Bluetooth in Industrial Environment

Bluetooth in Industrial Environment

Urban Bilstrup Per-Arne Wiberg

Electrical Engineering Electrical Engineering Halmstad University Halmstad University

School of Information Science, Computer and School of Information Science, Computer and

Box 823,301 18 Halmstad, Sweden Urbiixl .bil strtjp @ cui~.hh .se

Box 823,301 18 Halmstad, Sweden per-n me.wi berr @ cca.hh.se

Abstract

In this paper an initial study of the use of Bluetooth in industrial environment is presented. The tests have been performed at a paper-mill, and in an of ice environment at Halmstad University. It shows the possibility to use Bluetooth for wireless short range communication in an industrial environment.

1. Introduction

This work is an initial study of the performance of Bluetooth [ l ] in industrial and office environment. The motivation for this project is the fact that short-range radio communication in factories can reduce the use of cables and are in some case a good alternative to wired connections. When transmitting or receiving from rotating and moving machine parts, it is of great advantage to use wireless communication. Another type of application is an autonomous guided vehicle, which is dependent on some sort of wireless communication for communicating with surrounding machines. All type of industries, that have small production series and often have to reorganise their production line, have much to win by using wireless communication to satisfy demands on just in time production. In the future, machines assembling products might be able to communicate with the product without physical connection during the assembly sequence, performing tests and downloading software to the products. Other applications are support and surveillance functions without wired connections.

The Bluetooth short-range radio systems call for a research effort, before it can be used effectively in factories. Especially when there are high demands for safety and real-time performance, most applications in the industry have these demands [2]. We have performed some initial field tests to verify the performance of the Bluetooth system in a paper-mill, which is a complex process industry.

Chapter two is an introduction to the Bluetooth standard. Chapter three presents related work done in the area of

WFCS-2OO0, September 6-8, Porto, Portugal 0-7803-6500-3/00/$10.00 @2000 IEEE 239

wireless communication in factories and Bluetooth short range radio. In chapter four the test method is explained. Results are presented in chapter five. Conclusions about the tests and some enhancements recommended for the Bluetooth standard in industrial applications is done in chapter six.

2. Bluetooth standard

Bluetooth is a standard for short-range radio and personal area networks [3], developed by the Bluetooth Special Interest Group(S1G). The main function with the Bluetooth standard is cable replacement in office environment. The Bluetooth SIG is supported by a number of large multinational companies (Ericsson, Intel, IBM, NOKIA, Toshiba, Microsoft and others). This group of companies co-operate, to make the Bluetooth standard accepted and compatible worldwide. Anyone that is interested in using Bluetooth, can become an adopter of the standard. Different profiles are being specified for different applications to make the standard compatible; all products using the Bluetooth standard have to be certified, to be allowed to use the Bluetooth trade mark.

The Bluetooth standard uses the Industrial Scientific Medicine radio band (ISM) 2.45 GHz and Frequency Hopped Spread Spectrum (FHSS)[4] as medium access. The required and nominal range is 10m (transmit power 0 dBm, 1 mw) and an external power amplifier can be added which gives an extended range up to loom (transmit power 20 dBm 100 mw). The raw bit rate is approx. lMbps, according to the specification. The modulation in use is Gaussian Frequency Shift Keying (GFSK) [5,6,7].

Bluetooth supports point to point and point to multi-point connections. The establishment of connections are done ad hoc between the Bluetooth units. An ad hoc network (piconet) is established for example when a laptop enters a room which contains a printer then the laptop is able to connect wireless to the printer. At the same time the laptop can connect to an internet gateway which gives

access to a wired LAN. A Bluetooth unit can establish a piconet with one up to seven other Bluetooth units.

