AD AD6622PCB

FEATURES
Wideband Digital IF Parallel Output
Wideband Digital IF Parallel Input
Allows Cascade of Chips for Additional Channels
Programmable IF and Modulation for Each Channel
Programmable Interpolating RAM Coefficient Filter
High-Speed CIC Interpolating Filter
NCO Frequency Translation
Worst Spur Better than 100 dBc
Tuning Resolution Better than 0.02 Hz
Real or Complex Outputs
Digital Summation of Channels
Clipped or Wrapped Overrange
Two’s Complement or Offset Binary Output
Separate 3-Wire Serial Data Input for Each Channel
Microprocessor Control
JTAG Boundary Scan
FUNCTIONAL BLOCK DIAGRAM
CH A
SPORT
RCF
CIC
FILTER
NCO
CH B
SPORT
RCF
CIC
FILTER
NCO
CH C
SPORT
RCF
CIC
FILTER
NCO
RCF
CIC
FILTER
NCO
CH D
SPORT
JTAG
18
SUMMATION
a
Four-Channel, 75 MSPS Digital
Transmit Signal Processor (TSP)
AD6622
18
␮PORT
APPLICATIONS
Cellular/PCS Base Stations
Micro/Pico Cell Base Stations
WBCDMA
Wireless Local Loop Base Stations
Phase Array Beam Forming Antennas
PRODUCT DESCRIPTION
The AD6622 comprises four identical digital Transmit Signal
Processors (TSPs) complete with synchronization circuitry and
cascadable wideband channel summation. An external digitalto-analog converter (DAC) is all that is required to complete a
wide band digital up-converter. On-chip tuners allow the relative
phase and frequency for each RF carrier to be independently
controlled.
Each TSP has three cascaded signal processing elements: a
RAM-programmable Coefficient interpolating Filter (RCF), a
programmable Cascaded Integrator Comb (CIC) interpolating
filter, and a Numerically Controlled Oscillator/tuner (NCO).
The outputs of the four TSPs are summed and scaled on-chip.
In multichannel wideband transmitters, multiple AD6622s may
be combined using the chip’s cascadable output summation stage.
Each channel provides independent serial data inputs that may
be directly connected to the serial port of DSP chips. User programmable FIR filters can be used to filter linear inputs.
All control registers and coefficient values are programmed through
a generic microprocessor interface. Two microprocessor bus
modes are supported. All inputs and outputs are LVCMOS
compatible. All outputs are LVCMOS and 5 V TTL compatible.
REV. 0
Information furnished by Analog Devices is believed to be accurate and
reliable. However, no responsibility is assumed by Analog Devices for its
use, nor for any infringements of patents or other rights of third parties
which may result from its use. No license is granted by implication or
otherwise under any patent or patent rights of Analog Devices.
One Technology Way, P.O. Box 9106, Norwood, MA 02062-9106, U.S.A.
Tel: 781/329-4700
World Wide Web Site: http://www.analog.com
Fax: 781/326-8703
© Analog Devices, Inc., 2000
AD6622–SPECIFICATIONS
RECOMMENDED OPERATING CONDITIONS
Parameter
Test
Level
Min
VDD
TAMBIENT
IV
IV
2.4
–40
Test
Level
Min
AD6622AS
Typ
3.0
+25
Max
Unit
3.3
+70
V
°C
Max
Unit
VDD + 0.3
+0.8
10
10
V
V
µA
µA
pF
ELECTRICAL CHARACTERISTICS
AD6622AS
Typ
Parameter (Conditions)
Temp
LOGIC INPUTS (5 V TOLERANT)
Logic Compatibility
Logic “1” Voltage
Logic “0” Voltage
Logic “1” Current
Logic “0” Current
Input Capacitance
Full
Full
Full
Full
Full
25°C
IV
IV
IV
IV
V
2.0
–0.3
LOGIC OUTPUTS
Logic Compatibility
Logic “1” Voltage (IOH = 0.25 mA)
Logic “0” Voltage (IOL = 0.25 mA)
Full
Full
Full
IV
IV
VDD – 0.05
IDD SUPPLY CURRENT
CLK = 60 MHz, 3.3 V1
CLK = GSM Example
CLK = IS-136 Example
CLK = WBCDMA Example
Sleep Mode
POWER DISSIPATION
CLK = 60 MHz, 3.3 V1
CLK = GSM Example
CLK = IS-136 Example
CLK = WBCDMA Example
Sleep Mode
3.0 V CMOS
Full
Full
Full
Full
1
1
4
VDD – 0.035
0.02
0.05
IV
V
V
V
IV
506
2972
2402
2092
0.1
5661
IV
V
V
V
IV
1.77
0.892
0.722
0.6272
0.33
1.87
0.5
1.65
V
V
mA
mA
mA
mA
mA
W
W
W
W
mW
NOTES
1
This specification denotes an absolute maximum supply current for the device. The conditions include all channels active, minimum interpolation in both CIC stages,
maximum switching of input data, and maximum VDD of 3.3 V. In an actual application the power will be less; see the Thermal Management section of the data sheet
for further details.
2
GSM interpolation = 120 at 65 MHz, 4 channels active, IS-136 interpolation = 2560 at 62.208 MHz, 4 channels active. WBCDMA interpolation = 64, 4 channels
interleaved at 61.44 MHz.
Specifications subject to change without notice.
–2–
REV. 0
AD6622
TIMING CHARACTERISTICS1 (C
LOAD
= 40 pF, all outputs unless specified)
Temp
Test
Level
Min
AD6622AS
Typ
Max
Unit
CLK Timing Requirements:
tCLK
CLK Period
CLK Width Low
tCLKL
tCLKH
CLK Width High
Full
Full
Full
IV
IV
IV
13.3
5.5
5.5
0.5 × tCLK
0.5 × tCLK
ns
ns
ns
RESET Timing Requirements:
tRESL
RESET Width Low
Full
IV
30.0
ns
Input Wideband Data Timing Requirements:
tSI
Input to CLK Setup Time
tHI
Input to CLK Hold Time
Full
Full
IV
IV
0.5
3.5
ns
ns
Parallel Output Switching Characteristics:
tSO
CLK to Output Setup Time
tHO
CLK to Output Hold Time
tZO
Output Three-State Time
Full
Full
Full
IV
IV
V
4.1
SYNC Timing Requirements:
tSS
SYNC to CLK Setup Time
tHS
SYNC to CLK Hold Time
Full
Full
IV
IV
2.6
1.5
Serial Port Timing Requirements:
tDSCLK
CLK to SCLK Delay
tDSDFS
SCLK to SDFS Delay
tSSI
SDI to SCLK Setup Time
SDI to SCLK Hold Time
tHSI
tSCS
Serial Clock Skew
Full
Full
Full
Full
Full
V
IV
IV
IV
IV
–1.2
8.5
5.5
MODE INM Write Timing:
tHWR
WR(R/W) to RDY(DTACK) Hold Time
tSAM
Address/Data to WR(R/W) Setup Time
Address/Data to RDY(DTACK) Hold Time
tHAM
tDRDY
WR(R/W) to RDY(DTACK) Delay
tACCFAST
WR(R/W) to RDY(DTACK) High Delay
tACCMEDIUM WR(R/W) to RDY(DTACK) High Delay
tACCSLOW
WR(R/W) to RDY(DTACK) High Delay
Full
Full
Full
Full
Full
Full
Full
IV
IV
IV
IV
IV
IV
IV
0
0
0
MODE INM Read Timing:
Address to RD(DS) Setup Time
tSAM
tHA
Address to Data Hold Time
Data Three-State Delay
tZD
tDD
RDY(DTACK) to Data Delay
tDRDY
RD(DS) to RDY(DTACK) Delay
RD(DS) to RDY(DTACK) High Delay
tACCFAST
tACCMEDIUM RD(DS) to RDY(DTACK) High Delay
tACCSLOW
RD(DS) to RDY(DTACK) High Delay
Full
Full
Full
Full
Full
Full
Full
Full
IV
IV
IV
IV
IV
IV
IV
IV
0
0
3.4
Name
Parameter (Conditions)
12
5
ns
ns
ns
ns
ns
8.5
+2.4
7
ns
ns
ns
ns
ns
MICROPROCESSOR PORT, MODE INM (MODE = 0)
REV. 0
–3–
2 × tCLK
3 × tCLK
4 × tCLK
2 × tCLK
3 × tCLK
4 × tCLK
7
10.2
3 × tCLK
4 × tCLK
5 × tCLK
ns
ns
ns
ns
ns
ns
ns
10.5
tCLK – 10
10.2
3 × tCLK
4 × tCLK
5 × tCLK
ns
ns
ns
ns
ns
ns
ns
ns
AD6622
Temp
Test
Level
MODE MNM Write Timing:
tHDS
DS(RD) to DTACK(RDY) Hold Time
tHRW
R/W(WR) to DTACK(RDY) Hold Time
Address/Data to R/W(WR) Setup Time
tSAM
tHAM
Address/Data to R/W(WR) Hold Time
tDDTACK
DS(RD) to DTACK(RDY) Delay
R/W(WR) to DTACK(RDY) Low Delay
tACCFAST
tACCMEDIUM R/W(WR) to DTACK(RDY) Low Delay
tACCSLOW
R/W(WR) to DTACK(RDY) Low Delay
Full
Full
Full
Full
Full
Full
Full
Full
IV
IV
IV
IV
IV
IV
IV
IV
0
0
0
0
MODE MNM Read Timing:
tSAM
Address to DS(RD) Setup Time
Address to Data Hold Time
tHA
Data Three-State Delay
tZD
tDD
DTACK(RDY) to Data Delay
DS(RD) to DTACK(RDY) Delay
tDDTACK
tACCFAST
DS(RD) to DTACK(RDY) Low Delay
tACCMEDIUM DS(RD) to DTACK(RDY) Low Delay
tACCSLOW
DS(RD) to DTACK(RDY) Low Delay
Full
Full
Full
Full
Full
Full
Full
Full
IV
IV
IV
IV
IV
IV
IV
IV
0
0
0
Name
Parameter (Conditions)
Min
AD6622AS
Typ
Max
Unit
MICROPROCESSOR PORT, MODE MNM (MODE = 1)
2 × tCLK
3 × tCLK
4 × tCLK
2 × tCLK
3 × tCLK
4 × tCLK
1 × tCLK
3 × tCLK
4 × tCLK
5 × tCLK
ns
ns
ns
ns
ns
ns
ns
ns
tCLK – 10
1 × tCLK
3 × tCLK
4 × tCLK
5 × tCLK
ns
ns
ns
ns
ns
ns
ns
ns
NOTES
1
All Timing Specifications valid over VDD range of 2.4 V to 3.3 V.
Specifications subject to change without notice.
tCLK
CLK
tCLKL
tSI
CLK
tHI
tCLKH
IN[17:0],
QIN
tSO
Figure 3. Wideband Input Timing
OUT[17:0],
QOUT
tHO
tZO
tZO
CLK
OEN
tSS
Figure 1. Parallel Output Switching Characteristics
tHS
SYNC
Figure 4. SYNC Timing Inputs
CLK
tDSCLK
SCLK
CLKn
tDSDFS
tDSDFS
SDFS
tSSI
SDI
tHSI
DATAn
Figure 2. Serial Port Switching Characteristics
–4–
REV. 0
AD6622
RD (DS)
tHWR
WR (R/W)
CS
tHAM
tSAM
VALID ADDRESS
A[2:0]
tHAM
tSAM
VALID DATA
D[7:0]
tDRDY
RDY
(DTACK)
tACC
1. tACC ACCESS TIME DEPENDS ON THE ADDRESS ACCESSED. ACCESS TIME IS MEASURED FROM THE FE OF WR TO THE RE OF RDY.
2. tACCFAST REQUIRES A MAXIMUM OF THREE CLK PERIODS AND APPLIES TO A[2:0] = 7, 6, 5, 3, 2, 1
3. tACCMEDIUM REQUIRES A MAXIMUM OF FOUR CLK PERIODS AND APPLIES TO A[2:0] = 4 AND 0 IF THE ACCESS IS TO A CONTROL REGISTER
VERSUS A RAM REGISTER.
4. tACCSLOW REQUIRES A MAXIMUM OF FIVE CLK PERIODS AND APPLIES TO A[2:0] = 0 WHEN ACCESSING RAM REGISTERS.
Figure 5. INM Microport Write Timing Requirements
RD (DS)
WR (R/W)
CS
tSAM
A[2:0]
VALID ADDRESS
tHA
tDD
tZD
tZD
VALID DATA
D[7:0]
tDRDY
RDY
(DTACK)
tACC
1. tACC ACCESS TIME DEPENDS ON THE ADDRESS ACCESSED. ACCESS TIME IS MEASURED FROM THE FE OF WR TO THE RE OF RDY.
2. tACCFAST REQUIRES A MAXIMUM OF THREE CLK PERIODS AND APPLIES TO A[2:0] = 7, 6, 5, 3, 2, 1
3. tACCMEDIUM REQUIRES A MAXIMUM OF FOUR CLK PERIODS AND APPLIES TO A[2:0] = 4 AND 0 IF THE ACCESS IS TO A CONTROL REGISTER
VERSUS A RAM REGISTER.
4. tACCSLOW REQUIRES A MAXIMUM OF FIVE CLK PERIODS AND APPLIES TO A[2:0] = 0 WHEN ACCESSING RAM REGISTERS.
Figure 6. INM Microport Read Timing Requirements
REV. 0
–5–
AD6622
tHDS
DS (RD)
tHRW
R/W (WR)
CS
tSAM
tHAM
VALID ADDRESS
A[2:0]
tSAM
tHAM
VALID DATA
D[7:0]
tDDTACK
DTACK
(RDY)
tACC
1. tACC ACCESS TIME DEPENDS ON THE ADDRESS ACCESSED. ACCESS TIME IS MEASURED FROM THE FE OF DS TO THE FE OF DTACK.
2. tACCFAST REQUIRES A MAXIMUM OF FOUR CLK PERIODS AND APPLIES TO A[2:0] = 7, 6, 3, 2, 1
3. tACCMEDIUM REQUIRES A MAXIMUM OF FIVE CLK PERIODS AND APPLIES TO A[2:0] = 4, 5, AND 0 IF THE ACCESS IS TO A CONTROL REGISTER
VERSUS A RAM REGISTER.
4. tACCSLOW REQUIRES A MAXIMUM OF SIX CLK PERIODS AND APPLIES TO A[2:0] = 0 WHEN ACCESSING RAM REGISTERS.
Figure 7. MNM Microport Write Timing Requirements
tHDS
DS (RD)
R/W (WR)
CS
tSAM
A[2:0]
VALID ADDRESS
tHA
tDD
tZD
tZD
VALID DATA
D[7:0]
tDDTACK
DTACK
(RDY)
tACC
1. tACC ACCESS TIME DEPENDS ON THE ADDRESS ACCESSED. ACCESS TIME IS MEASURED FROM THE FE OF DS TO THE FE OF DTACK.
2. tACCFAST REQUIRES A MAXIMUM OF FOUR CLK PERIODS AND APPLIES TO A[2:0] = 7, 6, 3, 2, 1
3. tACCMEDIUM REQUIRES A MAXIMUM OF FIVE CLK PERIODS AND APPLIES TO A[2:0] = 4, 5, AND 0 IF THE ACCESS IS TO A CONTROL REGISTER
VERSUS A RAM REGISTER.
4. tACCSLOW REQUIRES A MAXIMUM OF SIX CLK PERIODS AND APPLIES TO A[2:0] = 0 WHEN ACCESSING RAM REGISTERS.
Figure 8. MNM Microport Read Timing Requirements
–6–
REV. 0
AD6622
ABSOLUTE MAXIMUM RATINGS*
EXPLANATION OF TEST LEVELS
Supply Voltage . . . . . . . . . . . . . . . . . . . . . . . . –0.3 V to +3.6 V
Input Voltage . . . . –0.3 V to VDD +0.3 V (Not 5 V Tolerant)
IN[17:0], QIN, OEN
Input Voltage . . . . . . . . . . . . . –0.3 V to +3.6 V (5 V Tolerant)
CLK, RESET, DS, R/W, MODE, A[2:0], D[7:0], SYNC, TRST,
TCK, TMS, TDI, SDINA, SDINB, SDINC, SDIND
Output Voltage Swing . . . . . . . . . . . . –0.3 V to VDD + 0.3 V
Load Capacitance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 pF
Junction Temperature Under Bias . . . . . . . . . . . . . . . . . 125°C
Storage Temperature Range . . . . . . . . . . . . –65°C to +150°C
Lead Temperature (5 sec) . . . . . . . . . . . . . . . . . . . . . . . 280°C
I.
100% Production Tested.
II. 100% Production Tested at 25°C, and Sample Tested at
Specified Temperatures.
III. Sample Tested Only.
IV. Parameter Guaranteed by Design and Analysis.
V. Parameter is Typical Value Only.
VI. 100% Production Tested at 25°C, and Sample Tested at
Temperature Extremes.
*Stresses greater than those listed above may cause permanent damage to the
device. These are stress ratings only; functional operation of the devices at these
or any other conditions greater than those indicated in the operational sections of
this specification is not implied. Exposure to absolute maximum rating conditions
for extended periods may affect device reliability.
THERMAL CHARACTERISTICS
128-Lead MQFP:
θJA = 33°C/W, No Airflow
θJA = 27°C/W, 200 LFPM Airflow
θJA = 24°C/W, 400 LFPM Airflow
θJC = 5.5°C/W
Thermal measurements made in the horizontal position on a
2-layer board.
ORDERING GUIDE
Model
Temperature Range
Package Description
Package Option
AD6622AS
AD6622S/PCB
–40°C to +70°C (Ambient)
128-Lead MQFP (Metric Quad Flatpack)
Evaluation Board with AD6622 and Software
S-128A
CAUTION
ESD (electrostatic discharge) sensitive device. Electrostatic charges as high as 4000 V readily
accumulate on the human body and test equipment and can discharge without detection. Although
the AD6622 features proprietary ESD protection circuitry, permanent damage may occur on
devices subjected to high-energy electrostatic discharges. Therefore, proper ESD precautions are
recommended to avoid performance degradation or loss of functionality.
REV. 0
–7–
WARNING!
ESD SENSITIVE DEVICE
AD6622
65 GND
67 CLK
66 VDD
68 GND
69 GND
71 IN17
70 QIN
72 GND
73 GND
74 IN16
75 IN15
76 IN14
78 VDD
77 IN13
79 IN12
80 IN11
81 IN10
83 GND
82 IN9
85 GND
84 GND
86 IN8
88 IN6
87 IN7
90 VDD
89 IN5
91 IN4
93 IN2
92 IN3
95 GND
94 IN1
97 IN0
96 GND
99 GND
98 GND
101 TCK
100 TRST
102 GND
PIN CONFIGURATION
GND 103
VDD 104
63 GND
GND 105
62 SYNC
TMS 106
TDO 107
61 RESET
60 CS
59 VDD
64 GND
TDI 108
SCLKA 109
58 A0
VDD 110
SDFSA 111
56 A2
