As far as patents or other rights of third parties are concerned, liability is only assumed for components, not for
applications, processes and c irc uit s imp lemented within components or as s em blies.
The information describe s the t yp e of co m ponent and shall not be considered as ass ured characteristics.
Terms of delivery and rights to ch ange design reserved.
For questions on technology, delivery and prices please contact the Semiconductor Group Offices in Germany or
the Siemens Companies and R epresentatives worldwide (see address list).
Due to technical requireme nt s com ponents may contain dangerous substances. For informatio n on t he t y pes in
question please contact yo ur nearest Siemens Office, Semiconductor Group.
Siemens AG is an approved CECC manufacturer.
Packing
Please use the recycling ope rat ors k now n t o y ou. W e ca n als o help you – get in touch with your nearest sa les
office. By agreement we will take packing material back, if it is sorted. You must bear the costs of transport.
For packing material that is returned to us unsorted or which we are not obliged to accept, we shall have to invoice
you for any costs incurred.
Components used in life-support devices or systems must be expressly authorized for such purpose!
Critical components
systems
1 A critical component is a component used in a life-support device or system whose failure can reasonably be
2 Life support devices or system s are int ended (a) to be implanted in the human body, or (b) to support and/or
2
with the express written approv al of the Semiconductor Group of Siemens AG.
expected to cause the failure of that life-support device or system, or to affect its safety or effectiveness of that
device or system.
maintain and sustain human life. If th ey fail, it is rea so nable to assume that the health of the us er m ay be endangered.
1
of the Semiconductor Group of Siemens AG, may only be used in life-s upport devices or
Combined with an anal og front end th e PSB 4860 provide s a s olution for embe dde d or
stand alone answering ma chine applica tions. Together wi th a standard microcont roller
for analog telephones these two chi ps form the co re of a featurep hone with fu ll duplex
speakerphone and answering machine capabilities.
™
The chip features recording by DigiTape
Messages recorded with DigiTape
™
can be played back with variable speed without
pitch alteration. Messages recorded with a higher bitrate can be converted into
messages with a lower b itrate arbitrarily. Current m embers of DigiTape (TM) span the
range from 3.3 kbit/s to 10.3 kbit/s.
Furthermore the PSB 4860, V2.1 has a full duplex s peakerphone, a caller ID dec oder,
DTMF recognition and generation and call progress tone detection. The frequency
response of cheap m icro phones or lou dspeak ers can be correct ed by a programma ble
equalizer.
, a family of high performance algorithms.
Messages and user da ta can be stored in ARAM/DRAM or flash mem ory wh ich can be
directly connected to the PSB 4860. The PSB 4860 also supports a voice prompt
EPROM for fixed announcements.
®
The PSB 4860 provides an IOM
-2 compatible inte rface with t wo channels f or speech
data.
Alternatively to the IOM
®
-2 compatible interface the PSB 4 860 su pport s a si mple se rial
data interface (SSDI) with separate strobe sign als for each direction (linear PCM data,
one channel).
A separate interface is used for a glueless connection to the PSB 4851.
The chip is programmed by a simple four wire serial control interface and can inform the
microcontroller of new events by an interrupt sign al. For data retention the PSB 4860
supports a power down mode where only the real time clock and the memory refresh (in
case of ARAM/DRAM) are operational.
The PSB 4860 supports interface pins to +5 V levels.
Multiplexed address outputs for ARAM, DRAM
access.
Non-multiplexed address outputs for voice
prompt EPROM.
Auxiliary Parallel Port:
General purpose I/O.
L
L
L
L
L
L
L
L
L
PSB 4860
Overview
42
43
44
45
46
47
50
51
35
MD
MD
MD
MD
MD
MD
MD
MD
CAS
0
1
2
3
4
5
6
7
0
ALE
36
CAS
1
FCS
34RAS
/
FOE
33VPRD
FCLE
I/O
I/O
I/O
I/O
I/O
I/O
I/O
I/O
/
O
/
O
OH
/
OH
-
-
-
-
-
-
-
-
2)
H
2)
2)
Memory Data 0-7:
Memory (ARAM, DRAM, Flash Memory,
EPROM) data bus.
ARAM, DRAM:
Column address strobe for memory bank 0 or 1.
Flash Memory:
Address Latch Enable for address lines A
16-A23
Chip select signal for Flash Memory
ARAM, DRAM:
Row address strobe for both memory banks.
Flash Memory:
Output enable signal for Flash Memory.
ARAM, DRAM:
Read signal for voice prompt EPROM.
Flash Memory:
Command latch enable for Flash Memory.
.
Semiconductor Group1610.97
Table 1Pin Definitions and Functions
32W/FWEOH
2)
ARAM, DRAM:
Write signal for all memory banks.
Flash Memory:
Write signal for Flash Memory.
31FRDYI-Flash Memory Ready
Input for Ready/Busy signal of Flash Memory
PSB 4860
Overview
5
6
OSC
OSC
1
2
I
O
Z
Auxiliary Oscillator:
Oscillator loop for 32.768 kHz crystal.
8CLKI-Alternative AFECLK Source
13,824 MHz
2
3
37
38
XTAL
XTAL
SPS
SPS
I
1
O
2
0
1
O
O
Z
L
L
Oscillator:
XTAL
XTAL
: External clock or input of oscillator loop.
1
: output of oscillator loop for crystal.
2
Multipurpose Outputs:
General purpose, speakerphone, address lines
or status
56ROO-Reserved Output
Must be left open.
1)
These lines are driven low with 125 µA until the mode (address lines or auxi liary port ) is def ined.
2)
These lines are driven high with 70 µA during reset.
Semiconductor Group1710.97
1.4Logic Symbol
1
PSB 4860
Overview
PSB
RST
AFECLK
AFEFS
AFEDD
AFEDU
V
DD
V
DDA
V
SS
MA0-MA15MD0-MD
OSC
CLK
PSB 4860
7
CAS0/
ALE
CAS1/
FCS
Memory
1
OSC
XTAL1XTAL
2
RAS/VPRD/
FOE
W/
FWE
2
FCLE
DU/DX
DD/DR
DCL
FSC
DXST
DRST
SDX
SDR
SCLK
FRDY
IOM®-2
SSDI4851
INT
SCI
CS
Figure 2Logic Symbol of PSB 4860
Semiconductor Group1810.97
1.5Functional Block Diagram
PSB 4860
Overview
AFECLK
AFEFS
AFEDD
AFEDU
RST
Analog
Front End
Interface
1
OSC
XTAL1XTAL
2
OSC
Reset and Timing Unit
DSP
Memory Interface
2
DRST
DXST
Data
Interface
Control
Interface
DU/DX
DD/DR
DCL
FSC
INT
SDX
SDR
SCLK
CS
FRDY
MA
-MA15MD0-MD7CAS0/
0
ALE
CAS1/
FCS
RAS/VPRD/
FOE
W/
FWE
FCLE
Figure 3Block Diagram of PSB 4860
1.6System Integration
The PSB 4860 combined with an analo g fron t end (PSB 4 851 ) can be used in a variety
of applications. This combination offers outstanding features like full duplex
speakerphone and emergenc y operation. Some applicat ions are given in the followin g
sections.
1.6.1Analog Featurephone with Digital Answering Machine
Figure 4 shows an example of an analog telephone system. The telephone can operate
during power failure by line powering. In this case only the handset and ringer circuit are
active. All other parts of the chipset are shut down leaving enough power for the external
microcontroller to perform basic tasks like keyboard monitoring.
Semiconductor Group1910.97
PSB 4860
Overview
For answering machine operation the voice data is stored in ARAM or Flash Memory
devices. In addition, voice pro mpts can be played back from an optional vo ice prompt
EPROM. If flash memory is used the func tionality of the voice prompt EPROM c an be
realized by the flash memory devices. The microcontroller can use the memory attached
to the PSB 4860/PSB 4851 to store and retrieve binary data.
ARAM
Flash Memory
PSB 4860
PSB 4851
Voice Prompt
EPROM
077-3445
tip/
ring
line
Microcontroller
Figure 4Analog Full Duplex Speakerphone with Digital Answering Machine
Semiconductor Group2010.97
PSB 4860
Overview
1.6.2Featurephone with Digital Answering Machine for ISDN Terminal
Figure 5 shows an ISDN featureph one that takes full advantage of two simu ltaneous
connections. In this a pplication o ne channel o f the PSB 4851 interfa ces to the hands et
and speakerphone while the other provides an interface for an externa l analog device
(e.g. FAX machine).
Flash Memory
PSB 4860
PSB 4851
IOM®-2
Power Controller
PSB 2120/1
SCI
SLIC
POTS
077-3445
Microcontroller
PSB 2186
®
-S TE
ISAC
S
-BUS
0
Figure 5Featurephone with Answering Machine for ISDN Terminal
In addition, the two chann els of the PSB 485 1 ca n be use d for holdi ng two c on nec tion s
simultaneously. O ne connection can be switch ed to the handset and the other to the
speakerphone box. Local three party conferences are also possible.
Semiconductor Group2110.97
PSB 4860
Overview
1.6.3DECT Basestation with Integrated Digital Answering Machine
Figure 6 shows a DECT basestation based on the PSB 4860/PSB 4851 chi pset. In this
application it is possible to service both an external call and an internal call at the same
time. For programming the serial control interface (SCI) is used while voice data is
®
transferred via the strobed serial data interface (SSDI/IOM
Flash Memory
-2).
PSB 4860
PSB 4851
SSDI/IOM®-2
Antenna
077-3445
tip/
ring
Microcontroller
line
Figure 6DECT Basestation
SCI
Burstmode
Controller
DECT
HF
Semiconductor Group2210.97
PSB 4860
Functional Description
2Functional Description
Functional Units
Functional Units
The PSB 4860 contains several functional units that can be combined wit h almost no
restrictions to perform a given task. Figure 7 gives an overview of the important
functional units.
®
SSDI/IOM
-2IOM®-2
Channel 2Channel 1
loud-
speaker
microphone
line
out
line
in
S
4
S
2
I
1
I
2
I
3
S
3
S
9
DTMF
Generator
S
10
I
1
I
2
I
3
S
1
S
5
S
14
Universal
Attenuator
I
1
S
6
I1I2I
I1I
I1I
2
Line Echo
Canceller
S
15
3
S
2
acoustic side
Speaker-
phone
line side
S
12
I3I
S
11
4
I1I
AGC
16
S
8
S
I1I2I
7
S
3
13
Speech
Decoder
Memory
Speech
Coder
I1I
2
I
2
1
Equalizer
S
17
S
18
I
1
CNG
Detector
I
1
Alert Tone
Detector
I
1
CPT
Detector
I
1
CID
Decoder
I
1
DTMF
Detector
SCI
signal summation:s igna l sou rc es:
I
1
I
2
I
3
S
,...,S
1
18
Figure 7Functional Units - Overview
Semiconductor Group2310.97
PSB 4860
Functional Description
Each unit has one or more signa l inputs (denoted by I). Most units have at least one
signal output (denoted by S). Any input I can be co nnected to any signal output S. In
addition to the sign als shown in figure 7 there is also the signal S
useful at signal summation points. Table 2 lists the available signals within the PSB 4860
according to their reference points.
Table 2Signal Summary
SignalDescription
(silence), which is
0
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Silence
Analog line input (channel 1 of PSB 4851 interface)
Analog line output (channel 1 of PSB 4851 interface)
Microphone input (channel 2 of PSB 4851 interface)
Loudspeaker/Handset output (channel 2 of PSB 4851 interface)
Serial interface input, channel 1
Serial interface output, channel 1
Serial interface input, channel 2
Serial interface output, channel 2
DTMF generator output
DTMF generator auxiliary output
Speakerphone output (acoustic side)
Speakerphone output (line side)
Speech decoder output
Universal attenuator output
S
15
S
16
S
17
S
18
Semiconductor Group2410.97
Line echo canceller output
Automatic gain control output (after gain stage)
Automatic gain control output (before gain stage)
Equalizer output
PSB 4860
Functional Description
The following figures show the connections for two typical states during operation. Units
that are not needed are not shown. Inputs that are not needed are connected to S
provides silence (denoted by 0). In figure 8 a hands-free phone conversation is currently
in progress. The speech coder is used to record the signals of both parties. The alert tone
detector is used to detect an alerting tone o f an of f-hoo k c all er id request while the CID
decoder decodes the actual data transmitted in this case.
which
0
loud-
speaker
microphone
line
out
line
in
0
0
0
Speech
coder
acoustic side
Speaker-
phone
line side
AGC
0
0
0
Memory
Line Echo
Canceller
CID
decoder
Alert Tone
Detector
SCI
Figure 8Functional Units - Recording a Phone Conversation
Semiconductor Group2510.97
PSB 4860
Functional Description
In figure 9 a phone conv ersation using the speakerpho ne is in progress. One party is
using the base station of a DECT system while the other party is using a mobile handset.
