EAR 1000 / ADRA 1000 / 2000
Installation and Programming Manual
EAR 1000/2000 and ADRA 1000/2000 Installation and Programming Manual

Version 3

Release 3

February 2004

NOTICE
This publication refers to the EAR 1000/2000 and the ADRA 1000/2000, Release 3. Additional copies of this manual may be obtained from ITS Telecom Reproduction of this manual or parts thereof without written permission from ITS is strictly prohibited. ITS Telecom reserves the right to modify the hardware and software described in the manual without prior notice. However, changes made to the hardware or software described does not necessarily render this publication invalid.

WARRANTY

In the event that the product proves to be defective in workmanship or materials within a period of one year from date of shipment, ITS Telecom shall repair or replace the product at its discretion. Transportation will be the responsibility of the dealer/distributor. Under no circumstances shall ITS Telecom be liable for consequential or special damages, loss of revenue or user/dealer expenses arising out of or in connection with the use or performance of the product, whether based on contract, tort, or any other legal agreement. The following shall void the above warranty: malfunctions resulting from fire, accident, neglect, abuse, or acts of God; use of improper electrical power; or repair of, tampering with or alteration of the product by anyone other than ITS authorized personnel.

TABLE OF CONTENTS

1. INTRODUCTION.... 1-1
SECTION I EAR 1000/2000 INSTALLATION AND PROGRAMMING 2. 2.1 3. 3.1 3.1.1 3.1.2 3.2 3.2.1 4. 4.1 4.2 4.2.1 4.2.2 OVERVIEW OF EAR 1000/2000... 2-1 Features and Services.... 2-1 DESCRIPTION AND INSTALLATION.. 3-1 Physical Description.... 3-1 Bottom Panel.... 3-1 Front Panel... 3-2 Installation.... 3-3 Installing the EAR 1000/2000... 3-3 DTMF PROGRAMMING... 4-1 Entering and Exiting the Programming Mode.. 4-1 DTMF Programming Commands... 4-2 Script Messages... 4-2 PBX Parameters.... 4-4
EAR/ADRA Installation and Programming Manual

Table of Contents

4.2.3 4.2.4 4.2.5 5. 6.
Menus Handling.... 4-6 Busy Menu Handling... 4-8 Advanced Features.... 4-10 CHANGING THE OPENING GREETING... 5-1 PROGRAMMING EXAMPLE.. 6-1
SECTION II ADRA 1000/2000 INSTALLATION AND PROGRAMMING 7. 7.1 8. 8.1 8.1.1 8.1.2 8.2 8.2.1 9. 9.1 9.2
OVERVIEW OF ADRA 1000/2000... 7-1 Features and Services.... 7-1 DESCRIPTION AND INSTALLATION.. 8-1 Physical Description.... 8-1 Bottom Panel.... 8-1 Front Panel... 8-2 Installation.... 8-3 Installing the ADRA 1000/2000... 8-3 DTMF PROGRAMMING... 9-1 Entering and Exiting the Programming Mode.. 9-1 DTMF Programming Commands... 9-2
9.2.1 9.2.2 9.2.3 9.2.4 9.2.5 10.
Script Messages... 9-2 PBX Parameters.... 9-4 Greeting Handling.... 9-6 Busy Menu Handling.... 9-7 Advanced Features... 9-8 CHANGING THE OPENING GREETING..10-1
APPENDIX A EAR/ADRA SPECIFICATIONS...A-1

INTRODUCTION

The EAR 1000/2000 is a small standalone Automated Attendant system. The ADRA 1000/2000 is a small, stand alone Voice Announcer System. Both systems are developed by ITS, a leader in the field of voice processing systems and PBX peripheral products. The EAR 1000/2000 and ADRA 1000/2000 incorporate state of the art technology, including DSP, flash memory and SMT production. This guide provides installation and programming instructions for both the EAR 1000/2000 and ADRA 1000/2000. See Section I for instructions about EAR 1000/2000; see Section II for instructions about ADRA 1000/2000.
SECTION I: EAR 1000/2000 Installation and Programming

OVERVIEW OF EAR 1000/2000
The EAR 1000 is a 1-port auto attendant system. The EAR 2000 is a 2-port auto attendant system. Both systems have 9 minutes of recording time. The EAR 1000/2000 can be integrated with most types of PBX through the analog ports and programmed by a touch-tone telephone.
Figure 2-1. EAR 2000 General View

Features and Services

The EAR 1000/2000s features includes the following features: Opening Greetings: Day, Night, Holiday for each line. Call Transfer The system administrator can program the EAR 1000/2000 to transfer the calls to extensions in one of the following modes:
EAR 1000/2000 Installation and Programming Manual
Overview of EAR 1000/2000

Non-Supervised. The EAR 1000/2000 transfers the call immediately without verifying the status of the extension. Semi-Supervised. The EAR 1000/2000 checks for a busy signal before transferring the call to the extension. Supervised. The EAR 1000/2000 checks for a busy or answer signal before transferring the call to the extension.
Up to 9 minutes of recording time. High quality recording. Non-volatile memory (Flash memory). Adjustable flash time. Busy detect using call progress tone or DTMF codes. Busy menu play back. No answer menu play back. Remote programming and recording. Simple operation and maintenance.
DESCRIPTION AND INSTALLATION

Physical Description

The functional components of the EAR 1000/2000 are located in its bottom side panel. The LEDs are located on the front panel. The back panel has two indented holes for wall mounting.

3.1.1 Bottom Panel

Line 1 EAR 2000 Only Line 2
Figure 3-1. EAR 1000/2000 Bottom Panel
Description and installation
The following description corresponds to the labels in Figure 3-1. 1. Power Supply Connector Connects the EAR 1000/2000 to the external power supply 2. 1/2 RJ-11 Sockets Connects the EAR 1000/2000 to PBX extensions

3.1.2 Front Panel

The following figure and table describe the function of the three LEDS on the front panel.

STATUS

HOLIDAY

Day Mode

- Night

On Off Off Off Flashing

Off On Off Flashing Flashing

Off Off On Off Flashing

Night Mode Holiday Mode

- Holiday

System Error1 System Error
Please contact your local dealer.
Figure 3-2. LEDs on the Front Panel 3-2 EAR 1000/2000 Installation and Programming Manual

Installation

The EAR 1000/2000 is delivered completely assembled. It is designed for mounting on a wall, close to the PBX.
3.2.1 Installing the EAR 1000/2000

To install the EAR 1000/2000:
1. Mount the unit on a wall close to the PBX cabinet. Use the drill template to place the two screws. 2. Connect the RJ-11 connector on one end of the cables to the RJ-11 sockets on the bottom panel of the EAR 1000/2000. Connect the other end of the cables to one analog telephone line on the Main Distribution Frame (MDF) of the PBX (see Figure 3-3).
Main Distribution Frame of the BPX
EAR 1000/2000 Bottom Panel
4-pin RJ-11 socket 4-pin RJ-11 socket

Line 1

Line 2 (EAR 2000 Only)
Figure 3-3. Analog Line Connections
EAR 1000/2000 Installation and Programming Manual 3-3
3. On the bottom panel of the EAR 1000/2000, plug the 9VDC adapter jack into the power supply connector. 4. Plug the 9VDC adapter into the main power supply outlet to turn the EAR 1000/2000 on. The LEDs on the front panel blink one after another and then the LED indicating the status of the EAR 1000/2000 turns on. 5. Call each EAR 1000/2000 line from any extension and verify the replay. 6. Program the EAR 1000/2000 according to your PBX type and required applications (see Chapter 4).
Single Line Digital Telephones Telephones 1 or 2 voice channels Central Office EAR 1000/2000

Operator Console

Figure 3-4. System Installation 3-4 EAR 1000/2000 Installation and Programming Manual

DTMF PROGRAMMING

The EAR 1000/2000 is programmed by telephone using DTMF tones. For the EAR 2000, both lines can be programmed and recorded using DTMF tones from one of the lines. Note: A confirmation tone is heard every time a programming command is entered.
Entering and Exiting the Programming Mode
The EAR 1000/2000 does not handle calls when in the programming mode. To enter the programming mode: 1. Call the EAR 1000/2000 Ext. from any touch-tone telephone. 2. Wait until the EAR 1000/2000 answers and plays the opening greeting or plays a tone if no menu is recorded. Next, dial *900. 3. Dial the System Administrators password (the default password is 1234) to enter the programming mode. To exit the programming mode: Dial *900. When the programming mode is exited by dialing *900, the EAR 1000/2000 plays the opening greeting. You can then test the changes made to the system.