57 A1
SDINA 112
55 MODE
54 GND
SCLKB 113
AD6622
SDFSB 114
GND 115
GND 116
53 GND
52 GND
TOP VIEW
(Not to Scale)
51 R/W(WR)
GND 117
SDINB 118
SCLKC 119
50 DTACK(RDY)
SDFSC 120
SDINC 121
47 VDD
49 DS(RD)
48 D0
46 D1
VDD 122
SCLKD 123
44 D3
SDFSD 124
SDIND 125
42 GND
45 D2
43 D4
GND 126
VDD 127
41 VDD
40 D5
PIN 1
IDENTIFIER
D6 37
GND 38
GND 35
GND 36
D7 33
GND 34
GND 32
GND 31
QOUT 30
OUT17 29
OUT15 27
24
OUT13
OUT16 28
23
OUT12
VDD 26
22
OUT11
OUT14 25
21
GND
GND 20
GND 19
OUT9 17
OUT10 18
9
GND
OUT8 16
8
OUT2
OUT7 15
7
OUT1
VDD 14
6
OUT0
OUT6 13
5
GND
OUT5 12
4
GND
OUT4 11
3
GND
OUT3 10
1
2
OEN
39 GND
GND
GND 128
DENOTES I/O POWER PIN
DENOTES CORE POWER PIN
–8–
REV. 0
AD6622
PIN FUNCTION DESCRIPTIONS
Pin Number
Name
1, 3–5, 9, 19–21, 31, 32,
GND
34–36, 38, 39, 42, 52–54,
63–65, 68, 69, 72, 73, 83–85,
95, 96, 98, 99, 102, 103,
105, 115–117, 126, 128
2
OEN
Type
Description
P
Ground Connection
I
Active High Output Enable Pin (Actively Pulled Down If Not Connected)
(Not 5 V Tolerant)
Wideband Output Data
27–29, 22–25, 15–18, 10–13,
6–8
14, 26, 41, 47, 122
59, 66, 78, 90, 104, 110, 127
30
33, 37, 40, 43–46, 48
49
50
OUT[17:0]
O/T
VDD
VDD
QOUT
D[7:0]
DS (RD)
DTACK (RDY)
P
P
O/T
I/O/T
I
O
51
55
56–58
60
61
62
R/W (WR)
MODE
A[2:0]
CS
RESET
SYNC
I
I
I
I
I
I
67
70
CLK
QIN
I
I
71, 74–77, 79–82, 86–89,
91–94, 97
100
101
106
107
108
109
111
112
113
114
118
119
120
121
123
124
125
IN[17:0]
I
TRST
TCK
TMS
TDO
TDI
SCLKA
SDFSA
SDINA
SCLKB
SDFSB
SDINB
SCLKC
SDFSC
SDINC
SCLKD
SDFSD
SDIND
I
I
I
O
I
O
O
I
O
O
I
O
O
I
O
O
I
REV. 0
+3.0 V Supply (I/O Supply)
+3.0 V Supply (Core Supply)
Indicates Q Output Data (Complex Output Mode)
Microprocessor Interface Data
INM Mode: Read Signal, MNM Mode: Data Strobe Signal
Acknowledgment of a Completed Transaction (Signals when µP Port
Is Ready for an Access) Open Drain, Must Be Pulled Up Externally
Read/Write Line (Write Signal)
Sets Microport Mode: MODE = 1, MNM Mode; MODE = 0, INM Mode
Microprocessor Interface Address
Chip Select, Enable the Chip for µP Access
Active Low Reset Pin (Actively Pulled Up If Not Connected)
SYNC Signal for Synchronizing Multiple AD6622s (Actively Pulled
Down If Not Connected)
Input Clock (Actively Pulled Down If Not Connected)
Indicates Q Input Data (Complex Input Mode) (Actively Pulled Down
If Not Connected) (Not 5 V Tolerant)
Wideband Input Data (Allows Cascade of Multiple AD6622 Chips In
a System) (Actively Pulled Down If Not Connected) (Not 5 V Tolerant)
Test Reset Pin (Actively Pulled Up If Not Connected)
Test Clock Input (Actively Pulled Down If Not Connected)
Test Mode Select (Actively Pulled Up If Not Connected)
Test Data Output
Test Data Input (Actively Pulled Down If Not Connected)
Serial Clock Output Channel A
Serial Data Frame Sync Output Channel A
Serial Data Input Channel A (Actively Pulled Down If Not Connected)
Serial Clock Output Channel B
Serial Data Frame Sync Output Channel B
Serial Data Input Channel B (Actively Pulled Down If Not Connected)
Serial Clock Output Channel C
Serial Data Frame Sync Output Channel C
Serial Data Input Channel C (Actively Pulled Down If Not Connected)
Serial Clock Output Channel D
Serial Data Frame Sync Output Channel D
Serial Data Input Channel D (Actively Pulled Down If Not Connected)
–9–
AD6622
AD6622
SPORT
SDINB
SDFSB
SCLKB
SPORT
SDINC
SDFSC
SCLKC
SPORT
DATA
DATA
DATA
I
RCF
Q
I
RCF
Q
RCF
Q
I
I
CIC
FILTER Q
NCO
I
CIC
FILTER Q
NCO
I
CIC
FILTER Q
NCO
DATa
QIN
IN
[17:0]
DATb
SUMMATION
SDINA
SDFSA
SCLKA
DATc
SYNC
OEN
QOUT
SDIND
SDFSD
SCLKD
SPORT
DATA
I
RCF
I
CIC
FILTER Q
Q
NCO
DATd
OUT
[17:0]
CLK
RESET
CS
A[2:0]
R/W
MODE
DS
D[7:0]
TMS
TRST
TDI
TDO
TCK
DTACK
MICROPORT
JTAG
Figure 9. Functional Block Diagram
THEORY OF OPERATION
As digital-to-analog converters (DACs) achieve higher sampling
rates, analog bandwidth, and dynamic range, it becomes increasingly attractive to accomplish the first IF stage of a transmitter
in the digital domain. Digital IF signal processing provides
repeatable manufacturing, higher accuracy, and more flexibility
than comparable high-dynamic-range analog designs.
The AD6622 Four-Channel Transmit Signal Processor (TSP) is
designed to bridge the gap between DSPs and high-speed DACs.
The wide range of interpolation factors in each filter stage makes
the AD6622 useful for creating both narrowband and wideband
carriers in a high-speed sample stream. The high-resolution NCO
allows flexibility in frequency planning and supports both digital
and analog air interface standards. The RAM-based architecture allows easy reconfiguration for multimode applications.
The interpolating filters remove unwanted images of signals
sampled at a fraction of the wideband rate. When the channel of
interest occupies far less bandwidth than the wideband output
signal, rejecting out-of-band noise is called “processing gain.”
For large interpolation factors, this processing gain allows a
14-bit DAC to express the sum of multiple 16-bit signals sampled
at a lower rate without significantly increasing the noise floor
about each carrier. In addition, the programmable RAM coefficient stage allows anti-imaging, and static equalization functions
to be combined in a single, cost-effective filter.
The high-speed NCO can be used to tune a quadrature sampled
signal to an IF channel, or the NCO can be directly frequencymodulated at an IF channel. Multicarrier phase synchronization
pins and phase offset registers allow intelligent management of
the relative phase of the independent RF channels. This capability
supports the requirements for phased array antenna architectures and management of the wideband peak/power ratio to
minimize clipping at the DAC.
noise. The wideband ports can be configured for real or quadrature outputs. Quadrature sampled outputs (I and Q) are limited
to half the master clock rate on the shared output bus.
FUNCTIONAL OVERVIEW
The following descriptions explain the functionality of each of
the core sections of the AD6622. Detailed timing, application,
and specifications are described in detail in their respective portions of the data sheet.
SERIAL DATA PORT
The AD6622 has four independent Serial Ports (A, B, C, and
D) of which accepts data to its own channel (1, 2, 3, or 4) of
the device. Each serial port has three pins: SCLK, SDFS, and
SDIN. The SCLK and SDFS pins are outputs that provide
serial clock and framing. The SDIN pins are inputs that accept
channel data. The serial ports do not accept configuration or
control inputs. The serial ports do not accept external clock
or framing signals, although it is possible to synchronize the
AD6622 serial ports to meet an external timing requirement.
The serial clock output, SCLK, is created by a programmable
internal counter that divides down the master clock. When the
channel is reset, SCLK is held low. SCLK starts on the first
rising edge of CLK after Channel Reset is removed (D0 through
D3 of External Address 4). Once active, the SCLK frequency is
determined by the master CLK frequency and the SCLK divider,
according to the equation below. The SCLK divider is a 5-bit
unsigned value located in Channel Register 0x0D. The user must
select the SCLK divider to ensure that SCLK is fast enough to
accept full input sample words at the input sample rate. See the
design example at the end of this section. The maximum SCLK
frequency is 1/2 of the master clock frequency. The minimum
SCLK frequency is 1/64 of the master clock frequency.
The wideband input and output ports allow multiple AD6622s
to be cascaded into a single DAC. The master clock for the
entire system is based on the DAC clock rate (up to 75 MSPS).
The external 18-bit resolution reduces summation of truncation
–10–
f SCLK =
fCLK
2 × (SCLK DIVIDER + 1)
(1)
REV. 0
AD6622
The serial data frame sync output, SDFS, is pulsed high for one
SCLK cycle at the input sample rate. The input sample rate is
determined by the master clock divided by channel interpolation
factor. If the SCLK rate is not an integer multiple of the input
sample rate, the SDFS will continually adjust the period by one
SCLK cycle in order to keep the average SDFS rate equal to the
input sample rate. When the channel is in sleep mode, SDFS is
held low. The first SDFS is delayed by the channel reset latency
after the Channel Reset is removed. The channel reset latency
varies dependent on channel configuration.
The serial data input, SDIN, accepts 32-bit words as channel
input data. The 32-bit word is interpreted as two 16 bit two’s
complement quadrature words, I followed by Q, MSB first.
The first bit is shifted into the serial port starting on the second
rising edge of SCLK after SDFS goes high, as shown by the
timing diagram below.
PROGRAMMABLE INTERPOLATING RAM
COEFFICIENT FILTER (RCF)
Each channel has a fully independent RAM Coefficient Filter
(RCF). The RCF accepts data from the serial port, filters it, and
passes the result to the CIC filter. The RCF implements a FIR
filter with optional interpolation. The FIR filter can produce
impulse responses up to 128 output samples long. The FIR
response may be interpolated up to a factor of 128, although
the best filter performance is usually achieved if the RCF interpolation factor is confined to 8 or below.
FIR Filter Implementation
The RCF accepts quadrature samples from the serial port with a
fixed point resolution of 16 bits each, for I and Q.
SDFS
SCLK
SDIN
RCF
DATA
MEM
SERIAL
PORT
16,16
CLK
16,16
ACCUMULATOR
tDSCLK
IQ TO
CIC
FILTER
16,16
SCLK
CLKn
tDSDFS
RCF COARSE
SCALE
COEFFICIENT
MEM
tDSDFS
Figure 11. RCF Block Diagram
SDFS
tSSI
tHSI
DATAn
SDI
Figure 10. Serial Port Switching Characteristics
As an example of the serial port operation, consider a CLK frequency of 62.208 MSPS and a channel interpolation of 2560.
In that case, the input sample rate is 24.3 kSPS (62.208 MSPS/
2560), which is also the SDFS rate. Substituting, fSCLK ≥ 32 ×
fSDFS into the equation below and solving for SCLKDIVIDER,
we find the maximum value for SCLKDIVIDER according to
Equation 2.
SCLK DIVIDER ≤
fCLK
64 × fSDFS
–1
In conclusion, the SDFS rate is determined by the AD6622 master
clock rate and the interpolation rate of the channel. The SDFS
rate is equal to the channel input rate. The channel interpolation is equal to RCF interpolation times CIC5 interpolation,
times CIC2 interpolation
(3)
The SCLK rate is determined by the AD6622 master clock
rate and SCLKDIVIDER. The SCLK is a divided version of the
AD6622 master CLK. The SCLK divide ratio is determined by
SCLKDIVIDER as shown in Equation 2. The SCLK must be fast
enough to input 32 bits of data prior to the next SDFS. Extra
SCLKs are ignored by the serial port.
REV. 0
fIN ⴛ LRCF
fIN
LRCF
a
b
NRCF TAP fIN ⴛ LRCF
FIR FILTER
h[n]
c
Figure 12. RCF Interpolation
(2)
Evaluating this equation for our example, SCLKDIVIDER must be
less than or equal to 39. Since the SCLKDIVIDER channel register is a 5-bit unsigned number it can only range from 0 to 31.
Any value in that range will be valid for this example, but if it is
important that the SDFS period is constant, then there is another
restriction. For regular frames, the ratio fSCLK/fSDFS must be equal
to an integer of 32 or larger. For this example, constant SDFS
periods can only be achieved with an SCLK divider of 19.
L = LRCF × LCIC 5 × LCIC 2
The AD6622 RCF realizes a sum-of-products filter using a polyphase implementation. This mode is equivalent to an interpolator followed by a FIR filter running at the interpolated rate. In
Figure 12, the interpolating block increases the rate by the RCF
interpolation factor (LRCF) by inserting LRCF-1 zero valued samples
between every input sample. The next block is a filter with a finite
impulse response length (NRCF) and an impulse response of h[n],
where n is an integer from 0 to NRCF-1.
The difference equation for Figure 12 is written below, where
h[n] is the RCF impulse response, b[n] is the interpolated input
sample sequence at point “b” in Figure 12, and c[n] is the output sample sequence at point “c” in the Figure 12.
N RCF −1
c[n ] = ∑ h[k − n ] × b[n ]
(4)
k=0
This difference equation can be described by the transfer function from point “b” to “c” as shown Equation 5.
N RCF −1
H bc ( z ) = ∑ h[n ] × z − n
(5)
n=0
The actual implementation of this filter uses a polyphase
decomposition to skip the multiply-accumulates when b[n] is
zero. Compared to the diagram above, this implementation has
the benefits of reducing by a factor of LRCF both the time needed to
calculate an output and the required data memory (DMEM). The
price of these benefits is that the user must place the coefficients
into the coefficient memory (CMEM) indexed by the interpolation phase. The process of selecting the coefficients and placing
them into the CMEM is broken into three steps shown below.
–11–
AD6622
1. Select the Impulse Response Length (NRCF) and the Interpolation Factor (L RCF ). The Impulse Response Length
(NRCF) is limited in three ways: by the available calculation
time, by the data memory size (DMEM), and by the coefficient memory size (CMEM). The equation below shows
that NRCF is limited to the minimum of these three conditions.
Time
Restriction
↓
N RCF
–1
LRCF
k=0
CMEM
Restriction
↓
L

