The Impact of Nodes Embedded with Data Processing Unit on Energy Consumption in a Wireless BAN Academic research paper on "Electrical engineering, electronic engineering, information engineering"

Available online at www.sciencedirect.com

SciVerse ScienceDirect

Procedía Computer Science 19 (2013) 938 - 943

2013 International Workshop on Body Area Sensor Networks (BASNet-2013)

The Impact of Nodes Embedded with Data Processing Unit on Energy Consumption in a Wireless BAN

Yanqing Ai* and Chiu-Sing Choy*

Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong, China

Abstract

Several access modes and access methods, combined with new power management schemes and frame structures are defined in the latest IEEE standard 802.15.6. The configurations and supported four kinds of data rates determine energy efficiency of the whole wireless body area network (WBAN) system. To improve quality of service (QoS) in WBAN, node power efficiency is very critical, especially for medical applications with implanted devices. In this paper, we proposed a method to reduce the node energy consumption meeting with current standard, targeting at a usual case; and developed an analytical model to evaluate the power efficiency. Simulation results show that good power efficiency can be achieved when the transaction transfers in higher data rate. Also, the embedded processing unit saves node energy for most applications, in which the payload is higher than 70 bits per beacon period. When its compression factor is bigger than 2 and its power consumption does not exceed 40% of the transceiver, better energy efficiency always can be guaranteed by processing unit insertion.

© 2(013 The Author s. Published by El sevier B.V.

Selection and peer-review under responsibility of Elhadi M. Shakshuki

Keywords: Wireless Body Area Network; Energy Efficiency;

1. Introduction

Increasing demands for high quality community-based services have triggered the concept of novel wireless human body monitoring, to sample vital signals from Electroencephalogram (EEG), Electrocardiogram (ECG), Electromyography (EMG) and so on. Moreover, worldwide aging problem put a strain on novel wireless human body monitoring system. However, currently available low power protocols including Bluetooth, Zigbee and Wi-Fi have been proved inappropriate for medical applications[1]. Targeting at health-care application scenarios, IEEE 802.15 working group formed task

* Corresponding author. Tel.: +852-3943-3282; fax: +852-3943-3282. E-mail address: {yqai, cschoy}@ee.cuhk.edu.hk.

1877-0509 © 2013 The Authors. Published by Elsevier B.V. Selection and peer-review under responsibility of Elhadi M. Shakshuki doi:10.1016/j.procs.2013.06.129

group 6 (TG6) in November 2007 and released the first version of wireless body area network (WBAN) standard, IEEE 802.15.6, at the beginning of February 2012[2].

According to the standard, WBAN connects medical and non-medical devices that could be placed on skin or inside the body, and sends or receives data from sensor nodes and the Internet. The physical layer of the standard can be narrow band (NB), ultra wideband (UWB) and human body communications (HBC)[3]. The medium access control (MAC) sublayer supports four access methods: random access, improvised access, scheduled access and unscheduled access which can be employed in different application scenarios to achieve optimal power performance.

In this work, we proposed a method to reduce the node energy consumption meeting with current standard, targeting at a usual case. Analytical model was developed to evaluate the power efficiency. This paper is organized as follows. Section 2 highlights the main concerns that are relative to power optimization defined in the standard. Section 3 describes the proposed power consumption model. Section IV shows the analysis results and discussions. Afterwards, section 4 draws the conclusions.

2. Application Configurations Based on IEEE 802.15.6

Data needed to exchange inside WBAN can be 25 bits or lower[4], or reach to 2.46 Mbits to transfer images for a swallowable endoscope [5]. Here let Linfo be the data length in bit, and we will sweep this variable from 100 bits to 10kbits to study the performance of IEEE 802.15.6. In this paper, the analysis focuses on beacon mode with superframes.