DTMF Programming

DTMF Programming Commands
The following tables include the DTMF commands available for the EAR 1000/2000 system.

4.2.1 Script Messages

OPERATION Record a script message. In the Ear 2000 it is possible to define whether to use different greetings for lines 1 and 2 or to use the same greetings for these lines. Refer to command *112 below. COMMAND *100 + XX + Beep + Record + # where XX = 00; Day greeting XX = 10; Night greeting XX = 20; Holiday Mode greeting XX = 21; No answer greeting XX = 22; Busy greeting XX = 01; Transfer greeting The following are only available for the second line of EAR 2000 XX = 50; Day greeting XX = 60; Night greeting XX = 70; Holiday greeting *101 + XX where XX = Script message number (Same as *100) DEFAULT

Play a script message

OPERATION Delete a script message Deletes the message and returns its parameters to default values. Define the set of scripts to be used when the second line is called (see also *100). Note: For EAR 2000 only
COMMAND *102 + XX where XX = Script message number (Same as *100) *112 + X where X = 0; Uses scripts 00, 10, 20 for both lines X = 1; Uses scripts 00, 10, 20 for line 1 and 50, 60, 70 for line 2 0

DEFAULT

4.2.2 PBX Parameters
OPERATION Number of digits in extension. COMMAND *300 + X where X = Number of digits in extension (1-6) *310 + X + Y where X = Line number (1 or 2) Y = Number of rings (1-9) *330 + X where X = Digit to be dialed by caller (0-9) 3 DEFAULT
Number of rings before line is answered.
Line 1 Y = 1 Line 2 Y = 2
Operator ID digit. When this digit is dialed during any script, the caller is transferred to the operator. (See also *360). Non/Semi /Supervised transfer
*350 + X + Y where X = 1; All extensions except the operator X = 2; Operator extension only Y = 0; Non supervised Y = 1; Semi supervised Y = 2; Supervised

All Ext. Non Supervised

OPERATION Operator extension number
COMMAND *360 + X + Ext. + # where X = 1; Day operator X = 2; Night + Holiday operators Ext. = Operator Ext. number *370 + XXX XXX A 3 digit number (000-980) in steps of 20 ms. *371 + X + YYYY where X = 1; Busy off time X = 2; Busy on time X = 3; Disconnect off time X = 4; Disconnect on time YYYY = Cadence in ms (01003000) in steps of 20ms

DEFAULT Ext. = 0

Flash Time

600 ms

Busy, disconnect and DTMF on/off time.
0500 ms 0500 ms 0240 ms 0240 ms
OPERATION PBX Transfer code and Recall from Busy / No answer code.
COMMAND *380 + X + Code + # where X = 1; Transfer code. X = 2; Recall from busy code. X = 3; Recall from no answer code Code respective DTMF code that contains up to four digits including 0-9, #, * and A-D. Special digits *0 for Ext. *4 for # *7 for C *1 for Pause *5 for A *8 for D *2 for Flash *6 for B ** for *

Flash + Ext. Flash Flash

4.2.3 Menus Handling
OPERATION Transfer call to an extension at the end of the script message or when the caller dials a digit during a script playback. COMMAND *120 + XX + B + Ext. + # where XX = Script message number (Same as *100) B = Digit to be dialed by caller or * at the end of message playback. Ext = Extension (contains up to six
OPERATION digits) Transfer the call to an operator at the end of the script message. Operator Ext. is defined by command *360 Disconnect the call at the end of the script message.

COMMAND

*125 + XX + # where XX = Script message number (Same as *100)
*140 + XX + * + # where XX = Script message number (Same as *100)
Direct transfer to an extension. Enables the user to directly dial an extension during playback of a script message
*170 + XX + first digit(s) + # where XX = Script message number (Same as *100) first digit(s) = First digit(s) of extension. Up to four different digits can be entered. *190 + XX + # where XX = Script message number (Same
Reset Scripts to default values. This command does not delete the recorded
OPERATION script messages. Reset value in Script to default values. This command does not delete the recorded script messages. as *100)
*190 + XX + Y + # where XX = Script message number (Same as *100) Y = Specific digit value to reset in the script.
4.2.4 Busy / No answer Menus Handling

OPERATION Return to the opening greeting when caller dials return digit during playback of Busy/No answer script. COMMAND *115 + XX + B where XX script number B = Digit to be dialed by caller or * at the end of greeting playback Note: This option is valid only for the No answer menu (script 21) Busy menu (script 22). DEFAULT
OPERATION Place a call on hold. Enables the call to remain on hold. The caller is put on hold for ten seconds before trying to transfer the call again.
COMMAND *180 + B where B = Digit to be dialed by caller to put the call on hold or * to put the call on hold at the end of playback Note: This option is valid only for the Busy menu (script 22). *185 + B where B = Blind transfer digit to be dialed by the caller or * at the end of greeting playback Note: This option is valid only for the Busy menu (script 22).
Blind Transfer to a busy extension. An immediate attempt is made to transfer the call to the extension when the user dials the transfer digit.

4.2.5 Advanced Features

OPERATION Time for End Of Message action. COMMAND *105 + XX + Y + # where XX = Script message number (Same as *100) Y = 0-9 (sec) *220 + X where X = 0; Busy/Answer detect using Call Progress tones. X = 1; Busy/Answer detect using DTMF codes. Note: Option X=1 is valid only if the ports assigned to the Ear 1000/2000 in the PBX are defined as VM ports. 5 DEFAULT
Selecting DTMF code or Call Progress tone detection. This command is valid in Supervised transfer mode.

Call Progress

OPERATION Defining the DTMF code for the busy/no answer condition.
COMMAND *221 + X + Code + # where X = 1 Answer X = 2 - Busy Code Each DTMF code can contain up to four digits including 0-9. To enter #, *, A-D use the following DTMF combinations: *4 for # *7 for C *5 for A *8 for D *6 for B ** for * Note: This option is valid only if the ports assigned to the Ear 1000/2000 in the PBX are defined as VM ports. *333 + Code + # where Code = Disconnection code. Same as Code table in *221 *369 + X where X = Volume level (0-9) 9 is the highest volume level

Disconnection code

Greeting volume level
OPERATION Busy signal cadence setup. Setup of busy on/off-time cadence.
COMMAND *375 + XXXX + # where XXXX Busy extension number. Setting-up the Busy on/off-time cadence by dialing the busy extension number. *600 + * + Old password + New password + # Password must include 4 digits. Do not use the digits * and #. This password cannot be disabled. *601 + * + Old password + New password + # Password must include 4 digits. Do not use the digits * and #.