N RCF ≤ min  , 16 × LRCF , 128
2

↑
DMEM
Restriction
(6)
where:
L = LRCF × LCIC5 × LCIC2
2. The interpolation rate (LRCF) may be any integer of NRCF
ranging from 1 to 128, while meeting the above equation.
Most filter designs can be optimized by choosing the smallest LRCF that does not compromise the image rejection of
the subsequent CIC filter. The quality of an interpolating
filter is a strong function of the NRCF/LRCF ratio and a weaker
function of NRCF. The best filters are usually achieved by
maximizing NRCF/LRCF (no larger than 16) and then increasing
both NRCF and LRCF by the same ratio until the filter becomes
time or CMEM limited.
3. Once NRCF and LRCF are selected, Channel Register 0x0A
is programmed to NRCF – 1, and Channel Register 0x0C is
programmed to NRCF/LRCF – 1.
ChannelCenterGainp = 2− g × ∑ h[k × LRCF + p ]
6. The worst-case peak is calculated similarly to the channel
center gain, except that the input sequence swings from fullscale positive to full-scale negative to match the polarity of the
coefficient by which it will be multiplied, so that each product is positive. This results in a maximal that must be less
than one to guarantee no possibility of wrapping. Note that
when LRCF is greater than one, each phase may produce its
worst-case peak in response to a different input sequence.
7. Programming DMEM and CMEM. The DMEM must be
initialized to all zeros to avoid any unpredictable start-up
transients since a reset does not clear the memory. The
impulse response h[n] must be reordered by phase for the
CMEM as shown in the code below. Several filters with
impulse lengths that total less than 128 can be programmed
into the CMEM simultaneously and selected later using the
RCF offset pointer (ORCF) which is set by Channel Register
0x0B.
/* Reorder Fir Coefficients for AD6622 CMEM */
for (p=0; p<L_RCF; p++)
for (k=0; k<N_RCF/L_RCF; k++)
CMEM[O_RCF + p*N_RCF/L_RCF + k] = C[k*L_RCF +p];
/* End of routine */
Table I. RCF Control Registers
4. Determine the Impulse Response. The impulse response
relative to the RCF output rate can be calculated using ordinary FIR design techniques. In most cases, it is desirable to
precompensate the inband frequency roll-off of the CIC filter that follows. There are no symmetry requirements, so the
RCF can also be used for static phase equalization. The
impulse response must be quantized to 16-bit two’s complement numbers for the CMEM. The channel center gain and
worst-case peak can be calculated for each of the LRCF phases
(p) according to the equations below. A RCF coarse scale
factor (g) that ranges between 0 and 3 is provided to limit
the gain without excessive loss of resolution in the CMEM.
The coarse scale factor is located in Channel Register 0x0D.
N RCF
–1
LRCF
(8)
WorstCasePeakp = 2− g × ∑ |h[k × LRCF + p ]|
Channel
Address
Bit
Width
0x0A
8
0x0B
8
0x0C
8
0x0D
8
0x0E
0x0F
0x10
0x11
0x20–0x3F
0x80–0xFF
16
16
16
16
16
16
(7)
k=0
5. The channel center gain is the response to a constant fullscale input at every output phase. The summation is split
into phases because the interpolation of the data insures that
only NRCF/LRCF coefficients can be active for any single output.
For LRCF = 1, there is only one phase and the channel center
gain is the simple sum of all the coefficients, scaled by 2–g. If
the channel center gain is not the same for every value of p,
some or all of the images of the channel center will be
imperfectly rejected by the RCF.
–12–
Description
7: Reserved (Must Be Written to 0)
6–0: NRCF–1
7: Reserved (Must Be Written to 0)
6–0: ORCF
7–6: Reserved
5–4: Reserved (Must Be Written to 0)
3–0: NRCF/LRCF–1
7–6: RCF Coarse Scale:
00 = 0 dB
01 = –6 dB
10 = –12 dB
11 = –18 dB
5: Reserved (Must Be Written to 0)
4–0: Serial Clock Divider
15–0: Reserved
15–0: Reserved
15–0: Reserved (Must Be Written to 0)
15–0: Reserved (Must Be Written to 0)
15–0: Data Memory (DMEM)
15–0: Coefficient Memory (CMEM)
REV. 0
AD6622
CASCASDED INTEGRATOR COMB (CIC)
INTERPOLATING FILTER
The I and Q outputs of the RCF stage are interpolated in integer factors by two cascaded integrator comb (CIC) filters. The
CIC section is separated into three discrete blocks: a fifth order
filter (CIC5), a second order filter (CIC2), and a scaling block
(CIC Scaling). The CIC5 and CIC2 blocks each exhibit a gain
that increases with respect to their interpolation factors, LCIC5
and LCIC2. The product of these gains must be compensated for
in a shared CIC Scaling block.
2–CIC_SCALE
LCIC5
LCIC2
CIC_SCALE
CIC5
CIC2
Figure 13. CIC Data Path
CIC Scaling
The CIC5 and CIC2 stages have a baseband gain of LCIC54 ×
LCIC2. The CIC scaling block is used to avoid numeric overflow
in the CIC stages. The CIC scale block reduces the signal level
without truncation or loss of resolution. The overall gain of the
CIC section is given by Equation 9.
× LCIC 2 × 2

