•Program mab le trans mit , receiv e and si de-t one
gains
•DSP-based:
i) Speakerphone switching algorithm
ii) DTMF and single tone generator
iii) Tone Ringer
•Differential interf ace to telepho ny tra nsdu cers
•Differential audio paths
•Singl e 5 volt pow er su ppl y
•X.25 Level 2 HD LC data form at ting
Applications
•Fully f eatu red dig ital t eleph one set s
•Cellula r phone sets
•Local area com m unications stations
ISSUE 2May 1995
Ordering Information
MT9092AP44 Pin PLCC
-40°C to +85°C
Description
The MT9092 HPhone-II is a fully featured integrated
digital telephone circuit which includes an HDLC
data formatter. Voice band signals are converted to
digital PCM and vice versa by a switched capacitor
Filter/Codec. The Filter/Codec uses an ingenious
differential architecture to achieve low noise
operation over a wide dynamic range with a single
5V supply. A Digital Signal Processor provides
handsfree speaker-phone operation. The DSP is
also used to generate tones (DTMF, Ringer and Call
Progress) and control audio gains. Internal registers
are accessed through a serial microport conforming
to INTEL MCS-51™ specifications. The device is
fabricated in Mitel's low power ISO
technology.
2
-CMOS
DSTo
DSTi
F0i
C4i
VSSD
VDD
VSSA
VSS
SPKR
VBias
VRef
Digital Signal ProcessorFilter/Codec Gain
AAA
AAAA
AAAA
AAAA
AAAA
A
AAA
AAAA
AAAA
AAAA
AAA
AAAA
AAA
AAAA
22.5/-72dB
AAA
AAAA
AAA
AAAA
∆1.5dB
AAA
AAAA
AAA
AAAA
Tx & Rx
AAA
AAAA
AAA
AAAA
AAA
AAAA
C-Channel
Registers
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
A
AAA
AAAA
AAA
AAAA
ENCODER
AAA
AAAA
AAA
AAAA
AAA
AAAA
AAA
AAAA
DECODER
AAA
AAAA
AAA
AAAA
AAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
A
A
A
A
A
A
A
A
STATUS
Control
Registers
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
-7dB
AAAA
AAAA
AAAA
HDLC
Timing
Circuits
LCD Driver
S1S12
BPWD PWRST
Figure 1 - Functional Block Diagram
7dB
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
AAAA
Transd ucer
Interface
New Call
Tone
Generator
S/P &
P/S
Converter
IC
Serial
Port
MCS-51
(
Compatible)
MICMIC+
MM+
HSPKR+
HSPKRSPKR+
SPKR-
DATA 2
DATA 1
SCLK
CS
IRQ
7-3
MT9092
PWRST
IC
VBias
VRef
NCM-VSSA
M+
MIC+
MIC-
VSS SPKR
VSSD
1
4443424140
23
2425262728
S1
S3S4S5S6S7
S2
39
38
37
36
35
34
33
32
31
30
29
SPKR+
SPKRHSPKR+
HSPKRVDD
BP
S12
S11
S10
S9
S8
DSTi
DSTo
C4i
F0i
VSSD
IRQ
SCLK
DATA 2
DATA 1
CS
WD
65432
7
8
9
10
11
12
13
14
15
16
17
1819202122
IC
NC
NC
44 PIN PLCC
Figure 2 - Pin Connections
Pin Description
Pin
NameDescription
#
1M+Non-Inver tin g Micro ph one (Input). Non-inverti ng input to microphone am plif ier from the
handset microphone.
2NCNo Connect. No internal connecti on to this pin.
3V
Bias
4V
5ICInternal Connection. Tie externally to V
6PWRST
Bias Voltage (Outpu t). (VDD/2) volts is available at this pin for biasing external ampli fier s.
Connect 0.1 µF capacitor to V
Reference voltage for codec (Output). Nominally [(VDD/2)-1.5] volts. Used internally.
Ref
Connect 0.1 µF capacit or to V
SSA
SSA
.
.
for normal operation.
SS
Power-up Reset (Input). CMOS compatible input with Schmit t Trigger (active low).
7DSTiST-BUS Serial Stream (Input). 2048 kbit/s input stream composed of 32 eight bit channels;
the first four of which are used by the MT9092. Input level is TTL compat ibl e.
8DSToST-BUS Serial Stream (Output). 2048 kbit/s output stream composed of 32 eight bit
channels. The MT9092 sources digital signals during the appropriate channel, time coincident
with the channels used for DSTi.
9C4i
10F0i
4096 kHz Clock (Input). CMOS level compatible.
Frame Pul se (Inp ut ). CMOS level compatible. This input is the frame synchronization pulse
for the 2048 kbit/ s ST-BUS stream.
11V
SSD
12IRQ
Digital Ground . Nominally 0 volts.
Interrupt Request (Open Drain Output). An active low output indicating an unmasked HDLC
interrupt event. Req uires 1 kΩ pull-up to V
DD
.
13SCLKSerial Po rt Syn chro no us Cl ock (In pu t). Data clock for MCS-51 compa tibl e micropo rt. TTL
level compatible.
7-4
Pin Description (continued)
MT9092
Pin
NameDescription
#
14DATA 2 Serial Data Transmit. In an alternate mode of operation, this pin is used for data transmit
from MT9092. In the default mode, serial data transmi t and receive are performed on the
DATA 1 pin and DATA 2 is tri-stated.
15DATA 1 Bidirectional Serial Data. Port for microprocessor serial data transfer compatible wit h MCS-
51 standard (default mode). In an alternat e mode of operati on , this pin becomes the data
receive pin only and data transmit is performed on the DATA 2 pin. Input level TTL compatible.
