dynamic range control (drc) and music/speech control (msc

15
56 EBU Technical Review Autumn 1994 Hoeg et al. Dynamic Range Control (DRC) and Music/Speech Control (MSC) Programme–associated data services for DAB W. Hoeg (DBP–Telekom – FTZ) N. Gilchrist (BBC) H. Twietmeyer (IRT) H. Jünger (Jünger Audio) 1. Introduction The Eureka 147 Digital Audio Broadcasting (DAB) system [1, 2, 3] provides not only a very high audio quality for the listener in general, but also includes provision for many additional data services. Most of these data services are indepen- dent of the audio programme, but some of them are closely related to the audio signal. The latter form the so–called Programme Associated Data (PAD), which includes the Dynamic Range Control (DRC) System and the Music/Speech loudness Control (MSC) system. Some other special fea- tures will be provided by the DAB system to tailor the reproduced audio signal to the individual re- quirements of each listener in his individual recep- tion environment, and so improve the acceptability of future audio reception in general. In order to offer an overview of data services for DAB, a general review is given. This is followed by the basic requirements, the strategy and the principles of DRC for DAB, based on the require- ments of Eureka 147, the International Telecom- munication Union (ITU) [2] and the European Telecommunication Standards Institution [3], to- gether with systems for the real–time implementa- tion of DRC for DAB. There will always be the need for broadcasters to apply dynamic range compression to some types of programme material, either because of the limitations of the broadcasting medium or because of listeners’ requirements. The Eureka 147 DAB digital audio broad casting system enables broadcasters to transmit programmes with a relatively wide dynamic range, accompanied by a dynamic range control (DRC) signal which the listener may use to effect unobtrusive compression of the programme dynamics, if required. A music/speech control (MSC) signal, which is also transmitted, will enable the listener to balance the loudness of different types of programme according to taste. The techniques used for the optional compression of programme dynamics in DAB may also be used to control the dynamic range of programmes unobtrusively for conventional VHF/FM broadcasting, but without control data being transmitted. Original language: English Manuscript received 2/9/94. The DAB logo has been registered by a member of the Eureka 147 – DAB consortium.

Upload: dangkien

Post on 10-Feb-2017

224 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Dynamic Range Control (DRC) and Music/Speech Control (MSC

56 EBU Technical Review Autumn 1994Hoeg et al.

Dynamic Range Control (DRC) andMusic/Speech Control (MSC)Programme–associated data services for DAB

W. Hoeg (DBP–Telekom – FTZ)

N. Gilchrist (BBC)

H. Twietmeyer (IRT)

H. Jünger (Jünger Audio)

1. Introduction

The Eureka 147 Digital Audio Broadcasting(DAB) system [1, 2, 3] provides not only a veryhigh audio quality for the listener in general, butalso includes provision for many additional dataservices. Most of these data services are indepen-dent of the audio programme, but some of them areclosely related to the audio signal. The latter formthe so–called Programme Associated Data (PAD),which includes the Dynamic Range Control(DRC) System and the Music/Speech loudnessControl (MSC) system. Some other special fea-tures will be provided by the DAB system to tailorthe reproduced audio signal to the individual re-quirements of each listener in his individual recep-tion environment, and so improve the acceptabilityof future audio reception in general.

In order to offer an overview of data services forDAB, a general review is given. This is followedby the basic requirements, the strategy and theprinciples of DRC for DAB, based on the require-ments of Eureka 147, the International Telecom-munication Union (ITU) [2] and the EuropeanTelecommunication Standards Institution [3], to-gether with systems for the real–time implementa-tion of DRC for DAB.

There will always be the need forbroadcasters to apply dynamic rangecompression to some types ofprogramme material, either becauseof the limitations of the broadcastingmedium or because of listeners’requirements.The Eureka 147 DAB digital audiobroad casting system enablesbroadcasters to transmit programmeswith a relatively wide dynamic range,accompanied by a dynamic rangecontrol (DRC) signal which the listenermay use to effect unobtrusivecompression of the programmedynamics, if required. A music/speechcontrol (MSC) signal, which is alsotransmitted, will enable the listener tobalance the loudness of different typesof programme according to taste.The techniques used for the optionalcompression of programme dynamicsin DAB may also be used to controlthe dynamic range of programmesunobtrusively for conventionalVHF/FM broadcasting, but withoutcontrol data being transmitted.

Original language: EnglishManuscript received 2/9/94.

The DAB logo has been registeredby a member of theEureka 147 – DAB consortium.

Page 2: Dynamic Range Control (DRC) and Music/Speech Control (MSC

57EBU Technical Review Autumn 1994Hoeg et al.

2. Data facilities in DAB

2.1. Architecture of the DABensemble

Although the DAB system is designed primarilyfor the broadcasting of audio programmes, its ar-chitecture is intended to offer the greatest flexibil-ity for non–audio applications, so that the best usecan be made of the limited resources of the RFspectrum and the overall data capacity of the DABsignal. As an international standard the systemmust be readily adaptable to specific national orlocal needs. It must also be adaptable to futureapplications, such as multimedia services.

The ensemble of audio programmes and data ser-vices carried by a DAB signal can be described ac-cording to a hierarchy of three levels:

– The service level telling the listener which ser-vices are available.

– The service component level informing the lis-tener of the components which comprise theservice.

– The channel level. The DAB capacity is parti-tioned in a set of sub–channels. Each audio ser-vice component resides in a sub–channel of itsown. A data service component can reside in asub–channel of its own as a continuous streamin multiples of 8 kbit/s, segmented in 24 msframes(stream mode), in the so–called FastInformation Channel (FIC) at low rates, of100 bit/s or less, or it can share a sub–channelwith other data service components in a packetmultiplex (packet mode).

Included in the FIC is the Multiplex ConfigurationInformation (MCI). This comprises all the in-formation needed to get fast access to every serviceand its components. The MCI describes in detailhow each service component can be accessed (i.e.in which sub–channel it is to be found), the grossand net bit rate, and any other details (e.g. the errorprotection scheme).

2.2.Categories of data in DAB

The DAB ensemble can contain four differentcategories of data:

– General data services, independent from otherdata or audio services.