 L
× f 
sin  π CIC 5


f


1 
CIC 5
CIC5( f ) =
LCIC 5 

f  
 
 sin  π L



CIC 5 
(13)
10
–10
–30
(9)
– CIC _ Scale
5
As an example, we will consider an input from the RCF whose
bandwidth is 0.141 of the RCF output rate, centered at baseband. Interpolation by a factor of five reveals five images, as
shown in Figure 14.
–50
dB
CIC _ Gain =
4
LCIC 5
The frequency response of the CIC5 can be expressed as follows.
The initial 1/LCIC5 factor normalizes for the increased rate, which is
appropriate when the samples are destined for a DAC with a
zero order hold output. The maximum gain is (LCIC5)4 at baseband, but internal registers peak in response to various dynamic
inputs. As long as LCIC5 is confined to 32 or less, there is no
possibility of overflow at any register.
The value CIC_Scale may range from 0 to 25, and can be independently programmed for each channel at Control Register
0x06. CIC_Scale may be safely calculated according Equation 10
to ensure the net gain through the CIC stages.
–70
–90
–110
–130
4
CIC_Scale = ceil (log 2( LCIC 5 × LCIC 2 ))
(10)
–150
The ceil function is the next highest integer. While this normally
constitutes a small loss, it can be recovered in the RCF scaling.
Likewise, if the RCF output level is known to be less than full
scale, the CIC gain can be increased by reducing CIC_Scale.
CIC5
The CIC5 is a fifth order interpolating cascaded integrator comb
whose impulse response is completely defined by its interpolation factor, LCIC5. The value L CIC5 –1 can be independently
programmed for each channel at location 0x09. While this control register is 8-bits wide, LCIC5 should be confined to the range
from 1 to 32 to avoid the possibility of internal overflow for
full-scale inputs. The transfer function of the CIC5 is given
by the following equations with respect to the CIC5 output
sample rate, fSAMP5.
1 – z – L 
CIC5(z ) = 

 1 – z –1 
–2
–1
0
1
2
Figure 14. Unfiltered CIC Interpolation Image
The CIC5 rejects each of the undesired images while passing
the image at baseband. The images of a pure tone at channel
center (dc) are nulled perfectly, but as the bandwidth increases
the rejection is diminished. The lower band edge of the first
image always has the least rejection. In this example, the CIC5
is interpolating by a factor of five and the input signal has a bandwidth of 0.141 of the RCF output sample rate. The plot below
shows –110 dBc rejection of the lower band edge of the first
image. All other image frequencies have better rejection.
10
–10
–30
5
CIC 5
(11)
dB
–50
This polynomial fraction can be completely reduced as follows,
demonstrating a finite impulse response with perfect phase linearity for all values of LCIC5.
L

CIC 5( z ) =  ∑ z − k 


 k=0

5
LCIC 5–1
CIC 5–1
=
∑
k =1
k

j 2π
 z −1 − e L


CIC 5




–70
–90
–110
5
–130
(12)
–150
–2
–1
0
1
2
Figure 15. Filtered CIC5 Interpolation Images
REV. 0
–13–
AD6622
Table II lists maximum bandwidth that will be rejected to various
levels for CIC5 interpolation factors from 1 to 32. Figure 15
corresponds to the listing in the –110 dB column and the LCIC5
= 5 row. It is worth noting that the rejection of the CIC5 improves
as the interpolation factor increases.
Table III. Maximum LCIC2 Limits
Table II. CIC5 Alias Protection
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
–110 dB
–100 dB
–90 dB
–80 dB
–70 dB
Full
0.101
0.126
0.136
0.141
0.143
0.144
0.145
0.146
0.146
0.147
0.147
0.147
0.147
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
0.148
Full
0.127
0.159
0.170
0.175
0.178
0.179
0.180
0.181
0.182
0.182
0.182
0.183
0.183
0.183
0.183
0.183
0.183
0.183
0.184
0.184
0.184
0.184
0.184
0.184
0.184
0.184
0.184
0.184
0.184
0.184
0.184
Full
0.160
0.198
0.211
0.217
0.220
0.222
0.224
0.224
0.225
0.226
0.226
0.226
0.226
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.227
0.228
Full
0.203
0.246
0.262
0.269
0.272
0.275
0.276
0.277
0.278
0.278
0.279
0.279
0.279
0.280
0.280
0.280
0.280
0.280
0.280
0.280
0.280
0.280
0.280
0.281
0.281
0.281
0.281
0.281
0.281
0.281
0.281
Full
0.256
0.307
0.325
0.333
0.337
0.340
0.341
0.342
0.343
0.344
0.344
0.345
0.345
0.345
0.345
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
0.346
LCIC5
Max LCIC2
1–19
20
21
22
23
24
25
26
27
28
29
30
31
32
256
209
172
143
119
101
85
73
63
54
47
41
36
32
The transfer function of the CIC2 is given by the following
equations with respect to the CIC2 output sample rate, fOUT.
1 – z – L 
CIC 2( z ) = 

 1 – z –1 
2
CIC 2
(14)
This polynomial fraction can be completely reduced as follows,
demonstrating a finite impulse response with perfect phase linearity for all values of LCIC2.
2
L

CIC 2( z ) =  ∑ z − k  =


 k=0

CIC 2–1
LCIC 2–1
∑
k =1
k

j 2π
 z −1 − e L


CIC 2




2
(15)
The frequency response of the CIC2 can be expressed as follows.
The maximum gain is LCIC2 at baseband. The initial 1/LCIC2
factor normalizes for the increased rate, which is appropriate
when the samples are destined for a DAC with a zero order hold
output.