The medium access layer and physical layer have big impacts on the overall energy consumption, and hence, the lifetime of a node. Among the three physical layers, HBC has an easier handled packet structure, as shown in Fig. 1. It has several advantages over NB and UWB, such as smaller path loss, higher security and smaller size, which reduce power consumption of whole system[5]. Different with general wireless communication, data transmission through modulated electric field in HBC is limited in the operation frequencies of 42 MHz. Four data rates are supported: DR1 - 164.1 kbps, DR2 - 328.1 kbps, DR3 - 656.3 kbps, DR4 - 1312.5 kbps.

PLCP Preamble SFD DRF/RI PLCP header PSDU

(256 bits) (64/76 bits) (32 bits)

MAC header MAC Frame Body FCS

(56 bits) (0~255 bytes) (16 bits)

Fig. 1. HBC Packet structure

As PHY payload on PHY layer, the data frames (PSDU) passed from MAC sublayer are wrapped into packets one by one by adding physical layer convergence protocol (PLCP) preamble, start-of-frame delimiter (SFD), PLCP header. To avoid blocking important vital signal collection, only burst mode (also named as RI mode) will be analyzed, resulting in HBC packet size with additional 364 bits. Herein PSDUs are assembled of various lengths, from 0 to 255 bytes, on MAC sublayer corresponding to different MAC frame body, such as management type frames, control type frames and data type frames. Therefore the information rate is from 0 when PSDU is 0 bit to 85.3% when PSDU is 255 bits.

Furthermore, two additional low-power modes are designed for nodes following the standard: 'sleep' and 'hibernate' to save energy during idle time. Nodes can be inactive for several allocation slots, a BP at most in sleep mode whereas several continuous BPs in hibernation mode.

3. Power Consumption Model

3.1. Topology and analysis base line

Basic network topology used for medium access includes one hub and a limited number of nodes, as shown in Fig. 2(a). Conceptually, a node can be divided into sensor, receiver (RCV), transmitter (TX) and an optional data processing unit, e.g. microcontroller (MCU), while the MCU can be built in the node or the hub, as shown in Fig. 2(b) and (c). For distributed MCU in the node, it can consume additional energy. Meanwhile, it acts as a digital signal processing core or data compressor, and save energy by reducing uplink payload size before transmission.

Node 1

Node 2

Node 1

Node 1

Node 2

Node 1

Node n

Node 1

Node n

Node 1

Node 2

Node n

Fig. 2. Network topology

TX ^ RCV Ud Sensor A MCU

k frames

pSIFS+GT

pMIFS+GT

pSIFS+GT

■■>■ <4-

twu+GT

Fig. 3. Data transmission with immediate acknowledgement

In following analysis, power efficiency of two schemes shown in Fig. 2(b) and (c) will be compared using typical characteristics of a node shown in Table 1[6]. In this table, 'type' is just a symbol used to represent all the relative items in an expression, i.e. column 'Func/Var.'.

Table 1. Power Relative Characteristics of Blocks

Func./Var. Fun. Block Type Value Situation/Mode

1 33.9mW Permission 0dBm

Transmitter 2 27mW Permission -6dBm

PTx,(type) Type: 1,2,3,4 (TX) 3 22.5mW Pemission - 12dBm

4 21mW Pemission=-18dBm

Receiver 5 39.3 mW Receive mode

(RCV) N/A 41.4 mW Listening mode

Transceiver N/A 26.7mW Start-up

(TX+RCV) N/A 2.7|iW Sleep mode

PMcu(type, f) Type: 1 Micro- 1 1.5mW/MHz Active mode

Controller N/A 7.5(j.W Start-up*

(MCU) N/A 3|iW Sleep/idle mode

tsu Start-up time N/A 3.13ms

»Notes: power of MCU in start-up period is scaled up from power in sleep mode by a coefficient of 2.4 derived in [7].

3.2. Analysis considerations

Before deducing the power consumption model, we make the following assumptions besides what we have done in section 2.

• The system operates in a steady state after power on and initialization finished.

• Study is conducted on the specified node which enters sleeping mode right after data transaction.

The data exchange details based on assumptions above are shown in Fig. 3, where pSIFS stands for the protocol parameter of short interframe space, pMIFS stands for the protocol parameter of minimum interframe space, GT stands for the guard time and I-Ack stands for the immediate acknowledgement.