The DAB system offers a flexible transport mecha-nism (with options for conditional access), that canbe used equally well for non–audio data services.The robustness of the DAB system and its RF char-

acteristics guarantee that DAB data services willbe well suited for reception in cars and portable de-vices. A stationary DAB receiver built into a PCcould serve as a data service terminal or be used asa multimedia receiver.

Data services from a small number of service pro-viders may be distributed as one single DAB en-semble to all private end users, e.g. the populationof a town or of a whole country. Typical applica-tions could be the distribution of computer pro-grammes, computer games, traffic guidance, pricelists, electronic newspapers and weather reports.

– Data as a service component of an audio pro-gramme

Nowadays radio and television programmes offerdata facilities that are linked (by technical mecha-nisms) to the selected programme. Traffic messageinformation (TMC) and teletext are examples ofthese. In the DAB system, different data servicecomponents may be linked either to an audio ser-vice or to another data service. By using this optionthe programme provider may increase the attrac-tion of his programme by offering additional faci-lities to his listeners. Applications may includetextual information on the programme schedule,detailed information about the topic addressed bythe audio, the manuscript of a lecture, maps, stillpictures referring to a news report, and travel in-formation. Several co–operating programme pro-viders may share the same service component. Alllisteners to the different co–operating programmeshave access to all of the information.

– Basic Service Information (SI)

Like the Radio Data System (RDS) in FM broad-casting, the DAB system offers basic service in-formation (carried in the FIC in machine–readableform) to assist the listener and his receiver to selecta service. It comprises programme service labelsdisplayed by the receiver, identification of the pro-gramme language and programme type, cross ref-erences to other programmes, transmitter identifi-cation for navigation purposes or for geographicalselection of information, and automatic control forpresenting special announcements (e.g. news, traf-fic information). The extent of these SI featuresgoes far beyond the features of RDS and is capableof meeting needs which will evolve in the future.

– Programme Associated Data (PAD)

Data travelling in the PAD channel is intimatelyrelated to the audio programme. It is assembledtogether with the coded audio in the same DABframe, so the time relation established at the sourceis maintained throughout the complete DAB trans-

Page 3: Dynamic Range Control (DRC) and Music/Speech Control (MSC

58 EBU Technical Review Autumn 1994Hoeg et al.

mission chain. So audio and PAD cannot be subjectto different transmission delays. Therefore PAD iswell suited to time critical data such as DRC data.

Because of the strong relation between audio andPAD, the programme associated data can be ac-cessed by the receiver only in conjunction with theaudio. The PAD channel consists of a fixed part(F–PAD, 2 Bytes per 24 ms frame, i.e. 0.66 kbit/s),and optionally of an extended part (X–PAD, com-prising 0, 4, or more than 4 bytes per 24 ms, de-pending upon the “free” data capacity in the audioframe). The first DAB receivers will be able tocope with up to 16 kbit/s of PAD.

The X–PAD may include textual informationabout the audio programme (title, schedule) andthe actual programme item (title, composer, opus,artist etc.). Some listeners will be interested in hav-ing lyrics, song or opera texts simultaneouslytranslated from the original language to the lan-guage commonly used in the coverage area of theDAB signal.

The F–PAD can convey codes to identify an itemof music (Universal Product Code, InternationalStandard Recording Code). However, the mainpurpose for the F–PAD is to carry time–criticalprogramme–related data. Transport mechanismsare defined for controlling additional devices (e.g.memory read out for still pictures), Music/SpeechControl (MSC) and Dynamic Range Control(DRC). In each audio frame, 6 bits of F–PAD (i.e.about 250 bit/s) can be assigned for the transmis-sion of DRC data.

Fig. 1 shows the structure of the PAD field withinthe DAB audio frame. More detailed information

on the total DAB multiplex and system supportfeatures can be found in [1] and [3].

3. Dynamic range control(DRC)

3.1. General

The dynamic range of an audio programme signal(sometimes termed the programme dynamic) is therange between the highest and the lowest usefulprogramme signal level. The problems associatedwith programmes having a wide dynamic range,and with achieving a satisfactory loudness balancebetween different parts of the radio programme(such as speech or music) are well known from ex-perience with VHF/FM [4, 5]. In many cases thesource programme dynamic range may be muchlarger than the available dynamic range of thetransmission channel; it may also be much largerthan the usable dynamic range in noisy environ-ments such as cars (the so–called reproductiondynamic). The reduction required in the dynamicrange of the programme may be 10 to 30 dB (ormore). Fig. 2 shows the dramatic differences in thereproduction dynamics to be expected (based onfigures given in [6]). In addition, there may be verydifferent requirements in the loudness balance be-tween music and speech programmes, dependingon the interests of the listener and the conditionsfor reproduction [5, 7]. During the last 30 yearsoperational practices have evolved to try and solvethese problems. These have included:

– reducing the dynamic range of sound record-ings and live transmissions by manual fadercontrol to within a maximum range of 40 dB inorder to match this range to the characteristicsof the available transmission channel (AM,FM), as well as to that of listening conditions ina typical living room [4, 8, 9];

Header CRC Sub–band samplesBit

allocation STUFFSCFSI X–PADScalefactors

SCFCRC

F–PAD

Audio data X–PAD SCF–CRC F–PAD

Variable length

4 bytes

Better protected part

Fixed length

ByteL–1

ByteL

Figure 1Structure of the PADfield in the DAB audioframe.

Page 4: Dynamic Range Control (DRC) and Music/Speech Control (MSC

59EBU Technical Review Autumn 1994Hoeg et al.

1525

35

25

40

>60

100km/h

50km/h

0km/h

Noisy(office,kitchen)

Quiet(livingroom)

Car Home Studio

90

70

50

30

10

0

40

60

CDDAB

AnalogueFM

Usableprogramme dynamic

Available dynamic range(dB, approx.)

Environmental noise floor

15

dBAS

– controlling different programme types (speech,symphonic music, chamber music etc.) to dif-ferent maximum levels in order to make the per-ceived loudness conform to a particular require-ment [7, 8, 10, 11].