16CS
Chip Select (Input). This input signal is used to select the device for microport data
transfers. Active low. (TTL level compatible .)
17WDWatchdog (Output). Watchdog timer output. Active high.
18 ICInternal Connection. Tie externally to V
19,
NCNo Connection. No internal connection to these pins.
for normal operation.
SS
20
21V
SSD
Digital Ground. Nominall y 0 volt s.
22-33S1-S12 Segment Drivers (Output). 12 independently controlled, two level, LCD segment drivers. An
in-phase signal, with respect to the BP pin, produces a non-energized LCD segment . An out-
of-phase signal, with respect to the BP pin, energizes its respective LCD segment.
34BPBackplane Drive (Output). A two-level output voltag e for biasing an LCD backplane.
35V
Positive Po wer Supply (Inp ut). Nominally 5 volt s .
DD
36HSPKR- I nvertin g Hand set Speaker (Outpu t). Output to the handset speaker (balanced).
37HSPKR+Non-Inver tin g Handset Sp eaker (Outp ut). Output to the handset speaker (balanced).
38SPKR-Inverting Speake r (Outpu t). Output to the speakerphone speaker (balanced).
39SPKR+ Non-Inverting Speaker (Outpu t). Output to the speakerphone speaker (balanced).
40V
Power Supply Rail for Analog Output Drivers. N o min ally 0 Volts.
SS
SPK R
41MIC-Inve rtin g Handsfree M icr oph on e (Inp ut). Handsfree microphone amplif ier invert ing input
pin.
42MIC+No n-inver tin g Hand sfree Micro ph on e (Inp ut). Handsfree microphone amplifier non-
inverting input pin.
43V
SSA
Anal og G round. Nominall y 0 V.
44M-Inve rtin g Micro ph on e (Inp ut). Inverting input to microphone am plif ier from the handse t
microphone.
NOTES:
Intel and MCS-51 are registered trademarks of Intel Corporation, Santa Clara, CA, USA.
7-5
MT9092
Overview
The functional block diagram of Figure 1 depicts the
main operations performed within the HPhone-II.
Each of these functional blocks will be described in
the sections to follow. This overview will describe
some of the end-user features which may be
implemented as a direct result of the level of
integration found within the HPhone-II.
The main feature required of a digital telephone is to
convert the digital Pulse Code Modulated (PCM)
information, be ing rece ived by the telephon e set, into
an analog electrical signal. This signal is then
applied to an appropriate audio transducer such that
the information is finally converted into intelligible
acoustic energy. The same is true of the reverse
direction where acoustic energy is converted first
into an electrical analog and then digitized (into
PCM) before being transmitted from the set. Along
the way if the signals can be manipulated, either in
the analog or the digital domains, other features
such as gain control, signal generation and filtering
may be added. More complex processing of the
digital signal is also possible and is limited only be
the processing power available. One example of this
processing power may be the inclusion of a complex
handsfree switching algorithm. Finally, most electroacoustic transducers (loudspeakers) require a large
amount of power to develop an effective acoustic
signal. The inclusion of audio amplifiers to provide
this power is required.
The HPhone-II features Digital Signal Processing
(DSP) of the voice encoded PCM, complete Analog/
Digital and Digital/Analog conversion of audio
signals (Filter/CODEC) and an analog interface to
the external world of electro-acoustic devices
(T ransducer Interface). These three functional blocks
combine to provide a standard full-duplex telephone
conversation utilizing a common handset. Selecting
transducers for handsfree operation, as well as
allowing the DSP to perform its handsfree switching
algorithm, is all that is required to convert the fullduplex handset conversation into a half-duplex
speakerphone conversation. In each of these
modes, full programmability of the receive path and
side-tone gains is available to set comfortable
listening levels for the user as well as transmit path
gain control for setting nominal transmit levels into
the network.
The ability to generate tones locally provides the
designer with a familiar method of feedback to the
telephone user as they proceed to set-up, and
ultimately, dismantle a telephone conversation. Also,
as the network slowly evolves from the dial pulse/
DTMF methods to the D-Channel protocols it is
essential that the older methods be available for
backward compatibility. As an example; once a call
has been established, say from your office to your
home, using the D-Channel signalling protocol it may
be necessary to use in-band DTMF signalling to
manipulate your personal answering machine in
order to retrieve messages. Thus the locally
generated tones must be of network quality and not
just a reasonable facsimile. The HPhone-II DSP can
generate the required tone pairs as well as single
tones to accommodate any in-band signalling
requirement.
Each of the programmable parameters within the
functional blocks is accessed through a serial
microcontroller port compatible with Intel MCS-51
specifications.
Functional Descripti on
In this section, each functional block within the
HPhone-II is described along with all of the
associated control/status bits. Each time a control/
status bit(s) is described it is followed by the address
register where it will be found. T he reader is r eferred
to the section titled ‘Register Summary' for a
complete listing of all address map registers, the
control/status bits associated with each register and
a definition of the function of each control/status bit.
The Register Summary is useful for future reference
of control/status bits without the need to locate them
within th e tex t o f th e f unctional des crip ti o ns.
Filter-CODEC
The Filter/CODEC block implements conversion of
the analog 3.3kHz speech signals to/from the digital
domain compatible with 64kb/s PCM B-Channels.
Selection of companding curves and digital code
assignment are register programmable. These are
CCITT G.711 A-law or µ-Law, with true-sign/
Alternate Digit Inversion or true-sign/Inverted
Magnitude coding, respectively. Optionally, signmagnitude coding may also be selected for
proprietary applications.