Change system administrator password.
Change operators password. Used by the operator to change the opening greeting (Day, Night, or Holiday). Resetting the EAR 1000/2000. Returns to default factory settings and deletes all script messages.
*654 + * + XXXX + # XXXX = System administrator password.
CHANGING THE OPENING GREETING
Changing the opening greeting between Day, Night and Holiday is performed in one of the following methods: 1. Press the Set button on the EAR 1000/2000 front panel. 2. Call one of the EAR 1000/2000 extensions and during the opening greeting dial: *8 + XXXX + Y where XXXX = Operator password (default=1234) Y = 0 ; Day Mode (default) Y = 1 ; Night Mode Y = 2 ; Holiday Mode Note: Operator password can be skipped if canceled using command *601.

PROGRAMMING EXAMPLE

The following paragraphs are examples for EAR 1000/2000 configuration. These examples consider a PBX with a 3 digit extension size, an extension from 400 to 599, and Hook Flash value of 600ms. Day Opening Greeting (Script 00) Welcome to ITS. If you know the extension number you require, please dial it now. For Help Desk dial 1. (Ext.440) For Marketing dial 2. (Ext. 450) For Management dial 3. (Ext. 460) For the operator dial 0 or wait on the line. Night Opening Greeting (Script 10) Welcome to ITS. If you know the extension number you need, please dial it now. Our offices are currently close, our working hours are: Monday to Friday 8am5pm. Thank you and goodbye. Extension No Answer Greeting (Script 21) The required extension does not answer. For another extension please dial 1. For the operator dial 0 or wait on the line.

Programming Example

Extension Busy Greeting (Script 22) The extension required is busy. To stay on hold, dial 1. For the operator dial 0 or wait.
EAR 1000/2000 DTMF Commands Command * * * * * 400 # * # * # * # * # * # * # Meaning Enter programming mode The Ext. numbers are 3 digits long The digit for operator ID is 0 Hook Flash value is set to 600ms In Day mode operator Ext. is 400 Enable direct call when either the digit 4 or 5 is dialed during script message 00 (Day) Transfer the call to Ext. 440 when the digit 1 is dialed during script message 00 (Day) Transfer the call to Ext. 450 when the digit 2 is dialed during script message 00 (Day) Transfer the call to Ext. 460 when the digit 3 is dialed during script message 00 (Day) At the end of script message 00 transfer call to operator Ext. Enable direct call when either the digit 4 or 5 is dialed during script message 10 (Night)
Command * * # * 1 * 0 * # * #

Meaning At the end of script message 10 (Night) disconnect the call Set all Ext. except operator Ext. to be semi-supervised Set operator Ext. to be non-supervised When an Ext. is busy, place the call on hold, if the digit 1 is dialed At the end of script message 22 transfer the call to the operator Ext.
SECTION II: ADRA 1000/2000 Installation and Programming
OVERVIEW OF ADRA 1000/2000
The ADRA 1000 is a 1-port AutoAnnouncer System. The ADRA 2000 is a 2-port AutoAnnouncer System. Both systems have up to 9 minutes of recording time. The ADRA 1000/2000 can be integrated with most types of PBX through their analog ports, and can be programmed by a touch-tone telephone.
Figure 7-1. ADRA 2000 General View
The ADRA 1000/2000 has the following features: Opening Greetings: Day, Night, Holiday for each line. Call Transfer The system administrator can program the ADRA 1000/2000 to transfer calls to extensions in one of the following modes:
ADRA 1000/2000 Installation and Programming Manual
Overview of ADRA 1000/2000
Non-supervised. The ADRA 1000/2000 transfers the call immediately, without verifying the status of the extension. Semi-supervised. The ADRA 1000/2000 checks for a busy signal before transferring a call to an extension. Supervised. The ADRA 1000/2000 checks for a busy or answer signal before transferring the call to the extension.
Up to 9 minutes of recording time. High quality recording. Non-volatile memory (Flash memory). Adjustable flash time. Busy status detection using call progress tone or DTMF codes. Busy Extension playback message. Remote programming and recording. Simple operation and maintenance.
The functional components of the ADRA1000/2000 are located on its bottom panel. The LEDs are on the front panel. The back panel has two indented holes for wall mounting.

8.1.1 Bottom Panel

Line 1 Line 2 ADRA 2000 Only
Figure 8-1. ADRA 1000/2000 Bottom Panel
The following description corresponds to the labels in Figure 8-1. 1. Power Supply Connector 2. 1/2 RJ-11 Sockets Connects the ADRA1000/2000 to the external powersupply. Connects the ADRA1000/2000 to PBX extensions.

8.1.2 Front Panel

STATUS Day Mode
DAY On Off Off Off Flashing
NIGHT Off On Off Flashing Flashing
HOLIDAY Off Off On Off Flashing
Night Mode Holiday Mode System Error1

System Error1

Figure 8-2. LEDs on the Front Panel 8-2
The ADRA 1000/2000 is delivered completely assembled. It is designed for mounting on a wall, close to the PBX.
8.2.1 Installing the ADRA 1000/2000
Figure 7-3 describes the system installation.
Single Line Digital Telephones Telephones

Telephone

1 or 2 voice channels Central Office EAR ADRA1000/2000
Figure 8-3. System Installation
To install the ADRA 1000/2000: 1. Mount the unit on a wall close to the PBX cabinet. Use the drill template to place the two screws. 2. Connect the RJ-11 connector to the RJ-11 socket on the bottom panel of the ADRA 1000/2000. Next, connect the other end of the cable to an analog telephone line on the Main Distribution Frame (MDF) of the PBX (see Figure 8-4).
ADRA 1000/2000 Bottom Panel

4-pin RJ-11 socket

Line 2 (ADRA 2000 Only)
Figure 8-4. Analog Line Connections
3. Plug the 9VDC adapter jack into the power supply connector on the bottom panel of the ADRA 1000/2000.
3. Plug the 9VDC adapter into the main power supply outlet. The LEDs on the front panel blink one after the other, then the LED indicating the status of the ADRA 1000/2000 turns on. 4. Call each ADRA 1000/2000 line from any extension, and verify a confirmation tone (short beep). 5. Program the ADRA 1000/2000 according to your PBX type and the required applications (see Chapter 9).
The ADRA 1000/2000 is programmed by telephone using DTMF tones. For the ADRA 2000, both lines can be programmed and recorded using DTMF tones from one of the lines. Note: A confirmation tone is heard every time a programming command is entered.
The ADRA 1000/2000 does not handle calls when in programming mode. To enter the programming mode: 1. Call the ADRA 1000/2000 Ext. from any touch-tone telephone. 2. Wait until the ADRA 1000/2000 answers and plays the opening greeting, or a clear tone is heard (if no greeting is recorded). Next, dial *900. 3. Dial the System Administrators password (the default password is 1234) to enter the programming mode. To exit the programming mode: Dial *900.

The following tables show the DTMF commands available for the ADRA 1000/2000 system.

9.2.1 Script Messages

OPERATION To Record an Opening Greeting. In the ADRA 2000 it is possible to define whether to use different greetings for lines 1 and 2 or to use the same greetings for both lines. Refer to command *112 below. Note: In case of rerecording, its recommended to first delete the existing message. COMMAND *100 + XX + Beep + Record + # where: XX = 00; Day greeting XX = 10; Night greeting XX = 20; Holiday greeting XX = 22; Busy greeting XX = 21; No answer greeting XX = 01; Transfer greeting The following are only available for the second line of ADRA 2000 XX = 50; Day greeting XX = 60; Night greeting XX = 70; Holiday greeting DEFAULT No message is recorded.
OPERATION Play an Opening Greeting.
COMMAND *101 + XX where: XX = Opening Greeting (same as *100) *102 + XX where: XX = Opening Greeting (same as *100) *112 + X where: X = 0; Uses greeting No. 00, 10, 20 for both lines. X = 1; Uses greeting No. 00, 10, 20 for line 1 and 50, 60, 70 for line 2
Delete an Opening Greeting.
Define the set of greetings to be used when the second line is called. Note: For ADRA 2000 only.