At the same time an external call is serviced by the answering machine. In the current
state a message (recorded or out going) is being played back. In this case the DTMF
detector is used t o detect sign als for remote a ccess while th e CPT detector is u sed to
determine the end of the external call.
®
SSDI/IOM
Channel 1
-2
loud-
speaker
microphone
line
out
line
in
0
0
0
0
00
0
acoustic side
Speaker-
phone
line side
Equalizer
Speech
decoder
0
Memory
Line Echo
Canceller
CPT
decoder
DTMF
Detector
SCI
Figure 9Functional Units - Simultaneous Internal and External Call
Semiconductor Group2610.97
PSB 4860
Functional Description
2.1Functional Units
In this section the functional units of the PSB 4860 are described in detail. The functional
units can be individually enabled or disabled.
2.1.1Full Duplex Speakerphone
The speakerphone unit (figure 10) is attached to four signals (microphone, loudspeaker,
line out and line in). The two input signals (microphone, line in) are preceded by a signal
summation point.
I
1
I
2
S
11
microphone
loudspeaker
a
c
o
u
s
t
Speakerphone
i
c
s
i
d
e
line out
l
i
n
e
s
i
d
e
line in
S
12
I
3
I
4
Figure 10 Speakerphone - Signal Connections
Internally, this unit can be divided into an echo cancellation unit and an echo suppression
unit (figure 11). The echo cancell ation unit provides the attenuation G
suppression unit provides the attenuation G
speakerphone is therefore ATT=G
C+Gs
.
. The total attenuation ATT of the
s
while the echo
c
Echo
Cancellation
G
loudspeakerline in
c
Echo
Suppression
G
S
line outmicrophone
Figure 11 Speakerphone - Block Diagram
The echo suppression unit can be enabled without the echo cancellation unit. If the echo
cancellation unit is disabled, the echo suppression unit still provides speakerphone
functionality, albeit onl y half duplex. As the echo cancellation must be dis abled during
recording or playback of speech data, this option allows for speakerphone operation
Semiconductor Group2710.97
PSB 4860
Functional Description
even if recording or playback is going on. The echo sup pression unit is also used to
provide additional attenuation if the echo cancellation unit cannot provide all of the
required attenuation itself.
2.1.2Echo Cancellation
A simplified block diagram of the echo cancellation unit is shown in figure 12.
microphone
line out
-
FIR
NLMSControl
loudspeaker
Filter
line in
Figure 12 Echo Cancellation Unit - Block Diagram
The echo cancellation unit consists of an finite impulse response filter (FIR) that models
the expected acoustic echo, an NLMS based adaption unit and a control unit. The
expected echo is subtracted from the actual input signal from the microphone. If the
model is exact and the echo does n ot excee d the len gth of th e filter t hen the e cho can
be completely cancelled. However, even if this ideal state can be achieved for one given
moment the acoustic echo usually changes over time. Therefore the NLMS unit
continuously adapts the coefficients of the FIR filter. This adaption process is steered by
the control unit. As an example, the adaption is inhibited as long as double talk is
detected by the control un it. Fu rthermore t he con trol unit informs the ech o supp ress ion
unit about the achieved echo return loss.
Table 3 shows the registers associated with the echo cancellation unit.
Table 3Echo Cancellation Unit Registers
Register# of BitsNameComment
SAELEN9LENLength of FIR filter
SAEATT15ATTAttenuation reduction during double-talk
SAEGS3GSGlobal scale (all blocks)
The length of the FIR filter can be varied from 127 to 511 taps (15.875ms to 63.875ms).
The taps are grouped into blocks. Each block contains 64 taps.
The performance of the FIR filter can be enhanced by p rescaling some or call of the
coefficients of the FIR filter. A coefficient is prescaled by multiplying it by a constant. The
advantage of prescaling is an enhanced precision and consequently an enhanced echo
cancellation. The disadvantage is a reduced echo cancellation performance if the signal
exceeds the maximal coefficient value. More precisely, if a coefficient at tap T
by a factor C
C
(Max: Maximum PCM value). As an example figure shows a typical room impulse
i
then the level of the echo (room impulse response) must not exceed Max/
i
response.
is scaled
i
A
0.5
0.25
t
0.25
t
Figure 13 Echo Cancellation Unit - Typical Room Impulse Response
First of all, the echo never exceeds 0.5 of the maximum value. Furthermore the echo
never exceeds 0.25 of the maximum value after time t
be scaled by a factor of 2 and all co efficients fo r taps correspon ding to times af ter t
. Therefore all coefficients can
0.25
0.25
can be scaled a factor of 4.
The echo cancellati on unit provides thre e parameters for scali ng coefficients. Th e first
parameter (GS) determines a scale for all coefficients. The second parameter (FB)
determines the first block for which an additional scale (PS) takes effect.
This feature can be used for different default settings like large or small rooms.
Semiconductor Group2910.97
PSB 4860
Functional Description
2.1.3Echo Suppression
The echo suppression unit can be in one of three states:
• transmit state
• receive state
• idle state
In transmit state the microphone signal drives the line output while the line input is
attenuated. In receive state the loudspeaker signa l is driven by the line inp ut while the
microphone signal is attenuated. In idle st ate both signal paths are a ctive with evenly
distributed attenuation.
idle state
microphone
loudspeaker
microphone
transmit state
loudspeaker
microphone
receive stat e
loudspeaker
Figure 14 Echo Suppression Unit - States of Operation
line out
line in
line out
line in
line out
line in
Semiconductor Group3010.97
PSB 4860
Functional Description
Figure15 shows the signal flow graph of the echo suppression unit in more detail.
LGAX
SDR
line outmicrophone
SCLSSCAS
AGCR
loudspeaker
AGCX
SDX
Attenuation
LGARline in
GHX
Control
GHR
Figure 15 Echo Suppression Unit - Signal Flow Graph
State switching is controlled by the speech comparators (SCAS, SCLS) and the speech
detectors (SDX, SDR). The amplifiers (AGCX, AGCR, LGAX, LGAR) are used to
achieve proper signal levels for each state. All blocks are programmable. Thus the
telephone set can be optimized and adjusted to the particular geometrical and acoustical
environment. The following sections discuss each block of the echo suppression unit in
detail.
Semiconductor Group3110.97
PSB 4860
Functional Description
2.1.3.1Speech Detector
For each signal source a speech detector (SDX, SDR) is available. The speech
detectors are identical but ca n be programm ed indiv idually . Figure 16 shows th e sig nal
flow graph of a speech detector.
OFF
-
LIMLP1PD
LP1
LIM
Signal Preprocessing
PDS
PDN
LP2
LP2S
LP2N
LP2L
Background Noise Monitor
Figure 16 Speech Detector - Signal Flow Graph
The first three units (LIM, LP1, PD) are used for preprocessing the signal while the actual
speech detection is performed by the background noise monitor.
Background Noise Monitor
The tasks of the noise monitor are to differentiate voice signals from background noise,
even if it exceeds the voice level, and to recognize voice signals without any delay.
Therefore the Background Noise Monitor consists of the Low-Pass Filter 2 (LP2) and the
offset in two separate branches. Basically it works on the burst-characteristic of the
speech: voice signals consist of short peaks with high power (bursts). In contrast,
background noise can be regarded approximately stationary from its average power.
Low-Pass Filter 2 provides different time constants for noise (non-detected speech) and
speech. It determines the average of the no ise reference level. In case of bac kground
noise the level at the output of LP2 is approximately the level of the input. As in the other
branch an additional offset OFF is added to the signal, the comparator signals noise. At
speech bursts the digital signals arriving at the comparator via the offset branch change
faster than those via the LP2-branch. If the difference exceeds the offset OFF, the
Semiconductor Group3210.97
PSB 4860
Functional Description
comparator signals speec h. Therefore the output of the backgrou nd noise monitor is a
digital signal indicating speech (1) or noise (0).
A small fade constant (LP2N) enables fast settling of LP2 to the average noise level after
the end of speec h re cog niti on. H oweve r, a to o s mall time constant for LP2N can c aus e
rapid charging to such a high level that after recognizing speech the danger of an
unwanted switchin g back to noise exists. It is recommended t o choose a large rising
constant (LP2S) so that spee ch itself charges the LP2 very sl owly. Generally, it is not
recommended to choos e an infini te LP2S because t hen appro aching the noi se level i s
disabled. Duri ng co nti nuou s speech or tones the LP2 will be charge d until the limitation
LP2L is reached. Then the value of LP2 is frozen until a break discharges the LP2. This
limitation permits transmission of continuous tones and “music on hold”.
The offset stage represents the estimated difference between the speech signal and
averaged noise.
Signal Preprocessing
As described in the preceding chapter, the background noise monitor is able to
discriminate betw een spee ch and noise. In very short spee ch p aus es e.g . between two
words, however, it changes immediately to non-speech, which is equal to noise.
Therefore a peak detection is required in front of the Noise Monitor.
The main task of the Peak Detector (PD) is to bridge the very short speech pauses during
a monolog so that this time constant has to be long. Furthermore, the speech bursts are
stored so that a sure speech detection is guaranteed. But if no speech is recognized the
noise low-pass LP 2 must be charged faster to the av erage noise lev el. In additi on, the
noise edges are to be smoothed. The refore two time constants are necessary. As the
peak detector is ve ry sensitive to spikes, the low-pass LP1 filters the inco ming signal
containing noise in a way that main spikes are eliminated. Due to the programmable time
constant it is possible to refuse high-energy sibilants and noise edges.
To compress the speech signals in their amplitudes and to ease the detection of speech,
the signals have t o be compand ed logarithmi cally. Hereby, the spee ch detector s hould
not be influenced by the s ystem noise w hich is always present but shoul d discriminate
between speech and background noise. The limitation of the logarithmic amplifier can be
programmed via the parameter LIM. LIM is related to the maximum PCM level. A signal
exceeding the limitation de fined by LIM is getting amplified logarithmic ally, while very
smooth system noi se bel ow is ne glecte d. It sh oul d be th e leve l o f the m ini mum sy stem
noise which is always existing; in the transmit path the noise generated by the telephone
circuitry itself and in receive direction the level of the first bit which is stable without any
speech signal at the receive path. Table 6 shows the parame ters for the s peech detect or.
Semiconductor Group3310.97
PSB 4860
Functional Description
Table 4Speech Detector Parameters
Parameter # of bytes RangeComment
LIM10 to 95 dBLimitation of log. amplifier
OFF10 to 95 dBLevel offset up to detected noise
PDS11 to 2000 msPeak decrement PD1 (speech)
PDN11 to 2000 msPeak decrement PD1 (noise)
LP111 to 2000 msTime constant LP1
LP2S12 to 250 sTime constant LP2 (speech)
LP2N11 to 2000 msTime constant LP2 (noise)
LP2L10 to 95 dBMaximum value of LP2
The input signal of the speech detector can be connected to either the input signal of the
echo suppression unit (as shown for SDX) or the output of the associated AGC (as
shown for SDR).
Semiconductor Group3410.97
PSB 4860
Functional Description
2.1.3.2Speech Comparators (SC)
The echo suppression un it has tw o ide ntic al s pee ch comp arato rs (SCAS, SCLS). Eac h
comparator can be programmed individually to accommodate the different system
characteristics of the acoust ic interface and the line interfa ce. As SCAS and SCLS are
identical, the following description holds for both SCAS and SCLS.
The SC has two input signals SX and SR, which map to microphone/loudsp eaker for
SCAS and line in/line out for SCLS.
In principle, the SC works according to the following equation:
ifSX > SR + V then switch state
Therefore, SCAS controls the switching to transmit state and SCLS controls the
switching to receive state. Switching is done only if SX exceeds SR by at least the
expected acoustic lev el enhancement V which is divid ed into two parts: G and GD. A
block diagram of the SC is shown in figure 17.
SX
SR
Log. Amp.