 LCIC2 × f  

 sin  π f


1 
OUT
CIC 2( f ) =

LCIC2 


f

 sin  π

 fOUT  

CIC2
The CIC2 is a second-order interpolating cascaded integrator
comb whose impulse response is completely defined by its interpolation factor, LCIC2. The value LCIC2–1 can be independently
programmed for each channel at location 0x08. While this control register is 8 bits wide, LCIC2 should be confined to the ranges
shown by the table below according to the interpolation factor
of the CIC5. Exceeding the recommended guidelines may result in
overflow for input sequences at or near full scale. While relatively
small values of LCIC5 allow for the larger overall interpolation
factors with minimal power consumption, LCIC5 should be maximized to achieve the best overall image rejection.
2
(16)
As an example, we will consider an input from the CIC5 whose
bandwidth is 0.0033 of the CIC5 rate, centered at baseband.
Interpolation by a factor of five reveals five images, as shown
below.
–14–
REV. 0
AD6622
Table IV lists maximum bandwidth that will be rejected to various
levels for CIC2 interpolation factors from 1 to 32. The example
above corresponds to the listing in the –110 dB column and
the LCIC2 = 5 row. It is worth noting that the rejection of the
CIC2 improves as the interpolation factor increases.
10
–10
–30
dB
–50
Table IV. CIC2 Alias Protection
–70
–90
–110
–130
–150
–2
–1
0
1
2
Figure 16. Unfiltered CIC2 Interpolation Images
The CIC2 rejects each of the undesired images while passing
the image at baseband. The images of a pure tone at channel
center (dc) are nulled perfectly, but as the bandwidth increases
the rejection is diminished. The lower band edge of the first
image always has the least rejection. In this example, the CIC2
is interpolating by a factor of five and the input signal has a bandwidth of 0.0033 of the CIC5 output sample rate. Figure 17
shows –110 dBc rejection of the lower band edge of the first
image. All other image frequencies have better rejection.
10
–10
–30
dB
–50
–70
–90
–110
–130
–150
–2
–1
0
1
2
Figure 17. Filtered CIC2 Interpolation Images
REV. 0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
–15–
–110 dB
–100 dB
–90 dB
–80 dB
–70 dB
Full
0.0023
0.0029
0.0032
0.0033
0.0034
0.0034
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
0.0035
Full
0.0040
0.0052
0.0057
0.0059
0.0060
0.0061
0.0062
0.0062
0.0062
0.0062
0.0062
0.0062
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
0.0063
Full
0.0072
0.0093
0.0101
0.0105
0.0107
0.0108
0.0109
0.0110
0.0110
0.0110
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0111
0.0112
0.0112
0.0112
0.0112
0.0112
0.0112
0.0112
0.0112
0.0112
Full
0.0127
0.0165
0.0179
0.0186
0.0189
0.0192
0.0193
0.0194
0.0195
0.0195
0.0196
0.0196
0.0196
0.0197
0.0197
0.0197
0.0197
0.0197
0.0197
0.0197
0.0197
0.0197
0.0197
0.0198
0.0198
0.0198
0.0198
0.0198
0.0198
0.0198
0.0198
Full
0.0226
0.0292
0.0316
0.0328
0.0334
0.0338
0.0341
0.0343
0.0344
0.0345
0.0346
0.0346
0.0347
0.0347
0.0347
0.0348
0.0348
0.0348
0.0348
0.0348
0.0348
0.0348
0.0348
0.0349
0.0349
0.0349
0.0349
0.0349
0.0349
0.0349
0.0349
AD6622
PHASE
OFFSET
16
16
I DATA
FROM CIC5
NCO
FREQUENCY
WORD
32
D
Q
32
ICOS
32
D
Q
ANGLE TO
CARTESIAN
CONVERSION
32
PHASE
ON
OFF
AMPLITUDE
PN
GEN.
ON
OFF
PN
GEN.
CLK
QSIN
Q DATA
FROM CIC5
Figure 18. NCO Block Diagram
NUMERICALLY CONTROLLED OSCILLATOR (NCO) TUNER
Each channel has a fully independent tuner. The tuner accepts
data from the CIC filter, tunes it to a digital Intermediate Frequency (IF), and passes the result to a shared summation block.
The tuner consists of a 32-bit quadrature NCO and a Quadrature
Amplitude Mixer (QAM). The NCO serves as a local oscillator
and the QAM translates the interpolated channel data from
baseband to the NCO frequency. The worst-case spurious signal
from the NCO is better than –100 dBc for all output frequencies.
The tuner can produce real or complex outputs as requested by
the shared summation block.
energy. If the phase truncation spurs are small, phase dither
will not be effective in reducing them further, but a slight elevation in total error energy will occur.
Amplitude Dither
In the complex mode, the NCO serves as a quadrature local
oscillator running at fCLK/2, capable of producing any frequency
between –f CLK /4 and +f CLK /4 with a resolution of fCLK /2 33
(0.0087 Hz for fCLK = 75 MHz).
Amplitude dither can also be used to improve spurious performance of the NCO. Amplitude dither is enabled by writing a one
to Bit 4 of Channel Register at 0x01. When enabled, amplitude
dither can reduce spurs due to truncation at the input to the QAM.
If the entire frequency word is close to a fraction that has a
small denominator, the spurs due to amplitude truncation will
be large and amplitude dither will spread these spurs effectively.
Amplitude dither also will increase the total error energy by
approximately 3 dB. For this reason amplitude dither should be
used judiciously.
Phase Offset
In the real mode, the NCO serves as a quadrature local oscillator running at fCLK, capable of producing any frequency between
–fCLK/2 and +fCLK/2 with a resolution of fCLK/232 (0.017 Hz for
fCLK = 75 MHz). The quadrature portion of the output is discarded. Negative frequencies are distinguished from positive
frequencies solely by spectral inversion.
The phase offset (Channel Register 0x04) adds an offset to the
phase accumulator of the NCO. This is a 16-bit register that is
interpreted as a 16-bit unsigned integer. Phase offset ranges
from 0 to nearly 2 π radians with a resolution of π/32768 radians.
This register allows multiple NCOs to be synchronized to produce sine waves with a known phase relationship.
The digital IF is calculated using Equation 17 below.
NCO Frequency Update and Phase Offset Update Hold-Off
Counters
f IF = f NCO ×
NCO_ frequency
232
(17)
where:
NCO_frequency is the value written to 0x02, fIF is the desired
intermediate frequency, and fNCO is fCLK/2 for complex outputs
and fCLK for real outputs.
Phase Dither
The AD6622 provides a phase dither option for improving the
spurious performance of the NCO. Phase dither is enabled by
writing a one to Bit 3 of Channel Register 0x01. When phase
dither is enabled, spurs due to phase truncation in the NCO are
randomized. The choice of whether phase dither is used in a
system will ultimately be decided by the system goals and the
choice of IF frequency. The 18 most significant bits of the phase
accumulator are used by the angle to Cartesian conversion. If
the NCO frequency has all zeroes below the 18th bit, then phase
dither has no effect. If the fraction below the 18th bit is near a
1/2 or 1/3, etc., of the 18th bit, then spurs will accumulate separated from the IF by 1/2 or 1/3, etc., of the CLK frequency. The
smaller the denominator of this residual fraction, the larger the
spurs due to phase truncation will be. If the phase truncation spurs
are unacceptably high for a given frequency, the phase dither can
reduce these at the penalty of a slight elevation in total error
The update of both the NCO Frequency and Phase Offset can
be synchronized with internal hold-off counters. Both of these
counters are 16-bit unsigned integers and are clocked at the
master CLK rate. These hold-off counters, used in conjunction
with the frequency or phase offset registers, allow Beam Forming and Frequency Hopping. See the Synchronization section of
this data sheet for additional details. The NCO phase can also
be cleared on Sync (set to 0x0000) by setting Bit 2 of Channel
Register 0x01 high.
NCO Output Scale
The output of the NCO can be scaled in four steps of 6 dB each
via Channel Register 0x01, Bits 1–0. Table V is a table of the
control scale. The NCO always has loss to accommodate the
possibility that both the I and Q inputs may reach full-scale simultaneously, resulting in a 3 dB input magnitude.
Table V. Control Scale
0x01 Bit 1
0
0
1
1
–16–
0x01 Bit 0
0
1
0
1
NCO Output Level
–6 dB
–12 dB
–18 dB
–24 dB
REV. 0
AD6622
SUMMATION BLOCK
The summation block of the AD6622 serves to combine the outputs of each channel to create a composite multicarrier signal.
The four channels are summed together and the result is then
added with the 18-bit wideband input bus (IN[17:0]). The final
summation is then driven on the 18-bit wideband output bus
(OUT[17:0]) on the rising edge of the high speed clock. If the
OEN input is high, this output bus is three-stated. If the OEN
input is low, this bus will be driven by the summed data. The OEN
is active high to allow the wideband output bus to be connected
to other buses without using extra logic. Most other buses (like
374-type registers) require a low output enable, which is opposite
the AD6622 OEN, thus eliminating extra circuitry.
The wideband output bus may be interpreted as a two’s complement number or as an offset binary number as defined by Bit 1
of the Summation Mode Control Register at address 0x000.
When this bit is high, the wideband output is in two's complement mode and when it is low it is configured for offset binary
output data.
The MSB (Bit 17) of the wideband output bus is typically used as
a guard bit for the purpose of clipping the wideband output
bus when Bit 0 of the Summation Mode Control Register at
address 0x000 is high. If clip detection is enabled, Bit 17 of the
output bus is not used as a data bit. Instead, Bit 16 will become the
MSB and be connected to the MSB of the DAC. Configuring
the DAC in this manner gives the summation block a gain of 0 dB.
When clip detection is not enabled and Bit 17 is used as a data
bit, then the summation block will have a gain of –6.02 dB.
There are two data output modes. The first is offset binary. This
mode is used only when driving offset binary DACs. Two’s complement mode may be used in one of two circumstances. The first is
when driving a DAC that accepts two’s complement data. The
second is when driving another AD6622 in cascade mode.
When clipping is enabled, the two’s complement mode output
bus will clip to 0x0FFFF for output signals more positive than the
output can express, and it will clip to 0x10000 for signals more
negative than the output can express. In offset binary mode the
output bus will clip to 0x1FFFF for output signals more positive
than the output can express, and it will clip to 0x00000 for signals
more negative than the output can express.
out of the system by use of the start hold-off counter. The preceding AD6622 in a cascaded chain can be started two high-speed
clock cycles before the following AD6622 is started and the data
from each AD6622 will arrive at the DAC on the same clock
cycle. In systems where the individual signals are not correlated, this is usually not necessary.
The AD6622 is capable of outputting both real and complex
data. When in real mode, the QIN input is tied low signaling that
all inputs on the wideband input bus are real and that all outputs
on the wideband output bus are real. The wideband input bus
will be pulled low and no data will be added to the composite
signal if this port is unused (not connected).
If complex data is desired, there are two ways this can be obtained.
The first method is simply to set the QIN input of the AD6622
high and set the wideband input bus low. This allows the AD6622
to output complex data on the wideband output bus. The I
data samples would be identified when QOUT is low and the Q
data samples would be identified when QOUT is high. The
second method of obtaining complex data is to provide a QIN
signal that toggles on every rising edge of the high-speed clock.
This could be obtained by connecting the QOUT of another
AD6622 to QIN. In a cascaded system the QIN of the first AD6622
in the chain would typically be tied high and the QOUT of the first
AD6622 would be connected to the QIN of the following part.
All AD6622s will synchronize themselves to the QIN input so
that the proper samples are always paired and the wideband output bus represents valid complex data samples.
Table VII. QIN, QOUT Functionality
QIN
Wideband
Input IN[17:0]
Output Data Type
OUT[17:0]
QOUT
Low
High
Pulsed
Real
Zero
Complex
Real
Complex
Complex
Low
Pulse
Pulse
TWO'S COMPLEMENT,
CLIPPING DISABLED
LOGIC1
QIN
OFFSET BIN,
CLIPPING
ENABLED
QOUT
QIN
OUT
[17:0]
IN
[17:0]
AD6622
AD6622
LOGIC0
Table VI. Numerical Data Representation
Number Represented
Output Representation
+Full-Scale Two’s Complement
–Full-Scale Two’s Complement
+Full-Scale Offset Binary
–Full-Scale Offset Binary
0x0FFFF
0x10000
0x1FFFF
0x00000
The wideband input is always interpreted as an 18-bit two’s
complement number and is typically connected to the wideband
output bus of another AD6622 in order to send more than four
carriers to a single DAC. The Output Bus of the proceeding
AD6622 should be configured in two's complement mode and
clip detection disabled. The 18-bit resolution ensures that the
noise and spur performance of the wideband data stream does
not become the limiting factor as large numbers of carriers are
summed.
There is a two-clock cycle latency from the wideband input
bus to the wideband output bus. This latency may be calibrated
REV. 0
IN
[17:0]
OUT
[16:3]
14-BIT
DAC
Figure 19. Cascade Operation of Two AD6622s
SYNCHRONIZATION
Three types of synchronization can be achieved with the AD6622.
These are Start, Hop, and Beam. Each is described in detail below.
The synchronization is accomplished with the use of a shadow
register and a hold-off counter. See Figure 20 for a simplified
schematic of the NCO Shadow Register and NCO Freq HoldOff Counter to understand basic operation. Enabling the clock
(AD6622 CLK) for the Hold-Off Counter can occur with either
a Soft Sync (via the Microport), or a Pin Sync (via the AD6622
sync pin, Pin 62). The functions that include shadow registers
to allow synchronization include:
1. Start
2. Hop (NCO Frequency)
3. Beam (NCO Phase Offset)
–17–
AD6622
Start refers to the start-up of an individual channel, chip, or
multiple chips. If a channel is not used, it should be put in the
Sleep Mode to reduce power dissipation. Following a hard reset
(low pulse on the AD6622 RESET pin), all channels are placed
in the Sleep Mode.
NCO
FREQUENCY
REGISTER
32
D
Q
NCO
REGISTER
32
D
Q
MICROPROCESSOR INTERFACE
16
D
Q
32
3. Write the Start bit and the SyncX(s) bit high (External
Address 5).
4. This starts the Start Update Hold-Off Counter counting
down. The counter is clocked with the AD6622 CLK signal.
When it reaches a count of one the Sleep bit of the appropriate channel(s) is set low to activate the channel(s).
HOLD-OFF
COUNTER
16
D
C=1
Start with Pin Sync
PL C = 0
A sync pin is provided on the AD6622 to provide the most
accurate synchronization, especially between multiple AD6622s.
Synchronization of start with an external signal is accomplished
with the following method.
ENA
HOP SYNC
START
COUNTER
START
HOLD-OFF
16
D
D
Q
16
D
Q
1. Set the appropriate channels to sleep mode (a hard reset
to the AD6622 RESET pin brings all four channels up in
sleep mode).
2. Write the Start Update Hold-Off Counter(s) (0x00) to the
appropriate value (greater than 1 and less than 216–1). If the
chip(s) is not initialized, all other registers should be loaded
at this step.
NCO PHASE
ACCUMULATOR
ENA
HOP
HOLD-OFF
Hold-Off Counter delays the start of a channel(s) by its value
(number of AD6622 CLKs). The following method is used to
synchronize the start of multiple channels via microprocessor
control.
C=1
SLEEP
1. Set the appropriate channels to sleep mode (a hard reset to
the AD6622 RESET pin brings all four channels up in
sleep mode).
SET
PL C = 0
ENA
2. Write the Start Update Hold-Off Counter(s) (0x00) to the
appropriate value (greater than 1 and less than 216–1). If the
chip(s) is not initialized, all other registers should be loaded
at this step.
START SYNC
EXTERNAL
ADDRESS 4
CLK
3. Set the start on pin sync bit and the appropriate sync pin
enable high (0x001).
RESET
PIN
4. When the sync pin is sampled high by the AD6622 CLK, it
enables the countdown of the Start Update Hold-Off Counter.
The counter is clocked with the AD6622 CLK signal. When
it reaches a count of one, the sleep bit of the appropriate
channel(s) is set low to activate the channel(s).
Figure 20. NCO Shadow Register and Hold-Off Counter
Start with No Sync
If no synchronization is needed to start multiple channels or
multiple AD6622s, the following method can be used to initialize the device.
1. To program a channel, it must first be set to the Program
Mode (bit high) and Sleep Mode (bit high) (External Address
4). The Program Mode allows programming of data memory
and coefficient memory (all other registers are programmable
whether or not in Program Mode). Since no synchronization
is used all sync bits are set low (External Address 5). All
appropriate control and memory registers (filter) are then
loaded. The Start Update Hold-Off Counter (0x00) should
be set to 0.
2. Set the appropriate program and sleep bits low (External
Address 4). This enables the channel. The channel must have
Program and Sleep Mode low to activate a channel.
Hop is a jump from one NCO frequency to a new NCO frequency.
This change in frequency can be synchronized via microprocessor control or an external sync signal as described below.
To set the NCO frequency without synchronization the following method should be used.
Set Freq No Hop
1. Set the NCO Freq Hold-Off Counter to 0.
2. Load the appropriate NCO frequency. The new frequency
will immediately be loaded to the NCO.
Hop with Soft Sync
Start with Soft Sync
The AD6622 includes the ability to synchronize channels or chips
under microprocessor control. One action to synchronize is the
start of channels or chips. The Start Update Hold-Off Counter
(0x00) in conjunction with the start bit and sync bit (External
Address 5) allow this synchronization. Basically the Start Update
The AD6622 includes the ability to synchronize a change in
NCO frequency of multiple channels or chips under microprocessor control. The NCO Freq Hold-Off Counter (0x03), in
conjunction with the hop bit and the sync bit (Ext Address 5),
allow this synchronization. Basically the NCO Freq Hold-Off
Counter delays the new frequency from being loaded into the
NCO by its value (number of AD6622 CLKs). The following
method is used to synchronize a hop in frequency of multiple channels via microprocessor control.
–18–
REV. 0
AD6622
1. Write the NCO Freq Hold-Off (0x03) Counter to the appropriate value (greater than 1 and less then 216–1).
2. Write the NCO Frequency Register(s) to the new desired
frequency.
3. Write the hop bit and the sync(s) bit high (Ext Address 5).
4. This starts the NCO Freq Hold-Off Counter counting down.
The counter is clocked with the AD6622 CLK signal. When
it reaches a count of one, the new frequency is loaded into
the NCO.
Hop with Pin Sync
A sync pin is provided on the AD6622 to provide the most
accurate synchronization, especially between multiple AD6622s.
Synchronization of hopping to a new NCO frequency with an
external signal is accomplished with the following method.
1. Write the NCO Freq Hold-Off Counter(s) (0x03) to the
appropriate value (greater than 1 and less than 216–1).
2. Write the NCO Frequency register(s) to the new desired
frequency.
3. Set the hop on pin sync bit and the appropriate sync pin
enable high (0x001).
4. When the sync pin is sampled high by the AD6622 CLK this
enables the countdown of the NCO Freq Hold-Off Counter.
The counter is clocked with the AD6622 CLK signal. When
it reaches a count of one the new frequency is loaded into the
NCO.
Beam is a change in phase for a particular channel and can be
synchronized with respect to other channels or AD6622s. This
change in phase can be synchronized via microprocessor control
or an external sync signal as described below.
To set the amplitude without synchronization the following
method should be used.
Set Phase No Beam
1. Set the NCO Phase Offset Update Hold-Off Counter (0x05)
to 0.
2. Load the appropriate NCO Phase Offset (0x04). The NCO
Phase Offset will be immediately loaded.
Beam with Pin Sync
A sync pin is provided on the AD6622 to provide the most
accurate synchronization, especially between multiple AD6622s.
Synchronization of beaming to a new NCO Phase Offset with an
external signal is accomplished with the following method.
1. Write the NCO Phase Offset Hold-Off (0x05) counter(s) to
the appropriate value (greater than 1 and less than 216–1).
2. Write the NCO Phase Offset register(s) to the new desired
phase and amplitude.
3. Set the beam on pin sync bit and the appropriate sync pin
enable high (0x001).
4. When the sync pin is sampled high by the AD6622 CLK, it
enables the countdown of the NCO Phase Offset Hold-Off
Counter. The counter is clocked with the AD6622 CLK signal. When it reaches a count of one, the new phase is loaded
into the NCO registers.
JTAG INTERFACE
The AD6622 supports a subset of IEEE Standard 1149.1
specifications. For additional details of the standard, please
see “IEEE Standard Test Access Port and Boundary-Scan
Architecture,” IEEE-1149 publication from IEEE.
The AD6622 has five pins associated with the JTAG interface.
These pins are used to access the on-chip Test Access Port and
are listed in Table VIII.
Table VIII. JTAG Pin List
Name
Pin Number
Description
TRST
TCK
TMS
TDI
TDO
100
101
106
108
107
Test Access Port Reset
Test Clock
Test Access Port Mode Select
Test Data Input
Test Data Output
The AD6622 supports four op codes as shown in Table IX.
These instructions set the mode of the JTAG interface.
Table IX. JTAG Op Codes
Beam with Soft Sync
The AD6622 includes the ability to synchronize a change in
NCO phase of multiple channels or chips under microprocessor
control. The NCO Phase Offset Update Hold-Off Counter, in
conjunction with the beam bit and the sync bit (Ext Address 5),
allow this synchronization. Basically the NCO Phase Offset
Update Hold-Off Counter delays the new phase from being
loaded into the NCO/RCF by its value (number of AD6622
CLKs). The following method is used to synchronize a beam-in