First Set

9.2.2 PBX Parameters
OPERATION Number of rings before line is answered. COMMAND *310 + X + Y where: X = Line number (1 or 2) Y = Number of rings (1-9) *350 + X + Y where: X = 1; For all extensions except the operators extension. X = 2; Operator extension only Y = 0; Non-supervised. Y = 1; Semi-supervised. Y = 2; Supervised *360 + X + Ext. + # where: X = 1; Day operator X = 2; Night + Holiday operators Ext. = Operator Ext. number DEFAULT Line 1 Y = 1 Line 2 Y = 2
Non/Semi/Supervised transfer. Note: When the required extension is the Operator or another extension.

All Ext. Nonsupervised.

Operator extension number.

Ext. = 0

OPERATION Flash Time.
COMMAND *370 + XXX XXX A 3 digit number (000-980) in steps of 20 ms. *371 + X + YYYY, where: X = 1; Busy off time X = 2; Busy on time X = 3; Disconnect off time X = 4; Disconnect on time YYYY = Cadence in ms (01003000) in steps of 20ms *380 + X + Code + # where: X = 1; Transfer code. X = 2; Recall from busy code. X = 3; Recall from no answer code Code respective DTMF code that contains up to four digits including 0-9, #, * and A-D. Special digits ** for * *4 for # *1 for Pause *5 for A *6 for B *2 for Flash *7 for C *8 for D *0 for Ext.

DEFAULT 600 ms

Busy and disconnect DTMF on/off time.
PBX Transfer code and Recall from Busy / No answer code.

9.2.3 Greeting Handling

OPERATION Transfer a call to an extension at the end of the Opening Greeting. COMMAND *120 + XX + * + Ext. + # where: XX = Opening Greeting number (same as *100) Ext = Extension (contains up to six digits). *125 + XX + # where: XX = Opening Greeting number (same as *100) DEFAULT

Transfer the call to the operators ext. at the end of the Opening Greeting. Operator Ext. Is defined by command *360 Disconnect the call at the end of the Opening Greeting.
*140 + XX + * + # where: XX = Opening Greeting number (same as *100) *190 + XX + # where XX = Opening Greeting number (same as *100)
Reset Opening Greeting to default values. This command does not delete the recorded script messages.

9.2.4 Busy Menu Handling

OPERATION Place a call on hold at the end of a BUSY greeting message (script 22). Enables the call to remain on hold. The caller is put on hold for 10 seconds before trying to transfer the call again. Blind Transfer to a busy extension. An immediate attempt is made to transfer the call to the extension. COMMAND *180 + * where: * = Puts the call on hold at the end of a Busy greeting message DEFAULT
*185 + * Blind transfer at the end of a Busy greeting message (script 22).

9.2.5 Advanced Features

OPERATION Selecting DTMF code or Call Progress tone detection. COMMAND *220 + X where: X = 0; Busy/Answer detect using Call Progress tones. X = 1; Busy/Answer detect using DTMF codes. Note: This option is valid only if the ports assigned to the ADRA 1000/2000 in the PBX are defined as VM ports. DEFAULT Call Progress
COMMAND *221 + X + Code + # where: X = 1 Answer X = 2 - Busy Code Each DTMF code can contain up to four digits including 0-9, #, * and A-D. Special digits ** for * *4 for # *5 for A *6 for B *7 for C *8 for D Note: This option is valid only if the ports assigned to the ADRA1000/2000 in the PBX are defined as VM ports. *333 + Code + # where: Code = Disconnection code. Same as Code table in *221 *369 + X where: X = Volume level (0-9) 9 is the highest volume level

Disconnection code.

Greeting volume level.
OPERATION Busy signal cadence learn. Checks the busy signal by dialing the busy extension number and implements the recorded values. Change system administrator password.
COMMAND *375 + XXXX + # where: XXXX Busy extension number
*600 + * + Old password + New password + # Password must include 4 digits. This password cannot be disabled. Warning: Do not use the * or # keys.
Change operators password. Used by the operator to change the Day, Night, and Holiday greetings.
*601 + * + Old password + New password + # Password must include 4 digits. Warning: Do not use the * or # keys.
OPERATION Resetting the ADRA 1000/2000. Returns to default factory settings and deletes all opening greetings.
COMMAND *654 + * + XXXX + # XXXX = System administrator password.

10. CHANGING THE OPENING GREETING
Changing the opening greeting between Day, Night, and Holiday, is possible in one of the following methods: 1. Using the set button on the units front panel. 2. Dialing to one of the units extensions and when the current greeting is played using the following DTMF code: *8 + XXXX + Y where: XXXX = Operator password (default=1234). Y = 0; Day greeting (default). Y = 1; Night greeting. Y = 2; Holiday greeting.
APPENDIX A EAR/ADRA SPECIFICATIONS
The following are the EAR 1000/2000 and ADRA 1000/2000 specifications: DC Power Supply (Input) 220VAC, 50 Hz (Input) 110VAC, 60 Hz (Output) 9 VDC/800 mA 24 to 72 VDC < 10 A 0 to 100 VDC > 5 M 100 to 200 VDC > 30 K 500 VAC/50 Hz > 20 K 100 VAC/25 Hz > 100 K 0.47 F 10% @ 50 VDC, 40 Vac/25 Hz > to 100 VAC/16 to 60 Hz 24 to 66 VDC @ 20 to 100 mA 300 to 3400 Hz 500 to to 3400 Hz > 46 db 300 to 3400 Hz > 18 db < 700 A +1.5% -6 to -8 dBm -8 to -10 dBm 70 to 80 ms 9 Minutes
Line Voltage DC Leakage Current On-hook Insulation Resistance between Line Terminal and the Ground Ring Capacitor On-hook Impedance Ring Detect DC Resistance (off-hook) Impedance (off-hook) Imbalance Ratio Return Loss Current during Break DTMF Transmission: Frequency Tolerance Frequency Level (High Group) Frequency Level (Low Group) Inter-digit Pause in Tone Dialing Recording Time