Log. Amp.Base GainGain ReservePeak Decrement
G
GDS
GDN
Peak Decrement
PDS
PDN
PDS
PDN
Figure 17 Speech Comparator - Block Diagram
At both inputs, logarithmic amplifiers compress the signal range. Hence after the
required signal processing for con trolling the acoustic echo, pure logarithmic levels on
both paths are compared.
The main task of the comparator is to control the echo. The internal coupling due to the
direct sound and mechanical resonances are covered by G. The external coupling,
mainly caused by the acoustic feedback, is controlled by GD/PD.
Semiconductor Group3510.97
PSB 4860
Functional Description
The base gain (G) correspo nds to th e terminal couplin gs of the complete telephone: G
is the measured or calculated level enhancement between both receive and transmit
inputs of the SC.
To control the acoustic feedback two parameters are necessary: GD represents the
actual reserve on the measured G. Together with the Peak Decrement (PD) it simulates
the echo behavior at the acoustic s ide: After speech has ended there is a short time
during which hard couplings through the mechanics and resonances and the direct echo
are present. Till the end of that time (∆
to G to prevent clipping caused by these internal couplings. Then, only the acoustic
feedback is present. This coup ling, however, is reduced by air atte nuation. For this in
general the longer the dela y, the smaller the echo being valid. This echo behavior is
featured by the decrement PD.
t) the level enhancement V must be at least equal
dB
GD*
PD*
GD
G
RX-Speech
PD
G
RX-noise
∆t
Figure 18 Speech Comparator - Interdependence of Parameters
t
According to figure 18 , a compromis e between the re serve GD and th e decrement PD
has to be made: a smaller reserve (GD) above the level enhancement G requires a
longer time to decrease (PD). It is easy to overshout the other side but the
intercommunication is harder because after the end of the speech, the level of the
estimated echo has to be exceeded. In contrary, with a higher reserve (GD*) it is harder
to overshout continuous speech or tones, but it enables a faster intercommunication
because of a stronger decrement (PD*).
Semiconductor Group3610.97
PSB 4860
Functional Description
Two pairs of coefficients, GDS/PDS when speech is detected, and GDN/PDN in case of
noise, offer a different echo handling for speech and non-speech.
With speech, even if very strong resonances are present, the performance will not be
worsened by the high GDS needed. Only when speech is detected, a high reserve
prevents clipping. A time period ET [ms] after speech end, the parameters of the
comparator are switched to the “noise” values. If both sets of the parameters are equal,
ET has no function.
Table 5Speech Comparator Parameters
Parameter # of bytes RangeComment
G1– 48 to + 48 dBBase Gain
GDS10 to 48 dBGain Reserve (Speech)
PDS10.025 to 6 dB/msPeak Decrement (Speech)
GDN10 to 48 dBGain Reserve (Noise)
PDN10.025 to 6 dB/msPeak Decrement (Noise)
ET10 to 992 msTime to Switch from speech to noise
parameters
2.1.3.3Attenuation Control
The attenuation control unit controls the attenuation stages GHX and GHR and performs
state switching. The programmable attenuation ATT is completely switched to GHX
(GHR) in receive state (transmit state). In idle state bot h GHX and GHR attenuate by
ATT/2.
In addition, a ttenuation is also influ enced by the auto matic ga in control stage s (AGCX,
AGCR).
State switching depends on the signals of one speech comparator and the
corresponding speec h detector. While each state is as sociated with the programmed
attenuation, the time is takes to reach th e steady-state atte nuation after a state switch
can be programmed (T
SW).
If the current state is either transmit or receive and n o speech on e ither side has been
detected for time T
then idle state is entered. To smoothen the transition, the
W
attenuation is incremented (decremented) by DS until the evenly di stribution ATT/2 for
both GHX and GHR is reached.
Table 6 shows the paramete rs for the attenua tion unit. Note that T
the current attenuation by the formula .
Semiconductor Group3710.97
T
sw
SWATT×=
is dependent on
SW
Table 6Attenuation Control Unit Parameters
Parameter # of bytes RangeComment
PSB 4860
Functional Description
TW116 ms to 4 sT
to return to idle state
W
ATT10 to 95 dBAttenuation for GHX and GHR
DS10.6 to 680 ms/dBDecay Speed (to idle state)
SW10.0052 to 10 ms/dB Decay Rate (used for T
SW
)
Note: In addition, attenua tion is also influenced by the Automa tic Gain Control stages
(AGCX, AGCR) in order to keep the total loop attenuation constant.
2.1.3.4Echo Suppression Status Output
The PSB 4860 can report the current state of the echo suppression unit to ease
optimization of the parameter set of the echo suppression unit. In this case the SPS
SPS
Furthermore the controller can read the current value of the SPS pins by reading register
SPSCTL.
2.1.3.5Loudhearing
The speakerphone unit can also be used for controlled loudhearing. If enabled in
loudhearing mode, the loudspeaker amplif ier of the PSB 4851 (ALS) is used instead of
GHR (figure 15) when appropriate to avoid oscillation. In order to enable this feature, the
PSB 4851 must be programmed to a llow ALS override. The ALS field within the AFE
control register AFECTL defines the value sent to the PSB 4851 if attenuation is
necessary (see specification of the PSB 4851).
2.1.3.6Automatic Gain Control
The echo suppression unit has two identical automatic gain control units (AGCX,
AGCR).
Semiconductor Group3810.97
PSB 4860
Functional Description
Operation of the AGC depends on a threshold level defined by the parameter COM
(value relative to the maximum PCM-value). The regulation speed is controlled by
SPEEDH for signal amplitudes above the threshold and SPEEDL for amplitudes below.
Usually SPEEDH will be chosen to be at least 10 times faster than SPEEDL. The bold
line in Figure 19 depicts th e steady-state output level of the AGC as a function of the
input level.
-10 dB-20 dB
AG_ATT
Example:
COM
AG_GAIN
AG_ATT
=
=
=
AGC input level
-30 dB
15 dB
20 dB
AG_GAIN
Figure 19 Echo Suppression Unit - Automatic Gain Control
max. PCM
-10 dB
-20 dB
COM
AGC
output
level
For reasons of physiological acceptance the AGC gain is automatically reduced in case
of continuous backg round noi se (e.g. by ventil ators). The reduc tion is programme d via
the NOlS parameter. When the noise level exceeds the threshold determined by NOIS,
the amplification will be reduced by the same amount the noise level is above the
threshold. The current gain/attenuation of the AGC can be read at any time (AG_CUR).
An additional low pass with time constant LP is provided to avoid an immediate response
of the AGC to very short signal bursts.
If SDX detects noise , AGCX is not working. In this case the last gain setting is used.
Regulation starts with this value as soon as SDX detects speech.
Likewise, if SDR detects noise, AGCR is not working. In this case the last gain setting is
used. Regulation starts with this value as soon as SDR detects speech. When the AGC
has been disabled the initial gain used immediately after enabling the AGC can be
programmed. Table 8 shows the parameters of the AGC.
Semiconductor Group3910.97
PSB 4860
Functional Description
Table 8Automatic Gain Control Parameters
Parameter # of Bytes RangeComment
AG_INIT1-95 dB to 95dBInitial AGC gain/attenuation
COM10 to – 95 dBCompare level rel. to max. PCM-value
AG_ATT10 to -95 dBAttenuation range
AG_GAIN10 to 95 dBGain range
AG_CUR1-95 dB to 95 dBCurrent gain/attenuation
SPEEDL10.25 to 62.5 dB/sChange rate for lower levels
SPEEDH10.25 to 62.5 dB/sChange rate for higher levels
NOIS10 to – 95 dBThreshold for AGC-reduction
by background noise
LP10.025 to 16 msAGC low pass time constant
Note: There are two sets of parameters, one for AGCX and one for AGCR.
Note: By setting AG_GAIN to 0 dB a limitation function can be realized with the AGC.
2.1.3.7Fixed Gain
Each signal path features an additional amplifier (LGAX, LGAR) that can be set to a fixed
gain. These amplifiers should be used for the basic amplification in order to avoid
saturation in the preceding stages. Table 9 shows the only parameter of this stage.
Table 9Fixed Gain Parameters
Parameter # of Bytes RangeComment
LGA1-12 dB to 12 dB always active
2.1.3.8Mode Control
Table 10 shows the registers used to determine the signal sources and the mode.
Table 10Speakerphone Control Registers
Register# of BitsNameComment
SCTL1ENSEcho suppression unit enable
SCTL1ENCEcho cancellation unit enable
SCTL1MDSpeakerphone or loudhearing mode
SCTL1AGXAGCX enable
Semiconductor Group4010.97
Table 10Speakerphone Control Registers
SCTL1AGRAGCR enable
SCTL1SDXSDX input tap
SCTL1SDRSDR input tap
AFECTL4ALSALS value for loudhearing
SSRC15I1Input signal 1 (microphone)
SSRC15I2Input signal 2 (microphone)
SSRC25I3Input signal 3 (line in)
SSRC25I4Input signal 4 (line in)
PSB 4860
Functional Description
Semiconductor Group4110.97
PSB 4860
Functional Description
2.1.4Line Echo Canceller
The PSB 4860 contains an adaptive line echo cancella tion unit for the cancellation of
near end echoes. The unit h as two modes: normal an d extended. In normal mod e, the
maximum echo length is 4 ms. This mode is alw ays available. In extended mode , the
maximum echo length is 24 ms. Extended mode cannot be used while the speech
encoder, the echo cancellation unit or slow playback is active.
The line echo cancellation unit is especially useful in front of the various detectors
(DTMF, CPT, etc.). A block diagram is shown in figure 20.
I
2
+
Σ
S
15
-
Adaptive
Filter
I
1
Figure 20 Line Echo Cancellation Unit - Block Diagram
The line echo cancelle r provides only one ou tgoing signal (S
signal would be identical with the input signal I
Input I
is usually connected to the l ine inpu t w hile inp ut I1 is connected to the outgoing
2
.
1
) as the other outgoing
15
signal.
In normal mode the adaption process can be controlled by three parameters: MIN, ATT
and MGN. Adaption takes only place if both of the following conditions hold:
I1MIN>
1.
I1I2–ATTMGN+–0>
2.
With the first con diti on ada ption to sma ll signal s can be avoid ed. The s econd condit ion
avoids adaption during double talk. The parameter ATT represents the echo loss
provided by external circuitry. The adaption stops if the power of the received signal (I2)
exceeds the power of the expected signal (I1-ATT) by more than the margin MGN.
Semiconductor Group4210.97
PSB 4860
Functional Description
Table 11 shows the registers associated with the line echo canceller.
Table 11Line Echo Cancellation Unit Registers
Register# of Bits Name CommentRelevant
Mode
LECCTL1ENLine echo canceller enableboth
LECCTL1MDLine echo canceller mode
LECCTL5I2Input signal selection for I
LECCTL5I1Input signal selection for I
LECLEV15MINMinimal power for signal I
2
1
1
LECATT15ATTExternally provided attenuation (I
to I2) normal
1
both
both
normal
LECMGN15MGNMargin for double talk detectionnormal
Semiconductor Group4310.97
PSB 4860
Functional Description
2.1.5DTMF Detector
Figure 21 shows a block dia gram of the D TMF det ector. The results of the detector are
available in the status register and a dedicated resul t register that can be re ad via the
serial control interface (SCI) by the external controller. All sixteen standard DTMF tones
are recognized.
I
1
DTMF
Recognition
SCI
Figure 21 DTMF Detector - Block Diagram
Table 12 to 14 show the associated registers.
Table 12DTMF Detector Control Register
Register# of BitsNameComment
DDCTL1ENDTMF detector enable
DDCTL5I1Input signal selection
As soon as a v ali d DTMF tone is reco gni zed, the status w ord a nd the DTMF tone c ode
are updated (table 13).
Table 13DTMF Detector Results
Register# of BitsNameComment
STATUS1DTVDTMF code valid
DDCTL5DTCDTMF tone code
DTV is set when a DTMF tone is recognized and reset when no DTMF tone is recognized
or the detector is disabled. The code for the DTMF tone is placed into the register
DDCTL. The registers DDTW and DDLEV hold parameters for detection (table 14).