The HPhone-II’s HDLC block is easy to use in
proprietary signalling protocols such as those within
PABXs and Key Systems. A fully interrupt driven
interface, buffered by 19 byte FIFOs in each
direction, simplifies the microcontroller's
asynchronous ac cess to the D-Cha n nel in fo rm at ion.
7-6
The Filter/CODEC block also implements transmit
and receive audio path gains in the analog domain.
These gains are in addition to the digital gain pad
provided in the DSP section and provide an overall
path gain resolution of 0.5dB. A programmable gain,
MT9092
voice side-tone path is also included to provide
proportional transmit speech feedback to the
handset receiver so that a dead sounding handset is
not encountered. Figure 3 depicts the nominal halfchannel and side-tone gains for the HPhone-II.
On PWRS T
(pin 6) the Filter/CODEC defaults such
that the side-tone path, dial tone filter and 400Hz
transmit filter are off, all programmable gains are set
to 0dB and µ-Law companding is selected. Further,
the Filter/CODEC is powered down due to the PuFC
bit (Transducer Control Register, address 0Eh) being
reset. This bit must be set high to enable the Filter/
CODEC.
The internal architecture is fully differential to provide
the best possible noise rejection as well as to allow a
SERIAL
PORT
DSP GAIN*
FILTER/CODEC
wide dynamic range from a single 5 volt supply
design. This fully differential architecture is
continued into the Transducer Interface section to
provide full chip realization of these capabilites.
A reference voltage (V
), for the conversion
Ref
requirements of the CODER section, and a bias
voltage (V
sections, are both generated on-chip. V
), for biasing the internal analog
Bias
Bias
is also
brought to an external pin so that it may be used for
biasing any external gain plan setting amplifiers. A
0.1µF capacitor must be connected from V
analog ground at all times. Likewise, although V
Bias
to
Ref
may only be used internally, a 0.1µF capacitor from
the V
pin to ground is required at all times. It is
Ref
suggested that the analog ground reference point for
these two cap ac ito r s b e ph y si cal ly the same p o int.
TRANSDUCER INTERFACE
µ-Law –6.3 dB
Α-Law –3.7 dB
-6 dB
HSPKR+
Handset
Receiver
(150Ω)
PCM
PCM
Receive
–72 to
+22.5 dB
(1.5dB
steps)
DTMF,
Tone
Ringer &
Handsfree
–72 to
+22.5 dB
(1.5dB
steps)
Transmit
Receive
Filter Gain
0 to –7 dB
(1 dB steps)
Side-tone
–9.96 to
+9.96d B
(3.32 dB steps)
Side-tone
Nominal
Gain
µ-Law –11 dB
Α-Law –18.8 dB
Transmit
Filter Gain
0 to +7dB
(1 dB steps)
-6 dB
Speaker Gain
0 to –24 dB
(8 dB steps)
µ-Law 6.1dB
Α-Law 15.4dB
Transmit
Gain
Receiver
Driver
Speaker
Phone
Driver
0.2dB*
Tone
Ringer
(input
from DSP)
M
U
X
HSPKR–
SPKR+
SPKR–
MIC+
MIC–
M+
M–
75
75
Speakerphone
Speaker
(40Ω nominal)
(32Ω min)
Handsfree
mic
Transmitter
microphone
DIGITAL DOMAIN
Internal to DeviceExternal to Device
Note: *gain the same for A-Law and µ−Law
ANALOG DOMAIN
Figure 3 - Audio Gain Partitioning
7-7
MT9092
To facili tate this the V
Ref
and V
pins are situated
Bias
on adjacent pins.
The transmit filter is designed to meet CCITT G.714
specifications. The nominal gain for this filter path is
0dB (gain control = 0dB). An anti-aliasing filter is
included. This is a second order lowpass
implementation with a corner frequency at 25kHz.
Attenuation is better than 32dB at 256 kHz and less
than 0.01dB within the passband.
An optional 400Hz high-pass function may be
included into the transmit path by enabling the Tfhp
bit in the Transducer Control Register (address 0Eh).
This option allows the reduction of transmitted
background noise such as motor and fan noise.
The receive filter is designed to meet CCITT G.714
specifications. The nominal gain for this filter path is
0 dB (gain control = 0dB). Filter response is peaked
to compensate for the sinx/x attenuation caused by
the 8 kH z sam p li ng rat e.
The Rx filter function can be altered by enabling the
DIAL EN control bit in the Transducer Control
Register (address 0Eh). This causes another
lowpass function to be added, with a 3dB point at
1000Hz. This function is intended to improve the
sound quality of digitally generated dial tone
received a s PCM.
Transmit sidetone is derived from the Tx filter and is
subject to the gain control of the Tx filter section.
Sidetone is summed into the receive path after the
Rx filter gain control section so that Rx gain
adjustment will not affect sidetone levels. The sidetone path may be enabled/disabled with the SIDE
EN bit located in the Transducer Control Register
(address 0Eh). See also STG
-STG2 (address 0Bh).
0
Transmit and receive filter gains are controlled by the
TxFG0-TxFG2 and RxFG0-RxFG2 control bits
respectively. These are loc ated in the FCODEC Gain
Control Register 1 (address 0Ah). Transmit filter gain
is adjustable from 0dB to +7dB and receive filter gain
from 0dB to -7dB, both in 1dB increments.
Side-tone filt er gain is controlled by the STG
-ST G
0
control bits located in the FCODEC Gain Control
Register 2 (address 0Bh). Side-tone gain is
adjustable from -9.96dB to +9.96dB in 3.32dB
increments.