Bluetooth supports both asynchronous and synchronous transmission channels. Ericsson's Bluetooth module (ROK 101 007/1) [4] supports synchronous (SCO) voice channels (up to three 64 kbit/s channels in one piconet) and asynchronous (ACL) data channels (up to 723.2 kbit/s forward and 57.6 kbit/s reverse in one piconet). There are six types of ACL data channels, according to the standard, using the packets presented in table 2.1 below.

Type

HVl

HV2

HV3

Type Payload FEC CRC Slots

Payload FEC Channel allocation kb/s

10 byte Repetition 100.00% 64

20 byte (15,lO)Hamming 50.00% 64

30 byte None 33.33% 64

Table 2.1 Possible types of packets to use in an ACL data channel.

3. Related work

In [9] measurements of received signal strength at a pulp factory is presented, the propagation model 3.1 is used , m is found to be 1.1 by regression at LOS. This is a very low value pointing out the effects of reflections. The theoretical m is 2, for LOS without reflections.

Pr=EIRP+Gr+ 1 Olog( U4dn)" (3.1)

Pr is the received power, ERIP is Effective Radiated Isotropic Power, Gr is the gain of the receiver antenna, h is the wavelength and d the distance. We found m to be 1.7 LOS in the office environment. In the industrial environment have we not been able to perform signal strength measurements.

In figure 2.1 the BER is a function of Eb/No for an AWGN channel (GFSK) as presented in [6,10].

Figure 2.1 BER for an AWGN channel, GFSK, modulation index=0.32, BTz0.5.

In [ 111 T.S. Rappaport gives an overview of future use of radio communication in factories. One important thing is that even if the interference in factories is significant it falls of rapidly above 1GHz. This is in favour for Bluetooth that uses the 2.45 GHz ISM band.

Microwave ovens operate in this band [12], can radiate up to 30 dBm (1 w). The interfering bandwidth for a microwave oven is smaller than the ISM bandwidth. Therefore only a part of the available frequency channels are affected. A paper written by Haartsen and Zurbes treats Bluetooth voice and data performance in 802.1 1 DS WLAN environment [13] showing that the throughput reduction is in the area of 22% when a direct sequence W-LAN is present.

In the papers [9,10] there is also pointed out that most interference is multipath interference created by multiple reflections from the building structure and surrounding machine inventory. Even if path loss increase more rapidly with distance in a factory environment it is not as severe as in partitioned office environment [14]. The tests performed in this paper are done inside and close to a paper machine, which is very obstructed and partitioned.

240

There are no interference among nodes in one piconet because of the Time Division Duplex (TDD) scheme, but when multiple piconets operates in the same area they interfere with each other. The tests in this paper have been single piconets without any interfering piconets present. According to calculations the aggregated throughput with multiple piconets in an area with a radius of 10 meters should be in the range of 10 Mbps (figure 3.2). The calculation is based on a model for non slotted aloha[ 151 media access.

,, - .. ' \

g= 1600*N*P (3.3)

p 5IIC (3.4)

Pfd= 1 -Psuc (3.5)

The probability (P) that one frequency is chosen by zero or one piconet, is calculated with equation 3.1. Where n is the total number of piconets and k is the number of piconets which can choose the same frequency, in this case k<2(no collision). V is the number of frequencies each piconet chose (one), S is the other 78 possible frequencies. Equation 3.2 calculates the total probability for frequency collision (P'). The total numbers of slots that interfere (g) are calculated in equation 3.3. The model for non slotted aloha is used, to calculate the probability for success (Psuc) in time (equation 3.4), T is 366 pS (interval for a slot). Equation 3.5 calculates the probability for collisions in time and frequency (Pfail).

The graph in figure 3.2 is calculated according to eq 3.1- 3.5, it shows the maximal aggregated throughput and how the system performance softly degrades when the load in a local area increases (the number of active piconets).

Figure 3.2 Aggregated throughput as a function of number of piconets (no consider taken to out

of band emissions).

The Bluetooth standard has inherit much of the technology from the IEEE 802.1 1 Wireless LAN standard [16], especially the FHSS and GFSK modulation specification for the radio. According to our knowledge no performance test has been done in industrial environment with Bluetooth.