Table 14DTMF Detector Parameters
Register# of BitsNameComment
DDTW15TWISTTwist for DTMF recognition
DDLEV6MINMinimum signal level to detect DTMF tones
Semiconductor Group4410.97
PSB 4860
Functional Description
2.1.6CNG Detector
The calling tone (C NG) detector c an detect the standard calling tone s of fax m achines
or modems. This helps to distinguish voice messages from data transfers. The result of
the detector is available in the status register that can be read via the serial control
interface (SCI) by the exte rnal con troller. The CNG d etector cons ists of tw o band-pas s
filters with fixed center frequency of 1100 Hz and 1300 Hz.
CNG Detector
I
1
1100 Hz 1300 Hz
SCI
Figure 22 CNG Detector - Block Diagram
Table 15 shows the available parameters.
Table 15CNG Detector Registers
Register# of BitsNameComment
CNGCTL1ENCNG detector enable
CNGCTL5I1Input signal selection
CNGLEV16MINMinimum signal level
CNGBT16TIMEMinimum time of signal burst
CNGRES16RESInput signal resolution
Both the programmed mini mum time and the mi nimum signal level m ust be exceeded
for a valid CNG tone. Furthermore the input signal resolution can be reduced by the RES
parameter. This can be useful in a noisy env ironment at low signal levels although the
accuracy of the detect ion decreases. As soo n as a valid tone is recognized, t he status
word of the PSB 4860 is updated. The status bits are defined as follows:
Table 16CNG Detector Result
Register# of BitsNameComment
STATUS1CNGFax/Modem calling tone detected
Semiconductor Group4510.97
PSB 4860
Functional Description
2.1.7Alert Tone Detector
The alert tone detector can detect the st andard alert tones (2130 Hz and 2750 Hz) for
caller id protocols. The results of the detector are available in the status register and the
dedicated regist er ATDCTL 0 that c an be re ad via the ser ial cont rol int erface (SCI) by the
external controller.
I
1
Detector
SCI
Figure 23 Alert Tone Detector - Block Diagram
Table 17Alert Tone Detector Registers
Register# of BitsNameComment
ATDCTL01ENAlert Tone Detector Enable
ATDCTL05I1Input signal selection
ATDCTL11MDDetection of dual tones or single tones
ATDCTL11DEVMaximum deviation (0.5% or 1.1%)
ATDCTL18MINMinimum signal level to detect alert tones
Alert Tone
As soon as a valid alert tone is recognized, the status word of the PSB 4860 and the code
for the detected combination of alert tones are updated (table 18).
Table 18Alert Tone Detector Results
Register# of BitsNameComment
STATUS1ATVAlert tone detected
ATDCTL02ATCAlert tone code
Semiconductor Group4610.97
PSB 4860
Functional Description
2.1.8CPT Detector
The selected signal is monitored continuously for a call progress tone. The CPT detector
consists of a band-pass and an optional timing checker (figure 24).
Band-pass
1
300-640 Hz
SCI (Status)I
Timing
Checker
Figure 24 CPT Detector - Block Diagram
The CPT detector can be used in two modes: raw and cooked. In raw mode, the
occurrence of a signal within the frequency range, time and energy limits is directly
reported. The timing checker is bypassed and therefore the PSB 4860 does not interpret
the length or interval of the signal.
In cooked mode, the number and duration of sig nal bursts are interpreted by th e t imin g
checker. A signal burst f ollowed by a gap is called a cycle. Cooked mode req uires a
minimum of two cycles. The CPT flag is set with the first burst after the programmed
number of cycles has been detected. The CPT flag remains set until the unit is disabled,
even if the conditions are not met anymore. In this mode the CPT is modelled as a
sequence of identical bursts separated by gaps with identical length. The PSB 4860 can
be programmed to accept a range for bo th the burst and the gap. It is also poss ible to
specify a maximum aberration of two consecutive bursts and gaps. Figure 25 shows the
parameters for a single cycle (burst and gap).
t
Bmin
Bmax
t
Gmin
t
Gmax
t
Figure 25 CPT Detector - Cooked Mode
The status bit is defined as follows:
Semiconductor Group4710.97
PSB 4860
Functional Description
Table 19CPT Detector Result
Register# of BitsNameComment
STATUS1CPTCP tone currently detected [340 Hz; 640 Hz]
CPT is not affected by reading the status word. It is automatically reset when the unit is
disabled. Table 20 shows the control register for the CPT detector.
Table 20CPT Detector Registers
Register# of Bits NameComment
CPTCTL1ENUnit enable
CPTCTL1MDMode (cooked, raw)
CPTCTL5I1Input signal selection
CPTMN8MINBMinimum time of a signal burst (t
CPTMN8MINGMinimum time of a signal gap (t
Gmin
CPTMX8MAXBMaximum time of a signal burst (t
CPTMX8MAXGMaximum time of a signal gap (t
Bmin
Bmax
Gmax
)
)
)
)
CPTDT8DIFBMa xim um diffe renc e betw een con sec uti ve burst s
CPTDT8DIFGMaxim um diffe renc e between consecutive gaps
CPTTR3NUMNumber of cycles (cooked mode), 0 (raw mode)
CPTTR8MINMinimum signal level to detect tones
CPTTR4SNMinimal signal-to-noise ratio
If any condition is violated during a sequen ce of c ycles t he timing checke r is reset and
restarts with the next valid burst.
Note: In cooked mode CPT is se t with the first burst after the pro grammed number of
cycles has been detected.
Note: The number of cycles must be set to zero in raw mode.
Semiconductor Group4810.97
PSB 4860
Functional Description
2.1.9Caller ID Decoder
The caller ID decoder i s basically a 1200 baud modem (FSK, demodu lation only). Th e
bit stream is formatted by a subsequent UART and the data is available in a data register
along with status information (figure 26).
I
1
FSK demod.
(Bellcore, V.23)
UART
SCI (Status, Data)
Figure 26 Caller ID Decoder - Block Diagram
The FSK demodulator supports two modes according to table 21. The appropriate mode
is detected automatically.
Table 21Caller ID Decoder Modes
ModeMark
(Hz)
Space
(Hz)
Comment
112002200Bellcore
213002100V.23
The CID decoder does not interpret the data received. Each byte received is placed into
the CIDCTL register (table 23). The status byte of the PSB 4860 is updated (table 22).
Table 22Caller ID Decoder Status
Register# of BitsNameComment
STATUS1CIACID byte received
STATUS1CDCarrier Detected
CIA and CD are cleared when the unit is disabled. In addition, CIA is cleared when
CIDCTL0 is read.
Table 23Caller ID Decoder Registers
Register# of Bits NameComment
CIDCTL01ENUnit enable
CIDCTL05I1Input signal selection
CIDCTL08DATALast CID data byte received
Semiconductor Group4910.97
PSB 4860
Functional Description
Table 23Caller ID Decoder Registers
Register# of Bits NameComment
CIDCTL15NMSSNumber of mark/space sequences necessary for
successful detection of carrier detect
CIDCTL16NMBNumber of mark bits necessary before space of first
byte after carrier detect
CIDCTL15MINMinimum signal level for CID detection
When the CID unit is enabl ed, it first waits for a channel sei zure signal consisti ng of a
series of alternating space and mark signals. The number of spaces and marks that have
to be received without errors before the PSB 4860 reports a carrier det ect by setting
status bit CD can be programmed.
Channel seizure must be followed by at least 16 continuous mark signals. The first space
signal detected is then regarded as the start bit of the first message byte.
The interpretation of the data, including message type, length and checksum is
completely left to the controller. The CID unit should be disabled as soon as the complete
information has been received as it cannot detect the end of the transmission by itself.
Note: Some caller ID mechanis m may require additional external compon ents for DC
decoupling. These tasks must be handled by the controller.
Note: The controller is responsible for selecting and storing parts of the CID as needed.
Semiconductor Group5010.97
PSB 4860
Functional Description
2.1.10DTMF Generator
The DTMF generator can g enerate single or dual to nes with programmable frequency
and gain. This unit is pr imari ly used to gen erate the co mmon D TMF ton es but ca n also
be used for signalling or other user defined tones. A block diagram is shown in figure 27.
f
1
generator
f
2
generator
gain1
gain2
att1
att2
S
9
S
10
Figure 27 DTMF Generator - Block Diagram
Both generators and amplifiers are identical. There are two modes for programming the
generators, cooked mode and raw mode. In cooked mode, the standard DTMF
frequencies are generated by programming a single 4 bit code. In raw mode, the
frequency of each generator/amplifier can be programmed i ndividually by a separate
register. The unit h as two outputs which provide the same signal but with indivi dually
programmable attenuation. Table 24 shows the parameters of this unit.
Table 24DTMF Generator Registers
Register# of BitsNameComment
DGCTL1ENEnable for generators
DGCTL1MDMode (cooked/raw)
DGCTL4DTCDTMF code (cooked mode)
DGF115FRQ1Frequency of generator 1
DGF215FRQ2Frequency of generator 2
DGL7LEV1Level of signal for generator 1
DGL7LEV2Level of signal for generator 2
DGATT8ATT1Attenuation of S
DGATT8ATT2Attenuation of S
9
10
Note: DGF1 and DGF2 are undefined when cooked mode is used and must not be
written.
Semiconductor Group5110.97
2.1.11Speech Coder
PSB 4860
Functional Description
The speech coder (figure 28) has two input signals I
and I2. The first signal (I1) is fed to
1
the coder while the second signal (I2 ) is used as a reference signal fo r voice co ntrolled
recording. The si gnal I
can be coded by eith er a High Quality coder or a Long Play
1
coder.
I
1
MIN
I
2
LP
HQ
10300 bit/s
Memory
LP
3300 bit/s
Figure 28 Speech Coder - Block Diagram
In High Quality the ou tput data stream run s at a fixed rate of 1 0300 bit/s and p rovides
excellent speech qual ity. In Long Pla y mode, the ou tput data strea m is further reduc ed
to an average of 3300 bit/s while still maintaining good quality.
Data is written starting at the current file pointer and the file pointer is advanced as
needed. In case of any m emory e rror (e. g. mem ory full ) a fil e erro r is i ndicate d and the
coder is disabled. The controller must subsequently close the file.
The coder can be switched on the fly. However, it may take up to 60 ms until the switch
is executed. The controller must therefore wait for at least this time until issuing another
command that relies on the mode switch. No audio data is lost during switching.
The signal I
is first filtered by a low pass LP1 with programmable time constant and then
2
compared to a reference level MIN. If the filtered signal exceeds MIN, then the status bit
SD (table 25) is set immediately. If the filtered signal has been smaller than MIN for a
programmable time TIME then the status bit SD is reset.
The coder can be enabled in permanent mode or in voice recognition mode. In
permanent mode, the coder starts immediately and compresses all input data
continuously. The current state of the status bit SD does not affect the coder.
In voice recognition mode, the coder is automatically started on the first transition of the
status bit from 0 to 1. Once the coder has started it remains active until disabled.
Table 25Speech Coder Status
Register# of BitsNameComment
STATUS1SD Speech detected
Semiconductor Group5210.97
Functional Description
The operation of the speech coder is defined according to table 26.
Table 26Speech Coder Registers
Register# of BitsNameComment
SCCTL1ENEnable speech coder
SCCTL1HQHigh quality mode
SCCTL1VCVoice controlled recording
SCCTL5I1Input signal 1 selection
SCCTL5I2Input signal 2 selection
SCCT28MINMinimal signal level for speech detection
SCCT28TIMEMinimum time for reset of SD
SCCT38LPTime constant for low-pass
PSB 4860
Note: The peak data rate in LP mode is 4800 bit/s.
Note: Both HQ and LP mode will not produce identical bit streams after a coding/
decoding cycle.
Semiconductor Group5310.97
PSB 4860
Functional Description
2.1.12Speech Decoder
The speech decoder (figure 29) decompresses the data previously coded by the speech
coder unit and delivers a standard 128 kbit/s data stream.
HQ
10300 bit/s
Memory
LP
3300 bit/s
S
13
Figure 29 Speech Decoder - Block Diagram
The decoder supports fast (1.5 and 2.0 times) and slow (0.5 times) motion independent
of the selected quality. Th e decoder requests input data as n eeded at a variable rate.
Table 27 shows the signal and mode selection for the speech decoder.
Table 27Speech Decoder Registers
Register# of BitsNameComment
SDCTL1ENEnable speech decoder
SDCTL2SPEEDSelection of playback speed
Data reading starts at the l ocation of the current file po inter. The file p ointer is upda ted
during speech decoding. If the end of the file is reached, the deco der is automatically
disabled. The PSB 4860 automatically resets SDCTL:EN at this point.