Law selection for the Filter/CODEC is provided by
the A/µ companding control bit while the coding
scheme is controlled by the sign-mag/CCITT
bit.
Both of these reside in the General Control Register
(address 0Fh).
Digital Signal Processor
The DSP block is located, functionally, between the
serial ST-BUS port and the Filter/CODEC block. Its
main purpose is to provide both a digital gain control
and a half-duplex handsfree switching function. The
DSP will also generate the digital patterns required
to produce standard DTMF signalling tones as well
as single tones and a tone ringer output. A
programmable (ON/OFF) offset null routine may also
be performed on the transmit PCM data stream. The
DSP can generate a ringer tone to be applied to the
speakerphone speaker during normal handset
operation so that the existing call is not interrupt ed.
The main functional control of the DSP is through
two hardware registers which are accessible at any
time via the microport. These are the Receive Gain
Control Register at address 1Dh and the DSP
Control Register at address 1Eh. In addition, other
functional control is accomplished via multiple RAMbased registers which are accessible only while the
DSP is held in a reset state. This is accomplished
with the DRESET
bit of the DSP Control Register.
Ram-based registers are us ed to store transmit gain
levels (20h for transmit PCM and 21h for transmit
DTMF levels), the coefficients for tone and ringer
generation (addresses 23h and 24h), and tone ringer
warble rates (address 26h). All undefined addresses
below 20h are reserved for the temporary storage of
interim variables calculated during the execution of
the DSP algorithms. These undefined addresses
should not be written to via the microprocessor port.
The DSP can be programmed to execute the
following micro-programs which are stored in
instruction ROM, (see PS0 to PS2, DSP Control
Register, address 1Eh). All program execution
begins at the frame pulse boundary.
PS2
PS1PS0Micro-program
000Power up r e se t p ro g r am
001Transmit and receive gain control
program; with autonulling of the
transmit PCM, if the AUTO bit is
2
010DTMF generation plus transmit
set (see address 1Dh)
and receive gain control
program (autonull available via
the AUTO control bit)
011Tone ringer plus transmit and
receive gain control program
(autonull available via the
AUTO control bit )
7-8
MT9092
PS2PS1PS0Micro-program
100handsfree switching program
10 1
110Last three selections reserved
11 1
Note: For the DSP to function it must be selected to
operat e, in co njunction with the Filter/C odec, in
one of t he B-C hannel s. Th erefor e, on e of the B Channel enable bits must be set (see Timing
Contro l, a d d ress 1 5 h : bi ts CH
EN and CH3EN).
2
Power Up reset Pro gram
A hardware power-up reset (pin 6, PWRST) will
initialize the DSP hardware registers to the default
values (all zeros) and will reset the DSP program
counter. The DSP will then be disabled and the PCM
streams will pass transparently through the DSP. The
RAM-based registers are not reset by the PWRST
pin but may be initialized to their default settings by
programming the DSP to execute the power up reset
program. None of the micro-programs actually
require the execution of the power up reset program
but it is useful for pre-setting the variables to a
known condition. Note that the reset program
requires one full frame (125µSec) for execution.
Gain Contr ol Progra m
Gain control is performed on converted linear code
for both the receive and the transmit PCM. Receive
gain control is set via the hardware register at
address 1Dh (see bits B0 - B5) and may be changed
at any time. Gain in 1.5dB increments is available
within a range of +22.5dB to -72dB. Normal
operation usually requires no more than a +20 to -20
dB range of control. However, the handsfree
switching algorithm requires a large attenuation
depth to maintain stability in worst case
environments, hence the large (-72 dB) negative
limit. Transmit gain control is divided into two RAM
registers, one for setting the network level of transmit
speech (address 20h) and the other for setting the
transmit level of DTMF tones into the network
(address 21h). Both registers provide gain control in
1.5dB increments and are encoded in the same
manner as the receive gain control register (see
address 1Dh, bits B0 - B5). The power up reset
program sets the default values such that the receive
gain is set to -72.0 dB, the transmit audio gain is set
to 0.0dB and the transmit DTMF gain is set to -3.0
dB (equivalent to a DTMF output level of -4dBm0 into
the network).
Optional Offset Nulling
Transmit PCM may contain residual offset in the form
of a DC component. An offset of up to ±fifteen linear
bits is acceptable with no degradation of the
parameters defined in CCITT G.714. The HPhone-II
filter/CODEC guarantees no more than ±ten linear
bits of offset in the transmit PCM when the autonull
routine is not enabled. By enabling autonulling (see
AUTO in the Receive Gain Control Register, address
1Dh) offsets a re reduced to within ±one bi t of zero.
Autonulling circuitry was essential in the first
generations of Filter/Codecs to remove the large DC
offsets found in the linear technology. Newer
technology has made nulling circuitry optional as
offered in the HPhone-II.
DTMF and Gai n Cont rol Prog ram
The DTMF program generates a dual cosine wave
pattern which may be routed into the receive path as
comfort tones or into the transmit path as network
signalling. In both cases, the digitally generated
signal will undergo gain adjustment as programmed
into the Receive Gain Control and the Transmit
DTMF Gain Control registers. The composite signal
output level in both directions is -4dBm0 when the
gain controls are set to 2Eh (-3.0 dB). Adjustments to
these levels may be made by altering the settings of
the gain control registers. Pre-twist of 2.0dB is
incorporated into the composite signal. The
frequency of the low group tone is programmed by
writing an 8-bit coefficient into Tone Coefficient
Register 1 (address 23h), while the high group tone
frequency uses the 8-bit coefficient programmed into
Tone Coefficient Register 2 (address 24h). Both
coefficients are determined by the following
equation:
COEFF = 0.128 x Frequency (in Hz)
where COEFF is a rounded off 8 bit binary integer
A single frequency tone may be generated instead of
a dual tone by programming the coefficient at
address 23h to a value of zero. In this case the
frequency of the single output tone is governed by
the coefficient stored at address 24h.