phase of multiple channels via microprocessor control.
1. Write the NCO Phase offset Update Hold-Off Counter (0x05)
to the appropriate value (greater than 1 and less then 216–1).
2. Write the NCO Phase Offset Register(s) to the new desired
phase and amplitude.
3. Write the beam bit and the sync(s) bit high (External
Address 5).
4. This starts the NCO Phase Offset Update Hold-Off counter
counting down. The counter is clocked with the AD6622
CLK signal. When it reaches a count of one, the new phase
is loaded into the NCO.
REV. 0
Instruction
Op Code
IDCODE
BYPASS
SAMPLE/PRELOAD
EXTEST
10
11
01
00
The Vendor Identification Code can be accessed through the
IDCODE instruction and has the following format.
Table X. JTAG ID String
MSB
Version
Part
Number
Manufacturing
ID #
LSB
Mandatory
000
0010
0111
1000
0000
000 1110 0101
1
A BSDL file for this device is available from Analog Devices, Inc.
Contact Analog Devices Inc. for more information.
–19–
AD6622
SCALING
Proper scaling of the wideband output is critical to maximize the
spurious and noise performance of the AD6622. A relatively small
overflow anywhere in the data path can cause the spurious free
dynamic range to drop precipitously. Scaling down the output
levels also reduces dynamic range relative to an approximately
constant noise floor. A well-balanced scaling plan at each point
in the signal path will be rewarded with optimum performance.
The scaling plan can be separated into two parts: multicarrier
scaling and single-carrier scaling.
Multicarrier Scaling
An arbitrary number of AD6622s can be cascaded to create a
composite digital IF with many carriers. As the number of carriers
increases, the peak-to-rms ratio of the composite digital IF will
increase as well. It is possible and beneficial to limit the peak-torms ratio through careful frequency planning and controlled phase
offsets. Nevertheless, in most cases with a large number of carriers,
the worst-case peak is an unlikely event.
The AD6622 immediately preceding the DAC can be programmed
to clip rather than wrap around (see the Summation Block description). For a large number of carriers, a rare but finite chance
of clipping at the AD6622 wideband output will result in superior
dynamic range compared to lowering each carrier level until
clipping is impossible. This will also be the case for most DACs.
Through analysis or experimentation, an optimal output level of
individual carriers can be determined for any particular DAC.
Single-Carrier Scaling
Once the optimal power level is determined for each carrier, one
must determine the best way to achieve that level. The maximum
SNR can be achieved by maximizing the intermediate power
level at each processing stage. This can be done by assuming the
proper level at the output and working backwards along the signal
path: Summation, NCO, CIC, and finally, RCF.
The summation block is intended to combine multiple carriers,
with each carrier at least 6 dB below full scale. For this configuration, the AD6622 driving the DAC should have clip detection
enable. OUT17 becomes a clip indicator that reports clipping in
both polarities. If the DAC requires offset binary outputs, the
internal offset binary conversion should be enabled as well. Any
preceding cascaded AD6622s should disable clip detection and
offset binary conversion. The IN17–IN0 of the first AD6622 in the
cascade should be grounded. See the Summation Block section
for details. In this configuration, intermediate OUT17s will
serve as guard bits that allow intermediate sums to exceed full
scale. As long as the final output does not exceed 6 dB over full
scale, the clip detector will perform correctly.
If a single carrier needs to exceed –6 dB full scale, hardwired
scaling can be accomplished according to the table below. This
is most useful when the AD6622 is processing a Single Wideband Carrier such as UMTS or CDMA 2000.
Table XI. Output Bit Scaling
Max SingleCarrier Level
–12.04 dB
–6.02 dB
0 dB
+6.02 dB
+12.04 dB
+18.06 dB
+24.08 dB
Connect to
DAC MSB
OUT17
OUT16
OUT15
OUT14
OUT13
OUT12
OUT11
Clip
Detect
N/A
±
+Only
+Only
+Only
+Only
+Only
Offset Binary
Compensation
Internal
Internal
0x18000
0x1C000
0x1E000
0x1F000
0x1F800
The NCO/Tuner is equipped with an output scalar that ranges
from –6.02 dB to –24.08 dB below full scale, in 6.02 dB steps.
See the NCO/Tuner section for details. The best SNR will be
achieved by maximizing the input level to the NCO and using
the largest possible NCO attenuation. For example, to achieve
an output level –20 dB below full scale, one should set the CIC
output level to –1.94 dB below full scale and attenuate by
–18.06 dB in the NCO.
The CIC is equipped with an output scalar that ranges from
0 dB to –150.51 dB below full scale in 6.02 dB steps. This large
attenuation is necessary to compensate for the potentially large
gains associated with CIC interpolation. See the CIC section for
details. For example to achieve an output level of –1.94 dB below
full scale, with a CIC5 interpolation of 27 (114.51 dB gain) and
a CIC2 interpolation of 3 (9.54 dB gain), one should set the
CIC_Scale to 20 and the RCF output level to –5.59 dB below
full scale.
(18)
–1.94 – 9.54 – 114.51 + 20 × 6.02 = –5.59
The RCF is equipped with an output scalar that ranges from 0 dB
to –18.06 dB below full scale in 6.02 dB steps. This attenuation
can be used to compensate for filter gain in the RCF. For example,
if the desired RCF output is –5.59 dB and the maxim gain of the
RCF coefficients is 11.04 dB, then the RCF_Coarse_Scale should
be set to two and the coefficients should be scaled so that the
largest coefficient is –4.59 dB below full scale. The largest possible gain of the RCF coefficients is when the largest coefficient
of the impulse response is normalized to one. This means that
all of the coefficients are as large as possible so the sum of the
coefficients are as large as possible. This maximum gain will
determine the RCF_Coarse_Scale, which should be used to
make the total RCF gain between 0 dB and –6.02 dB. After the
RCF_Coarse_Scale is chosen, the coefficients can be rescaled, as
in the example, to set the total RCF gain to a desired level. See the
RCF section for additional information.
(19)
–5.59 – 11.04 + 2 × 6.02 = –4.59
Finally, as described in the RCF section, there may be a worstcase peak of a phase that is larger than the channel center gain. In
the preceding example, if the worst-case to channel center ratio
is larger than 4.59 dB (potentially overflowing the RCF), the
RCF_Coarse_Scale should be reduced by one and the CIC_Scale
should be increased by one. In the preceding example, if the worstcase to channel center ratio is larger than 5.59 dB (potentially
overflowing the RCF and CIC), the RCF_Coarse_Scale should be
reduced by one and the NCO_Output_Scale should be increased
by one.
MICROPORT INTERFACE
The Microport interface is the communications port between the
AD6622 and the host controller. There are two modes of bus
operation: Intel Nonmultiplexed Mode (INM), and Motorola
Nonmultiplexed Mode (MNM), which is set by hard wiring the
MODE pin to either ground or supply. The mode is selected
based on the use of the Microport control lines (DS or RD,
DTACK or RDY, R/W or WR) and the capabilities of the host
processor. See the timing diagrams for details on the operation of
both modes.
The External Memory Map provides data and address registers
to read and write the extensive control registers in the Internal
Memory Map. The control registers access global chip functions
and multiple control functions for each independent channel.
–20–
REV. 0
AD6622
Microport Control
All accesses to the internal registers and memory of the AD6622
are accomplished indirectly through the use of the microprocessor port external registers shown in Table XII. Accesses to the
External Registers are accomplished through the 3-bit address
bus (A[2:0]) and the 8-bit data bus (D[7:0]) of the AD6622
(Microport). External Address [3:0] provides access to data read
from or written to the internal memory (up to 32 bits). External
Address [0] is the least significant byte and External Address [3]
is the most significant byte. External Address [4] controls the
resets of each channel. External Address [5] controls the sync
status of each channel. External Address [7:6] determines the
Internal Address selected and whether this address is incremented
after subsequent reads and/or writes to the internal registers.
EXTERNAL MEMORY MAP
The External Memory Map is used to gain access to the Internal Memory Map described below. External Address [7:6] sets
the Internal Address to which subsequent reads or writes will
be performed. The top two bits of External Address [7] allow
the user to set the address to autoincrement after reads, writes,
or both. All internal data words have widths that are less than
or equal to 32 bits. Accesses to External Address [0] trigger
accesses to the AD6622’s internal memory map. Thus during
writes to the internal registers, External Address [0] must be
written last to ensure all data is transferred. Reads are the opposite in that External Address [0] must be the first data register
read (after setting the appropriate internal address) to initiate an
internal access.
External Address [5:4] reads and writes are immediately transferred to internal control registers. External Address [4] is the
reset register. The reset bits can be set collectively by the address.
The reset bits can be cleared by operation of start syncs (described
below).
External Address [5] is the sync register. These bits are write
only. There are three types of syncs: start, hop, and beam. Each
of these can be sent to any or all of the four channels. For example,
a write of X0010100 would issue a start sync to Channel C
only. A write of X1101111 would issue a beam sync and a hop
sync to all channels.
The internal address bus is 11 bits wide and the internal data
bus is 32 bits wide. External Address 7 is the Chan (Channel)
and stores the upper three bits of the address space in Chan[2:0].
Chan[7:6] define the autoincrement feature. If Bit 6 is high, the
internal address in incremented after an internal read. If Bit 7 is
high, the internal address is incremented after an internal write.
If both bits are high, the internal address in incremented after
either a write or a read. This feature is designed for sequential
access to internal locations. External Address 6 is the Addr
(Address) and stores the lower eight bits of the internal address.
External Addresses 3 through 0 store the 32 bits of the internal
data. All internal accesses are two clock cycles long.
bits of the internal address (it does not matter if the Addr is
written before the Chan as long as both are written before the
internal access). Since the data width of the internal address is
16 bits, only Data Register 1 and Data Register 0 are needed.
Data Register 1 must be written first because the write to Data
Register 0 triggers the internal access. Data Register 0 must
always be the last register written to initiate the internal write.
Reading from the Microport is accomplished in a similar manner.
The internal address is first written. A read from Data Register
0 activates the internal read, thus register 0 must always be read
first to initiate an internal read. This provides the 8 LSBs of the
internal read through the Microport (D[7:0]). Additional bytes
are then read by changing the external address (A[2:0]) and
performing additional reads. If Data Register 3 (or any other)
is read before Data Register 0, incorrect data will be read. Data
Register 0 must be read first in order to transfer data from the
core memory to the external memory locations. Once data register
is read, the remaining locations may be examined in any order.
The Microport of the AD6622 allows for multiple accesses
while CS is held low (CS can be tied permanently low if the
Microport is not shared with additional devices). The user can
access multiple locations by pulsing the WR or RD line and
changing the contents of the external 3-bit address bus. Access
to the external registers of Table XII is accomplished in one
of two modes using the CS, RD, WR, and MODE inputs. The
access modes are Intel Nonmultiplexed Mode and Motorola
Nonmultiplexed Mode. These modes are controlled by the
MODE input (MODE = 0 for INM, MODE = 1 for MNM).
CS, RD, and WR control the access type for each mode.
Intel Nonmultiplexed Mode (INM)
MODE must be tied low to operate the AD6622 Microport
in INM Mode. The access type is controlled by the user with
the CS, RD (DS), and WR (R/W) inputs. The RDY (DTACK)
signal is produced by the Microport to communicate to the
user the Microport is ready for an access. RDY (DTACK) goes
low at the start of the access and is released when the internal
cycle is complete. See the timing diagrams for both the read and
write modes in the specifications.
Motorola Nonmultiplexed Mode (MNM)
MODE must be tied high to operate the AD6622 microprocessor
in MNM mode. The access type is controlled by the user with
the CS, DS (RD), and R/W (WR) inputs. The DTACK (RDY)
signal is produced by the Microport to acknowledge the completion of an access to the user. DTACK (RDY) goes low when an
internal access is complete and then will return high after DS
(RD) is deasserted. See the timing diagrams for both the read
and write modes in the Specifications.
The DTACK pin is configured as an open drain so that multiple
devices may be tied together at the microprocessor/microcontroller
without contention.
Writing to an internal location with a data width of 16 bits is
achieved by first writing the upper three bits of the address to
Bits 2 through 0 of the Chan. (Bits 7 and 6 of the Chan are
written to determine whether or not the auto increment feature is enabled.) The Addr is then written with the lower eight
REV. 0
–21–
AD6622
Table XII. External Memory Map
External
Address
7: Chan
6: Addr
5: Sync
4: Reset
3: Byte3
2: Byte2
1: Byte1
0: Byte0
D7
D6
D5
External Data
D4
D3
D2
D1
D0
Wrinc
IA7
Rdinc
IA6
Beam
Prog C
ID30
ID22
ID14
ID6
IA5
Hop
Prog B
ID29
ID21
ID13
ID5
IA4
Start
Prog A
ID28
ID20
ID12
ID4
IA10
IA2
Sync C
Sleep C
ID26
ID18
ID10
ID2
IA9
IA1
Sync B
Sleep B
ID25
ID17
ID9
ID1
IA8
IA0
Sync A
Sleep A
ID24
ID16
ID8
ID0
Prog D
ID31
ID23
ID15
ID7
IA3
Sync D
Sleep D
ID27
ID19
ID11
ID3
External Address 7 Upper Address Register (Chan)
External Address 4 Reset
Sets the three most significant bits of the internal address, effectively selecting channels 1, 2, 3, or 4 (D2:D0). The autoincrement
of read and write are also set (D7:D6).
Bits in this register determine how the chip is programmed and
enables the channels. The program bits (D7:D4) must be set
high to allow programming of CMEM and DMEM for each
channel. Sleep bits (D3:D0) are used to activate or sleep channels.
These can be used manually by the user to bring up a channel
by simply writing the required channel high. These bits can also
be used in conjunction with the Start and Sync signals available in External Address 5 to synchronize the channels. See
the Synchronization section of the data sheet for detailed explanation of different modes.
External Address 6 Lower Address Register (Addr)
Sets the internal address 8 LSBs (D7:D0).
External Address 5 Sync
This register is read only. Bits in this address control the synchronization of the AD6622 channels. If the user intends to bring up
channels with no synchronization requirements, then all bits of
this register should be written low. Two types of sync signals
are available with the AD6622. The first is Soft Sync. Soft Sync
is software synchronization enabled through the Microport. The
second synchronization method is Pin Sync. Pin Sync is enabled
by a signal applied to the sync pin (Pin 62). See the Synchronization section of the data sheet for detailed explanations of the
different modes.
External Address 3:0 (Data Bytes)
These bits set the internal address to be accessed for a read or
write.
–22–
REV. 0
AD6622
INTERNAL CONTROL REGISTERS AND ON-CHIP RAM
Listed below is the mapping of internal AD6622 registers.
Table XIII. Internal Memory Map
Address
Bit Width
Name
Notation
Description
Common Function Registers (Not Associated with a Particular Channel)
0x000
8
Summation MODE Control
0x001
8
Sync MODE Control
0: Clip Wideband Output
1: Offset Binary Wideband Output
2: Reserved, Must Be Set High
3–7: Reserved, Should Be Set Low
0: Ch. A Sync Pin Enable
1: Ch. B Sync Pin Enable
2: Ch. C Sync Pin Enable
3: Ch. D Sync Pin Enable
4: Start on Pin Sync
5: Hop on Pin Sync
6: Beam Steer on Pin Sync
7: First Sync Only
Channel Function Registers (0x1XX = Ch. A, 0x2XX = Ch. B, 0x3XX = Ch. C, 0x4XX = Ch. D )
0x100
0x101
16
8
Start Update Hold-Off Counter
NCO Control
0x102
0x103
0x104
0x105
0x106
32
16
16
16
8
NCO Frequency
NCO Freq Hold Off
NCO Phase Offset
NCO Phase Hold Off
CIC Scale
0x107
0x108
0x109
0x10A
8
8
8
8
Reserved
CIC2 Interpolation-1
CIC5 Interpolation-1
RCF Coefficient Count
NRCF-1
0x10B
8
RCF Coefficient Offset
ORCF
0x10C
8
Channel MODE Control 1
NRCF/LRCF-1
0x10D
8
Channel MODE Control 2
0x10E
0x10F
0x110
0x111
0x112–0x11F
0x120–0x13F
0x140–0x17F
0x180–0x1FF
16
16
16
16
16
16
16
Reserved
Reserved
Reserved
Data Memory
Reserved
Coefficient Memory
Start Update Hold Off Counter
1-0: Ch. A NCO Output Scale
2: Ch. A NCO Clear Phase Accum on Sync
3: Ch. A NCO Phase Dither Enable
4: Ch. A NCO Amp Dither Enable
7–5: Reserved
Ch. A NCO Frequency Value
Ch. A NCO Frequency Update Hold-Off Ctr
Ch. A NCO Phase Offset
Ch. A NCO Phase Offset Update Hold-Off Ctr
4–0: Ch. A CIC Scale
7–5: Reserved
7–0: Reserved
Ch. A CIC2 Interpolation Factor-1
Ch. A CIC5 Interpolation Factor-1
6–0: Ch. A RCF Coefficient Count, N RCF–1
7: Reserved
6–0: Ch. A RCF Coefficient Offset
7: Reserved
3–0: Ch. A NRCF/LRCF–1
5–4: Ch. A Input Format:
00 = FIR
6: Reserved
7: Reserved
4–0: Ch. A Serial Clock Divider
5: Ch. A Phase EQ Enable
7–6: Ch. A RCF Coarse Scale:
00 = 0 dB
01 = –6 dB
10 = –12 dB
11 = –18 dB
15–0: Reserved
15–0: Reserved
Reserved
Reserved
Reserved
Ch. A Data Memory
Reserved
Ch. A Coefficient Memory
Channel B
Channel C
Channel D
Ch. B Registers (Organized as Ch. A Above)
Ch. C Registers (Organized as Ch. A Above)
Ch. D Registers (Organized as Ch. A Above)
Additional Channels
0x200–0x2FF
0x300–0x3FF
0x400-0x4FF
REV. 0
Various
Various
Various
–23–
AD6622
(0x000) Summation Mode Control
(0xn05) NCO Phase Offset Update Hold-Off Counter
Controls functions in the summation block of the AD6622. When
set high, Bit 0 causes the output data to be clipped (no wraparound) when overrange of the output occurs. When Bit 0 is low,
overrange will result in wraparound. When set low, Bit 1 formats
the output data as two's complement. Bit 1 set high will format output data as offset binary.
The Hold-Off Counter is used to synchronize the change of NCO
phases. See the Synchronization section of the data sheet for
detailed explanation. If no synchronization is required, this
register should be set to 0.
(0xn06) CIC Scale
Bits 5:0 set the CIC scaling per the equation below.
(0x001) Sync Mode Control
4
CIC _ Scale = ceil (log2 ( LCIC 5 × LCIC 2 ))
(21)
Bits 3–0 when high enable synchronization of these channels.
See the Synchronization section of the data sheet for detailed
explanation.
See CIC section of the data sheet for details. Bits 7:6 are reserved
and should be set to 0.
Channel Function Registers
(0xn07) Reserved
The following registers are channel-specific. “0x” denotes that
these values are represented as hexadecimal numbers. “n” represents the specified channel. Valid channels are n = 1, 2, 3, and 4.
This register is reserved and should be set to 0.
(0xn00) Start Update Hold-Off Counter
The Start Update Hold-Off Counter is used to synchronize start
up of AD6622 channels and can be used to synchronize multiple
chips. The Start Update Hold-Off Counter is clocked by the
AD6622 CLK (master clock). See the Synchronization section
of the data sheet for detailed explanation. If no synchronization
is required, this register should be set to 0.
(0xn01) NCO Control
Bit 1:0 set the NCO scaling per the Table XIV.
0x01 Bit 0
NCO Output Level
0
0
1
1
0
1
0
1
–6 dB
–12 dB
–18 dB
–24 dB
This register sets the interpolation rate for the CIC2 filter stage
(unsigned integer). The programmed value is the CIC2 Interpolation – 1. Maximum interpolation is limited by the CIC scaling
available (See CIC section of the data sheet).
(0xn09) CIC5 Interpolation – 1
This register sets the interpolation rate for the CIC5 filter
stage (unsigned integer). The programmed value is the CIC5
Interpolation – 1. Maximum interpolation is limited by the CIC
scaling available (See CIC section of the data sheet).
(0xn0A) Number of RCF Coefficients – 1
This register sets the number of RCF Coefficients and is limited
to a maximum of 128. The programmed value is the number of
RCF Coefficients – 1.
Table XIV. Control Scale
0x01 Bit 1
(0xn08) CIC2 Interpolation – 1
(0xn0B) RCF Coefficient Offset
This register sets the offset for RCF Coefficients and is normally
set to 0. It can be viewed as a pointer that selects the portion of
the CMEM used when computing the RCF filter. This allows
multiple filters to be stored in the coefficient memory space,
selecting the appropriate filter by setting the offset.
Bit 2, when high, clears the NCO phase accumulator to 0 on
either a Soft Sync or Pin Sync (see Synchronization for details).
(0xn0C) Channel Mode Control 1
Bits 3:0 set NRCF/LRCF-1.
Bit 3, when high, enables NCO phase dither.
Bit 4, when high, enables NCO amplitude dither.
Bits 7:5 are reserved and should be written low.
Bits 5:4 set the channel input format as shown below.
Table XV. Filter Mode
(0xn02) NCO Frequency
This register is a 32-bit unsigned integer that sets the NCO
Frequency. The NCO Frequency contains a shadow register for
synchronization purposes. The shadow can be read back directly,
the NCO Frequency cannot.
f

