Improving the intelligibility of underground station public address/

voice alarm systems through a close proximity horizontal line array

Diego Cordes 1 Sustainable Acoustics - LSBU Unit 1, Gander Down Barns, Rodfield Lane, Ovington SO24 0HS Douglas Shearer 2 RBA Acoustics - LSBU 103 Borough Rd, London SE1 0AA

ABSTRACT

As London’s extensive underground network passenger communication is highly dependent

on its Public Address/ Voice Alarm (PA / VA) system, attention to its efficacy is of great

importance. In its current state, the system lacks 21 st- century performance regarding modern

technological capabilities. It has not been able to overcome the difficult acoustic challenge that

an underground station represents, struggling to achieve the minimal life safety intelligibility

requirements established by the British standard and bringing an uncomfortable experience to

its millions of daily passengers. A reinterpretation of the line array / planar source concept has

been proven to raise the intelligibility beyond the minimal requirements by achieving a

homogeneous sound distribution close to the public, with minimal room excitation thus

minimising reverberations destructive interference to intelligibility.

1 dcordes@sustainableacoustics.co.uk

2 doug.shearer@rba-acoustics.co.uk

1. INTRODUCTION

Communication is an essential part of human interaction where speech is the primary method used. It brings us together to form a society and makes the interaction between ourselves and the environment safe and efficient. From a technical point of view, it allows us to interact with industrial machinery and move faster and further than naturally possible, from flying around the globe to taking the tube to work. And communication is what makes this machine-human-environment coordination possible. When this link breaks, subjects become isolated from the system, and coordination is no longer possible, increasing the risk of conflicts.

2. THEORETICAL BACKGROUND

2.1. Speech Intelligibility The speech intelligibility of a communication system measures the proportion of the content of a speech message that can be correctly understood (1).

When speech is transmitted into a reverberant space, reflections and reverberation distort the waveform by smearing it in time, where the reverberant tail of one word can overhang the start of the next one and mask it, reducing the clarity and intelligibility. And in cases when the SNR ratio is too low, parts of words become lost, and intelligibility deteriorates (2).

2.2. Public Address Voice Alarm (PAVA)

Public Address and voice alarm (PA/VA) systems are essential means of communication with the public used for general information distribution and guidance regarding safety procedures in case of an emergency. The difficult acoustic conditions in an underground station represent a severe challenge, as high-volume spaces with little or no absorption create high reverberation levels that greatly complicate reaching the minimum life safety intelligibility requirements established by the authorities (0.50 min. STI, >65dB SPL, >10dB SNR, 90% coverage).

3. PROCEDURE

An average underground station from the London Underground has been modelled.

Fig. 1 An average London underground station. Absorption in green - Tunnel entrance in Black.

A standard PAVA system currently used by the London Underground was modelled into the station to “calibrate” the space to current standard parameters. Once the platform achieved the current station conditions, the new proposed line array (LA) PAVA system was modelled in the space to be compared with the current system.

The LA design used simulated a ribbon transducer along the whole extension of the platform at 2.3m height, through 66 modules of 1.7 m long passive array units each with 27 modular line elements of 8 cm, except for 2m free at the beginning. It is the closest representation of a continuous line source with a low processing power requirement (3), where the sound field from each transducer is combined to form a continuous wavefront and cover the entire horizontal volume, with limited comb lines formation due to minimal overlapping dispersion.

Fig. 2 Line array image from CATT-A directivity module.

Fig. 3 & 4 Line array position and angle in the platform station lateral wall.

It was heavily filtered to maximise intelligibility, as these modules were not designed for this use, as opposed to the Penton MCS20T, which are specifically designed for the application and perform their best without any equalisation.

Freq. 125Hz 250Hz 500Hz 1000Hz 2000Hz 4000Hz 8000Hz 16Khz

Gain -14 -88 -25 -10 -6 2 4 0

Table 1. Line array system filter used in the simulation (dBA).

6 receivers were located at 1.6 m height in accordance with the BS EN 50849:2017 for an acoustically different area (ADA) for the 355m2 of the platform. The positions also considered being off-axis regarding the current Penton PAVA system with speakers 5 m apart (4), and distances 0.5, 1,25 and 2m from the lateral wall are used to cover the entire representative public area. From the total 2.96 m wide of the platform, the last 0.42 m next to the track were considered a “no step” zone.

Fig. 5 London Underground station with current Penton system, receivers positions and head aim.