4. Test method

Ericsson Bluetooth Development Kit (EBDK) [ 171 has been used as receiver and transmitter in these tests. In each test position 200.000 bits have been transmitted with AUXl packets, which have no CRC or FEC. The packet is mentioned to perform tests with. According to the Bluetooth RF test specification [18] at least 1.6 millions bits should be transmitted. Because the fact that this is only a first study, to evaluate the possible use of Bluetooth in factories we have reduced the number of bits to 200.000. Another restriction is that only the bit errors in the payload have been studied and not the header errors.

Two different types of FEC methods are used according to the Bluetooth standard; a (15,lO) shortened Hamming code and a simple repetition code. The repetition code repeats the bit three times (only used in the header of ACL packets and in HV1 packets). If the probability of bit error is known, standard equations and definitions can be used [6, 71 to calculate the probability of packet error and throughput after diversity is introduced.

HV1 packets and ACL packet header

The HV1 packet uses repetition code, which is a very simple block code. Symbol error rate after N bit repetitions can be calculated with equation 4.1

N = number of repeats (3 repetitions in the Bluetooth standard) Pe = probability of bit error with repetition code word

The packet error rate can be calculated with equation 4.3 when the error rate for each repetition code word have been calculated, Pe' is in this case the code word error rate and L the number of code words in one packet.

HV2, DM1, DM3 and DM5 packets

HV2, DM1, DM3 and DM5 packets use the shortened (15,lO) Hamming codes. Shortened (15,lO) Hamming code can correct all single error and detect all double error in each code word, a code word contains 10 data bits and 5 redundancy bits, this gives a code word length of 15 bits. The code word error rate (Pew') is calculated according to equation 4.2.

24 1

Then the packet error rate can be calculated with equation 4.3 where L is the number of code words and Pe is the code word error rate.

HV3, DHl, DH3 and DH5 packets

The HV3, DH1, DH3 and DH5 packets use no FEC at all so the packet error rate (Pp) can directly be calculated with equation 4.3. Pe is the bit error rate without any diversity and L is the number of bits in one packet.

Packet error rate

Pe is the bit error or code word error rate. L is the number of bits or code words in one packet (se table 2.1) and Pp is the packet error rate (PER).

pp= 1 -( 1 -PelL (4.3)

ARQ throughput reduction

Automatic repeat request is very efficient for delivering reliable data, but it reduces the data throughput according to equation 4.4 and the time delay are not deterministic.

The reduced throughput part of maximum throughput is calculated with equation 4.4, Pr is the reduced part of the throughput. Pp is the probability that a packet contains errors and is calculated with equation 4.3.

Reduced throughput part:

PI= 1 -Pp ( 4.4)

The actual throughput (Ra) is calculated with equation 4.5, Rs is derived from the specification of the different types of packets (table 2.1), that are possible to use according to the specification [5 ] . Ra is the actual throughput and is not considering the packet loss caused by header errors.

Ra=Rs*Pr ( 4.5)

Symmetric throughput performance

Symmetric throughput performance is presented as a function of BER for different packet types in figure 4.1.

Figure 4.1 Throughput as a function of BER for different packet types.

5. Test results

We have conducted these measurements at Stora Enso Hylte at PM1 and in an office environment at Halmstad University. The measurements are performed closed to potential sources of disturbance and at potential placement of sensors, a plan is included for each test position. The test platform used is Ericsson Bluetooth Development Kit (EBDK)[ 141.

Test No 1 Line Of Sight (LOS) in office environment A reference LOS test has been done in an office corridor at Halmstad University presented below (Figure 5.1 Office). The walls and ceiling are made out of plaster tiles with a frame of wood. The transmitted signal fades because of the reflections in walls, ceiling and floor. This causes the irregularity in the BER Vs distance in Figure 5.l(Office).