The dynamic range commonly used by the broad-casters (ca. 40 dB) may well be too large for poorlistening conditions (e.g. in a moving car) and thenumber of recordings with a much more extendeddynamic range (taken from CD or other digital re-cording media) is increasing dramatically. So theneed for dynamic range control is increasing. Fur-thermore, the balancing of different types of pro-grammes by broadcasters may not be ideal for alllisteners. This means that these problems can onlybe solved at the receiver, taking into account thelistening conditions and the requirements of thelistener. For this to be feasible, additional informa-tion must be provided by the broadcaster concern-ing the current type of programme and the gain ad-justments which may be needed to reduce thedynamic range.

3.2. State of the art

In addition to the methods used by broadcasters tosolve dynamic range problems by operationalpractices, proposals for using a type of compandersystem (compression at the studio side with expan-sion at the receiving side) have been made [4], butthere has been no practical implementation of this.

A method for “Variable Dynamic Range” has beenproposed, using a dynamic control signal derivedeither from a fader unit operated by a sound engi-neer, or from a dynamic range processing unit [12].

The control signal, transmitted in parallel with theaudio programme signal, may be used at the re-ceiving side in order to compress or to expand thedynamic range of the reproduced audio signal. Theintention is to effect a very smooth change of thegain, similar to careful control with a manual fader,and the facility for setting the balance in loudnessbetween speech and music programmes is in-cluded. However, the hardware realisation [13]needs a relatively long look–ahead–time (5–10 se-conds), and a relatively high data rate to transmitthe dynamic control signal.

Solutions fulfilling the Eureka 147 requirementsfor DRC data for DAB have been proposed by theBBC [14] and German Telekom [15]. Both sys-tems need a significant look–ahead time. This istypically from 30 ms up to 3 seconds for the BBCsystem, and about 24 ms for the German Telekomsystem. More details of these systems are given inSections 4 and 5 below.

Self–contained compressors, for use in receiversand operating without the need for transmittedcontrol data, have been proposed by several au-thors [16, 17]. One of these, developed by the IRT[17], is based on scale factor weighting in theMPEG1–Layer 2 source decoder (the so–calledMUSICAM–DRC system).

Table 1 gives an overview of possible dynamicrange control system concepts. The existence of apractical system is indicated in the first column.The table shows, by the number of proposals, thatthere is significant interest in having dynamicrange control systems in broadcasting.

Figure 2Reproduced dynamic

range.

Page 5: Dynamic Range Control (DRC) and Music/Speech Control (MSC

60 EBU Technical Review Autumn 1994Hoeg et al.

3.3. Basic requirements

3.3.1. System requirements

In general, there are certain requirements which aDRC system must fulfil in order to guarantee ahigh audio quality:

– Universality: The system should be suitable forall kinds of future broadcast programme (radioor television).

– Improvability: It should be possible to improveor to change the characteristics at the transmit-ting side without changes needed in receivers.

– Flexibility: The listener should have the optionof using DRC or not, with control over thedegree of compression.

– Reversibility: A requirement for a system ofpre–compression, for use with VHF/FM broad-casting, is that it should be possible at thereceiver to restore the original dynamic range ofthe source programme (i.e. remove completelyany changes in the dynamic range made at thetransmitting end).

– Compatibility: If the programme is compressedby the broadcaster, it must be usable withoutany further processing at the receiver.

– Low complexity: The measures needed at the re-ceiver for use of the system should cause a mini-mum of additional cost.

– Delay: Any delay in the signal path introducedby dynamic range control should be minimised.However, most processes introduce some delay,and as far as possible this should occur in thebroadcasters’ equipment, and not in thereceiver.

3.3.2. Requirements for dynamicprocessors

In general, the intention is to compress the overallreproduced dynamic range for the listener whilstpreserving the impact of a dramatic change in theprogramme dynamics, as well as the detaileddynamic structure within the music. If a signalwith a reduced dynamic range is broadcast, withre–expansion at the receiver, the output signalshould be virtually indistinguishable from the un-compressed source signal (this is only relevant toVHF/FM, see reversibility).

There are a number of characteristics defining theoperation of a compressor or expander. The princi-pal characteristics are described below.

The (static) operational gain characteristic (some-times termed the gain law) indicates the output sig-

Table 1Overview of possibleDRC systemconcepts.

Page 6: Dynamic Range Control (DRC) and Music/Speech Control (MSC

61EBU Technical Review Autumn 1994Hoeg et al.

Input level (dB)

Out

put l

evel

(dB

)

0

0 Input level (dB)

Out

put l

evel

(dB

)

0

0

Input level (dB)

Out

put l

evel

(dB

)

0

0 Input level (dB)

Out

put l

evel

(dB

)

0

0

a b

c d

Gain law with limiting ofextreme levels

Gain law givingcompression over theentire dynamic range

Gain law combining thecharacteristics a and b

Gain law showinganother combination ofcharacteristics a and b

0 dBrel : relative to Permitted Maximum Signal level (ITU–R).

nal level as a function of the input signal level.Generally, this indicates that quiet signals aremade louder, and loud signals quieter. Some exam-ples of possible gain laws are shown in Fig. 3. Thelaw shown in Fig. 3a would provide limiting atboth extremes of the dynamic range; the law in Fig.3b would compress all dynamics; that in Fig. 3ccombines both of the previous laws. Fig. 3d showsanother gain law derived from Fig. 3b, with nocompression for extremely low levels.

For the simple characteristics such as those shownin Figs. 3b and 3d, it is possible to define a ratio ofthe input and output dynamic ranges, which ap-plies over whole characteristic, or over any part ofthe characteristic:

Ratio r = �Lin / �Lout

where:�Lin is the input dynamic range�Lout is the output dynamic range

A static characteristic cannot fully describe the dy-namic working of a compressor. It shows princi-pally the gain values which are reached in a quasi–stationary state of the system. The realcompression effect will be influenced by the dy-namic parameters of the system, such as the attacktime and the release time.

The attack time mainly influences the reactiontime of the system to signal peaks, in order to re-duce the gain in time to avoid reproducing exces-sively loud sounds, or causing over–modulation.

The release time determines the rate at which thegain can increase following loud programme ma-terial. A long release time gives gentle control ofthe programme level; a short release time enablesthe processing to follow the dynamics of the sourceprogramme more closely, compressing to greatereffect but at the risk of introducing obtrusive ef-fects such as “gain ducking” (or “pumping”) andnon–linear distortion.

Figure 3Principal operational

characteristics forDRC systems.