Semiconductor Group5410.97
PSB 4860
Functional Description
2.1.13Analog Front End Interface
There are two identical interfaces at the analog side (to PSB 4851) as shown in figure 30.
Channel 2
S
4
IG4
I
1
I
2
I
3
S
3
S
line out
line inIG1
2
Channel 1
IG2
HPIG3HP
I
1
I
2
I
3
S
1
loudspeaker
microphone
Figure 30 Analog Front End Interface - Block Diagram
For each signal a n ampl ifier is pr ovided for le vel a djustme nt. The in coming signa ls ca n
be passed through an optional high-pass (HP). This high-pas s (f
=20 Hz) is useful for
g
blocking DC offsets and should be enabled by default. Furthermore , up to thre e sig nals
can be mixed in order to generate the outgoing signals (S
). Table 28 shows the
2,S4
associated registers.
Table 28Analog Front End Interface Registers
Register# of BitsNameComment
IFG116IG1Gain for IG1
IFG216IG2Gain for IG2
IFS11HPHigh-pass for S
1
IFS15I1Input signal 1 for IG2
IFS15I2Input signal 2 for IG2
IFS15I3Input signal 3 for IG2
IFG316IG3Gain for IG3
IFG416IG4Gain for IG4
IFS21HPHigh-pass for S
3
IFS25I1Input signal 1 for IG4
IFS25I2Input signal 2 for IG4
IFS25I3Input signal 3 for IG4
Semiconductor Group5510.97
PSB 4860
Functional Description
2.1.14Digital Interface
There are two almost ident ical interfaces at the digital side as shown in fi gure 31. The
only difference between these two inte rfaces is that only channel 1 sup ports the SSDI
mode.
Each outgoing signal c an b e the s um of tw o sign als w ith no a tten uation a nd one sig nal
with programmable attenu ation (ATT). The attenuator can be u sed for artificial echo if
there is none externally provided (e.g. ISDN application). Each input can be passed
through an optional high-pass (HP). The associated registers are shown in table 29.
Table 29Digital Interface Registers
Register# of BitsNameComment
IFS35I1Input signal 1 for S
IFS35I2Input signal 2 for S
IFS35I3Input signal 3 for S
IFS31HPHigh-pass for S
IFS45I1Input signal 1 for S
IFS45I2Input signal 2 for S
IFS45I3Input signal 3 for S
IFS41HPHigh-pass for S
Semiconductor Group5610.97
6
6
6
5
8
8
8
7
PSB 4860
Functional Description
Table 29Digital Interface Registers
Register# of BitsNameComment
IFG58ATT1Attenuation for input signal I3 (Channel 1)
IFG58ATT2Attenuation for input signal I3 (Channel 2)
Semiconductor Group5710.97
PSB 4860
Functional Description
2.1.15Universal Attenuator
The PSB 4860 contains an universal attenuator that can be connected to any signal (e.g.
for sidetone gain in ISDN applications).
Table 30Universal Attenuator Registers
Register# of BitsNameComment
UA8ATTAttenuation for UA
UA5I1Input signal for UA
S
14
Semiconductor Group5810.97
PSB 4860
Functional Description
2.1.16Automatic Gain Control Unit
In addition to the universal attenuator with programmable but fixe d gain the PSB 4860
contains an amplifier with auto matic gain control (AGC). The AGC is preceeded by a
signal summation point for two input signals. One of the input signals can be attenuated.
I
1
I
2
ATT
AGC
S
16
S
17
Figure 33 Automatic Gain Control Unit - Block Diagram
Furthermore the signal after the summation point is available. Besides providing a
general signal summation (S
provides the input signal for the speech coder. In this case S
not used) this signal is esp ecially useful if t he AGC unit
16
can be used as a
17
reference signal for voice controlled recording.
The operation of the AGC is similar to AGCX (ACCR) of the speakerphone. The
differences are as follows:
• No NOIS parameter
• Separate enable/disable control
• Slightly different coefficient format
Furthermore the AGC contains a co mparator that starts and stops the gain regulation.
The signal after the summation point (S17) is filtered by a peak detector with time
constant DEC for decay. Then the signal is compared to a programmable limit LIM.
Regulation takes only place when the filtered signal exceeds the limit.
Table 31 shows the associated registers.
Table 31Automatic Gain Control Registers
Register# of BitsNameComment
AGCCTL1ENEnable
AGCCTL5I1Input signal 1 for AGC
AGCCTL5I2Input signal 2 for AGC
AGCATT15ATTAttenuation for I
2
AGC18AG_INITInitial AGC gain/attenuation
AGC18COMCompare level rel. to max. PCM-value
Semiconductor Group5910.97
Functional Description
Table 31Automatic Gain Control Registers
Register# of BitsNameComment
AGC28SPEEDLChange rate for lower levels
AGC28SPEEDHChange rate for higher level
AGC38AG_ATTAttenuation range
AGC37AG_GAI NGain range
AGC47DECPeak detector time constant
AGC48LIMComparator minimal signal level
AGC57LPAGC low pass time constant
PSB 4860
Semiconductor Group6010.97
PSB 4860
Functional Description
2.1.17Equalizer
The PSB 4860 also provides an equalizer that can be inserted into any signal path. The
main application for t he e qua liz er is the adaption to the frequenc y c hara cte ri sti cs of the
microphone, transducer or loudspeaker.
The equalizer consists of an IIR filter followed by an FIR filter as shown in figure 34.
-1
I
A1A2A9
z
-1
z
D1D2D17
-1
z
-1
z
-1
z
IIR
S
18
FIR
C2
-1
z
B2B9
C1
-1
z
-1
z
-1
z
Figure 34 Equalizer - Block Diagram
The coefficients A
C
belong to the FIR filter. Table 32 shows the registers associated with the equalizer.
2
, B2-B9 and C1 belong to the IIR filter, the coefficients D1-D17 and
1-A9
Table 32Equalizer Registers
Register# of BitsNameComment
FCFCTL1ENEnable
FCFCTL5IInput signal for equalizer
FCFCTL6ADRFilter coefficient address
FCFCOF16Filter coefficient data
Semiconductor Group6110.97
PSB 4860
Functional Description
Due to the multitude of coefficie nts the use s an indirec t address ing sche me for reading
or writing an individual coefficient. The ad dress of the coeffic ient is given by ADR and
the actual value is read or written to register FCFCOF.
In order to ease programming the PSB 4860 automatically increments the address ADR
after each access to FCFCOF.
Note: Any access to an out-of-range address automatically resets FCFCTL:ADR.
Semiconductor Group6210.97
PSB 4860
Functional Description
2.2Memory Management
Memory Management
Memory Management - General
This section describes the mem ory management provided by the PSB 4860. As figure
35 shows, three units can access the exte rnal memory. During recording, the speech
coder can write compressed spee ch data into the external memory. For playba ck, the
speech decoder reads compressed speech data from external memory. In addition, the
microcontroller can directly access the memory by the SCI interface.
Speech Decoder
MemorySCI
Speech Coder
Figure 35 Memory Management - Data Flow
The memory is organize d as a f ile sy stem. For e ach me mory spa ce (R/W-m emory an d
voice prompt memory) the PSB 4860 maintains a directory with 255 file descriptors
(figure 36).
file descriptor 1
file descriptor n
file descriptor 255
file descriptor (R/W)directory
length (0-65535)
user data (1 6 bits)
RTC1 (16 bits)
RTC2 (16 bits)
Figure 36 Memory Management - Directory Structure
The directories must be created after each power failure for volatile R/W-memory. All file
descriptors are cleared (all words zero). For non-volatile memory, the directories have to
Semiconductor Group6310.97
PSB 4860
Functional Description
be created only once. If the directories already exist, the memory has just to be activated
after a reset. The file descriptors are not changed in this case.
All commands that access the other fields or involve a write access must not be used in
voice prompt memory space.
2.2.1File Definition and Access
A file is a linear sequence of units and can be accessed in two modes: binary and audio.
In binary mode, a unit is a w ord. In audio mode, a uni t is a variable number of words
representing 30 ms of uncompressed speech. A file can contain at most 65 535 units.
Figure 37 shows an audio file containi ng 100 aud io units. The len gth of the message is
therefore 3 s.
3 s
Hi Jack, this is Tom. Please call me back tomorrow.
099
Figure 37 Audio File Organization - Example
Figure 38 shows a binary file of 11 words containing a phonebook (with only two entries).
There is one special file i n the voice prompt directory (referenced by file number 255)
which is intended for a large number of phrases and hence has a different
organization.This file exists only in the directory for the voice prompt memory. It consists
of up to 2048 phrases of arbitrary individual length. The actual number of units within an
individual phrase is determined during creation and cannot be altered afterwards.
Phrases can be combined in any sequence without intermediate noise or gaps.
Semiconductor Group6410.97
PSB 4860
Functional Description
Figure 39 shows a phrase file containing a total of five phrases.
one two you have messag es leftfriday
014
Figure 39 Phrase File Organization - Example
Before an access to a file can take place, the file must be opened with the following
information:
1. memory space
2. file number
3. access mode
These parameters remain effect ive until the next ope n comma nd is giv en or, in c ase of
the file pointer, until a file access. All other files are closed and cannot be accessed. The
file with file number 0 is not a physical file. Opening this file closes all physical files.
The PSB 4860 provides four regi sters for file access and two bits within the STATUS
register. Table 33 shows these registers.
Table 33Memory Management Registers
Register# of BitsComme nt
FCMD16Command to execute
FCTL16Access mode and file number
FDATA16Data transfer and additional parameters
FPTR16 (11)File pointer (phrase selector)
STATUS16Busy and Error indication
The status register contains two flags (table 34) to indicate if currently a file command is
under execution and if the las t file com man d terminated without error. A new command
must not be written to FCMD while the last one is still running (STATUS: BSY=1). The
only command that can be aborted is Compress File.
Table 34Memory Management Status
Register# of BitsNameComment
STATUS1BSYFile command or decoder/encoder still running
STATUS1ERRFile command completed/aborted with error
Semiconductor Group6510.97
PSB 4860
Functional Description
Writing to FCMD also resets the error bit in the status register.
Table 35 shows the parameters de fining the acces s mode and the acc ess location. All
parameters can only be written when no file command is currently running. They become
effective after the completi on of an open comma nd. If another unit (e.g. speech coder)
accesses the file, the file pointer is updated automatic ally. Therefore the con troller can
monitor the progress of recording or playing by reading the file pointer.
Table 35Memory Management Parameters
Register# of BitsNameComment
FCTL1MSMemory space (R/W or voice prompt)
FCTL1MDAccess mode (audio or binary)
FCTL1TSWrite timestamp (file open only)
FCTL8FNOFile number (active file)
FPTR16File pointer or phrase selector
Commands are written to the FCMD register. The busy bit in the STATUS register is set
within 125µ s. The command may start execution after a del ay, however (see section
2.2.5). Some commands require additional parameters which are written prior to the
command into the specified registers. Data transfer is done by the register FDATA (both
reading and writing).
2.2.2User Data Word
The user data word consists of 12 bits that can be read or written by the user, two bits
(R) that are reserved for future use and two read-only bits (D,M) which indicate the status
of a file.
150
DMRRUser Definable
If D is set, the file is m arked for deletion and should not be used any more. This bit is
maintained by the PSB 4860 for housekeeping.
Semiconductor Group6610.97
PSB 4860
Functional Description
2.2.3High Level Memory Management Commands
This section describes each of the high level memory management commands in detail.
These commands are sufficient for normal operation of an answering machine. In
addition, there are four low level c ommand s (sectio n 2.2. 4). Thes e comma nds ar e only
required for special tasks like in-system reprogramming of the voice prompt area.
Memory Management - Commands
2.2.3.1Initialize
This command creates a directory, sets the external memory configuration and delivers
the size of usable memory in 1 kByte blocks. Furthermore the voice prompt memory
space is scanned for a valid directory. The PSB 4860 can either create an empty
directory from scratch or leave the first n files o f an existing directory u ntouched while
deleting the remaining files (ARAM/DRAM only). This option is useful if due to an
unexpected event (e.g. power loss during recording) some data is corrupted. In that case
vital system information can still be recovered if it has been stored in the first files.