Figure 5.1 BER as function of distance (LOS).

The BER causes retransmissions if ACL connection is used. This means that for the ACL connection the throughput will be reduced when retransmissions are done. The length of a packet has a large impact on the PER, more bits increase the probability of error in the

242

packet. FEC reduces the PER because of the possibility of correcting one bit error in each code word.

but a higher error rate have to be accepted in the transmitted data, because no ARQ is present then the throughput will always be 64kbps.

Figure 5.2 PER for DH and DM packets (LOS in office environment).

Figure 5.2 presents the PER for different packets that can be used in an ACL connection. Calculated PER is done, according to chapter four, for the different ACL packets.

As can be seen in figure 5.2 the PER for DH packets are high already at short distance and LOS. This due to the fact that DH packets have no FEC. It has only a CRC that checks for bit errors. The PER for DM packets, which have FEC, are presented in figure 5.2 and show the decreased PER compared with DH packets.

The actual throughput for the different packets, LOS in an office environment is presented in figure 5.3 below. Only the symmetric throughput has been considered, if the results are compared with the offered throughput according to the standard (table 2.1) the FEC manage to keep the throughput up. It is possible to increase the throughput by using an asymmetric ACL channel, which use different packet types in different directions, but this has not been studied in this paper.

I" :_I 5 ... -- ............. ;:., . ....

............................ . . . . ~-: ~~ -~.~... rw __-_

Figure 5.3 Throughput (kbit/s) in LOS in an off ice environment.

If it is possible to transmit data over a SCO channel without any re-transmission the packet error rate should be as presented table 5.1. In this case the packets are transmitted periodical. The channel is time deterministic

....... ......... -.

;,,"

-. , _ , . .- ,

Figure 5.4 PER for HV packets in LOS in an office environment.

Test No 2 LOS test in a factory environment

This test is done in the basement of the paper-mill under the paper machine. Close to two 600 kW drivers for the mixing pumps (figure 5 . 9 , the large flat metal surfaces and pillars of reinforced concrete are giving severe radio signal reflections.

Figure 5.5 Sketch of the test area in the basement of the paper mill.

Only symmetric DM packets and HV packets have been considered in this test, because of the better performance in the sense of PER and retransmissions. HV packets are not retransmitted at all, and are symmetric according to the BT standard [3]. The BER for the tests at different distances are given in figure 5.1 (Test No 2).

The BER in the rough industrial environment is higher than in the office environment(figure 5.1). This is caused by the radio signal reflected by the pumps, electrical engines and water pipes. In this environment the PER increases rapidly between 5m and 10m even when the DMl packets are in use (table 5.1).

243

Position DMl

C12(5m) 6,7*10-2

DM3 DM5

0,35 0,54

Table 5.2 Throughput (kbit/s) DM packets LOS( pump station).

Between 5m and 10m only DM1 packets should be considered. Longer range then 10m are not suitable for Bluetooth if not an extemal amplifier is used [ 161.

C13(10m)

C14(15m)

C 14( 15m)

0,39 0,95 190

0,72 1 ,oo 130

Table 5.3 PER for HV packets LOS (pump station).

The PER for the synchronous SCO channel is acceptable for voice up to 10m if HV1 packets are used and up to 5m if HV2 packets are used (table 5.3). Error sensitive data, transmitted under these circumstances, suffer by severe damage.

A293(3.9)

-- U

3.9*10-3

Figure 5.6 Sketch of the test area paper machine.

3 F E \J (C B

Figure 5.7 Sketch of the test area paper machine..

A clutter of solid metal machine parts in this case reflects the signal. The disturbed signal is reflected by the high BER values in test A291 and A292 (table 5.4). The test point A292 transmission line goes through a concrete pillar. In test A293 the transmitting unit is placed at a point in floor height, close to a large DC-engine. The cooler equipment to the DC-engine is reflecting and shadowing the radio signal.