Page 7: Dynamic Range Control (DRC) and Music/Speech Control (MSC

62 EBU Technical Review Autumn 1994Hoeg et al.

Sourceencoder

(ISO Layer II)

PAD – MUX

DRC datagenerator

DABreceiver

Sourcedecoder

(ISO Layer II)

Gaincontrol

DRCdecoder

(compression)

PADDEMUX

Transmission Reception

Audioinput

Uncompressed audio+ DRC data

6–bit DRC data6–bit

DRC data

Compressedaudio

In practice, different time constants may be used,depending upon the content and history of the au-dio signal (analysed during an appropriate look–ahead time) in combination with a defined staticcharacteristic.

4. DRC systems for DAB (withtransmitted control data)

4.1. Requirements of the DABspecification.

The requirement is to provide the potential foradapting the dynamic range of the broadcast pro-gramme to the individual needs of each listener.This can be realised by means of a Dynamic RangeControl (DRC) system with transmitted controldata, incorporated in the DAB signal [2, 3]. Fig. 4shows a simplified block schematic diagram of aDRC system for DAB.

At the broadcaster’s premises a DRC signal is gen-erated, which describes the audio gain to be ap-plied in the receiver, as a succession of values. ThisDRC signal is transmitted in a coded form (“DRCdata”) together with the audio signal. It is a require-ment of the DAB specification [13] that the audio

signal is transmitted with its original programmedynamic, without any pre–compression.

The audio signal is bit–rate reduced by an ISOLayer 2 encoder. The DRC data is incorporated inthe ISO Layer 2 bit–stream as programmeassociated data (PAD). Within the PAD, the data isformatted as shown in Fig. 1. A complete specifi-cation of the DAB data format, including the DRCdata, is contained in the draft European Telecom-munication Standard [3].

In the receiver, the regenerated DRC signal may beused to control the audio gain in order to match thedynamic range of the received audio programme tothe requirements of the listener, or to improveaudibility in difficult conditions.

4.2. The Telekom DRC system

Telekom FTZ and Jünger Audio Inc. have jointlydeveloped a Dynamic Range Control (DRC) sys-tem for DAB, conforming to the main require-ments of the Eureka 147 DAB specification [3].

Fig. 5 shows a simplified block schematic diagramof the DRC generator. The audio source pro-gramme is analysed and DRC data is generated.Depending on the processing time (including a

Time delay(2 x 24 ms)

Signalanalysis

Compressorsimulation

Limitersimulation

Gainreductionalgorithm

AES/EBUdata

insertion

DRC datageneration

AES/EBUaudioinput AES/EBU

audio output

Tosourcedecoder

To PAD

DABframeclock

6–bitDRC data

(DRC data)

Figure 4Block schematicdiagram of a DRCsystem for DAB.

Figure 5Block schematicdiagram of theTelekom DRC signalgenerator.

Page 8: Dynamic Range Control (DRC) and Music/Speech Control (MSC

63EBU Technical Review Autumn 1994Hoeg et al.

look–ahead time which is needed for an appropri-ate signal analysis), some delay must beintroduced into the signal path. The Telekom sys-tem needs a delay equivalent to two 24ms–frames.

The DRC signal is generated by a characteristicwhich is linear over most of the dynamic range,with an appropriate compression ratio r. This isshown in Fig. 6. An overall range of gain variationof about 15dB is linearly coded in a 6–bit digitalword, for each 24ms frame of the broadcast audiosignal.

The intention is to provide long–term compressionof the dynamic range whilst preserving the impactof dramatic changes in the programme dynamics.Therefore the generating algorithms use a specificset of time characteristics controlled by the audioprogramme, as mentioned above; these are not thesubject of any specification. Optionally, it could beadjusted by a pre–set function to different types ofprogramme.

The DAB audio programme signal is transmittedwithout any dynamic compression applied afterthe production/post–production stage. During theproduction of some types of programmes, com-pression may be applied for “artistic” reasons. Pro-gramme signals compressed at this stage aretreated in the same way as uncompressed signalsby the DRC system. The DRC data is sent outsimultaneously with the audio signal to the receiv-er, and may be used optionally for controlling a(further) compression process in the receiver if thelistener so wishes.

4.3. The BBC DRC system

Recognising that a suitably–trained operator (astudio manager in radio, or a sound supervisor intelevision) can produce very good results whencontrolling the audio programme level, the BBChas developed a control algorithm to run in digitalsignal processing (DSP) equipment which imitatesthe action of the studio manager. When compres-sing the dynamic range of musical material, thestudio manager discreetly raises the level of quietpassages in the music, and anticipates an ap-proaching fortissimo by slowly (and, hopefully,unobtrusively) reducing the level prior to the musi-cal climax. The intention is to compress the dy-namic range whilst preserving the impact of a dra-matic change in the programme dynamics (e.g.that caused by a crescendo or subito fortissimo ).A DSP has neither the studio manager’s knowl-edge of music nor the ability to read a musical

score, but it can gain prior experience of the pro-gramme content by delaying the programme inmemory whilst making the necessary level adjust-ments.

The original algorithm looked 3 seconds aheadinto the audio data. With such an arrangement, itcould anticipate impending changes in level andstart to make gradual gain changes. The figure of3 seconds was chosen as a compromise betweenbetter anticipation and ease of operational use. Ifa shorter delay is used there is less warning of thechanges in level, so more abrupt reductions in gainsometimes have to be made. However, the require-ments of “live” broadcasting are likely to precludethe possibility of using delays as long as 3 seconds,and subsequent development of the algorithm hasenabled delays as short as 30 milliseconds to beused.

The DRC system was originally conceived as aself–contained dynamic range controller, for usewith conventional FM transmitters [14]. In DAB,it comprises a broadcaster’s DRC generator and areceiver gain control operating in the manner al-ready described for the Telekom DRC system, inthe situation shown in the block schematic dia-gram of Fig. 4. The audio samples are transmittedunmodified and the DRC generator supplementsthese with the DRC signal, comprising the gain

� Lin

� Lout

Maximumcompression

gain

–60

–50

–40

–30

–20

–10

0 dB rel.–60 –50 –40 –30 –20 –10

Input level (db)

Out

put l

evel

(dB

)

Compression ratio ro �

�Lin�Lout

� 1.3

0 dBrel is relative to the Permitted Maximum Signal level (ITU–R).