NCOFrequency = 232 ×  channel
 CLK 
(20)
(0xn03) NCO Frequency Update Hold-Off Counter
The Hold-Off Counter is used to synchronize the change of
NCO frequencies. See the Synchronization section of the data
sheet for detailed explanation. If no synchronization is required,
this register should be set to 0.
Bit 5
Bit 4
Input Mode
0
0
1
1
0
1
0
1
FIR
Reserved
Reserved
Reserved
Bit 6 Reserved.
Bit 7 Reserved.
(0xn0D) Channel Mode Control 2
Bits 4:0 set the SCLKDIVIDER which determines the serial clock
frequency based on the following equation.
(0xn04) NCO Phase Offset
This register is a 16-bit unsigned integer that is added to the phase
accumulator of the NCO. This allows phase synchronization of
multiple channels of the AD6622(s). See the Synchronization
section of the data sheet for details. The NCO Phase Offset contains a shadow register for synchronization purposes. The shadow
can be read back directly, the NCO Phase Offset cannot.
–24–
f SCLK =
CLK
2 × (SCLK DIVIDER + 1)
(22)
REV. 0
AD6622
Bit 5 Reserved. Must be set low.
Bits 7:6 set the RCF Coarse Scale as shown below.
READ PSEUDOCODE
Void Read_Micro(ext_address);
Main()
Table XVI. RCF Scaling
Bit 7
Bit 6
RCF Coarse Scale
0
0
1
1
0
1
0
1
0 dB
–6 dB
–12 dB
–18 dB
(0xn0E) Reserved
(0xn0F) Reserved
(0xn10) Reserved (Must Be Written to 0)
(0xn11) Reserved (Must Be written to 0)
(0xn12–0xn1F) Reserved
(0xn20–0xn3F) Data Memory
This group of registers contain the RCF Filter Data. See the RCF
section of the data sheet for additional detail.
(0xn40–0xn7F) Reserved
(0xn80–0xnFF) Coefficient Memory
This group of registers contain the RCF Filter Coefficients. See
the RCF section of the data sheet for additional detail.
WRITE PSEUDOCODE
Void Write_Micro(ext_address, int data);
{
/* This code shows the reading of the NCO frequency register
using the Read_Micro function defined above. The variable
address is the External Address A[2:0]
Internal Address = 0x102, channel 1
*/
/*Holding registers for NCO byte wide access data*/
int d3, d2, d1, d0;
/*NCO frequency word (32 bits wide)*/
/*write Chan */
Write_Micro(7, 0x01);
/*write Addr*/
Write_Micro(6,0x02);
/*read Byte 0, all data is moved from the Internal Registers to
the interface registers on this access, thus Byte 0 must be accessed
first for the other Bytes to be valid*/
d0=Read_Micro(0) & 0xFF;
/*read Byte 1*/
d1=Read_Micro(1) & 0xFF;
/*read Byte 2*/
d2=Read_Micro(2) & 0xFF;
/*read Byte 0 */
d3=Read_Micro(3) & 0xFF;
}
Main()
{
/* This code shows the programming of the NCO frequency
register using the Write_Micro function defined above. The
variable address is the External Address A[2:0] and data is the
value to be placed in the external interface register.
Internal Address = 0x102, channel 1
*/
/*Holding registers for NCO byte wide access data*/
int d3, d2, d1, d0;
/*NCO frequency word (32 bits wide)*/
NCO_FREQ=0x1BEFEFFF;
/*write Chan */
Write_Micro(7, 0x01);
/*write Addr */
Write_Micro(6,0x02);
/*write Byte 3*/
d3=(NCO_FREQ & 0xFF000000)>>24;
Write_Micro(3,d3);
/*write Byte 2*/
d2=(NCO_FREQ & 0xFF0000)>>16;
Write_Micro(2,d2);
/*write Byte 1*/
d1=(NCO_FREQ & 0xFF00)>>8;
Write_Micro(1,d1);
/*write Byte 0, Byte 0 is written last and causes an internal write
to occur*/
d0=NCO_FREQ & 0xFF;
Write_Micro(0,d0);
}
REV. 0
APPLICATIONS
The AD6622 provides considerable flexibility for the control of
the synchronization, relative phasing, and scaling of the individual
channel inputs. Implementation of a multichannel transmitter
invariably begins with an analysis of the output spectrum that
must be generated.
DIGITAL-TO-ANALOG CONVERTER (DAC) SELECTION
The selection of a high-performance DAC depends on a number
of factors. The dynamic range of the DAC must be considered
from a noise and spectral purity perspective. The 14-bit AD9754
and AD9772 are the best choices for overall bandwidth, noise,
and spectral purity.
In order to minimize the complexity of the analog interpolation
filter which must follow the DAC, the sample rate of the master
clock is generally set to at least three times the maximum analog
frequency of interest.
In the case where a 15 MHz band of interest is to be up-converted
to RF, the lowest frequency might be 5 MHz and the upper
band edge at 20 MHz (offset from dc to afford the best image
reject filter after the first digital IF). The minimum sample rate
would be set to 75 MSPS.
Consideration must also be given to data rate of the incoming
data stream, interpolation factors, and the clock rate of the DSP.
–25–
AD6622
Time
Restriction
↓
MULTIPLE TSP OPERATION
Each of the four Transmit Signal Processors (TSPs) of the
AD6622 can adequately reject the interpolation images of narrow bandwidth carriers such as AMPS, IS-136, GSM, EDGE,
and PHS. Wider bandwidth carriers such as IS-95 and UMTS
require a coordinated effort of multiple processing channels.
L