Table 36Initialize Memory Parameters
Register# of BitsNameComment
FCMD5CMDInitialize command code
FCMD1INConfirmation for Initialization
FCTL8FNO0: delete no file
1: delete all files
n: delete starting with file n
CCTL2MTType of R/W memory (DRAM, Flash)
CCTL1MQQuality of R/W memory (Audio, Normal)
CCTL1MVScan for voice prompt directory
Table 37Initialize Memory Results
Register# of BitsNameComment
FDATA16Number of usable 1kByte blocks in R/W memory
Possible Errors:
• no R/W memory found
• more than 59 bad blocks (flash and ARAM)
• voice prompt directory requested, but not detected
Note: This command must be given only once for flash devices.
Semiconductor Group6710.97
PSB 4860
Functional Description
2.2.3.2Activate
This command activates an ex isting directory, sets the external memory configuration
and delivers the size of usable memory in 1 kByte blocks. Furthermore the voice prompt
memory space is scanned for a valid directory. Upon activation the PSB 4860 checks (in
case of ARAM/DRAM only) the consistency of the directory in R/W memory space. It
returns the first file that contains corrupted data (if any). If corrupted data is detected an
initialization should be performed with the same file number as an input parameter.
Table 38Activate Memory Parameters
Register# of BitsNameComment
FCMD5CMDActivate command code
CCTL2MTType of R/W memory (DRAM, Flash)
CCTL1MQQuality of R/W memory (Audio, Normal)
CCTL1MVVoice prompt directory available
Table 39Activate Memory Results
Register# of BitsNameComment
FDATA16Number of usable 1 kByte blocks in R/W memory
FCTL8FNOn: number of first corrupted file
Possible error conditions:
• no memory connected
• no directory found
• device ID wrong (flash only)
• corrupted files found (see FCTL:FNO)
• directory corrupted
This command can have three types of result as shown in table 40.
Table 40Activate Memory Result Interpretation
ResultSTATUS:
ERR
no error00Command successful, memory activated.
soft error1nThe first n-1 files are O.K. The memory is activated.
hard error 11The memory is not activated due to a hard error.
Semiconductor Group6810.97
FCTL:
FNO
Comment
PSB 4860
Functional Description
2.2.3.3Open File
A specific file is opened for subsequent accesses with the specified access mode.
Opening a new file automatically closes the currently open file and clears the file pointer.
Opening file number 0 can be used to close all physic al files. If the TS flag is set, the
current content of RTC1 and RTC2 is written to the appropriate fields of the file descriptor
in order to provide a timestamp.
Table 41Open File Parameters
Register# of BitsNameComment
FCMD5CMDOpen command code
FCTL1MSMemory space (R/W, voice prompt)
FCTL1MDAccess mode (audio or binary)
FCTL1TSWrite timestamp
FCTL8FNOFile number <fno>
Possible error conditions:
• selected file marked for deletion, but not yet deleted by garbage collection
• memory space invalid
• new file selected, but memory full
• <fno> exceeds number of prompts (in voice prompt space only)
• wrong access mode selected for existing file
Note: In case of flash memory existing ones in the entries RTC1/RTC2 of the file
descriptor cannot be altered. Therefore TS should be set only once during the
lifetime of a file.
2.2.3.4Open Next Free File
The next free file is ope ned for subsequent write accesses wi th the specified access
mode. The search starts at the spec ified file number. If the TS flag is set, the current
content of RTC1 and RTC2 is written to the appropriate fields of the file descriptor in
order to provide a timestamp. If a free file has been found, the file is opened and the file
number is returned in FCTL:FNO. Otherwise an error is reported.
Table 42Open Next Free File Parameters
Register# of BitsNameComment
FCMD5CMDOpen Next Free File command code
FCTL1MDAccess mode (audio or binary)
Semiconductor Group6910.97
Table 42Open Next Free File Parameters
Register# of BitsNameComment
FCTL1TSWrite timestamp
FCTL8FNOStarting point (>0)
:
Table 43Open Next Free File Results
Register# of BitsNameComment
FCTL8FNOFile number
Possible error conditions:
• no unused file found
• memory full
PSB 4860
Functional Description
Note: In case of flash memory existing ones cannot be altered. Therefore TS should be
set only once during the lifetime of a file.
Note: R/W-memory must be selected. Otherwise the result is unpredictable.
2.2.3.5Seek
The file pointer of the curren tly open ed file i s set to the spe cified p osition . If t he current
file is the phrase file the PSB 4860 starts the speech decoder immediately after the seek
is finished. This is done by simply enabling the decoder. All other settings of the decoder
remain unaffected. The BSY bit is first set during the file command. It is then reset for a
short period until the speech decoder is enabled internally. It is then set again while the
decoder is running and finally reset when the phrase is finished.
Table 44Seek Parameters
Register# of BitsNameComment
All units starting wit h the unit a ddressed by the file pointe r are removed from the file. If
all units are deleted th e file is marked for deletion (see user data word). Howev er, the
associated file descriptor and memory space are released only after a subsequent
garbage collection.
Table 45Cut File Parameters
Register# of BitsNameComment
FCMD5CMDCut command code
FPTR16Position of first unit to delete
Possible error conditions:
• file pointer out of range
• voice prompt memory selected
2.2.3.7Compress File
An audio file that has been recorde d in HQ mode ca n be recode d using LP mod e. This
reduces the file si ze to approxim ately one third of the ori gina l size. Th e spee ch quality ,
however, is somewhat lo wer compared to a si gnal that has been recorded in LP mode
in the first place. This comm and can b e aborte d at any t ime an d resumed l ater w ithout
loss of information. Prior to this command all files must be closed. Table 46 shows the
parameters for this command.
.
Table 46Compress File Parameters
Register# of BitsNameComment
FCMD5CMDCompress command code
FCTL8FNOFile number <fno>
Possible error conditions:
• <fno> invalid
• another file currently open
• binary file selected
Semiconductor Group7110.97
PSB 4860
Functional Description
2.2.3.8Memory Status
This command returns the number of available 1 kB blocks in R/W memory space.
Table 47Memory Status Parameters
Register# of BitsNameComment
FCMD5CMDMemory status code
Table 48Memory Status Results
Register# of BitsNameComment
FDATA16FREENumber of free blocks
Possible error conditions:
• file open
2.2.3.9Garbage Collection
This command initiates a garbage collection. Until a garbage collection files that are
marked for deletion stil l occupy the associ ated file d escriptor and memory sp ace. After
the garbage collection these file descriptors and the associated memory space are
available again. This command can optionally remap the directory. In this mode the
remaining file descriptors are remapped to form a contiguous block starting with file
number 1. The original order is preserved. This command requires that all files are
closed, i.e. file 0 is opened. Inde pendently of th e selected di rectory only th e read/write
directory is used.
Table 49Garbage Collection Parameters
Register# of BitsNameComment
By this command the length, user data word and RTC1/RTC2 of a file descriptor can be
read. The user data word can also be written. The file or the other entries of the file
descriptor are not affected by this command.
Semiconductor Group7210.97
PSB 4860
Functional Description
Table 50Access File Descriptor Parameters
Register# of BitsNameComment
FCMD5CMDRead Access or Write Access command code
FDATA16User data (write access only)
Table 51Access File Descriptor Results
Register# of BitsNameComment
FDATA16Content of selected entry (read access only)
Possible error conditions:
• none
Note: In case of flash memory bits already set to 1 cannot be altered.
Note: Do not use this command with the phrase file (fno = 255).
2.2.3.11Read Data
This command can be u sed in binary access mode onl y. A single word is read at th e
position given by the file pointer. The file pointer can be set by the Seek command. The
file pointer is advanced by one word automatically.
Table 52Read Data Parameters
Register# of BitsNameComment
FCMD5CMDRead Data Command Code
Table 53Read Data Results
Register# of BitsNameComment
FDATA16Data word
Possible error conditions:
• file pointer out of range
• phrase file selected
• audio file selected
Semiconductor Group7310.97
PSB 4860
Functional Description
2.2.3.12Write Data
This commands can be used in binary access mode only. A single word is written at the
position of the file pointer. The file pointer is advanced by one word automatically. Note,
that for FLASH memory only zeroes can be overw ritte n by ones . This restric tio n occ urs
only if an already used value within an existing file is to be overwritten.
Table 54Write Data Parameters
Register# of BitsNameComment
FCMD5CMDAccess Mode Command Code (including mode)
FDATA16Data word
Possible error conditions:
• file pointer out of range (for existing files only)
• voice prompt memory selected
• memory full
• audio file selected
Semiconductor Group7410.97
PSB 4860
Functional Description
2.2.4Low Level Memory Management Commands
These commands allo w the direct access of any l ocation (single word) of th e external
memory. Additionally it is possible to erase any bloc k in case of a flash device . These
commands should not be us ed during normal operation as they may interfere with the
file system. No file must be open when one of these commands is given.
The primary use of these commands is the in-system programming of a flash device with
voice prompts. Please refer to the appropriate Application Notes.
2.2.4.1Set Address
This command sets the 24 bit add ress pointer APTR. Only the address bi ts A
set, the address bits A
are automatically cleared.
0-A7
8-A23
are
Table 55Set Address Parameters
Register# of BitsNameComment
FCMD5CMDSet Address command code
FDATA16ADRAddress bits A
of address pointer APTR
8-A23
Possible error conditions:
• file open
2.2.4.2DMA Read
This command reads a single word addressed by APTR. After the read access APTR is
automatically incremented by one. Table 56 shows the parameters for this command.
Table 56DMA Read Parameters
Register# of BitsNameComment
FCMD5CMDDMA Read command code
Table 57DMA Read Results
Register# of BitsNameComment
FDATA16DATAData read from address APTR.
Possible error conditions:
• file open
Semiconductor Group7510.97
PSB 4860
Functional Description
2.2.4.3DMA Write
This command writes a single word to the locat ion add ressed by APTR. After the write
access APTR is automaticall y increm ent ed by one. Tabl e 58 shows the param eters fo r
this command.
Table 58DMA Write Parameters
Register# of BitsNameComment
FCMD5CMDDMA Write command code
FDATA16DATAData to be written to APTR
Possible error conditions:
• file open
Note: If flash memory is connected the actual write is only performed when the last word
within a page is writte n. U nti l th en t he data is merely buffered in the flas h d evi ce.
Please check the flash memory data sheets on page size.
2.2.4.4Block Erase
This command erases the physical block which includes the address given by APTR.
The actual amount of memory erased by this command depends on the block size of the
flash device. Table 59 shows the parameters for this command.
Table 59Block Erase Parameters
Register# of BitsNameComment
FCMD5CMDBlock Erase command code
Possible error conditions:
• file open
• ARAM/DRAM configured
Semiconductor Group7610.97
PSB 4860
Functional Description
2.2.5Execution Time
The execution time of the file commands is determined by four factors:
1. Internal state of the PSB 4860
2. Memory configu ratio n
3. Memory state
4. Individual characteristics of the memory devices
Therefore there is no genera l formula for an exac t calculation of the execution time for
file commands. For ARAM/DRAM items three and four are not significant as the memory
access timing is al ways fixed and no additional delay i s incurred for erasing memory
blocks. However, the amount of memory ha s significant impact on the initialization in
case of ARAM and flash.
For flash devices the particular location of a write access in combination with the internal
organization of th e memory device may result in a block erase and subsequent write
accesses in order to copy data. In this ca se the indiv idual erase and writ e timing of the
attached devices also prolongs the execution time.
The first factor, the internal state of the PSB 4860, can influence all file commands
regardless of the memory type attached. In general the PSB 4860 may delay any file
command by up to 30 ms. However, it is possible to skip this delay if the following
conditions hold:
1. The command is not
initialize/activate
2. Neither the DTMF detector nor the speech coder nor the speech decoder are running
If neither condition is violated then the PSB 4860 can be forced to start command
execution immediat ely. This is done by setti ng the EIE bit in the FCMD register along
with the command code.
Table 60 gives an indication of the execution time for two typical memory configurations.