Figure 6Operational

characteristic ofTelekom system.

Page 9: Dynamic Range Control (DRC) and Music/Speech Control (MSC

64 EBU Technical Review Autumn 1994Hoeg et al.

values derived by the processor, which are suitablyencoded for transmission with the audio. A dia-gram of the DRC generator is shown in Fig. 7.

The DSP in the DRC generator is programmed toanalyse the undelayed audio samples and derive asuccession of peak–programme meter (PPM) levelindications, one every quarter of a second. Thevalue of gain (or attenuation) that would need to beapplied to the input signal to bring the peak levelin the time window provided by the delay to the de-sired level is derived from the gain law. This lawis based upon the simple characteristics shown inFigs. 3a to 3c, but is significantly more compli-cated. Although the gain law may be specified atonly a few discrete points, the DSP interpolateslinearly between them to calculate the desired out-put signal level for any input signal level, and indi-cates the required gain values to the receiver viathe DRC signal. The DSP aims to have the gain atthe required value by the time the programme peakin the window appears at the output of the delay;this will not always be achieved, however.

The maximum rate at which the gain can changeunobtrusively to that required is preset in the DSPprogram. A figure of just over 1 dB per second wasused initially in the BBC’s experimental imple-mentation. Listening tests have shown that the in-creases in gain intended to raise the level of quietsignals were more obtrusive than reductions ingain. In one item of choral music the gradual in-crease in gain applied to a quiet passage gave thedisturbing impression that the whole choir wascreeping forwards. It was necessary to reduce therate of gain increase to about 0.5 dB per second toovercome this effect. The maximum rate of de-

crease in gain remains at about 1 dB per second.Once calculated, the gain values are smoothed sothat they do not continuously follow the smallestvariations in signal level. The limit on the maxi-mum rate of reduction of gain is ignored if the out-put level would otherwise exceed a predeterminedmaximum level, to ensure that the listener is notsubjected to excessively high levels in the repro-duced programme. If the system is being used as aself–contained controller (e.g. for the pre–compression of programmes for VHF/FM broad-casting) this would prevent overloading or“clipping” later in the broadcasting chain, or over–deviation of the transmitter.

A limit is set to the maximum gain which can beapplied to the very quietest signals. There are tworeasons for this. The first is that quiet sources, suchas soft singing, can sound unnatural when repro-duced at an excessively high level. The second isthat background noise may be raised to the levelwhere it becomes obtrusive.

The algorithm implemented as described aboveworks well. However, some impairments werenoticed on certain critical types of material. Theseled to the addition of some secondary characteris-tics to the algorithm using the longer delay, to re-move the impairments.

One item of music which revealed unnaturalsounding impairments was a passage of piano mu-sic containing a gradual diminuendo. As the levelof the input signal fell, the gain applied was in-creased. In this case, a peculiar sustaining of thedecay of individual piano notes was noticed, dur-ing the diminuendo. Each decay took fractionallylonger to extinction than was expected, and after ashort period of listening the effect became verynoticeable. Similarly, when there were very smallcrescendi in an item of music in which the pro-gramme level was generally rising, the effect of thegradual reduction in gain which was occurringtended to obliterate them.

The secondary characteristics which were addedwere as follows. Whilst the gain is being increased,if two or more successive quarter–second blocksof samples show decreasing level, then the gain isnot increased during those blocks. While the gainis being decreased, if two successive quarter–second blocks show increasing level, then the gainis not reduced during those blocks (although prior-ity is given to ensuring that the output signal doesnot exceed the predetermined level at which down-stream equipment would introduce limiting).

The most obvious effects of the processing whichremain occur when a very loud passage follows a

Gain law

Processing

Secondarycharacteristics

Out

In

1 2 3 4 5 6 7

PPMvalues

Digital signalprocessor

Digital audio delay lineAudioinput

Audiooutput

Gain valuesNon–delayed

samples

Figure 7Block schematicdiagram of the BBCDRC signal generator.

Page 10: Dynamic Range Control (DRC) and Music/Speech Control (MSC

65EBU Technical Review Autumn 1994Hoeg et al.

very quiet one. Under these circumstances, thedynamic range controller has to change from max-imum gain to maximum attenuation during thelook–ahead time provided by the delay, to preventthe signal peaks exceeding the limiting level ofequipment later in the broadcasting chain. To dothis it has to override the limit on the maximum rateof gain change. Even when this happens, the ad-justment of the programme level is relatively un-obtrusive, and because the gain can increase atonly a low rate there are none of the repeatedchanges of gain at the programme peaks whichcharacterise fast–acting limiters or compressorsand give rise to some of the more objectionable im-pairments associated with this type of device.

4.4. The receiving side

According to the DAB Specification [3], a com-mon and simple receiver concept realises the nec-essary DRC functions. These functions do not de-pend upon the type of algorithm in the DRC signalgenerator used at the transmitting side. The sameoptions for listening to the uncompressed pro-gramme, the programme with the nominal degreeof compression and the programme with a greaterdegree of compression are always available. Asmentioned above, the more compression that is ap-plied to the programme, the greater the risk of im-pairing the sound quality. This should be borne inmind together with factors such as the reproduc-tion conditions when determining the degree ofcompression to be applied.

4.4.1. Regeneration of the DRC signal

The DRC data is de–multiplexed from the PADarea, and a linear interpolation regenerates a con-tinuous signal to control the gain of the audio chan-nel, in order to match the dynamic range of the re-ceived programme to the individual listener’sneeds. Using characteristics as described above,

the maximum gain change in the compressionmode will be about 15 dB. If the error protectionapplied to the PAD fails, specifically defined limit-ing characteristics for the rate of gain change willprotect the compressed audio signal against severeimpairments.

4.4.2. Controlling the dynamic rangeusing gain adjustments

There are at least two possible ways to realise thegain variation under the control of the DRC signal.

The conventional way is to use a simple multi-plication in the digital domain, or a voltage–controlled or digitally–contriolled amplifier (VCAor DCA) at the analogue end of the receiver to real-ise all necessary gain adjustments (DRC, manualand automatic volume control and other gain–related features such as the balancing of music andspeech loudness).