N RCF ≤ min  , 16 × LRCF , 128
2

This section demonstrates how to coordinate multiple TSPs
to create wider bandwidth channels without sacrificing image
rejection. As an example, a UMTS carrier is modulated using
four TSPs (an entire AD6622). The same principals can be
applied to different designs using more or fewer TSPs. This section does not explore techniques for using multiple TSPs to
solve problems other than Serial Port or RCF throughput.
Designing filter coefficients and control settings for deinterleaved
TSPs is no harder than designing a filter for a single TSP. For
example, if four TSPs are to be used, simply divide the input data
rate by four and generate the filter as normal. For any design, a
better filter can always be realized by incrementing the number
of TSPs to be used. When it is time to program the TSPs, only
two small differences must be programmed. First each channel
is configured with exactly the same filter, scalars, modes and
NCO frequency. Since each channel receives data at 1/4 the
data rate and in a staggered fashion, the Start Hold-Off Counters
must also be staggered (see Programming Multiple TSPs section
below). Second, the phase offset of each NCO must be set to
match the demultiplexed ratio (1/4 in this example). Thus the
phase offset should be set to 90 degrees (16384, which is 1/4
of a 16-bit register).
Determining the Number of TSPs to Use
There are three limitations of a single TSP that can be overcome by deinterleaving an input stream into multiple TSPs:
Serial Port bandwidth, the time restriction to the RCF impulse
response length (NRCF), and the DMEM restriction to NRCF.
If the input sample rate is faster than the Serial Port can accept
data, the data can be deinterleaved into multiple Serial Ports.
Recalling from the Serial Port description, the SCLK frequency
(fSCLK) is determined by the equation below. To minimize the
number of processing channels, SCLKDIVIDER should be set
as low as possible to get the highest fSCLK that the serial data
source can accept.
f SCLK =
fCLK
2 × (SCLK DIVIDER + 1)
(23)
(25)
↑
DMEM
Restriction
where:
L = LRCF × LCIC5 × LCIC2 =
N TSP × fCLK
f IN
Deinterleaving the input data into multiple TSPs will extend the
time restriction and may possibly extend the DMEM restriction,
but will not extend the CMEM restriction. Deinterleaving the
input stream to multiple TSPs divides the input sample rate to
each TSP by the number of TSPs used (NTSP). To keep the output rate fixed, L must be increased by a factor of NCH, which
extends the time restriction. This increase in L may be achieved
by increasing any one or more of LRCF, LCIC5, or LCIC2 within their
normal limits. Achieving a larger L by increasing LRCF instead
of LCIC5 or LCIC2, will relieve the DMEM restriction as well.
In a UMTS example, NTSP = 4, fCLK = 61.44 MHz, and fIN =
3.84 MHz, resulting in L = 64. Factoring L into LRCF = 8, LCIC =
8, and LCIC2 = 1, results in a maximum NRCF = 32 due to the time
restriction. Figure 22 shows an example RCF impulse response
that has a frequency response as shown in Figure 23 from
0 Hz to 7.68 MHz (fIN × LRCF/NTSP). The composite RCF and
CIC frequency response is shown in Figure 24, on the same frequency scale. This figure demonstrates a good approximation to
a root-raised-cosine with a roll-off factor of 0.22, a pass-band
ripple of 0.1 dB, and a stopband ripple better than –65 dB until
the lobe of the first image which peaks at –50 dB about 5.6 MHz
from the carrier center. This lobe could be reduced by shifting
more of the interpolation towards the RCF, but that would
sacrifice near-in performance. As shown, the first image can easily
be rejected by an analog filter further up the signal path.
Scaling must be considered as normal with an interpolation
factor of L, to guarantee no overflow in the RCF, CIC, or NCOs.
The output level at the summation port should be calculated
using an interpolation factor of L/NTSP.
Programming Multiple TSPs
A minimum of 32 SCLK cycles are required to accept an input
sample, so the minimum number of TSPs (NTSP) due to limited
Serial Port bandwidth is a function of the input sample rate (fIN),
as shown by the equation below.
 32 × f IN 
NTSP ≥ ceil 

 fSCLK 
CMEM
Restriction
↓
(24)
For a sample UMTS system, we will assume fCLK = 61.44 MHz,
and the serial data source can drive data at 30.72 MBPS
(SCLKDIVIDER = 0). To achieve fIN = 3.84 MHz, the minimum NTSP is 4. (This is TSP channels, not TSP ICs.)
Multiple TSPs are also required if the RCF does not have enough
time or DMEM space to calculate the required RCF filter. Recalling the maximum NTAPS equation from the RCF description, are
three restrictions to the RCF impulse response length, NRCF.
Configuring the TSPs for deinterleaved operation is straightforward. All of the Channel Registers and CMEM of each TSP
are programmed identically, except the Start Hold-Off Counters
and NCO Phase Offset.
In order to separate the input timing to each TSP, the HoldOff Counters must be used to start each TSP successively in
response to a common Start SYNC. The Start SYNC may originate from the SYNC pin or the Microport. Each subsequent
TSP must have a Hold-Off Counter value L/NTSP larger than its
predecessor’s. If the TSPs are located on cascaded AD6622s,
the Hold-Off Counters of the upstream device should be incremented by an additional one.
In the UMTS example, L = 64 and NTSP = 4, so in order to
respond as quickly as possible to a Start SYNC, the Hold-Off
Counter values should be 1, 17, 33, and 49.
–26–
REV. 0
AD6622
Driving Multiple TSP Serial Ports
10
When properly configured, the AD6622 will drive each SDFS
out of phase. Each new piece of data should be driven only into
the TSP that pulses its SDFS pin at that time.
RAM COEFFICIENT FILTER
0
–10
–20
In the UMTS example, L = 64 and NTSP = 4, so each serial port
need only accept every 4th input sample. Each serial port is
shifting at peak capacity, so sample 1, 2, and 3 begin shifting
into Serial Ports B, C, and D before Sample 0 is completed into
Serial Port A.
dBc
–30
–40
–50
–60
–70
SDFSA
SDFSB
0
–80
4
5
1
–90
SDFSC
6
2
SDFSD
–100
0
7
3
8.196MSPS 65.536MSPS
RAM I
COEF Q
CIC
NCO
FILTER
1.024
MCPS
32
–20
6000
7000
8000
HCIC(f)
( LEVEL
)
dB
–30
–40
5
SPEC( f )
–50
dBc
8.196MSPS 65.536MSPS
RAM I
CIC
NCO
COEF Q
FILTER
32
5000
–10
SUMMATION
32
DATA
REFORMATTER
4.096
MCPS
4000
0
AD6622
1.024
MCPS
3000
Figure 24. Typical FIR Frequency Response for WBCDMA
8.196MSPS 65.536MSPS
RAM I
COEF Q
CIC
NCO
FILTER
32
2000
kHz
Figure 21. SDFS Timing for WBCDMA
1.024
MCPS
1000
65.536
MSPS
–60
–70
DAC
–80
dB(AD6624(f))
–90
–100
–110
8.196MSPS 65.536MSPS
RAM I
CIC
NCO
COEF Q
FILTER
1.024
MCPS
32
–120
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0
f
MHz
Figure 25. Typical Composite Frequency Response for
WBCDMA
COMPLEX SIGNAL 32 BITS (16 I, 16 Q)
REAL OR IMAGINARY SIGNAL
Figure 22. Block Diagram for WBCDMA
THERMAL MANAGEMENT
The power dissipation of the AD6622 is primarily determined
by three factors: the clock rate, the number of channels active,
and the distribution of interpolation rates. The faster the clock
rate the more power dissipated by the CMOS structures of
the AD6622; the more channels active, the higher the overall
power of the chip. Low interpolation rates in the CIC stages
(CIC5, CIC2) results in higher power dissipation. All these factors
should be analyzed as each application has different thermal
requirements.
1.0
0.5
COEF – J
0
0
5
10
15
20
25
30
The AD6622 128-lead MQFP is specially designed to provide
excellent thermal performance. To achieve the best performance,
the power and ground leads should be connected directly to
planes on the PC board. This provides the best thermal transfer
from the AD6622 to the PC board.
Figure 23. Typical Impulse Response for WBCDMA
REV. 0
–27–
AD6622
OUTLINE DIMENSIONS
Dimensions shown in inches and (mm).
0.685 (17.40)
0.669 (17.00)
0.555 (14.10)
0.547 (13.90)
0.134 (3.40)
MAX
0.041 (1.03)
0.031 (0.78)
C3772–8–5/00 (rev. 0) 00968
128-Lead MQFP (Metric Quad Flatpack)
(S-128A)
128
1
103
102
SEATING
PLANE
0.791 (20.10)
0.783 (19.90)
TOP VIEW
(PINS DOWN)
0.921 (23.40)
0.906 (23.00)
0.003 (0.08)
MAX
0.010 (0.25)
MIN
38
39
0.020 (0.50)
BSC
0.011 (0.27)
0.007 (0.17)
PRINTED IN U.S.A.
0.110 (2.80)
0.102 (2.60)
65
64
–28–
REV. 0