7-9
MT9092
Table 1 gives the standard DTMF frequencies, the
coefficient required to generate the closest
frequency, the actual frequency generated and the
percent deviation of the generated tone from the
nominal.
The sum of harm onic and no is e po w er in th e
freque nc y ba nd from 50Hz to 3 50 0H z i s ty pi ca ll y
more t ha n 30 dB below th e p ower in the to n e p ai r. All
individual harmonics are typica lly more than 40dB
below the level of the low group tone.
Tone Ring er and G ain Co ntrol P rogram
A locally generated alerting (ringing) signal is used to
prompt the user when an incoming call must be
answered. The DSP uses the values programmed
into Tone Coefficient Registers 1 and 2 (addresses
23h and 24h) to generate two different squarewave
frequencies in PCM code. The amplitude of the
squarewave frequencies is set to a mid level before
being sent to the receive gain control block. From
there the PCM passes through the decoder and
receive filter, replacing the normal receive PCM data,
on its way to the loudspeaker driver. Both
coefficients are determined by the following
equation:
COEFF = 8000/Frequency (Hz)
where COEFF is a rounded off 8 bit binary integer.
The ringer program switches between these two
frequencies at a rate defined by the 8-bit coefficient
programmed into the Tone Ringer Warble Rate
Register (address 26h). The warble rate is defined
by the equation:
Tone duration (warble frequency
in Hz) = 500/COEFF
An alternate method of generating ringer tones to the
speakerphone speaker is available. With this method
the normal receive speech path through the decoder
and receive filter is uninterrupted to the handset,
allowing an existing conversation to continue. The
normal DSP and Filter/CODEC receive gain control
is also retained by the speech path. When the OPT
bit (DSP Control Register address 1Eh) is set high
the DSP will generate the new call tone according to
the coefficients programmed into registers 23h, 24h
and 26h as before. In this mode the DSP output is no
longer a PCM code but a toggling signal which is
routed directly through the New Call Tone gain
control section to the loudspeaker driver. Refer to
the secti o n ti tle d ‘Ne w Call To ne ’.
Handsfree Program
A half-duplex speakerphone program, fully contained
on chip, provides high quality gain switching of the
transmit and receive speech PCM to maintain loop
stability under most network and local acoustic
environments. Gain switching is performed in
continuous 1.5dB increments and operates in a
complimentary fashion. That is, with the transmit
path at maximum gain the receive path is fully
attenuated and vice versa. This implies that there is
a mid position where both transmit and receive paths
are attenuated equally during transition. This is
known as the idle state.
Of the 64 possible attenuator states, the algorithm
may rest in only one of three stable states; full
receive, full transmit and idle. The maximum gain
values for full transmit and full receive are
programmable through the microport at addresses
20h and 1Dh respectively, as is done for normal
handset operation. This allows the user to set the
maximum volumes to which the algorithm will
adhere. The algorithm determines which path should
maintain control of the loop based upon the relative
levels of the transmit and receive audio signals after
the detection and removal of background noise
energy. If the algorithm determines that neither the
transmit or the receive path has valid speech energy
then the idle state will be sought. The present state
of the algorithm plus the result of the Tx vs. Rx
decision will determine which transition the algorithm
will take toward its next stable state. The time
durations required to move from one stable state to
the next are parameters defined in CCITT
Recommendation P.34 and are used by default by
this algorith m (i.e ., b u ild- u p time , h a ng- ov e r time a nd
switching time).
where 0 < COEFF < 256, a warble rate of 5-20Hz is
suggested.
7-10
MT9092
Quiet Code
The DSP can be made to send quiet code to the
decoder and receive filter path by setting the
RxMUTE bit high. Likewise, the DSP will send quiet
code in the transmit (DSTo) path when the TxMUTE
bit is high. Both of these control bits reside in the
DSP Control Register at address 1Eh. When either
of these bits are low, their respective paths function
normally.
HDLC
The High-level Data Link Control (HDLC) block is
located, functionally, between the serial ST-BUS port
and the serial Microcontroller port. This functional
block handles the bit oriented protocol requirements
of layer 2 X.25 packet switching and Q.921 link
access protocols defined by CCITT. The HDLC is
dedicated to D-Channel operation at 16kb/s and
offers buffered access to the serial D-Channel data
through separate 19 byte transmit and receive
FIFOs.
The HDLC generat es and det ects the flags, various
link channel states and abort sequences as well as
performing a cyclic redundancy check on data
packets according to the CCITT defined polynomial.
Lastly, the protocol functions may be disabled to
provide transparent access, of the serial port DChannel, to the microport.
generates flags and appends them to the packet to
be transmitted. The receiver searches the incoming
data stream for flags on a bit-by-bit basis to establish
frame synchronization. The receiver uses flags for
synchronization only and does not transfer them to
the Rx FIFO.
Address F ield
The address field consists of one or two 8-bit bytes
directly following the opening flag. Address, Control
and Information fields are known collectively as the
Data field.
Control Fi eld
The control field consists of one 8-bit byte directly
following the address field. The HDLC does not
distinguish between the control field and the
informatio n field.
Information Field
The information field immediately follows the control
field and consists of N bytes of data where one by te
contains 8 bits. A packet does not need to contain an
information field to be valid. The HDLC does not
distinguish between the control field and the
informatio n field.