Table 60Execution Times
CommandARAM (4 MBit) KM29LV040
Initialize40 s
1)
<11 s
Activate< 10 ms3 s
Open File /Open Next Free File<10 ms<26 ms
Seek (within 4 MBit File)<0.5 s<0.5 s
Seek (wit hin phrase f ile)<1 ms<1 ms
Cut File<5 ms<5 ms
Compress File#units * 30 ms#units * 30 ms
Access File Descriptor<10 ms<10 ms
Semiconductor Group7710.97
PSB 4860
Functional Description
Table 60Execution Times
CommandARAM (4 MBit) KM29LV040
Memory Status<10 ms<10 ms
Read/Write Data<10 ms<10 ms
Garbage Collection<20 ms3 s
1)
less than 20 ms for DRAM
2.2.6Special Notes on File Commands
1. No MMU comma nds must be inserted be tween opening a file an d writing data to it,
either by writing data to a binary file or by enabling the coder for audio files.
Therefore reading or writing the file descriptor is only allowed after all data writing has
happened.
2. If an audio file has been opened for replay, a Write File Descriptor Command must be
followed by a Seek command before the decoder can be enabled.
Semiconductor Group7810.97
PSB 4860
Functional Description
2.3Miscellaneous
Miscellaneous
Miscellaneous
2.3.1Real Time Clock
The PSB 4860 supplies a real time c lock which maintains time with a resolution of a
second and a range of up to a year. There are two registe rs which contain the current
time and date (table 61).
Table 61Real Time Clock Registers
Register# of BitsNameComment
The real time clock maintains time during normal mode and power down mode only if the
auxiliary oscillator OSC is running and the RTC is enabled.
Note: Writing out-of-range values to RTC1 and RTC2 results in undefine d operation of
the RTC
2.3.2SPS Control Register
The two SPS outputs (SPS
, SPS1) can be used as either general purpose outputs,
0
speakerphone status outputs, extended address outputs for Voice Prompt EPROM or as
status register outputs. Table 62 shows the associated register.
Table 62SPS Registers
SPSCTL1SP0Output Value of SPS
SPSCTL1SP1Output Value of SPS
0
1
SPSCTL3MODEMode of Operation
SPSCTL4POSPosition for status register window
When used as status register outputs, the status register bit at position POS appears at
SPS
and the bit at position POS+1 appears at SPS1. This mode of operation can be
0
used for debugging purposes or direct polling of status register bits.
2.3.3Reset and Power Down Mode
The PSB 4860 can be in either reset mode , power down mode or active mode. During
reset the PSB 4860 clears the hardwa re configurat ion registers an d st ops both in ternal
Semiconductor Group7910.97
PSB 4860
Functional Description
and external activity. The address lines MA0-MA15 provide a weak low until they are
actually used as address lines (strong outputs) or auxiliary port pins (I/O). In reset mode
the hardware configuration registe rs can be rea d and written. Wit h the first access to a
read/write register the PSB 4860 enters active mode. In this mode the main oscillator is
running and normal o perat ion tak es pl ace. By setting the po we r do wn bit (PD) the PSB
4860 can be brought to power down mode.
Table 63Power Down Bit
Register# of BitsNameComment
CCTL1PDpower down mode
In power down mode the main oscillator is stopped and, depending on
HWCONFIG2:PPM), the memory c ontro l lin es are released (weak high). Depen din g on
the configuration (ARAM/DRAM, APP) the PSB 4860 may still generate external activity
(e.g. refresh cycles). The PSB 4860 enters active mode again upon an access to a read/
write register. Figure 40 shows a state chart of the modes of the PSB 4860.
Reset
Mode
R/W reg. access
RST=1RST=1
Active
Mode
CCTL.PD=1
Power Down
Mode
R/W reg. access
Figure 40 Operation Modes - State Chart
2.3.4Interrupt
The PSB 4860 can generate an interrupt to inform the host of an update of the STATUS
register according to table 64. An interrupt mask register (INTM) can be used to disable
or enable the interrupting capability of each bit of the STATUS register except ABT
individually.
Semiconductor Group8010.97
Table 64Interrupt Source Summary
PSB 4860
Functional Description
STATUS
(old)
RDY=0RDY=1Command completedCommand issued
CIA=0CIA=1New Caller ID byte availableCIDCTL0 read
CD=0CD=1Carrier detectedCarrier lost
CD=1CD=0Carrier lostCarrier detected
CPT=0CPT=1Call progress tone detectedCPT lost
CPT=1CPT=0Call progress tone lostCPT detected
CNG=0CNG=1Fax calling tone detectedCNG lost
DTV=0DTV=1DTMF tone detectedDTMF tone lost
DTV=1DTV=0DTMF tone lostDTMF tone detected
ATV=0ATV=1Alert tone detectedAlert tone lost
ATV=1ATV=0Alert tone lostAlert tone detected
BSY=1BSY=0File command completedNew command issued
SD=0SD=1Speech activity detectedSpeech activity lost
An interrupt is internally generated if any combina tion of these events occurs and the
interrupt is not masked. The interrupt is cleared when the host reads the STATUS
register. If a new event occurs while the host reads the status register, the status register
after
is updated
immediately after the access has ended.
the current access is terminated and a new interrupt is generated
Note: If the internal interrupt oc curs after the controller h as alrea dy se lec ted the de vic e
but not yet read the STATU S word, then the STATUS word is upd ated and the
internal interrupt is cleared. Therefore the cont roller should always evaluate the
STATUS word when read.
2.3.5Abort
If the PSB 4860 cannot continue the current operations in progress (e.g. due to a
transient loss of power) it st ops operation and initializ es all read/write reg isters to their
reset state. After that it sets the ABT bit of the STATUS register and generates an
interrupt. The PSB 4860 discards all commands with the exception of a write command
to the revision register whi le ABT is set. Only after the write comma nd to the revision
register (with any value) the ABT bit is reset and a reinitialization can take place.
Semiconductor Group8110.97
PSB 4860
Functional Description
2.3.6Revision Register
The PSB 4860 contains a revision register. This register is read only and does not
influence operation in any way. A write to the revision regi ster clears the ABT bit of the
STATUS register but does not alter the content of the revision register.
2.3.7Hardware Configuration
The PSB 4860 can be adapted to various external hardware configurations by four
special registers: HWCONFIG0 to HWCONFIG3. These registers are usually only
written once during initialization and must not be changed while the PSB 4860 is in active
mode. It is mandatory that the programmed configuration reflects the external hardware
for proper operation. Special care must be taken to avoid I/O conflicts or excess current
by enabling inputs without an external driving source. Table 65 can be used as a
checklist.
HWCONFIG0PFRDY
HWCONFIG0OSC1OSC1/2 must be connected to a crystal
HWCONFIG0ACS1CLK must not float (tie low if no clock present)
HWCONFIG1MFS1FSC must not float (tie low if no clock present)
HWCONFIG1ACT1FSC must not float (tie low if no clock present)
2.3.8Frame Synchronization
The PSB 4860 locks itsel f to either an externally suppli ed clock or frame syn c signa l or
generates the frame sync signal itself. This internal reference frame sync signal is called
master frame sync (MFSC). In addition, the PSB 4860 can derive the AFECLK and
AFEFSC from either the main oscillator or an auxiliary clock input. Table 66 shows how
AFECLK and MFSC are derived by the PSB 4860. The bits ACS and MFS are contained
in the hardware configuration registers.
The PSB 4860 can adjust AFECLK and AFEFSC dynamically to a sli ghtly varyi ng FSC
if AFECLK and AFEFSC are derived from the main oscillator (XTAL). This mode requires
that both AFEFSC and FSC are nominally running at the same frequency (8 kHz).
This feature is especially useful when the FSC signal is not derived from the same clock
source as AFECLK (ISDN application).
2.3.10Dependencies of Modules
There are some restrict ions concerning the modu les that can be enabled at the same
time (table 67). A checked cell indic ates that the two module s (defined by t he row and
the column of the cell) must not be enabled at the same time.
There are three classes of file commands denoted by the letters B, O and I. Table 68
shows the definition of these classes:
Table 68File Command Classes
ClassDescription
BBackground commands (Activate, Recompress, Garbage Collection, Initialize)
OOpen Commands (Open, Open Next Free File)
IAny command executed with EIE=1 (i.e. immediate execution)
Examples:
• The line echo canceller (in 24 ms mode) cannot be enabled when the speech decoder
is running at slow speed.
• If the DTMF detector is running, none of the background file command s (B) must be
executed. In addition, no file command must be executed w ith immediate execution
Semiconductor Group8310.97
PSB 4860
Functional Description
enabled (I). However, files my be opened and other commands (like read or write)
may be executed without immediate execution enabled.
Furthermore it may be necessary to restrict the length of the FIR filter of the echo
cancellation unit if sev eral other units are operating at the same time. T he sum of all
weights (table 69) of the simultaneou sly enabled mod ules must not exceed 100 at any
given time.
DTMF Generator2.2X
Echo Cancellation52.1127 taps (16 ms)
Echo Cancellation62.5255 taps (32 ms)X
Echo Cancellation72.9383 taps (48 ms)
Echo Cancellation83.3511 taps (64 ms)X
Line Echo Cancellation12.7X
Universal Attenuator0.2
Digital Interface1.7channel 1 or SSDIX
Digital Interface1.7channel 2
Analog Interface2.5XX
Clock Tracking0.6X
Miscellaneous8.0always activeXX
1)
The alert tone dete ctor would add another 2 .6, but can be disabled after the alert tone ha s been detected.
Therefore it can be left out of the ca lcu lat ion.
Example:
• For an analog phone echo c ancellation, DTMF tone genera tion, caller ID reception,
and line echo cancellation are ne cessary. The system uses the PSB 4851 and the
equalizer to linearize the loudspeaker. In this case the sum of all weights without echo
cancellation is 35.6. Therefore 255 taps can be used for a total of 98.1.
• In an ISDN phone echo cancellation, channel 1 of the digital interface, the analog
interface with clo ck tracking and the eq ualizer shall be enable d at the same time. In
Semiconductor Group8410.97
PSB 4860
Functional Description
this application the sum of all weights without echo cancellation is 15.6. Therefore 511
taps can be used for a total of 98.9.
Semiconductor Group8510.97
PSB 4860
Functional Description
2.4Interfaces
Interfaces
Interfaces
This section describes the interfaces of the PSB 4860. The PSB 4860 supports both an
®
-2 interface with single and double clock mo de and a strobe d serial data interface
IOM
(SSDI). However, these two interfaces cannot be used simultaneously as they share
some pins. Both interfaces are for data transfer only and cannot be used for
programming the PSB 4860. Table 70 lists the features of the two alternative interfaces.
The data stream is p artitioned into packets cal led frames. Each f rame is di vided into a
fixed number of timeslots. Each timeslot is us ed to transfer 8 bits. Figure 41 shows a
commonly used terminal mode (three channels ch
, ch1 and ch2 with four timeslots
0
each). The first timeslot (in fi gure 41: B1 ) is denoted by number 0, the s ec ond one (B2 )
by 1 and so on.
125 µs
FSC
DD/DU
B1M0B2
ch
CI0IC1M1IC2CI1
0
ch
1
ch
2
®
Figure 41 IOM
-2 Interface - Frame Structure
The signal FSC is used to indicate the start of a frame. Figure 42 shows as an example
*
two valid FSC-signals (FSC, FSC
clock cycle of a new frame (T
) which both indicate the same clock cycle as the first
).
1
Note: Any timeslot (including M0, CI0, ...) can be used for data transfer. However,
programming is not supported via the monitor channels.
Semiconductor Group8610.97
PSB 4860
Functional Description
DCL
FSC
FSC
T
1
*
T
2
Figure 42 IOM®-2 Interface - Frame Start
The PSB 4860 supports both single clock mode and double clock mode. In single clock
mode, the bit rate is equal to the c lock rate. Bits are shifted ou t with the rising edge of
DCL and sampled at the fal ling edge . In dou ble cloc k mode, the cloc k runs at twice the
bit rate. Therefore for each bit there are two clock cycles. Bits are shifted out with the
rising edge of the first clock cycle and sampled with the falling edge of the second clock
cycle. Figure 43 shows the timing for single c loc k mod e and figure 44 shows th e timing
for double clock mode.
DCL
DU/DX
DD/DR
Figure 43 IOM
T
1
bit 0bit 1bit 2
bit 0bit 1bit 2
®
-2 Interface - Single Clock Mode
T
2
Semiconductor Group8710.97
PSB 4860
Functional Description
DCL
DU/DX
DD/DR
T
1
bit 0bit 1bit 2
T
2
bit 0bit 1
T
3
T
4
T
5
Figure 44 IOM®-2 Interface - Double Clock Mode
The PSB 4860 supports up to two channels simul taneously for data transfer. Both the
coding (PCM or linear) and the data dire ction (DD/DU assignme nt for transmit/ receive)
can be programmed individually for each channel. Table 71 shows the registers used for
®
configuration of the IOM
Table 71IOM
®
-2 Interface Registers
-2 interface.