Position BER

Test No 3, Non Line Of Sight (NLOS) a sensor on a paper Machine

Test No 3 is performed at a potential sensor position at the dry end of the paper machine close to the pope. One node is placed 8m away and 4m above the floor which gives a distance of 8.9m (A291 and A292 see figure 5.6 and 5.7).

244

Position D M l

r l ( 8 . 9 m I 0:;l 1 0,69 I 0;; 1 A292(9.3) 0,99

A293(3.9) 2,4*10-2 0,14 0,24

DM3 DM5

Table 5.5 PER DM packets (paper machine).

At test point A292 is hardly no throughput achieved at all(tab1e 5.6), even if DM1 packets are in use. Only at test point A293 can DM3 and DM5 packets be considered.

Position BER

Table 5.6 Throughput (kbit/s) DM packets (paper machine).

Position

G181(6m)

D M l DM3 DM5

0,11 OS2 0,74 Position D M l DM3 DM5

A291(8.9m) 90,l 78,5 33,3

A292(9.3m) 4 9 3 1 3 090

A293(3.9m> 106,l 219,6 216,9

Table 5.7 PER for HV packets(paper machine).

Test No 4 NLOS transmitting from the inside of the paper machine

This test is a short-range test when the transmitting unit is placed inside the machine in an inspection alley (figure 5.8). The receiving unit is NLOS directly outside the paper machine.

G182(8m)

Figure 5.8 Sketch of the test area inside paper machine.

4,0*10-2 0,22 0,3675

Position DMl DM3

G181(6m) 96,7 122,8

G182(8m) 1043 199,O

G 18 1 (6m) 8.7" 10-3 + G182(8m) 5.0*10-3

DM5

75,3

181,3

Table 5.8 BER NLOS (inside a paper machine).

Position

A291(8.9m)

HVl HV2 HV3

2,89*10-2 0,21 0,93

A292(9.3m>

A293(3.9m:

Table 5.10 Throughput DM packets (inside a paper machine).

0,12 0,63 1 ,o 3,60*10-3 3,0*10-2 0,61

Position

G181

Table 5.1 1 PER for HV packets (inside a paper machine).

HVl HV2 HV3

1,79*10-2 0,13 0,88

6. Conclusion

This work is a first study of the performance of Bluetooth in an industrial environment. The short-range radio communication in factories can reduce the use of cables and are in some cases a good alternative to wired connections. Especially when sensors are placed on rotating and moving machine parts.

We have tried to stress the Bluetooth radio to the upper limit to see what possible use Bluetooth can be in an industrial environment. As can be concluded from the first reference test done in an office environment the throughput will not reach the specified throughput according to the standard (table2.1). DH packets that are supposed to give high throughput according to the standard, can not be used for longer distance than four meters. This is caused by the fact that there is no FEC in these type of packets, errors are only detected by CRC. The large number of bits in each packet indicates that it is impossible to use them without any forward error

245

correcting. If we take a closer look in figure 4.2 we can see that the throughput reduces fast when the BER increases.

The LOS test1 and test2 shows that it is possible to use all DH packets up to 5 meters, but then the throughput reduces rapidly between 4m and 6m (figure 5.3). When the communicating units are NLOS the performance is reduced in a very unpredictable way (table 5.6 and table 5.10), demanding for dynamic use of FEC. Even if the Bluetooth standard [5] supports different types of packets, the only FEC possible to use in an ACL data channel is the (15,lO) Hamming code, the dynamic is in the length of the packet (number of bits). We believe that a more dynamic range of different FEC codes should be provided by the standard, to guarantee the safety of the data and reducing the need for re-transmissions. This should increase the number of applications that Bluetooth can be used in.

