DATASHEET

HSP50214B
®
Data Sheet
May 1, 2007
FN4450.4
Programmable Downconverter
Features
The HSP50214B Programmable Downconverter converts
digitized IF data into filtered baseband data which can be
processed by a standard DSP microprocessor. The
Programmable Downconverter (PDC) performs down
conversion, decimation, narrowband low pass filtering, gain
scaling, resampling, and Cartesian to Polar coordinate
conversion.
• Up to 65MSPS Front-End Processing Rates (CLKIN) and
55MHz Back-End Processing Rates (PROCCLK)
Clocks May Be Asynchronous
The 14-bit sampled IF input is down converted to baseband
by digital mixers and a quadrature NCO, as shown in the
Block Diagram. A decimating (4 to 32) fifth order Cascaded
Integrator-Comb (CIC) filter can be applied to the data
before it is processed by up to 5 decimate-by-2 halfband
filters. The halfband filters are followed by a 255-tap
programmable FIR filter. The output data from the
programmable FIR filter is scaled by a digital AGC before
being re-sampled in a polyphase FIR filter. The output
section can provide seven types of data: Cartesian (I, Q),
polar (R, θ), filtered frequency (dθ/dt), Timing Error (TE), and
AGC level in either parallel or serial format.
• Processing Capable of >100dB SFDR
• Up to 255-Tap Programmable FIR
• Overall Decimation Factor Ranging from 4 to 16384
• Output Samples Rates to ≅ 12.94MSPS with Output
Bandwidths to ≅ 982kHz Lowpass
• 32-Bit Programmable NCO for Channel Selection and
Carrier Tracking
• Digital Resampling Filter for Symbol Tracking Loops and
Incommensurate Sample-to-Output Clock Ratios
• Digital AGC with Programmable Limits and Slew Rate to
Optimize Output Signal Resolution; Fixed or Auto Gain
Adjust
• Serial, Parallel, and FIFO 16-Bit Output Modes
• Cartesian to Polar Converter and Frequency Discriminator
for AFC Loops and Demodulation of AM, FM, FSK, and
DPSK
• Input Level Detector for External I.F. AGC Support
• Pb-Free Plus Anneal Available (RoHS Compliant)
Applications
• Single Channel Digital Software Radio Receivers
• Base Station Rx’s: AMPS, NA TDMA, GSM, and CDMA
• Compatible with HSP50210 Digital Costas Loop for PSK
Reception
• Evaluation Platform Available
Ordering Information
PART NUMBER
HSP50214BVC
PART MARKING
HSP50214BVC
HSP50214BVCZ (Note) HSP50214BVCZ
TEMP. RANGE (°C)
PACKAGE
PKG. DWG. NO.
0 to +70
120 Ld MQFP
Q120.28x28
0 to +70
120 Ld MQFP (Pb-free) Q120.28x28
HSP50214BVI
HSP50214BVI
-40 to +85
120 Ld MQFP
Q120.28x28
HSP50214BVIZ (Note)
HSP50214BVIZ
-40 to +85
120 Ld MQFP (Pb-free) Q120.28x28
NOTE: Intersil Pb-free plus anneal products employ special Pb-free material sets; molding compounds/die attach materials and 100% matte tin plate
termination finish, which are RoHS compliant and compatible with both SnPb and Pb-free soldering operations. Intersil Pb-free products are MSL
classified at Pb-free peak reflow temperatures that meet or exceed the Pb-free requirements of IPC/JEDEC J STD-020.
1
HSP50214B
Block Diagram
MICROPROCESSOR
READ/WRITE
C(7:0)
CONTROL
AGC LOOP FILTER
AGC
255-TAP
FIR FILTER
255-TAP
FIR FILTER
I OUT
POLYPHASE
FIR AND
HALFBAND
FILTERS
POLYPHASE
FIR AND
HALFBAND
FILTERS
SEROUTA
CARTESIAN
TO
POLAR
COORDINATE
CONVERTER
MAG.