The input signal for both systems was “Lv_voice_IECmale” and calibrated to reach 90dB SPL at receiver position to guarantee the minimum 10 dB signal-to-noise ratio established by the BS 5839- 1:2017. The absorption coefficients and scattering values were based on previous research (5)(6) of the underground station’s materials and adapted to this particular case.

100.000 rays were used for the current Penton system simulation with a 3000 ms impulse response length, and 1.000.000 rays with a 3000 ms impulse response were used for the LA system. In both cases, the impulse response and amount of rays were increased until a late echo was cleared , although still slightly present in the echograms and auralisations.

The simulation was done using Algorithm 1, max split-order 1, E averages 1, the diffraction step was not used, air absorption step was used, and lost rays were 0.00%. Background noise (Gang) of 52dBA.

One significant limitation is the exclusion of the pedestrian access routes in the model, as it was considered that they would make the model more complex and difficult to standardise.

4. RESULTS

STI: 0.43 vs 0.51 (avg. at receivers)

Fig. 6 & 7 STI Mapping Penton & LA.

EDT: 2.47 vs 1.73 sec. (avg. at receivers)

Fig. 8&9 EDT Mapping Penton vs LA.

SPL – Direct Only: 78 vs 80dB (statistical). 11 vs 7dB difference with SPL (89 vs 87dB)

Fig. 10&11 Penton vs LA, SPL Mapping - Direct sound only.

A direct correlation SPL - STI was observed, where higher SPL meant higher STI, and an inverse one with T20, where higher reverberation time meant less STI, as this are the two most important parameters for the overall intelligibility (1).

As the acoustic power of the LA was closer to the receivers, improving the DRR as supported by the T20 decrease from 2.65 sec. to 2.2 sec., the EDT from 2.47 sec. to 1.73 sec. and the larger direct field shown by the higher proportion of direct SPL towards the platform, the total amount of energy into the space is not only reduced but also better distributed at the receiver positions, leaving much less spillage to excite the room and create destructive interference to diminish the intelligibility of the PAVA signal, which results in a considerable increase in STI from 0.43 to 0.51, bringing the London Underground station above the minimum life safety requirements of 0.50 STI.

5. CONCLUSIONS

The proposed LA system showed a performance increase in all areas of comparison against the current Penton system.

By creating a line source emitting a plane wave along the full extension of the platform at close distance to the receiver, the direct-to-reverberant ratio is maximised with a very homogeneous sound distribution making the whole platform become a complete on-axis area at receiver height, as energy concentration is maximised while room excitation and destructive interference are minimised.

Thanks to the extreme spatial distribution of energy of the line array module, close proximity to the source can be achieved without being uncomfortable, as each individual transducer produces very small amounts of acoustic energy (~50dB), when the system aims to reach 90dB SPL at receiver position for adequate PAVA communication.

Although this simulation was done with ribbon transducers, which would be too sensitive and uneconomical for an underground station PAVA system, it aimed to prove the theory and was done with a transferable system in mind. According to Thibault Guillaume, this kind of plane wave could be replicated through a plane wave radiator economically and feasibly, which would also have a considerable resistance advantage against fire hazards, fewer maintenance requirements and better aesthetics.

6. AKNOWLEDGEMENTS Graeme Littleford, Thibault Guillaume, Peter Mapp, Luis Gomez Agustina, Douglas Shearer and

Steven Dance.

7. REFERENCES

1. British Standards Institute (2013) BS 5839-8:2013 Fire detection and fire alarm systems for

buildings - Part 8: Code of practice for the design, installation, commissioning and maintenance of voice alarm systems. London: BSI.

2. Lyons, A. and Ballou, G. M. (2008) Handbook for sound engineers. Elsevier Science &

Technology. 3. Dalenbäck, B.I. (2007), CATT-Acoustic User’s manual, version 8.0g, www.catt.se, Sweden.

4. British Standards Institute (2011) BS EN 60268-16:2020 Sound system equipment. Part 16:

Objective rating of speech intelligibility by speech transmission index. London: BSI. 5. Gomez Agustina, Luis. (2012) Design and optimisation of voice alarm systems for

underground stations. Ph.D. thesis, London South Bank University.

6. Littleford, Graeme (2019) The Effect of Passenger Occupancy on the Acoustic Performance

of VA Systems in Underground Stations. MSc dissertation, London South Bank University.