Frame Checking Sequence Field
A power up reset (PWRST
via RST (address 0Fh) will cause the HDLC
transceiver to be initialized. This results in the
transmitter and receiver being disabled and all HDLC
registers defaulting to their power reset values.
HDLC Frame Stru cture
A valid HDLC frame begins with an opening flag,
contains at least 16 bits of address, control or
information, ends with a 16 bit FCS followed by a
closing flag. Data formatted in this manner is also
referred to a s a " packet". Refer to Fi gu re 4 .
FLAGDATA FIELDFCSFLAG
One
Byte
n Bytes
(n≥2)
Figure 4 - Frame Format
, pin 6) or a softwar e reset
Two
Bytes
One
Byte
Flag Sequ ence
All HDLC frames start and end with a unique
sequence of 8 bits. This sequence is 0111 1110
(7Eh). The closing flag of one frame can be the
opening flag of the next frame. The transmitter
The 16 bits preceding a closing flag are the FCS
field. A cyclic redundancy check utilizing the CRCCCITT standard generator polynomial X
+1 produces the 16-bit FCS. In the transmitter the
FCS is calculated on all bits of the address, control
and information fields. The complement of the FCS
is transmitted, most significant bit first, in the FCS
field. The receiver calculates the FCS on the
incoming packet's address, control, information and
FCS fields and compares the result to 'F0B8'. This
result verif ies no transm ission errors occu rred. If the
packet, between flags, is also at least 32 bits in
length then t he address, control and information field
data are entered into the receive FIFO minus the
FCS which is discarded.
16
+ X12 +X
Order of Bit Transmission
Address, control and information field data are
entered into the transmit FIFO. This data is then
transmitted and received on the serial bus least
significant bit first. The FCS is sent most significant
bit first on the serial bus. Note that it is the
complement of the calculated FCS which is
transmitted. The HDLC does not distinguish
ADDRESS/CONTROL/INFORMATION bytes except
5
7-11
MT9092
to determine if the packet is of minimum valid length.
These fields are transferred transparently through
the FIFO's.
Data Transparency (Ze ro inserti on/d eletio n)
Transparency ensures that the contents of a data
packet do not imitate a flag, go-ahead, frame abort
or idle channel. The content s of a transmitted frame,
between the flags, is examined on a bit-by-bit basis
and a 0 bit is inserted after all sequences of five
contiguous 1 bits (including the last five bits of the
FCS). Upon receiving five contiguous 1s within a
frame the re ce ive r d ele te s th e fo llow i n g 0 bit.
Invalid Frames
A frame is invalid if one of the following four
conditions exists. Inserted zeros are not part of a
valid bit count:
1. If the FCS pattern generated from the received
data does not match the 'F0B8' pattern then the
last data byte of the packet is written to the
receive FIFO with a 'bad packet' indication.
2. A short frame exists if there are less than 25 bits
between the flags. Short frames are ignored by
the receiver and nothing is written into the
receive FIFO.
3. P ackets which are at least 25 bits in length but
less than 32 bits (between the flags) are also
invalid. In this case the data is written to the
FIFO but the last byte is tagged with a 'bad
packet' indication.
4. If a frame abort sequence is detected the packet
is invalid. Some or all of the current packet will
reside in the receive FIFO, assuming the packet
length before t he abort sequence was at least 26
bits long.
Frame Ab ort
The transmitter will abort a current packet by
substituting a zero followed by seven contiguous 1s
in place of the normal data. The receiver will abort
upon reception of seven contiguous 1s occurring
between the flags of a packet which contains at least
26 bits.
Note that should the last receive byte before the
frame abort end with contiguous 1s, these are
included in the seven 1s required for a receiver
abort. This means that the location of the abort
sequence in the receiver may occur before the
location of the abort sequence in the originally
transmitted packet. If this happens, t hen the last data
written to the receive FIFO will not correspond
exactly with the last byte received before the frame
abort.
Interframe T ime F ill and L ink Cha nne l State s
When the HDLC transmitter is not sending packets it
will wait in one of two states.
Interframe Time Fill: This is a continuous series of
flags occurring between
frames indicating that the
channel is active but that no
data is being sent.
Idle:An idle channel occurs when at least fifteen
contiguous 1s are transmitted or received.
In both cases the transmitter will exit the wait state
when data is loaded into the transmit FIFO.
Go-Ahead
A go-ahead is defined as the pattern ‘011111110’
(contiguous 7F’s) and is the occurrence of a frame
abort sequence followed by a zero, outside of the
boundaries of a normal packet. Being able to
distinguish a proper (in packet) frame abort
sequence from one occurring outside of a packet
allows a higher level of signalling protocol which is
not part of the HDLC specifications.
Transmitter
Following initialization and enabling, via the HTxEN
bit (address 03h), the transmitter is in the Idle
Channe l State (Mark Idle ). Interf rame time fill may
be selected by setting the Mark Idle bit (address 03h)
high. The transmitter remains in its programmed
state until data is written to the Tx FIFO. The
transmitter will then proceed as follows:
1)If the transmitter is in the idle state the present
byte of ones will be completely transmitted
before the opening flag and packet data is
sent.
2)If the transmitter is in the interframe time fill
state the flag currently being transmitted will be
used as the opening flag followed by the
packet data.
To assist in loading multiple packets into the transmit
FIFO the last packet byte is tagged with either EOP
(to indicate the end of the current packet) or FA.
Control Register 1 (address 03h) bits EOP (end of
7-12
MT9092
packet) and FA (frame abort) are set before writing
the last packet byte to the Tx FIFO. The act of
loading the last packet byte will then automatically
reset the EOP and FA bits. Tx FIFO bytes are
continuously transmitted until the FIFO is empty, by
which time an EOP or FA tag should have been
encounte r ed b y the transmitter.