Another solution which could readily be imple-mented in Layer 2 integrated circuits [17] uses theexisting functions of the source decoder to modifythe scale factors in accordance with the DRC data.Fig. 8 shows in more detail those receiver func-tions needed for dynamic range control.

The listener may choose to use the DRC data un-modified, and accept the degree of compressionfor which the broadcaster’s DRC data generator isadjusted, or may modify the DRC data to increaseor decrease the degree of compression. If no com-pression is selected, the audio programme is repro-duced with the dynamic range as broadcast. If thelistener selects the nominal compression, the audioprogramme is reproduced with the degree ofdynamic compression set by the broadcaster. It islikely that the maximum compression ratio avail-able to the listener makes sense only for poorreproduction conditions (e.g. in a car) and for cer-tain types of programme, because some impair-

Source decoder(ISO Layer II)

DRC datainterpolation

DRC datamodify

PADDEMUX

M/S ratiostore

Gaincontrol

D/Aconverter

Audio inputwith DRC data

DRC decoder

Volumecontrol

Compression ratio /M/S ratio switch

Userinterface

Figure 8Dynamic range control

for DAB: Receivingblock diagram.

Page 11: Dynamic Range Control (DRC) and Music/Speech Control (MSC

66 EBU Technical Review Autumn 1994Hoeg et al.

ment of the sound quality is inevitable. A receiverwhich is required to compress the dynamic rangeneeds only to provide the gain signalled to it via theDRC data. Thus the same receiver is suitable foruse with both Telekom and BBC systems.

4.5. Benefits

Benefits of a DRC system with transmitted controlsignal are:

– because the control data generation algo-rithm(s) will not be specified in detail, there willalways be the opportunity for the broadcaster toimprove the performance of the DRC data gen-eration without the need to change the receivingequipment (“improvability”);

– the broadcaster is always able to check theresults expected at the receiving end by moni-toring at the studio end;

Herbert Jünger studied electrical engineering at the University of Rostock (Germany) and received aMasters degree (Diplom–Ingenieur) in 1974. He joined the electronics industry as a design engineer inmeasurement instruments.

Based on his interest in music and sound recording, he has been involved in many aspects of professionalaudio since 1985. His first products in this field were developed for the public broadcast organizationin the former East Germany.

In 1990 he founded Jünger Audio GmbH in Berlin as a privatecompany specializing in the development of high–performancedynamic range processors and related products. He has devel-oped new principles of inaudible dynamic range reduction for ap-plications with digital signal processors.

Wolfgang Hoeg studied electrical engineering and acoustics at the Technische Hochschule Dresden(Germany) and received a Masters degree (Diplom–Ingenieur) in 1959, followed by a postgraduate de-gree as a specialist engineer for automation of telecommunication technology in 1974. In 1959 he joinedthe Rundfunk– und Fernsehtechnisches Zentralamt (RFZ) in Berlin where he was in many research proj-ects in the fields of psychoacoustics, sound studio technology, sound reinforcement and sound transmis-sion systems, and especially with the development and introduction of stereophonic broadcasting andautomation of studio processes. Since 1991, he has been with the research and development center ofTelekom (FTZ Berlin), in the Research Group for New Sound Transmission Systems. Now he is dealingwith multi–channel sound systems and special aspects of DAB.

For many years, Mr. Hoeg was involved in the activities of international standardization organizations,including the former OIRT, the ITU Radiocommunication Bureau, and the Eureka 147 DAB project. Heis Chairman of EBU Specialist Group G1/LIST (Listening conditions).

Neil Gilchrist joined BBC Designs Department in 1965, after graduating from Manchester University(UK) with an Honours degree in Physics and Electronic Engineering. He moved to Research Departmentin 1967, working initially on various radio–frequency projects. Since 1976 he has been working mainlyin the field of digital audio, on analogue/digital conversion, NICAM, sampling–frequency synchroniza-tion, standardization activites and collaborative projects in RACE and Eureka. He is currently ProjectManager in Studio Section of Research and Development Department.

Since 1981, Mr. Gilchrist has represented the BBC on EBU Sub–group V3 (Sound). He is active in theBritish Section of the AES, was formerly Chairman of CCIR Interim Working Party 10/6 dealing withthe international exchange of sound programmes, and he represents the United Kingdom in WorkingParty 10C of the ITU Radiocommunication Bureau.

In 1989 he joined the Eureka 147 Digital Audio Broadcasting project, where he represents the BBC inthe group concerned with audio source coding.

Heinrich Twietmeyer studied physics at the universities of Hamburg and Münster (Germany). In 1975he joined the IRT where, since 1985, he has been the head of the section for Sound and Data Transmis-sion Systems/Channel Coding. His current research interests are concentrated on new data servicesfor digital broadcast systems such as DAB and on contribution and distribution technologies relatedto DAB.

Mr. Twietmeyer is involved in the Eureka 147 DAB project, especially in the field of data services.

Page 12: Dynamic Range Control (DRC) and Music/Speech Control (MSC

67EBU Technical Review Autumn 1994Hoeg et al.

– there is the potential to use the same algorithmor DRC signal to compress the programme sig-nals at the studio side for transmission via othersystems, for instance FM or AM broadcasting(“universality”).

– the complexity at the receiving side will be verylow, because no “intelligent” analysis is needed,and most of the processing needed is alreadyavailable in the receiver.

4.6. Music/Speech loudnesscontrol

In addition to the DRC system described, a Music/Speech loudness control (MSC) can balance therelative programme loudness to the listener’s indi-vidual taste. There is the option to signal four statesof the so–called M(usic)/S(peech) flag:

1 programme content = music

2 programme content = speech

3 programme content = not signalled

4 reserved for future use (e.g. for a programmecontaining music and speech with artisticallywell–balanced levels, for instance drama or amusical).

This M/S flag information can be generated at thestudio side automatically (e.g. derived from the cor-responding channel fader at the mixing console). Asan item of programme–related data, the informationis transported within the PAD, and may be used atthe receiver to control the output signal level (vol-ume) in a predetermined manner, probably in a levelrange of about 0 to –15 dB, either to enhance themusic, or the speech (or nothing).