Register # of BitsNameComment
SDCONF 1ENInterface enable
SDCONF 1DCLSelection of clock mode
SDCONF 6NTSNumber of timeslots within frame
SDCHN11ENChannel 1 enable
SDCHN16TSFirst timeslot (channel 1)
SDCHN11DDData Direction (channel 1)
SDCHN11PCM8 bit code or 16 bit linear PCM (channel 1)
SDCHN11PCD8 bit code (A-law or µ-law, channel 1)
SDCHN21ENChannel 2 enable
SDCHN26TSFirst timeslot (channel 2)
SDCHN21DDData Direction (channel 2)
SDCHN21PCM8 bit code or 16 bit linear PCM (channel 2)
SDCHN21PCD8 bit code (A-law or µ-law, channel 2)
In A-law or µ-law mode, only 8 bits are transferred and therefore only one timeslot is
needed for a channel. In line ar mode, 16 bits are needed for a single cha nnel. In this
mode, two consecut iv e t ime slo ts are used for data transfe r. Bits 8 to 15 are trans ferred
Semiconductor Group8810.97
PSB 4860
Functional Description
within the first timeslot and bits 0 to 7 are transfe rred within the next timeslot. The fi rst
timeslot must have an even number. The most significant bit is always transmitted first.
Semiconductor Group8910.97
PSB 4860
Functional Description
2.4.2SSDI Interface
The SSDI interface is intended for seamless connection to low-cost burst mode
controllers (e.g. PMB 27251) a nd s up ports a s ing le c han nel in eac h direction. The data
stream is partitioned into frames. Within each frame one 16 bit value can be sent and
received by the PSB 4860. Th e start of a frame is i ndicated b y the risin g edge of FSC.
Data is always sampled at the falling edge of DCL and shifted out with the rising edge of
DCL.
The SSDI transmitter and receiver are operating independently of each other except that
both use the same FSC and DCL signal.
2.4.2.1SSDI Interface - Transmitter
The PSB 4860 indicates outgoing data (on signal DX) by activating DXST for 16 clocks.
The signal DXST is activated with the same rising edge of DCL that is used to send the
first bit (Bit 15) of the data. DXST is deactivated with the first rising edge of DCL after the
last bit has been transferred. The PSB 4860 drives the sign al DX only when DXST is
activated. Figure 45 shows the timing for the transmitter.
125 µs
FSC
DXST
DCL
DU/DX
bit 15bit 14bit 1bit 0
Figure 45 SSDI Interface - Transmitter Timing
2.4.2.2SSDI Interface - Receiver
Valid data is indicated by an active DRST pulse. Each DRST pulse must last for exactly
16 DCL clocks. As there may be more th an on e DRST pul ses wi thin a sing le fram e the
PSB 4860 can be programmed to listen to the n-th pulse with n ranging from 1 to 16. In
order to detect the first pulse pro perly, DRST must not be active at the ri sing edge of
FSC. In figure 46 the PSB 4860 is listening to the third DRST pulse (n=3).
Semiconductor Group9010.97
FSC
DRST
Figure 46 SSDI Interface - Active Pulse Selection
Figure 47 shows the timing for the SSDI receiver.
125 µs
FSC
PSB 4860
Functional Description
active pulse (n=3)
DRST
DCL
DD/DR
bit 15bit 14bit 1bit 0
Figure 47 SSDI Interface - Receiver Timing
Table 72 shows the registers used for configuration of the SSDI interface.
Table 72SSDI Interface Register
Register # of BitsNameComment
SDCHN14NASNumber of active DRST strobe
Semiconductor Group9110.97
PSB 4860
Functional Description
2.4.3Analog Front End Interface
®
The PSB 4860 uses a four wire interface similar to the IOM
information with the analog front end (PSB 4851). The main difference is that all
timeslots and the channel assignments are fixed as shown in figure 48.
.
125 µs
AFEFS
-2 interface to exchange
AFEDD
AFEDU
Channel C
16 bit16 bit8 bit
Channel C
1
000OV
Channel C
2
3
ALS
unused
Figure 48 Analog Front End Interface - Frame Structure
Voice data is transferred in 16 bit linear coding in two bidirectional channels C
An auxiliary channel C
is used to transfer the current setting of the loudspeaker
3
and C2.
1
amplifier ALS to the PSB 4860. The remaining bits are fixed to zero. In the other direction
transfers an override value for ALS from the PSB 4860 to the PSB 4851. An additional
C
3
override bit OV determines if the currently transmitted value should override the
AOAR:LSC
1)
setting. The AOAR:LSC setting is not affected by C3:ALS override. Table
73 shows the source control of the gain for the ALS amplifier.
Table 73Control of ALS Amplifier
AOPR:OVREC
:OVGain of A LS amplifi er
3
0-AOAR:LSC
10AOAR:LSC
11C
:ALS
3
Furthermore the AFE interface can be enabled or disabled according to table 74.
Table 74Analog Front End Interface Register
Register# of BitsNameComment
AFECTL1ENInterface enable
1)
See specification of PSB 485 1, aut om at ic ally s et by the PSB 4860 in loudhearing mode .
Semiconductor Group9210.97
PSB 4860
Functional Description
AFECLK
AFEFS
T
1
T
2
Figure 49 Analog Front End Interface - Frame Start
Figure 49 shows the synchronization of a frame by AFEFS. The first clock of a new frame
) is indicated by AFEFS switching from low to high before the falling edge of T1.
(T
1
AFEFS may remain high during subsequent cycles up to T
AFECLK
AFEDU
T
1
bit 0bit 1bit 2
T
2
32
.
AFEDD
bit 0bit 1bit 2
Figure 50 Analog Front End Interface - Data Transfer
The data is shifted out with th e rising e dge of AFEC LK and s ampled at the fal ling edg e
of AFECLK (figure 50). If AOPR:OVRE is not set, the channel C
4851. All values (C
, C2, C3:ALS) are transferred MSB first. The dat a clock (AFECLK)
1
is not used by the PSB
3
rate is fixed at 6.912 MHz. Table 75 shows the clock cycles used for the three channels.
Table 75Analog Front End Interface Clock Cycles
Clock CyclesAFEDD (driven by PSB 4860)AFEDU (driven by PSB 4851)
T
1-T16
T
17-T32
T
33-T40
T
41-T864
C1 dataC1 data
C2 dataC2 data
C3 dataC3 data
0tristate
Semiconductor Group9310.97
2.4.4Serial Control Interface
PSB 4860
Functional Description
The serial control interface (SCI) uses four lines: SDR, SDX, SCLK and CS
. Data is
transferred by the lines SDR and SDX at the rate given by SCLK. The falling edge of CS
indicates the beginning of an access. Data is sampled by the PSB 4860 at the rising edge
of SCLK and shifted out at the falling edge of SCLK. Each access must be terminated by
a rising edge of CS
. The accesses to the PSB 4860 can be divided into three classes:
1. Configuration Read/Write
2. Status/Data Read
3. Register Read/Write
If the PSB 4860 is in power down mode, a read access to the st atus register does not
deliver valid data with th e exception of the R DY bit. After the statu s has been read the
access can be either terminated or extended to read data from the PSB 4860. A register
read/write access can onl y be performed when th e PSB 4860 is ready. The RDY bit in
the status register provides this information.
Any access to the PSB 4860 starts with the transfer of 16 bits to the PSB 4860 over line
SDR. This first word specifies the access class, access type (read or write) and, if
necessary, the register accessed. If a configuration register is written, the first word also
includes the data and the access is terminated. Likewise, if a register read is issued, the
access is terminated a fter the first w ord. All other a ccesses contin ue by the transfer of
the status register from the PSB 4860 over line SDX. If a register (excluding
configuration) is to b e written, the next 16 bits containing the data are tra nsferred over
line SDR and the access is terminated. Figures 51 to 54 show the t iming diagrams fo r
the different access classes and types to the PSB 4860.
CS
SCLK
SDRc
SDX
INT
c
s
15
15
,..,c0:
,..,s0:
command word for status regis t er read
status register
15c14
:
c1c
0
s15s
14
s1s
:
0
Figure 51 Status Register Read Access
Semiconductor Group9410.97
CS
SCLK
PSB 4860
Functional Description
SDRc
15c14
c1c
SDX
,..,c0:
c
15
,..,s0:
s
15
d15,..,d0: data to be re ad
command word for data read
status register
:
:
Figure 52 Data Read Access
CS
SCLK
SDRc
15c14
c1c
0
0
s15s
:
14
s1s
0
d15d
d15d
14
14
d1d
d1d
0
0
SDX
c
s
15
15
,..,c0:
,..,s0:
command word for register write
status register
:
d15,..,d0: data to be writ t en
s15s
14
:
:
s1s
0
Figure 53 Register Write Access
Semiconductor Group9510.97
CS
SCLK
PSB 4860
Functional Description
SDRc
SDX
,..,c0:
c
15
,..,s0:
s
15
d15,..,d0: data to be read
15c14
command word for configurat ion register read
status register
:
:
c1c
0
s15s
14
s1s0d15d
:
14
Figure 54 Configuration Register Read Access
Configuration registers at even adresses use bit positions d
registers at odd adresses use bit positions d
CS
SCLK
15-d8
.
d1d
0
while configuration
7-d0
SDRc
c15,..,c0: command word for configuration register write
15c14
or register read
:
c1c
0
:
Figure 55 Configuration Register Write Access or Register Read Command
The internal interrupt signa l is cleared when the first bit of the s tatus register is put on
SDX. However, external ly the sign al INT
is deactivated as long as CS stays low. If the
internal interrupt signal is not cleared or another event causing an interrupt occurs while
the microcontroller is already reading the status belonging to the first event t hen INT
goes low again immediately after CS
is removed. The timing is shown in figure 51. Table
76 shows the formats of the different command words. All other c ommand words are
reserved.
In case of a configuration register read, R determines which pair of configuration
registers is to be read (table 78):
Table 78Address Field R for Configuration Register Read
9Register pai r
0HWCONFIG 0 / HWCONFIG 1
1HWCONFIG 2 / HWCONFIG 3
Note: Reading any register except the status register or a hardware configuration
register requires at least two accesses. The first access is a register read
command (figure 55). With this a ccess the registe r address is transferre d to the.
After that access data read accesses (figure 52) must be executed. The first data
read access with STATUS:RDY=1 delivers the value of the register.
Semiconductor Group9710.97
PSB 4860
Functional Description
2.4.5Memory Interface
The PSB 4860 supports either Flash Memory or ARAM/DRAM as external memory for
storing messages. If ARAM/DRAM is used, an EPROM can be added optionally to
support read-only messages (e.g. voice prompts). Table 79 summarizes the different
configurations supported.
8ARAM/DRAM1Mx41Mx4
16ARAM/DRAM4Mx4-2k or 4k refresh
16ARAM/DRAM2Mx82k refresh
32ARAM/DRAM4Mx44Mx42k or 4k refresh
32ARAM/DRAM2x2Mx82k refresh
64ARAM/DRAM16Mx4-4k or 8k refresh
64ARAM/DRAM8Mx84k or 8k refresh
128ARAM/DRAM16Mx416Mx44k or 8k refresh
4-128FLASH512kx8 devicesKM29N040
16-128FLASH2Mx8 devicesKM29N16000
If ARAM/DRAM is used, the total amount of memory must be a power of two and all
devices must be of the same type. The pin FRDY must be tied high.
For flash devices, the PSB 4860 supports in-circuit programming of voice prompts by
releasing the control lines during rese t and (op tionall y) pow er down . Instea d of ac tively
driving the lines FCS
, FOE, FWE, FCLE and ALE these lines are pulled high by a weak
pullup during reset and (optionally) power down.
Semiconductor Group9810.97
PSB 4860
Functional Description
2.4.5.1ARAM/DRAM Interface
The PSB 4860 supports up to two b anks of memory which may be 4 bit or 8 bit wide
(Figure 56). If both banks are used they must be populated identically.
The timing of the ARAM/DRAM interface is shown in figures 57 to 59. The timing is
derived form the internal memo ry clock MCLK* which runs at a quarter of the system
clock.