires you either use UDP fragmentation or tell the encoder not to generate NALs larger than MTU-X. Mode 1 lets you do fragmentation. Details of how to set up a fragmentation unit packet can be found in the RFC. Basically the fragmentation information is on the front. Small NAL units, e.g. SPS and PPS packets, can be aggregated together using single-time aggregation packets (STAPs). Each packet requires normal RTP headers with incremented sequence numbers but the same timestamp. Though a mark on the last RTP packet of a frame is expected,  it is not guaranteed. Mode 2 lets you fragment, combine, and interleave the transmission order to change how a burst loss will affect a stream, among other things.

Friday, March 2, 2012

Evolving Random Access Channel for IoT: III Slotted ALOHA Models

Access Channel Enhancements for 1x Rel. F
One Eighth Rate Access Probes for Smart Terminals
Evolving Random Access Channel for IoT: I Introduction
Evolving Random Access Channel for IoT: IV Dumb Access Probes and Smart Access Probes

It is well known that the access channel design of 3G cellular mobile network is based on ALOHA models. More specifically it is based on a single-channel single-hub slotted ALOHA plus open-loop power control.  Indeed, there are many possible variations of slotted ALOHA.  Considering access channel configuration, there are the cases with a single shared access channel and the cases with multiple access channels.  Considering network configuration, there are single hub models and multi-hub models. Though multi-channel slotted ALOHA models have been intensively studied in computer engineering domain, a ALOHA model with more than one participating hub and more than one access channel hasn't received enough attention so far. I guess one reason is the original ALOHA models were proposed for satellite communications, where typically it is a single-cell or single-hub service scenario. Another reason I guess is the current computer engineering focuses more on network topology and MAC layer above issues instead of related channel modeling and PHY layer challenges. My feeling is the study of a multi-hub model with PHY-layer macro-diversity usually belongs to the stronghold of communication engineering. Conceptually the access channel of a cellular network is also more closely related to a multi-hub slotted ALOHA model, which has either single acess channel or multiple access channels.

Access Channel Design Challenges

It is well-known that there is a tradeoff between access capacity and throughput in ALOHA models, including both slotted and pure ALOHA.  Basically say, after a certain point, more access load will result in lower access success rate, which usually in turn results in more access delay due to retransmission. This tradeoff can be illustrated in Figure 1.
Figure 1. The Access Channel Access Capacity and Throughput Dilemma
In addition, a traditional CDMA2000 access channel can be modeled as a classic slotted ALOHA, i.e., single-channel single-hub slotted ALOHA. In this design, each base station only detects the access probes with its own access channel mask. Though it is well-known that additional macro-diversity gain is achievable with multiple base stations detect one access probe and the standards also doesn't prevent a base station chipset manufacture from implementing a base station which can detects the access probes to both itself and neighbor base stations, the resulted implementation complexity increase, in both PHY layer, MAC and upper layers, can be very high. One simple illustration of this tradeoff of diversity gain and complexity increase can be shown in Figure 2.
Figure 2. The Access Channel Search Complexity and Achievable Macro-Diversity Dilemma

Then one question is, what is the big deal of macro-diversity for access channel? It helps reduce transmit power, therefore, reduce interference and improve battery life. It helps mitigate network imbalance. It can also help network positioning.

Macro-diversity can help reduce network imbalance. There are two kinds of related network imbalance issues. One is forward link and reverse link imbalance. The other one is the reverse link load imbalance among base stations. Network imbalance is not only because of the non-uniform distribution of terminals but also because it is hard for a terminal to make a good decision on which base station it should point its access probes to without knowing base station side rise over thermal (RoT) condition.  This is important especially when a network user capacity is high, the cell coverage is large and the network is heavily and non-uniformly loaded.  Further more, since the existing access procedure simply asks a mobile to point its access probes to a sector typically with the strongest forward link pilot channel (F-PICH), this approach apparently is suboptimal. Therefore a new network load aware access probing mechanism is necessary in this case.

Access Channel Design Models

Table 1. A Comparison of Different Slotted ALOHA Models for  Mobile Network Access Channel Design

New Access Channel Proposal

Table 2. A Comparison of Different Access Channel Designs for Mobile Network