www.its-tel.com

I82000019

2002 ,rerepmelK luaP c

.syasse eseht tuoba snoitseggus lufpleh sih rof dna ,snoitcua G3 eht fo yduts eht ni noitaroballoc ruo rof izzongaP ocraM ot lufetarg yllaicepse ma I dna ,tfard reilrae na dewohs I mohw ot noitcua namreG eht ni devlovni smr eht fo emos fo sevitatneserper eht dna rettetsfloW ramlE ,sregrB namliT morf stnemmoc lufpleh rof lufetarg ma I.swal ro selur elbacilppa yna setaloiv woleb dessucsid ruoivaheb eht fo yna taht tseggus ot dnetni ton od I.enola enim era repap siht ni desserpxe sweiv eht tub ,noitcua esnecil enohp-elibom KU eht nar dna dengised hcihw ,ycnegA snoitacinummocoidaR stnemnrevog KU eht gnisivda tsiroeht noitcua lapicnirp eht saw I :stnemegdelwonkcA.gniddiB ,STMU ,G3 ,snoitcuA murtcepS ,snoitacinummoceleT ,snoitcuA :sdrowyeK.)snoitacinummocelet( 69L ,)snoitcua( 44D :soN LEJ ).gro.rerepmelkluap.www ta osla dna 2002 lanruoJ cimonocE sesneciL moceleT G3 hsitirB eht fo elaS eht :revE noitcuA tseggiB ehT dna gro.rerepmelkluap.www ta dna 2002 weiveR cimonocE naeporuE snoitcuA moceleT G3 naeporuE eht :snoitcuA nuR ot )toN( woH gnidaer tsr dnemmocer I ,tcej bus eht ot wen sredaer esoht roF(.snoitcua mocelet eht no ecnerefnoc ofI SEC 1002 rebmeceD eht ta detneserp srepap owt no stnemmoc ym seldnub repap sihT.tneice neeb evah ot smees noitcua hsitirB eht fo emoctuo eht tub ,tnatropmi neeb evah yam noitasimixam ecnamrofrep-evitaleR.snoitcua mocelet G3 namreG dna hsitirB 0002 raey eht ni gniddib gnilzzup yltnerappa eht rof snoitanalpxe tseggus I tcartsbA 2002 rebmetpeS :tfard sihT ku.ca.xo.scimonoce@rerepmelk.luap :liame 44+ :xaF tnI 44+ :leT tnI dnalgnE FN1 1XO drofxO ytisrevinU drofxO ,egelloC dleuN rerepmelK luaP gro.rerepmelkluap.www//:ptth ta eb lliw ,lairetam detaler dna ,repap siht fo noisrev tsetal eht
snoitcuA moceleT G3 namreG dna hsitirB eht no snoitavresbO emoS
secirp kcots mocelet dna seunever atipac rep snoitcua murtceps elibom G3 naeporuE :erugiF.3
.snoitcua G3 rehto lareves tuo ba seussi no yeirb tnemmoc osla I.yroeht gnisimixam-ecnamrofrep-evitaler a gnidulcni ,snoitanalpxe evitanretla tseggus dna ,yrotcafsitas ylluf noitanalp -xe rieht dn ton od I yhw ssucsid re pap rieht no stnemmoc esehT.noitcua mocelet G3 namreG 0002 raey eht ni gniddib gnilzzup yltnerappa eht rof noitan -alpxe gniugirtni na depoleved evah )2002( rettetsfloW dna ledeiR ,mmirG tcartsbA
)2002( rettetsfloW dna ledeiR ,mmirG no stnemmoc :noitcua mocelet G3 namreG eht no snoitavresbo emoS.2
.tneice neeb gnivah emoctuo eht htiw tnetsisnoc era stluser )2002( snnamtsuD dna sregrB taht evresbo dna ,noitcua mocelet G3 hsitirB 0002 raey eht ni gniddib eht fo emos rof noitanalpxe na reo I tcartsbA
)2002( nnamtsuD dna sregrB no stnemmoc :noitcua mocelet G3 hsitirB eht no snoitavresbo emoS.1
Some Observations on the British 3G Telecom Auction: Comments on Brgers and Dustmann
By Paul Klemperer Contents
I. Introduction II. Efficiency of the UK Auction III. BTs Bidding Behavior
I offer an explanation for some of the bidding in the year 2000 British 3G telecom auction, and observe that Brgers and Dustmanns (2002) results are consistent with the outcome having been efficient.
I. Introduction Brgers and Dustmann (2002a), henceforth B -D, is a very valuable and insightful paper that is full of useful detail about the actual bidding in the UK 3G auction and will become a key reference for anyone studying it. As discussed in Klemperer (2002 a,c) and Binmore and Klemperer (2002), the UK auction was one of the most successful of the western European 3G auctions. Indeed in terms of revenue raised per capita it was the most successful of all the auctions, and it is therefore appropriate to examine, as B-D do, whether the auction's outcome was also as efficient as is often claimed. Furthermore, BD draw attention to many previously unnoted features of the bidding in the UK auction which do not fit well with standard theory, and which may have important implications for future auctions. I have learnt a lot from B- Ds analysis. In what follows, I discuss just two is sues about which my interpretation is slightly different. 1

Acknowledgements: I was the principal auction theorist advising the UK governments Radiocommunications Agency, which designed and ran the UK mobile-phone license auction discussed here, but the views expressed in this paper are mine alone. I do not intend to suggest that any of the behaviour discussed below violates any applicable rules or laws. I am very grateful to Tilman B rgers for useful comments, and to Marco Pagnozzi for our collaboration in the study of the 3G auctions and his helpful su ggestions about this essay. 1 I was the principal auction theorist advising the Radiocommunications Agency wh ich designed and ran the UK auction, but the views expressed in this paper are mine alone.

Paul Klemperer

II. Efficiency of the UK Auction B- Ds analysis makes clear that an ascending auction like the UKs runs the risk of an at least slightly inefficient outcome arising in some circumstances. However, it also seems clear that the actual outcome of the UK auction was efficient, or very close to efficient, in the sense of maximising the sum of the valuations of the licence holders. Klaus Schmidts (2002) excellent comment explains that the evidence from the bidding in the auction itself suggests that the UK auction was probably efficient. Evidence subsequent to the auction supports the same claim. It seems clear after the fact and especially after the other European auctions that the four incumbents had the highest valuations, 2 so were efficient winners. And there is no evidence that any losing entrant had a value for a license that exceeded TIW -Hutchisons. Finally, the evidence subsequent to the auction, as well as from within it (including the interpretation of the bidding offered below), suggests Vodafone had a higher incremental value for a large license than did any other incumbent, and therefore that the allocation of licenses among winners was also correct. In short, all the available evidence suggests that the UK auctions outcome was efficient in the sense claimed. 3
III. BTs Bidding Behavior B- D also suggests that some of the behavior they document is very hard to rationalise, but I conjecture that doing sufficient research into the environment in which the auction took place will yield good explanations, as I will illustrate by examining the main puzzle- BTs bidding.4 BTs bidding was such that the prices bid for the large (2 x 15 MHz) B and small (2 x 10 MHz) C, D, and E licences differed by roughly a constant in the early stages of the auction (phase 1 of the auction in B- D's term inology), and then switched to differing by roughly a fixed proportion (fifty per cent of the price level of the small licenses) in the later stages of the auc tion (phases 2 and 3 in B- Ds terminology). 5 This pattern seems unusual, but reviewing analysts reports

2 See van Damme (2002) and Fortis (2000) for evidence and discussion of these value differences. (Indirect evidence is also provided by the fact that only one out of the thirty incumbent bidders in the eight western European ascending auctions failed to win a license---and even this single failure was attributed to collusion or organizational strife within the bidder, rather than to the incumbent having a low value, see Klemperer 2002a). 3 Cable et al (2001) use stockmarket data to argue that there is no evidence that the outcome of the auction was anything but efficient. 4 However, B-D are to be congratulated on having already explained so much; they also looked at evidence from outside the auction to explain behaviour within it. 5 That BTs behaviour in the later stages of the auction can be described in this way was observed independently, by B-D and myself, after the conference in Munich to which their paper was contributed. The details are reported in B-Ds companion paper, Brgers and Dustmann (2002b).
Comments on Brgers and Dustmann
suggests a clue: some analysts assumed the value of the large license must be
times the value of a small license (reflecting an assumption that 1 times the amount of spectrum would allow offering 1 times the service ), while 2 1
several other analysts insisted the large license was worth a fixed sum more than a small one (reflecting the additional costs base stations, etc. required to run the same service with a smaller licence), and it was clearly well understood in the industry that different bidders might make different choices between these two different valuation models. Of course, if one or more bidders valued the large licence at 1