5. DRC hardware

5.1. Control algorithm problems

The basic conflict in a dynamic range control sys-tem is to handle rapid changes of the input signalwhilst minimising side effects such as gain “pump-ing” and signal distortion, as explained earlier. Theadvantage of the systems described in this paper isthat the dynamic parameters are variable and areadapted to the incoming programme signal. Thecontrol algorithms implemented in the digital sig-nal processor are very complex, and they continu-ally optimise a multitude of different parameters.

Because the DRC data is transmitted discontinu-ously in time intervals of one DAB frame (24 ms),to avoid overshoots (i.e. brief periods in which thereproduced signal is excessively loud) in the output

from the receiver, it is necessary to have a look–ahead time of one complete DAB frame in which toanalyse the level of the signal. Self–contained dy-namic range controllers for use with VHF/FMtransmission do not have to work with discontinu-ous control signals. However, a delay is still neededin the signal path so that the incoming audio signalmay be examined prior to the application of anychanges in gain. This delay is 24 ms in the TelekomDRC process; the BBC process requires 30 ms, butmay be implemented with a longer delay.

5.2. The Telekom DRC signalgenerator and decoderhardware

The hardware realisation of the Telekom DRC sig-nal generator and decoder was first presented at the2nd Radio Montreux Symposium, 1994 [15]. Sim-ilar items of equipment are used for DAB and forFM broadcasting. They are based on the well–known professional digital dynamics processorhardware2. The equipment is designed to processdigital audio signals delivered in AES/EBU for-

2. Digital dynamics processor model d01 (Jünger AudioInc. Berlin, Germany) [18].

Table 2Typical technical

characteristics of theTelekom/Jünger DRC

coder and decodermodel for DAB

Page 13: Dynamic Range Control (DRC) and Music/Speech Control (MSC

68 EBU Technical Review Autumn 1994Hoeg et al.

mat. The input signal arriving via the XLR connec-tor and a balanced transformer is synchronized au-tomatically, and the sample rate is detected anddisplayed on a corresponding LED. A 32–bit float-ing–point DSP performs the processing. The DSPcarries out the function of DRC data generation,including the necessary delaying of the signal. Itmeasures the digital input and output signal levelsand displays the gain change. In the bypass mode,the digital signal is switched directly to the output,without any processing. As this function bypassesthe internal processor, no delay is introduced inthis mode of operation.

Additionally, an analogue output signal is avail-able, which allows audio monitoring of the dynam-ic range processing. It is generated by a digital–to–analogue converter with a resolution of 18 bits persample, and fed to balanced output drivers.

Table 2 shows some important technical character-istics of the existing device. Fig. 9 shows the frontpanel of the DRC signal generator.

5.3. The BBC DRC signalgenerator and decoderhardware

The BBC dynamic range controller, which hasbeen termed “DRACULA”3, has been implement-ed in two 32–bit processor systems. It is capable offunctioning as a DAB DRC signal generator or asa self–contained dynamic range controller. Thelaboratory implementation is in a general–purposeDSP32C–based floating–point system4, and this iscapable of running the algorithms with either along delay or a short delay. A DRACULA proces-sor based on a DSP 56000 series DSP and runningthe short–delay algorithm, is being manufacturedby a British company5.

In order to enable operators and experimenters toadjust the parameters of the laboratory version ofDRACULA with ease, controlling software hasbeen developed to run on a laptop PC (PersonalComputer). The PC is connected to the dynamicrange controller via an RS 232 serial interface.

The user is presented with a screen display on thePC comprising three windows, a graph, meter dis-play and a status window. These are shown in Fig.10. The meter display, at the bottom of the screen,gives input and output stereo PPM level readingsin the form of bar–graph indicators and also as nu-merical values. It also indicates the gain, expressedin dB, being applied to the audio signal. The statuswindow at the top right–hand side shows the look–ahead delay being used, and indicates which of thefacilities listed are being used. At the top left–handside, the graph window displays the compressionlaw which is in use. At the very top of the screenin Fig. 10 is a menu bar. The user has the option tochange parts of the display or some of the parame-ters of the controller, using the menu [14].

3. Dynamic Range Audio Controller with UnobtrusiveLevel Adjustment.

4. The laboratory hardware was made by the Fraun-hofer Institute for Integrated Circuits, Erlangen, Germany.

5. Pro–Bel Ltd. Reading, Berkshire, UK.

Figure 10User interface screendisplay of thelaboratory version ofDRACULA.

Figure 9DAB DRC coder(Telekom/JüngerAudio).

Page 14: Dynamic Range Control (DRC) and Music/Speech Control (MSC

69EBU Technical Review Autumn 1994Hoeg et al.

6. Other applications

The DRC system with transmitted control data hasbeen developed principally for use in a Eureka 147DAB system. As the additional data capacity re-quired for the DRC signal is very low, this systemcould also be used for other digital audio transmis-sion systems, such as:

– digital video broadcasting;

– digital medium–wave broadcasting;

– multichannel sound transmission systems (e.g.MPEG2) with or without accompanying pic-ture;

– other future digital transmission or recordingformats.

A self–contained controller for dynamic rangecontrol is needed for the existing VHF/FM broad-cast services, because of the smaller dynamicrange capability of the FM channel compared withthe source programme dynamic range. Either ofthe DRC systems described in this paper would besuitable for pre–compression of the audio pro-gramme. A compression ratio of 1.3 has beenfound suitable for the Telekom system, in this ap-plication [19], so that it gives about 15 dB reduc-tion in the dynamic range of the programme. Pre-liminary studies with the BBC’s DRACULAsystem have indicated that this, too, operates unob-trusively when effecting this degree of compres-sion.

7. Conclusions

For many years broadcasters have had to workwith a medium which does not have the capabilityto carry many types of source programme with theoriginal dynamic range. They have had to employskilled operators, and to use equipment which alltoo often impairs the audio quality in order to ef-fect the necessary reduction in dynamic range. Theintroduction of digital audio broadcasting (DAB),with the capability of conveying programmes tothe listener with a wider dynamic range, will notsolve the problem. Not all listeners will want, or bein a position to appreciate, the full dynamic rangeof a symphony orchestra or a choir, for example.There will still be the need for some form of dy-namic range control (DRC).