Values of pSIFS and pMIFS are defined in Table 2. According to the standard, GT and pMIFS should be inserted between continuous data frames; GT and pSIFS should be inserted between each data frame-acknowledgement pair; whereas time margin of wake-up time twu and GT should be left before next beacon.

According to the packet structure, total transmission time in RI mode is the sum of time spent on each component:

T _, ,, ,, ,, _LPLCPpreamble , LSFD-RI , LPSDUheader+LPSDU (* N

1frame=lPLCPpreamble+lSFD+lPSDUheader+lPSDU DR0 h DRg + DR (1)

where DR0 is data rate sending PLCP preamble and SFD, i.e. 5.25 Mbps, and DR can be any of DR^ shown in Table 1. Let l be the length of MAC frame body, as PSDU is constituted by MAC header (56 bits), FCS (16 bits) and MAC frame body, (1) can be transformed to:

Tframe (l DR)=0.063+d~

The duration of BP can be calculated with allocation slots (AS), as follows:

BP = AS xNslot

where, Nslolis the number of AS with range fro m 0 to 256 and AS = pAllocalionSM + L x pAllocalionSR. pAllocalionSM and pAllocalionSR are PHY layer parameters, as shown in Table 2. L is a parameter with range from 0 to 255 to config the length of AS. The smaller the BP you set, the higher compatibility can be achieved. Here we consider a general case with setting Nslol=16 and L=8, thereafter the BP is set to 68 ms to support most bio-sensors' requirements.

Table 2. MAC Sublayer Parameters Table 3. Power Relative Characteristics of Node

Parameter Value

General parameters mHubClockPPMLimit 40 ppm

mClockResolution 4 (is

mNominalSynchlnterval 8 x bp

mTimeOut 30 us

PHY-dependent parameters pAllocationSM 500 ns

pAllocationSR 500 us

pExtralFS 10 HS

pMaxFrameBodyLength 255 bytes

pMIFS 20 HS

pSIFS 75 HS

Power Var. Total power of general nodes Total power of relay nodes

Start-up PWRsu 26.71mW 26.7mW

Active mode PWRa Ptx(1~5)+ Pmcu(1, f) Ptx(1~5)

Listening mode PWRl 41.403mW 41.4mW

Sleep mode PWRs 5.7HW 2.7hW

Node power includes not only energy consumed during active periods, but also consumed in in-active intervals. It can be calculated as

PWR= — (PWR xlSU+PWRA xlA +PWRl xlL+PWRSxlS)

where PWRSU , PWRA , PWRL and PWRS are power consumed in different modes in one BP and the expressions are shown in Table 3. lSU, lA, lL and lS are time intervals consumed during start-up, active mode, listening and sleep, respectively. A full function node generally consumes more power than the relay node which is a function-reduced general node without MCU.

For the transaction shown in Fig. 3, frames that are exchanged in specified BP include k data frames, an immediate acknowledgment and a beacon frame. For the node, start-up is the state in which the node recovers from sleep; and active mode includes transmission and receiving periods. In the rest of the timeslots except the sleep interval, the node is working in listening mode.

Let lATX, lARCV be the time intervals consumed during transmission and receiving phases, k be the number of data frames, Linfo be the length of information that need to be sent. For the node without MCU, the time components are calculated as:

k=/2Lxs/ ®

tATX=(k-1)*Tframe (2040,DR)+Tframe (Llnfo-2040k+2040,DR)

-0.063+-

tL = (k-1) x (pMIFS+GT) +3 x(pSIFS+GT)

tS=BP~tSU~tATX~tARCV~tL

For the node with MCU, assuming the data compression factor is CF, (2) is transformed to:

v r Lmfo_ 7

l255x8xCFl

4. Simulation and Discussions

In this section we present the simulation results obtained by sweeping the information length, the CF/power of MCU and the transmission rate.