1 times the 2

value of the small license, this cannot on its own explain the price difference being a fixed proportion of the value of the small license. For example, if BTs private valuations for small and large licences were 4 billion and 6 billion, respectively, while Vodafones were 6 billion and 9 billion, respectively, and other bidders were closer to indifferent between small and large licences, then with straightforward bidding (in B-Ds terminology) the absolute value of the price difference would quickly move to equal 2 billion (since whenever the price difference was less than 2 billion, both BT and Vodafone would regard the large license as the best deal, and so would bid on it).7 However, it seems plausible that BT intrinsically valued a large license more than a smaller license by a fixed value that was considerably below 50% of the final price of a small license. BT may also have become very confident that Vodafone valued a large licence at 50% more than a small license. (Apart from any information from outside the auction, Vodafone never placed a bid on any license other than the large license in the auction.) Furthermore, BT may have wished to make Vodafone pay as much as possible for its license8 for at least tw o reasons. First, this would reduce Vodafones budget and so make Vodafone a weaker competitor in subsequent auctions (the British auction was the first of nine western European 3G auctions, and was also followed by others elsewhere in the world). And second, making Vodafone pay more would make the mar ket think Vodafone had not done better than BT in the auction. There is anec dotal evidence that BT was very concerned both about the stock- markets perceptions

The technology might actually allow offering slightly more than 1

1 times the service, 2

1. And even if, as I will argue, some of the early bidding was non-serious, the price difference would move to the fixed amount, 2 billion, as soon as the bidding became serious. 8 After the auction BT claimed it had deliberately pushed up the price that Vodafone had paid, and this was reported in the press, see Cane and Owen (2000). (At the time, this claim was pooh poohed by auction theorists as implausible, since it was hard to reconcile with the evidence without realising that BT and Vodafone might both have had different valuation models and also have had a reasonably clear idea of the other's valuation model.)
hence the value ratio might be slightly more than 1
of its performance, and about the wider markets view of its position relative to Vodafone. Allowing Vodafone to win the larger license at a lower per-MHz price than BT was paying might suggest BTs managers had got a bad deal. Or it might suggest that BT was not able to make effective use of a larger license in the way that Vodafone could, and hence that BT thought it was in a weak market position, while Vodafone was clearly number one. 9 So bidding up the large licenses price to 50% more than the current small license price may have seemed a reasonable risk to take, even given the small chance of ending up winning the large license at hundreds of millions of pounds more than BT valued it at. 10 Of course, even a small risk of winning the larg license might seem to have a e significant expected cost. But it was also possible that if BT did end up winning the large license, it might have been able to resell part of it at little or no loss, given that the auction prices would then have established a clear price per MHz. (The possibilities for resale were unclear, but Hutchison did in effect resell a fraction of the licence it won, very shortly after the auction, to KPN and Docomo at almost exactly the price per MHz that BT and Vodafone paid in the auc tion. 11 ) And, anyway, observers might not think BTs managers had made a bad decision, even if BT did end up winning (and keeping) the large license for 50% more than the price of a small license. 12 This theory, of course, leaves an important question unanswered. Why did BT not push up the price of the large license in the early stages of the auction? One reason is that much of the bidding in the early stages of the contest, when it was clear that there was no realistic chance of the auction ending very quickly (B Ds phase 1) does not seem to have been very serious.13 In fact, some bids were probably slightly frivolous, or designed to attract media attention. For example,

9 Klemperer (2002c) and Abbink et al (2002) discuss the importance of bidders concerns about relative performance in two other auctions in which BTs and Vodafones subsidiaries both competed, the German 3G auction and the previous years German DCS-1800 spectrum auction. (Strictly, Vodafone was not involved in the earlier auction, but Mannesman, which was a subsidiary of Vodafone by the time of the UK auction, did compete in the earlier auction.) 10 If BT was correct in its assessment that Vodafones valuation of a large licence was (at least) 50% more than that for a small license, the (only) risk that BT faced was that Vodafone would quit the auction altogether. But this outcome was completely implausible, since it would imply that Vodafones valuation for a small license was below that of Orange and One-2-One (which were both weaker incumbents) and at least one new entrant. The real risk would have been that BT had misjudged Vodafones valuation difference between the licenses, and BT perhaps knew this risk was small. 11 The UK Government now seems likely to make resale relatively easy, but this was unclear at the time of the auction, and actual resale of part of a license may in any case be unattractive since bringing a new competitor into the industry makes the remaining spectrum less valuable. Bringing new partners into a joint-venture as Hutchison did therefore seems the most relevant form of resale. 12 Of course, the arguments of this paragraph are in effect postulating that there may have been important common-value elements to valuations. Note that with common value elements it is plausible that the large license might be worth a fixed amount (say 500 million 1 billion) more than a small license at low prices, but a constant fraction (say 150%) of the small license at large prices. 13 Four bidders have informally confirmed this.
One-2-One raised its bid by slightly more than the minimum required in round 76 to bid 1,212,100,000!14 And BT did start pushing up the price difference between the large and small licences in round 99 when there were still 9 bidders left (so 4 more dropouts were still required to end the auction), and did not then stop pushing up the price difference until round 112 when the large license was more than 50% (and more than 1 licenses. A more serious reason why BT did not push up the price difference earlier is that BT may not have wanted to influence other bidders too early to think that license values were very high (since these other bidders might need time to adjust their views, and get extra money approved by their Boards, etc.). For example, if BTs valuation for a small license was 5 billion, it might have been conf ident that Vodafones value exceeded 4 billion for a small license, and therefore that Vodafone would pay at least 2 billion more for a large license. But pushing the price difference up to 2 billion immediately would have sent a very clear signal about what the ultimate prices might be at a time at which the auction prices for the smaller licences were still very low, and this could only have been damaging to BTs interests. A final possible reason why BT did not push up the price difference early on is that BT may not have become confident that Vodafones valuation of the large license was 1

To be sure, the figure flatters larger countries (especially Germany, conversely it underrates tiny Denmark), flatters centrally located countries (Ger many, again, and also Austria and Switzerland), flatters countries with lightly regulated telecom industries (Germany, again, amon g others) since larger, centrally located, lightly regulated markets are worth more but it also ig nores the fact that Germany and Austria sold more licences than other countries, reducing the total profitability of those markets. However, the more systematic discussion of the relative performance of the different auctions in Klemperer (2002a) comes to very similar conclusions.
1 Although efficiency was generally the primary objective, there is no evidence that efficiency differed much across the different countries auctions. Hence the focus on revenues.
Comments on Grimm, Riedel and Wolfstetter
So, since the German auction was both a success and was of a novel and complex design, it clearly deserves study, and Grimm, Riedel and Wolfstetters (2002) paper (henceforth GRW) would be welcome for that reason alone; their paper gives very valuable detail about the auction and will be a key reference for anyone studying it. But more than that, the paper is extremely interesting and makes acute observations about both the German and other 3G auctions. I have learned a lot from the paper, and agree with much of it. However, these comments will naturally focus on the disagreements. Section II summarises why I think GRWs explanation of the bidding in the German auction is not fully satis factory, and Section III develops this point more fully (the latter section can be omitted by readers who do not want too much detail). Section IV suggests other explanations for the bidding, and Section V develops a relativeperformance- maximising theory for it. Sections VI and VII briefly comment on some other 3G auctions, and stress (as GRW also do) the importance of attracting entry into an auction.
II. The German Auction In particular, I disagree with GRWs central claim that their model, as it stands, rationalises the behaviour of the two strongest bidders T- Mobil and Mannes man, or T and M, in GRWs terminology. 2 These bidders initially pushed up the price in the hope of driving out the sixth-strongest bidder, bidder 6 in GRWs terminology, but then gave up pushing the price up so that the auction did actually end with six winners but at a much higher price than was necessary to end the auction with this number of winners. This seems bizarre. To put the point simply, consider Ts and Ms decision about whether to end the auction with six winners at some given price, or whether to push the price up further. Raising the per - block price by 1 euro costs T and M 2 euros each, since they would each win 2 blocks in a six-w inner outcome. Their gain is the probability that bidder 6 quits, times their benefit from bidder 6 quitting. If it is worthwhile for T and M to push the price up in one round, but to stop pushing the price up in the next round, then the perceived probability of bidder 6 quitting in the next round must be both low, and also lower than it was in the last round. However, most observers thought the probability of bidder 6 quitting in the next round, conditional on not hav ing previously quit, was high and increasing around the time T and M ended the auction (when per capita price levels were approaching those achieved in the UK) and was much lower earlier (a six-player conclusion for the auction became possible at 55% of the UK price lev els). So any rationalisation of T and Ms behaviour must explain this apparently irrational behaviour of theirs. But GRWs model sidesteps this basic issue, as I now explain:

2 T-Mobil and Mannesman were subsidiaries of Deutsche Telekom and Vodafone, re spectively.
III. GRWs Analysis of the German Auction (This section can be omitted by reader s who do not want too much detail.) To understand GRWs argument and why I believe it is incomplete in this context consider the preferences of either one of the two strong bidders, M and T. At any point of time, it would like to end the auction only if it prefers this to waiting until the price has risen a small further amount, price, before the auction ends. The gain from waiting is the probability, 6 , that bidder 6 will quit in the price interval, price, times the value of driving bidder 6 out. This value is itself the benefit, , of winning a third block, including the benefits from excluding bidder 6 from the industry (leading to a more concentrated, and hence more profitable, market), less the current price of buying an additional block. That is, the total gain from waiting equals 6[ price]. The cost of waiting is the extra price, price, paid on the two units that the bidder will win anyway, i.e., 2( price). That is, the bidder would prefer to plan to end the auction if a further price rise of price fails to drive out bidder 6, rather than end it now if

6[ price] > 2[price]

The bidder would prefer to end the auction now if

6[ price] < 2[price].

In GRWs model, bidder 6s value can only take one of two possible values,
v6 (strong type) and v6 (weak type), and the auction cannot be ended before
price p (< v6 ), so the only conceivably sensible strategies for the strong bidders are a) b) c) to end the auction as soon as possible at p, or to push the price up a further price = v 6 p to v6 to drive out the weaker type of bidder 6, but then to end the auction, or to push the price up further still, by an additional price = v v6 more to 6 v6 , to drive out both types of bidder 6.
The observed behaviour in the actual auction corresponded to case b) of GRWs model. The condition for b) to be preferred to a) is the appropriate version of (1) or equivalently is GRWs equation (9). 3 The condition for b) to be preferred to c) is

3 In GRWs notation, = v13 + b for the case of su ccessfully driving out bidder 6 at price v6.
the appropriate version of (2), or equivalent ly GRWs equation (6).4 So these are the key conditions in GRWs Theorem. 5 If (6) holds there is an equilibrium in which outcome b) arises. (If (6) fails, both strong bidders prefer c), and either can unilaterally implement it.) If (9) holds, the equilibrium is unique. If (9) fails, outcome b) can still be an equilibrium of the model if (6) holds, since neither strong bidder can unilaterally end the auction, but the equilibrium is neither unique, nor plausible. 6 So for the observed play to correspond to a p lausible equilibrium of GRWs model, both (9) and (6) are required. 7 Noting that (9) and (6) are just my equations (1) and (2) suggests why GRWs theory seems unlikely to describe reality. Of course, (9) and (6) correspond to (1) and (2) evaluated at different values of 6 , , price and price, reflecting the different stages of the game at which (9) and (6) are computed. So the observed play can correspond to an equilibrium of GRWs model. But this requires that [ 6 / (price )] not be too much lower when the strong bidders could first have ended the auction (when (1) must be satisfied) than at the actual end of the auction (when (2) must be satisfied).8 Furthermore, the tension between conditions (9) and (6) is more severe when the game is generalised to many small stages since the values of [ 6 / (price)] ,
, and price to be substituted into (1) and (2) cannot then vary much between stages, and related conditions must then hold at all the stages an issue that GRW do not address.9 In particular, (1) must hold just before the auction ends,
4 In GRWs notation, = v13 + b for the case of driving out bidder 6 at price v6 , but 6 = 1 in GRWs model at stage c). 5 GRW rename (6) as " 2 0" in their statement of Theorem 1. pr 6 This equilibrium is not plausible if (9) fails because in this equilibrium both strong players prefer outcome a) to outcome b), but both follow the strategy corresponding to b) because each expects the other to do this. This logic can only hold in the two -stage model: with more stages, each strong player would know that if it followed the strategy corresponding to a), then the other strong player would follow just one round of the auction later (if (9) fails) that is, the players could trivially coordinate on strategy a) in the a ctual multi-round auction, which eliminates this equilibrium. This equilibrium is, of course, also Pareto dominated by the more natural equilibrium for the players in GRWs two -stage model, and GRW also eliminate this e quilibrium in their limited extension to multiple rounds (see note 9). However, an equilibrium of this kind becomes more plausible if M and T are each u ncertain that its rival shares its assessment of the parameters, or are uncertain about the rivals objectives see section V. 7 GRW note that Theorem 1 requires other conditions too. 8 It does not seem likely that ( price) ever became very small because includes both the value of a third block to a strong bidder and the value, b in GRWs terminology, of excluding bidder 6 from the industry, leading to a more concentrated and hence more profitable market. Therefore, must be greater than p2 + b , where p2 is the expected maximum of the value of a fourth block to a strong bidder and a third block to a less strong bidder. And, as GRW point out, p2 must itself be quite high for the GRW equilibrium to make sense the logic of GRWs equilibrium requires p2 to at least equal the final German auction price. (The very limited anecdotal evidence su ggests that p2 might have been, very roughly, in the region of the final German auction price.) 9 GRW do briefly consider extending their model to many rounds of bidding, but when they do this they maintain the extreme assumption that bidder 6s valuation can take only

A similar argument is that if all firms are known to be ordinary profit maximisers, but firms are unsure that their rival has the same estimates of parameters such as 6 , then firms may be unwilling to reduce demand first. 16 Abbink et al (2002), section 4, provides some evidence that relative-performance issues were important to these bidders in the German DCS-1800 auction.
to maximise both of their absolute profits. Indeed it seems that M did initially try to signal to T that they should do just this (see GRW and Klemperer 2002a,b). But T could not know whether M was sincere, and the firms apparently mistrusted each other's intentions (see Klemperer 2002b) and, as we have seen, there are very strong relative-performance arguments (it suffices that each feared that the other might maximise relative-performance) why neither was prepared to be the first to reduce demand while prices were still low. T then reduced demand later when prices were higher, perhaps for relative performance reasons 17 ,and/or because this could also improve its absolute performance if M failed (or was unable) to follow its demand reduction.18,19 And once T had reduced demand, there a several possible reasons why M re followed straightaway. First, M had a strong relative-performance incentive to follow mmediately, as explained above. Second, M may have wanted to i develop a rep utation for co-operative behaviour in which M and T parallel each other's behav iour a kind of relative-performance effect but strictly driven by Ms long-run absolute- performance goals (see Klemperer 2002b). Third, M might have been concerned only with its (short-run) absolute performance, but it might all along have taken the view that this would be maximised by M and T both reduc ing demand, and it might have stuck to this view (i.e., M may have been extremely pessimistic about driv ing out bidder 6 at low prices and, even though driving out bidder 6 seemed more likely at high prices, remained fairly pessimistic see Klemperer 2002a); this is consistent with Ms early behaviour
17 It might seem that a firm could protect its relative performance by following a strategy of quitting only if its rival quits when prices are above p2. Ho wever, it takes time to be sure the rival has quit (because the auctioneer gave the bidders only limited inform ation about their rivals bidding), and it also takes time to respond. Fu rthermore, some of the weaker players may have been staying in the auction in the hope of being a winner in a five-firm industry, which would have been the outcome if M and T had successfully driven one of them out in particular, each of Mobilcom and Group 3G might have hoped that the other (or possibly E-plus or Viag) was the bidder 6 who might have been driven out. In this case, when one of M and T quits bidding for a third block, these weaker players may expect the other of M and T to try to follow suit and may therefore try to quit first rather than find themselves stuck as winners in a much-less-profitable six-firm industry. So if e ither M or T failed to quit first when prices became high, it might have risked being stranded buying a third block at a higher price than its opponent, and achieving a very poor relative-performance. 18 If T thought M was not too concerned with relative performance, T could improve both its relative and absolute performance by r educing demand once the price exceeded p2 , and free-riding on M continuing to push price up to drive out bidder 6. And even if M was concerned with relative performance, there was the possibility that M would have been unable to follow T (see note 16). 19 Of course, there may be other reasons, such as T being influenced by stockmarket pressure or its government ownership (se e section IV). It might be argued that another possibility was that was not in fact that high. However, this seems less likely since must have substantially exceeded p2 (see note 8) and if p2 were low then both firms would have been willing to reduce demand early on for relative-, as well as absolute-, performance reasons.