After the last bit of the EOP byte has been
transmitted a 16 bit FCS is sent followed by a closing
flag. When multiple packets of data are loaded into
the Tx FIFO only one flag is sent between packets.
When the transmitter encounters a byte tagged FA
then a frame abort sequence is sent instead of the
tagged byte. All bytes previous to but not including
the tagged byte are sent.
The transm itter r eturns to its p rogramm ed wai t state
after concluding the transmission of EOP or FA if the
Tx FIFO is empty.
Transmit FIFO Status
The transmit FIFO is 19 bytes deep (address 02h).
As data is loaded into (from the microport) and
extracted from (via the serial port) the Tx FIFO the
present 'fill state' can be monitored using the Txstat1
and Txstat2 bits found in the HDLC Status Register
(address 04h). These states are encoded as shown
in Table 2. Note that the FIFO emptying threshold,
where an interrupt (TxFL if unmasked) will occur, can
be set to a low level 4 (default) or to a high level 14
by the Fltx bit in the HDLC Control Register 2
(address 05h).
Command/Address byte which indicates a microport
read of address 07h. Since all interrupts are
generated by the occurrence of an HDLC event (i.e.,
a transition), this register informs that an event has
occurred but does not guarantee that it is still valid.
To determine current validity the HDLC Status
Register (address 04h) should be read. Due to the
asynchronous nature of the interrupts an interrupt
occurring during a read of the Interrupt Status
register will be held until the read cycle is over,
unless it is an interrupt which is already valid.
There are three interrupts associated with the
transmitter.
TEOPTransmit End Of Packet:
Set when the transmitter has finished
sending the closing f lag of a packet or
after an abort sequence has been
completed.
TxFLTransmit FIFO Low:
Set when a transit ion from 5 to 4 bytes
in the Tx FIFO has occurred. This is an
early warning to the microprocessor
that the FIFO is emptying and should
be serviced before it empties
completely; a condition which will
result in a transmit underrun unles s an
EOP or FA byte has been written to the
FIFO. By setting the Fltx bit (address
05h) high the FIFO emptying condition
will occur at the transition from 15 to 14
bytes. This will allow the microport
more time to react to this interrupt
condition.
A Tx FIFO underrun occurs if the Tx FIFO empties
without the oc currence of an EOP or FA tagged byte.
A frame abort sequence is automatically transmitted
under this condition.
Transm it Interrupts
The HDLC Interrupt Enable Register (address 06h)
is used to select (unmask) only those interrupts
which are deemed important to the microprocessor.
After a PWRST
be cleared causing all interrupts to be masked.
All selected i n te rrup t e vents will c ause the IRQ pin to
become active. Unselected interrupt events will not
cau se IRQ
still be represented by the appropriate bit in the
HDLC Interrupt Status Register (address 07h). This
register must be read after receiving an IRQ
be polled at any time. The IRQ
coincident with the first SCLK falling edge following a
or software RST all enable bits will
to become active however, the event will
or may
output pin is reset
Txunder Transmit underrun:
Set when the Tx FIFO empties without
the occurrence of an EOP or FA
tagged byte. A frame abort sequence
is automatically transmitted under this
condition. Note that this register bit
position is shared with the frame abort
(FA) interrupt (see receive interrupts).
For this b it to r eflect T xund er t he Ints el
bit in Control Register 2 (address 05h)
must be set high.
Disabling, Res et, Transparent Operat ion an d CRC
Disabling the transm itter via the HTxEn bit will occur
after the current packet is completely transmitted.
The status and Interrupt registers may still be read
and the Tx FIFO and control registers written while
the transm it te r is d isa ble d .
7-13
MT9092
The Tx FIFO may be reset by setting the Txfrst bit in
the HDLC Control Register 2 (address 05h). The
HDLC Status Register will identify the Tx FIFO as
being empty although the actual data in the FIFO will
not be reset. Txfrst will be cleared by the next write
to the Tx FIFO.
Transparent data may be sent by setting the TRANS
bit (address 03h) high. The transmitter will no longer
generate the flag, abort and idle sequences, nor will
it insert zeros and append the FCS. Data will still be
transmitted LSB first. If there is no data in the Tx
FIFO or the Tx FIFO empties the last byte
transmitted will be repetitively sent until new data is
presented to the FIFO. It will t ake typically two STBUS frames, af ter writing TRANS, before this mode
begins. Note that CH
Transmission of the FCS field CRC may be inhibited
using the Tcrci (Transmit Crc Inhibit) bit at address
05h. W hile this bit is set the o pe n in g fl a g fo ll o we d b y
the data fields and closing flag is transmitted,
including zero insertion, but the calculated CRC is
not. This allows the processor to insert the CRC as
part of the data field. This usage is for V.120 terminal
adaptation for synchronous protocol sensitive UI
frames.
Receiver
Following initialization and enabling, via the HRxEN
bit at address 03h, the receiver begins clocking in
serial data checking for flags (0111 1110), go-aheads
( 0111 1111 0), and idl e c hannel states (at l eas t f ifteen
contiguous ones). Upon detecting a flag the receiver
synchronizes itself to the data stream and begins
calculating the CRC. If the packet length, between
the flags and after zero deletion, is less than 25 bits
the packet is ignored and nothing is written to the Rx
FIFO. If the packet length, after zero deletion, is
between 25 and 31 bits a last byte, bad packet
indication is written into the Rx FIFO.
EN must also be set.