Using modern digital signal processing (DSP)techniques, and drawing on the experience ob-tained by broadcasters over many years, it is nowpossible to effect the control of programme dy-namic range virtually unobtrusively, and without

the need for a human operator. Two approaches todynamic range control (DRC) have been de-scribed, and both are suitable for use either in aself–contained controller (e.g. for conventionalVHF/FM broadcasting) or in a DRC signal genera-tor providing optional compression at the receiverin a DAB system, via the broadcast programmeassociated data (PAD) channel.

The satisfactory balancing of the loudness of dif-ferent types of programme, particularly music andspeech, depends principally upon the requirementsof the listener. In conventional broadcasting it hasnever been possible to satisfy every listener. How-ever, with DAB it will be possible to let individuallisteners determine the balance to be reproduced intheir own environments, using music/ speech con-trol (MSC) information, signalled in the PAD.

Acknowledgements

The authors are indebted to Telekom FTZ , the IRTand the BBC for permission to publish this paper.They also wish to thank their colleagues in the Eu-reka 147 DAB project for their support and encour-agement.

Bibliography

[1] Riley, J.L.: The DAB multiplex and systemsupport featuresEBU Technical Review No. 259 (Spring1994).

[2] Draft Revision of Recommendation ITU–RBO.789, Document 10/74E, Geneva (Decem-ber 1993): Digital sound broadcasting tovehicular, portable and fixed receivers forBSS (sound)

[3] Draft European Telecommunication StandardprETS 300 401 (1994): Radio BroadcastingSystem; Digital Audio Broadcasting (DAB)to mobile, portable and fixed receivers

[4] Müller, K.: What dynamic range does thelistener wish to have?EBU Review 124–A (Technical), December,1970.

[5] Ilmonen, K.: Listener preferences for loud-ness balance of broadcasts, with specialconsideration to listening in noise.Research Report from Finnish BroadcastingCompany No. 12/1971, Helsinki (1971).

[6] Wagner, K.: Zu den Unterschieden derSchallwiedergabe im Auto und im Wohn-bereich (On the difference of sound repro-duction conditions in a car or in the home envi-ronment)Report 17. Tonmeistertagung, Karlsruhe(1992).

Page 15: Dynamic Range Control (DRC) and Music/Speech Control (MSC

70 EBU Technical Review Autumn 1994Hoeg et al.

[7] Steffen, E.: Untersuchungen zurlautstärkegerechten Aussteuerung vonRundfunk–Tonsignalen (Investigations onloudness–oriented level control for audiobroadcast programme signals)Tech. Mitt. des RFZ, 22 (1978) 3/4.

[8] ARD – Richtlinien für die Aussteuerung imTonrundfunk (Guidelines of the ARD for levelcontrol in audio broadcasting) Arbeitskommi-sion der Hörfunkbetriebsleiter der ARD(1972).

[9] Steinke, G.: Das Pegelprofil in der Tonstu-dio– und Rundfunkübertragungstechnik(Level profiles in sound studio technique andsound broadcast transmission chains)dB–Magazin für Studiotechnik, Cologne(1990)1/2.

[10] OIRT Report Nr. 28/1–II. Prague (April, 1988):Über das Verhältnis der Lautstärke vonMusik und Sprache bei Rundfunk– undFernsehsendungen (On the relation of theloudness levels between music and speech inradio and TV programmes)

[11] Hoeg, W.: Technologische Probleme beider Automatisierung der Programmab-wicklung im Hörrundfunk (Technologicalproblems concerning an automated pro-gramme continuity process in sound broad-casting)Submitted for a postgraduate degree; Hoch-schule f. Verkehrswesen, Dresden (1974).

[12] Plenge, G., Spikofski, G. and Theile, G.: Vari-able dynamic range: A method for improv-ing the sound broadcasting and televisionserviceEBU Review – Technical No. 218, August1986.

[13] Million, A., Krause, M.: Intelligenter automa-tischer Kompander mit großen Zeitkons-tanten (An intelligent automatic companderwith large look–ahead time constants)NTG–Fachberichte 91 (1986).

[14] Gilchrist, N.H.C.: Dynamic Range Controlfor DABProceedings, 2nd International DAB Sympo-sium, Toronto (March, 1994).

[15] Hoeg, W., Twietmeyer, H. and Jünger, H.:Additional data services for DAB: Dynam-ic Range Control (DRC)Proceedings, 2nd International Radio Sympo-sium Montreux (June 1994).

[16] Azizi, S.A.: Digital Dynamic Range Proces-sor for car audio applications94th AES Convention, Berlin (March, 1993),Preprint 3463.

[17] Theile, G. and Link, M.: Low–complexity Dy-namic Range Control system based onscalefactor weighting94th AES Convention, Berlin (March, 1993),Preprint 3563.

[18] Fa. JÜNGER AUDIO: Digital Dynamics Pro-cessor – Model d01: Operation ManualBerlin (November 1992)

[19] Wagner, K.: Dynamik von Musik bei Pro-duktion und Sendung (Dynamic of musicprogrammes in production and broadcast pro-cesses)J. radio fernsehen elektronik, Berlin40(1991)6.

IBC Award

Continuing a long–standing tradition of the International Broadcasting Convention, the 1994 IBC JohnTucker Award was presented to Dr. Mario Cominetti of RAI–Radiotelevisione Italiana. The Award wasmade in recognition of Dr. Cominetti’s outstanding contribution to the establishment of standards fordigital transmission systems for radio and television on terrestrial and satellite channels. His major

achievements in recent years have been concerned with digital HDTV trans-mission by satellite and his involvement in the European Digital Video Broad-casting (DVB) project, although his keen interest in the development of digitaltechnologies in broadcasting extends back many years, notably through hiswork on teletext and other forms of data broadcasting.

Dr. Cominetti graduated in physics from Turin University and followed thiswith a period at the Politecnico di Torino where he completed a course in elec-tronic engineering. After spending some time as a research engineer at theInstituto Nazionale Galileo Ferraris, and later with IBM, he joined the RAI Re-search Centre in Turin where he is now Head of RF Technologies Division.Dr. Cominetti is Chairman of EBU Sub–group V2 (Data broadcasting).

Dr. Mario Cominetti (left) receiving the IBC Award from Mr. Stanley Baron.