The SCO channel has three different kind of packets all giving a throughput of 64kbps and allocating different amount of the total channel capacity. The SCO channel is provided as a synchronous channel. Voice transmissions are not as highly dependent on bit errors as data transmissions. Voice channel is more delay sensitive than error sensitive. The SCO channel could be used as a data channel to provide synchronous data transmissions. As seen in figure 5.4 and tables 5.3, 5.7 and 5.1 1 the packet error rate is rather high and these errors must be accepted because no re-transmission is possible in a synchronous channel. This calls for dynamic use of different FEC’s. Today the standard only provides; no FEC, (15,lO) Hamming or repetition code. The cycle time for a synchronous data channel should also be possible to choose. This is caused by the different periods data have compared with voice. Today only three different periods are possible.

As a last concluding remark one can say the Bluetooth has the potential to offer a wireless transmission link even in quite hash industrial environment, but additional error correction schemes and more adaptive error correction protocols will increase the applicability.

References

[l] J. Haartsen, ”Bluetooth the universal radio interface for ad hoc, wireless connectivity”, Ericsson Review, No 3,

[2] J.R. Pimentel, Communication Networks for Manufacturing, first edition, Prentice-Hall inc., USA, 1990.

1 9 9 8 , ~ ~ . 110-117.

[3] Thomas M. Siep et alt, “Paving the Way for Personal Area Network Standards: An overview of the IEEE P802.15 Working Group for wireless Personal Area Networks”, IEEE Personal Communications, pp. 37-43, Feb, 2000.

[4] Yiu Ming Lang, P.H Wittke, ”Frequency-Hopped Spread- Spectrum Transmission with Band-Efficient Modulation and Simplified Noncoherent Sequence Estimation”, IEEE Transactions on Communication, pp. 21 84-2 195,vol. 38, No 12, Dec, 1990. [5] Telefonaktiebolaget LM Ericsson, ”Specification of the

Bluetooth system Core v l .Oa”, Telefonaktiebolaget LM Ericsson, 1999.

[6] Dr. Kamilo Feher, Wireless Digital Communication, Modulation & Spread Spectrum Applications, Prentice- Hall, New Jersey, 1995.

[7] John G. Proakis, Digital Communication, McGraw-Hill, Singapore, 1995.

[8] Ericsson Components AB, ”ROK 101 007/1 Bluetooth Module”, Ericsson Components AB, Kista, 1999.

[9] Snorre Kjesbu ”Bluetooth in Industry”, Bluetooth Congress 2000, Monte Carlo, June, 2000.

[lo] Seyed Morazavi, ”Fundamentals of RF propagation at 2.45 GHz”, Bluetooth Congress 2000, Monte Carlo, June, 2000.

[l 11 Theodore S . Rappaport, “Indoor Radio Communication for Factories in the future”, IEEE Communication Magazine, pp. 15-24, May, 1989. [ 121 Ad Kamerman and Nedime Erkocevic, ”Microwave Oven Interference on Wireless LANs Operating in the 2.4 GHz ISM band”, Personal, Indoor and Mobile Radio Communication PIMRC ‘97,1997. [13] J.C. Haartsen and Stefan Zurbes, Bluetooth voice and data

performance in 802.11 DS WLAN environment, Telefonaktiebolaget LM Ericsson, May, 1999.

[14] S.Y Seidel: “914 MHz Path Loss prediction Models for Indoor Wireless Communication in multiflored Buildings”, IEEE Transactions on Antennas and Propagation, vol40, No 2, Feb, 1992.

[15] Raphael Rom, Moshe Sidi, Multiple access Protocols Performance and Analysis, Springer-Verlag. New York, 1990.

[16] IEEE standard board, Wireless LAN medium access control and physical Layer specifications IEEE 802.1 1, IEEE standard board, June, 1997.

[17] Ericsson Mobile Communication AB: EBDK Technical Dokumentation, Ericsson Mobile Communication AB, 1999.

[18] Martin Roter: Test Specification RF 0.7, Document number 20.B 153/0.7 Bluetooth SIG, 1999.

246