PHASE
Q OUT
CARRIER
NCO
SOF
DISCRIMINATOR
RESAMPLING
NCO
CLKIN
PROCCLK
REFCLK
OUTPUT FORMATTER
COF
5TH
ORDER
CIC
FILTER
HALFBAND
FILTERS
GAIN
ADJ
(2:0)
INPUT
SECTION
IN(13:0)
5TH
ORDER
CIC
FILTER
HALFBAND
FILTERS
LEVEL DETECT
SEROUTB
AOUT(15:0)
BOUT(15:0)
FREQ
TIMING ERROR
Δ
2
FN4450.4
May 1, 2007
HSP50214B
Pinout
120
119
118
117
116
115
114
113
112
111
110
109
108
107
106
105
104
103
102
101
100
99
98
97
96
95
94
93
92
91
IN11
IN12
IN13
AGCGNSEL
VCC
REFCLK
GND
OEAH
AOUT15
AOUT14
AOUT13
AOUT12
AOUT11
AOUT10
GND
NC
AOUT9
AOUT8
AOUT7
AOUT6
AOUT5
VCC
NC
AOUT4
AOUT3
AOUT2
AOUT1
AOUT0
GND
OEAL
HSP50214B
(120 LD MQFP)
TOP VIEW
90
89
88
87
86
85
84
83
82
81
80
79
78
77
76
75
74
73
72
71
70
69
68
67
66
65
64
63
62
61
DATARDY
OEBH
BOUT15
BOUT14
VCC
NC
BOUT13
BOUT12
BOUT11
BOUT10
BOUT9
BOUT8
GND
GND
PROCCLK
VCC
MSYNCI
MSYNCO
GND
BOUT7
BOUT6
BOUT5
GND
BOUT4
NC
BOUT3
BOUT2
BOUT1
BOUT0
OEBL
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
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
SYNCOUT
INTRRP
WR
RD
GND
C7
C6
NC
C5
C4
VCC
C3
C2
C1
NC
C0
A2
A1
A0
GND
SEL2
SEL1
SEL0
GND
SEROUTA
SEROUTB
SERSYNC
SEROE
SERCLK
VCC
IN10
IN9
IN8
GND
IN7
NC
IN6
IN5
IN4
IN3
IN2
GND
IN1
IN0
VCC
CLKIN
GND
NC
ENI
GAINADJ2
GAINADJ1
GAINADJ0
COF
COFSYNC
GND
SOF
SOFSYNC
VCC
SYNCIN1
SYNCIN2
3
FN4450.4
May 1, 2007
HSP50214B
Pin Descriptions
NAME
TYPE
DESCRIPTION
VCC
-
Positive Power Supply Voltage.
GND
-
Ground.
CLKIN
I
Input Clock. This clock should be a multiple of the input sample rate. All input section processing occurs on the rising
edge of CLKIN. The frequency of CLKIN is designated fCLKIN .
IN(13:0)
I
Input Data. The format of the input data may be set to offset binary or 2’s complement. IN13 is the MSB (see Control
Word 0).
ENI
I
Input Enable. Active Low. This pin enables the input to the part in one of two modes, gated or interpolated (see
Control Word 0). In gated mode, one sample is taken per CLKIN when ENI is asserted. The input sample rate is
designated fS, which can be different from fCLKIN when ENI is used.
GAINADJ(2:0)
I
GAINADJ Input. Adds an offset to the gain via the shifter following the mixer. GAINADJ value is added to the shift
code from the microprocessor (μP) interface. The shift code is saturated to a maximum code of F. The gain is offset
by (6dB)(GAINADJ); (000 = 0dB gain adjust; 111 = 42dB gain adjust) GAINADJ2 is the MSB. See “Using the Input
Gain Adjust Control Signals” Section.
PROCCLK
I
Processing Clock. PROCCLK is the clock for all processing functions following the CIC Section. Processing is
performed on PROCCLK’s rising edge. All output timing is derived from this clock.
This clock may be asynchronous to CLKIN.
AGCGNSEL
I
AGC Gain Select. This pin selects between two AGC loop gains. This input is setup and held relative to PROCCLK.