—e— w/o MCU

20 -B— CF=2

-V— CF=4

15 —#— CF=8

—1— CF-16

10 20 30

Payload (kbit)

1.5 ) 4

o 3 p -> «

* 2 1.5

-Payload: 100 bits -Payload: 1000 bits Payload: 10000 bits

CCoompprreesssion Facttor

Fig. 4. (a) Payload sweeping; (b) Compression factor sweeping

Fig. 5. (a) Power of MCU sweeping; (b) Data rate sweeping

The effects on power of the payload Lmfo and compression factor CF are shown in Fig. 4(a), where Nsiot and L are set to 16 and 8 respectively, and the power of MCU is set to the default value in Table 1. As expected, the graph shows that with more data exchanged, more energy is consumed in a linear relationship. It also shows that power consumption is influenced by CF with a ratio inverse relationship. The improvement of energy efficiency is getting smaller with increasing CF and tends to be a constant, as proven in Fig. 4(b). It is noticeable that improvement can be achieved even though the payload is as light as 70 bits; it can be concluded that the inclusion of MCU is beneficial to most applications.

Fig. 5(a) shows the effects on energy of power of MCU in case of fixed payload (1000 bits per BP). Two trends are observed: the larger CF, the less influence caused by the power of MCU; the higher power of MCU, the larger CF should be designed otherwise more energy will be wasted. The trends are reasonable because when CF getting larger, the information need uploaded is scaled down more deeply. Therefore the total transmission time is reduced and more data can be transmitted in the given BP. When power of MCU is more than 26.3mW/MHz, two times of information compression factor is not enough to reduce the node power any more for this case.

Fig. 5(b) shows that higher data rate is proved to achieve higher power efficiency. The reason is that although the transmission power is getting larger in higher data rate, active time is shortened so as to reduce the node power. The power saving is 8.25 mW when CF=2 whereas 1.4 mW when CF=16, matching with the analysis of Fig. 4(a).

5. Conclusion

In this section we present the simulation results obtained by sweeping the information length, the CF/power of MCU and the transmission rate. Released in this year, the latest IEEE standard, 802.15.6, defines MAC sub layer and PHY layer for WBAN based on previous studies and drafts. Probably the first study of its kind we know of, this paper shows the relationships between energy efficiency and transaction rate, length of information and especially the power of MCU and data compression rate of this unit.

The method we proposed to reduce the node energy consumption by embedding a data processing unit has been verified. Following conclusions can be made: distributed MCUs can be used in most of the applications except for payload shorter than 70 bits per BP; the less power consumption of MCU, the higher power efficiency on nodes side can be achieved; it is good to transmit data in higher data rate when the transceiver is designed according to the standard.

We believe these conclusions are important in designing an ultra-low power WBNA system and choosing the architecture of the nodes and hub, especially for applications where lifetime is a key requirement.

References

[1] IEEE Standard for Local and metropolitan area networksPart 15.6: Wireless Body Area Networks. IEEE Std 802.15.6-2012, 2012: p. 1-271.

[2] Kwak, K.S., S. Ullah, and N. Ullah, An Overview of IEEE 802.15. 6 Standard. Arxiv preprint arXiv:1102.4106, 2011.

[3] IEEE Standard for Local and metropolitan area networksPart 15.6: Wireless Body Area Networks. IEEE Std 802.15.6-2012, 2012: p. 1-271.

[4] Park, C., et al., "An Ultra-Wearable, Wireless, Low Power ECG Monitoring System," Proceedings of IEEE BioCAS, 2006, pp. 241-244.

[5] Bang, S., et al., First clinical trial of the "MiRo" capsule endoscope by using a novel transmission technology: electric-field propagation. Gastrointestinal endoscopy, 2009. 69(2): p. 253-259.

[6] Chipcon AS, CC2420 IEEE 802.15.4/ZigBee Transceiver Data Sheet, http://www.chipcon.com/files/CC2420_Data_Sheet_1 _2.pdf, 2005.

[7] Chen, H., et al., Statistical power analysis for high-performance processors. Journal of Low Power Electronics, 2009. 5(1): p. 70-76.