(signalling T to reduce demand but not unilaterally reducing demand) if it feared that T might place a large weight on relative performance. Of course, there may be other reasons for the observed behaviour in the auction. For example, fear that one's rival has very different perceptions from ones own about the chance of driving out bidder 6 can have similar effects to fear that one's rival is a relative-performance maximiser 20 , and Klemperer (2002b) emphasises the mistrust and misunderstanding between the bidders. But the point is that the apparently puzzling behaviour can be e xplained by postulating only a limited concern with relative performance. To explain why M and T failed to reduce demand early on, it suffices that each firm thought its rival put some weight on relative performance; it is not necessary that either firm actually did so, and even the conjectured weights on relative performance need not have been large if firms were also uncertain about their rivals perceptions about bidder 6s b ehavior, etc. And not much more concern with relative performance is needed to explain the firms later behaviour in the auction.
VI. The Austrian and Swiss Auctions Turning to other 3G auctions, I disagree with GRWs assertion that the Austrian auction design was superior to the Swiss, except to the extent that the Austrian reserve price was somewhat more realistically chosen than the Swiss reserve. Neither auction attracted more bidders than there were winners, and neither involved any significant bidding. (Although there was a semblance of serious bidding in the Austrian auction, the bidders there were put under consider able pressure from the authorities to continue the bidding, and it was widely b elieved that the bidding only lasted the few rounds it did in order to create some public perception of genuine competition and reduce the risk of the gover nment changing the rules.) Neither auction achieved more than 11% more than the r serve price that had been set. The only important difference is that e the Swiss reserve price had been set ludicrously low at 20 euros per capita, while the Au strian reserve price, although still far lower than it should have been, was 90 euros per capita.21 But revenues in excess of 300 euros per capita should probably have been attainable in both auctions, see Klemperer (2002a). So both of these auctions were failures, and both were intensely embarrassing to their respective governments. Indeed there was no su ccessful European 3G auction after the UK and German auctions until the Danes switched to a sealedbid design. I have discussed all these auctions in more detail in Klemperer (2002a).

VII. The Importance of Entry, and the UK Auction
20 In particular, e ither fear can make a bidder unwilling to reduce demand first when prices are low, because of the perceived risk that the rival will not follow. See note 15. 21 Of course, Switzerland sold 4 licences while Austria sold 6, but the Swiss could obviously have used their same design to sell 6 ilcences if they had preferred that outcome.
Where I do agree very strongly with GRW is on the importance of attracting entry into an auction. 22 As GRW say, competition is not a free good and auctions must be designed with this in mind. However, this does not imply that there is any single best design. Often a sealed-bid design is best for attracting entry, as is suggested by the Danish example in the previous par agraph. But this need not be the case. The UK design was appropriate in its context, because the UK auction was the first 3G auction and was therefore unlikely to suffer from entry problems. (See Klemperer (2002a,b) for more discussion of why being first was so important.) Indeed the UK auction attracted 13 bidders compared with the 7 that entered the German auction. It seems improbable that the Ge man design would have usefully increased competition in the British r auction, and the British design had other advantages over the German design23 , see Binmore and Klemperer (2002) and Klemperer (2002a) (though my view may be coloured by my having been the principal auction theorist for the UK auction24 ). In another context, when Peter Cramton, Eric Maskin and I advised on the UKs 2002 auction for greenhouse gas emission reductions, we chose a uniform price ascen ding design as being most likely to attract small bidders who did not have the resources to work out how to bid correctly in a discriminatory price auc tion. 25 And nor, of course, is entry always the key issue. As I discuss further in Klemperer (2002d), good auction design is not one size fits all, but must always be tailored to its co ntext.
Grimm, Riedel and Wolfstetter (2002) have developed an intriguing explanation for the apparently puzzling bidding in the year 2000 German 3G telecom auction. These comments on their paper discuss why I do not find their explanation fully satisfactory, and suggest alternative explanations, including a relative-performance-maximising theory. I also comment briefly on issues about several other 3G auctions.

22 I emphasised this in Bulow and Klemperer (1996), and Klemperer (1998, 2000, 2002d). 23 One advantage is identified in note 16: in the German design a bidder might rationally follow a strategy that could mean that it felt sorry to have won as soon as the a uction finished. 24 I was the principal auction theorist advising the Radiocommunications Agency which designed and ran the UK auction. Ken Binmore had a leading role and supervised experiments testing the proposed designs. Other academic advisors i cluded Tilman n Brgers, Jeremy Bulow , Philippe Jehiel and Joe Swierzbinski. 25 Larry Ausubel and Jeremy Bulow were also involved in the implementation of this auction. Strictly this was a descending auction, since the auctioneer was buying reductions in emissions rather than selling permits to emit, but the auction corresponded to an ascending auction to sell emissions.
Abbink, K., B. Irlenbusch, B. Rockenbach, A. Sadrieh, and R. Selten (2002) The behavioural approach to the strategic analysis of spectrum auctions: the case of the German DCS -1800 auction, Working Paper, Universities of Nottingham, Erfurt, Tilburg, and Bonn. Binmore, K. and P. Klemperer (2002), The Biggest Auction Ever: the Sale of the British 3G Telecom Licences, Economic Journal, 112 (478), C74C96, and reproduced with fewer typos at www.paulklemperer.org. Bulow , J. and P. Klemperer (1996), Auctions vs. Negotiations, American Economic Review 86( 1), 180194. Ewerhart, C. and B. Moldovanu (2001), A Stylized Model of the German UMTS Auction, Working Paper, University of Mannheim, Germany. Grimm, V., F. Riedel, and E. Wolfstetter (2002), The Third Generation (UMTS)Spectrum Auction in Germany, ifo Studien, 48(1), Klemperer, P. (1998), Auctions with Almost Common Values, European Economic Review 42 (3-5), 757769. Klemperer, P. (2000), Why Every Economist Should Learn Some Auction Theory, in: M. Dewatripont, L. Hansen, and S. Turnovsky (Eds.), Advances in Economics and conometrics: Invited Lectures to Eighth World Congress of the Econometric Society, Cambridge, UK: Cambridge University Press, and available at www.paulklemperer.org. Klemperer, P. (2002a), How (Not) to Run Auctions: the European 3G Telecom Auctions, European Economic Review 46 (4-5), 829 -845. Klemperer, P. (2002b), Using and Abusing Economic Theory. 2002 Marshall Lecture to the European Economic Association, forthcoming at www.paulklemperer.org. Klemperer, P. (2002c), Some observations on the British 3G telecom auction: comments on Brgers and Dustmann, ifo Studien, 48 (1), and at www.paulklemperer.org. Klemperer, P. (2002d), What Really Matters In Auction Design, Journal of Economic Perspectives 16 (1), 169189.