Gain setting 1 is selected when AGCGNSEL = 1.
COF
I
Carrier Offset Frequency Input. This serial input pin is used to load the carrier offset frequency into the Carrier NCO
(see Serial Interface Section). The offset may be 8, 16, 24, or 32-bits. The setup and hold times are relative to CLKIN.
This input is compatible with the output of the HSP50210 Costas loop [1].
COFSYNC
I
Carrier Offset Frequency Sync. This signal is asserted one CLK before the most significant bit (MSB) of the offset
frequency word (see Serial Interface Section). The setup and hold times are relative to CLKIN. This input is
compatible with the output of the HSP50210 Costas loop [1].
SOF
I
Re-Sampler Offset Frequency Input. This serial input pin is used to load the offset frequency into the Re-Sampler
NCO (see Serial Interface Section). The offset may be 8, 16, 24, or 32-bits. The setup and hold times are relative
to PROCCLK. This input is compatible with the output of the HSP50210 Costas loop [1].
SOFSYNC
I
Re-Sampler Offset Frequency Sync. This signal is asserted one CLK before the MSB of the offset frequency word
(see Serial Interface Section). The setup and hold times are relative to PROCCLK. This input is compatible with the
output of the HSP50210 Costas loop [1].
AOUT(15:0)
O
Parallel Output Bus A. Two parallel output modes are available on the HSP50214B. The first is called the Direct
Output Port, where the source is selected through Control Word 20 (see the Microprocessor Write Section) and
comes directly from the Output MUX Section (see Output Control Section). The most significant byte of AOUT
always outputs the most significant byte of the Parallel Direct Output Port whose data type is selected via μP
interface. AOUT15 is the MSB. In this mode, the AOUT(15:0) bus is updated as soon as data is available. DATARDY
is asserted to indicate new data. For this mode, the output choices are: I, |r|, or f. The format is 2’s complement,
except for magnitude, which is unsigned binary with a zero as the MSB.
The second mode for parallel data is called the Buffer RAM Output Port. The Buffer RAM Output Port acts like a
FIFO for blocks of information called data sets. Within a data set is I, Q, magnitude, phase, and frequency
information; a data type is selected using SEL(2:0). Up to 7 data sets are stored in the Buffer RAM Output Port. The
LSBytes of the AOUT and BOUT busses form the 16-bits for the buffered output mode and can be used for buffered
mode while the MSBytes are outputting data in the direct output mode. For this mode, the output formats are the
same as the Direct Output Port mode.
BOUT(15:0)
O
Parallel Output Bus B. Two parallel output modes are available on the HSP50214B. The first is called the Direct
Output Port, where the source is selected through Control Word 20 (see the Microprocessor Write Section) and
comes directly from the Output MUX Section (see Output Control Section). The most significant byte of BOUT
always outputs the most significant byte of the Parallel Direct Output Port whose data type is selected via μP
interface. BOUT15 is the MSB. In this mode, the BOUT(15:0) bus is updated as soon as data is available. DATARDY
is asserted to indicate new data. For this mode, the output choices are: Q, φ, or |r|. The format is 2’s complement,
except for magnitude which is unsigned binary with a zero as the MSB.
The second mode for parallel data is called the Buffer RAM Output Port. The Buffer RAM Output Port acts like a
FIFO for blocks of information called data sets. Within a data set is I, Q, magnitude, phase, and frequency
information; a particular information is selected using SEL(2:0). Up to 7 data sets is stored in the Buffer RAM Output
Port. The least significant byte of BOUT can be used to either output the least significant byte of the B Parallel Direct
Output Port or the least significant byte of the Buffer RAM Output Port. See Output Section. For this mode the output
formats are the same as the Direct Output Port mode.
4
FN4450.4
May 1, 2007
HSP50214B
Pin Descriptions
(Continued)
NAME
TYPE
DESCRIPTION
DATARDY
O
Output Strobe Signal. Active Low. Indicates when new data from the Direct Output Port Section is available.