0
Idle Channel
When the receiver detects at least 15 contiguous
ones it declares an idle channel condition exists and
sets the IdleChan bit in the HDLC status register
high (address 04h). This bit remains set until the
received condition changes.
If address recognition is required, Receive Address
Recognition Registers 1 and/or 2 (addresses 00h
and 01h respectively) are loaded with the desired
address comparison information, the Adrec bit is set
high and A1EN and A2EN are set as required. Bit 0
(A1EN and A2EN) of both recognition registers is
used as an enable for that by te. When either of these
bits are low their respective address mask
information is ignored. In this way either or both of
the first two received bytes can be compared to the
expected mask values. Only those packets passing
the appropriate comparison test will be loaded into
the Rx FIFO. The appropriate comparison test
(single/dual byte address, All-call) is defined by the
logic state of bit 0 of the first byte received after the
opening flag.
Bit 0 of the first received address byte (address
extension bit) is monitored to determine if a single or
dual byte address is being received.
1. If the address extension bit is 1 then a single
byte address is being received. If A1EN is high
the stored bit mask (Adr11 - Adr16 and
sometimes Adr10) is compared to the received
first address byte. Any packet failing this
address comparison will not be stored in the Rx
FIFO except for the All-call condition. A1EN
must be set high for a single-byte All-call
(11111111) address to be recognized. The
second mask byte is ignored. Seven bits of
address comparison may be realized for single
byte recognition by setting the SEVEN bit
(address 05h) high. This mode will then include
Adr10 as part of the mask information. The first
received byte must also have bit 0 set to a 1
indicating single byte addressing.
2. If the address extension bit is 0 then a two byte
address is being received and the six most
significant bits of the first received byte are
compared. The seven most signif icant bits of the
second received byte are compared (Adr20 Adr26, note A2EN must be set high also). Any
packet failing this address comparison will not
be stored in the Rx FIFO. An All-call condition
(1111111x) is also monitored for in the s econd
received address byte and, if found, the first and
second byte masks are ignored (not compared
with the mask byte). Packets addressed with Allcall are written into the Rx FIFO.
Address Recognition
When Adrec (HDLC Control Register 1, address
03h) is low all valid received packets, regardless of
the address field inform ation, are loaded into the Rx
FIFO.
7-14
In CCITT Q.921 parlance the Adr11 - Adr16 bits are
defined as Sapi0 - Sapi5 (Service Access Point
Identifier n). Adr10 is defined as C/R (Command/
Response). Adr20 - Adr26 are defined as Tei0 - Tei6
(Terminal Endpoint Identifier n).
MT9092
Receive Byte Status
As each received pack et byte is wr itten into th e Rx
FIFO two bits are appended to indicate the status of
that byte. As these bytes are read from the Rx FIFO
the status bits are made available to the
microprocessor in the HDLC Status Register
(address 04h) as RxBS1 and RxBS2. Since the
information contained in RxBS1 & RxBS2 pertains to
the byte about to be read from the Rx FIFO, it is
important that this information be read before
reading the data byte from the FIFO. RxBS1 and
RxBS2 are encoded as shown in Table 2. A good
packet indication means a good FCS and no frame
abort whereas a bad packet indication means either
an incorrect FCS or a frame abort occurred.
Receive F IFO Status
The receive FIFO is 19 bytes deep (address 02h). A s
data is loaded into (from the serial port) and
extracted from (via the microport) the Rx FIFO the
present 'fill state' can be monitored using the Rxstat1
and Rxstat2 bits found in the HDLC Status Register
(address 04h). These states are encoded as shown
in Table 2. Note that the FIFO filling threshold, where
an interrupt (RxFf if unmasked) will occur, can be set
to a high level 15 (default) or to a low level 5 by the
Flrx bit in the HDLC Control Register 2 (address
05h).
In the case of an Rx FIFO overflow, an attempt by
the receiver to write data into an already full FIFO,
the receiver is disabled causing it to stop writing to
the Rx FIFO. The remainder of the current receive
packet is therefore ignored. The receiver will be reenabled when the next flag is detected but will
overflow again if the Rx FIFO level has not been
reduced to less than full. If two 'first byte' (RxBS1
and RxBS2) conditions are observed in the FIFO
without an intervening 'last byte' then an overflow
occurred for the first packet.
Receive Interrupt s
The HDLC Interrupt Enable Register (address 06h)
is used to select (unmask) only those interrupts
which are deemed important to the microprocessor.
After a PWRST
be cleared causing all interrupts to be masked.
or software RST all enable bits will
RxBS2,Are status bits from the Rx FIFO.
RxBS1RxBS2
Note-If two consecutive first byte signals are received without an inte rven ing last byte, then an
overflow has occurred and the first packet (or packets) are bad. A bad packet indicates that
either a frame abort had occurred or the FCS did not match.
-On power-up these bits are in an indeterminate state until the first byte is writte n to Rx FIFO.
Txstat2,
Txstat1These two bits are encoded to indicate the present state of Tx FIFO. This is an asynchronous
event.
Txstat2
00TxFULL
015 OR MORE BYTES (15 if Fltx set)
114 OR LESS BYTES (14 if Fltx set)
10TxEMPTY
Rxstat2,
Rxstat1These two bits are encoded to indicate the present state of Rx FIFO. This is an asynch ronous
event.
Rxstat2
00RxEMPTY
0114 OR LESS BYTES (4 if Flrx set)
1115 OR MORE BYTES (5 if Flrx set)
10RxOVERFLOW EXISTS
RxBS1Byte stat us
Txstat1Tx FIFO Status
Rxstat1Rx FIFO Status
Ta ble 2 - HDLC S tatus Bits
7-15
Loading...
+ 29 hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.