DATARDY is asserted for one PROCCLK cycle during the first clock cycle that data is available on the parallel out
busses. See Output Section.
OEAH
I
Output enable for the MSByte of the AOUT bus. Active Low. The AOUT MSByte outputs are three-stated when
OEAH is high.
OEAL
I
Output enable for the LSByte of the AOUT bus. Active Low. The AOUT LSByte outputs are three-stated when
OEAL is high.
OEBH
I
Output enable for the MSByte of the BOUT bus. Active Low. The BOUT MSByte outputs are three-stated when
OEBH is high.
OEBL
I
Output enable for the LSByte of the BOUT bus. Active Low. The BOUT LSByte outputs are three-stated when
OEBL is high.
SEL(2:0)
I
Select Address is used to choose which information in a data set from the Buffer RAM Output Port is sent to the
least significant bytes of AOUT and BOUT. SEL2 is the MSB.
INTRRP
O
Interrupt Output. Active Low. This output is asserted for 8 PROCCLK cycles when the Buffer RAM Output Port is
ready for reading.
SEROUTA
O
Serial Output Bus A Data. I, Q, magnitude, phase, frequency, timing error and AGC information can be sequenced
in programmable order. See Output Section and Microprocessor Write Section.
SEROUTB
O
Serial Output Bus B Data. Contents may be related to SEROUTA. I, Q, magnitude, phase, frequency, timing error
and AGC information can be sequenced in programmable order. See Output Section and Microprocessor Write
Section.
SERCLK
O
Output Clock for Serial Data Out. Derived from PROCCLK as given by Control Word 20 in the Microprocessor Write
Section.
SERSYNC
O
Serial Output Sync Signal. Serves as serial data strobes. See Output Section and Microprocessor Write Section.
SEROE
I
Serial Output Enable. When high, the SEROUTA, SEROUTB, SERCLK, and SERSYNC signals are set to a high
impedance.
C(7:0)
I/O
A(2:0)
I
Processor Interface Address Bus. See Microprocessor Write Section. A2 is the MSB.
WR
I
Processor Interface Write Strobe. C(7:0) is written to Control Words selected by A(2:0) in the Programmable Down
Converter on the rising edge of this signal. See Microprocessor Write Section.
RD
I
Processor Interface Read Strobe. C(7:0) is read from output or status locations selected by A(2:0) in the
Programmable Down Converter on the falling edge of this signal. See Microprocessor Read Section.
REFCLK
I
Reference Clock. Used as an input clock for the timing error detector. The timing error is computed relative to
REFCLK. REFCLK frequency must be less than or equal to PROCCLK/2.
MSYNCO
O
Multiple Chip Sync Output. Provided for synchronizing multiple parts when CLKIN and PROCCLK are
asynchronous. MSYNCO is the synchronization signal between the input section operating under CLKIN and the
back end processing operating under PROCCLK. This output sync signal from one part is connected to the MSYNCI
signal of all the HSP50214Bs.
MSYNCI
I
Multiple Chip Sync Input. The MSYNCI pin of all the parts should be tied to the MSYNCO of one part.
NOTE: MSYNCI must be connected to an MSYNCO signal for operation.
SYNCIN1
I
CIC Decimation/Carrier NCO Update Sync. Can be used to synchronize the CIC Section, carrier NCO update, or
both. See the Multiple Chip Synchronization Section and Control Word 0 in the Microprocessor Write Section. Active
High.
SYNCIN2
I
FIR/Timing NCO Update/AGC Gain Update Sync. Can be used to synchronize the FIR, Timing NCO update, AGC
gain update, or any combination of the above. See the Multiple Chip Synchronization Section and Control Words 7,
8, and 10 in the Microprocessor Write Section. Active High.
SYNCOUT
O
Strobe Output. This synchronization signal is generated by the μP interface for synchronizing multiple parts. Can be
generated by PROCLK or CLKIN (see Control Word 0 and Control Word 24 in the Microprocessor Write Section).
Active High.
Processor Interface Data Bus. See Microprocessor Write Section. C7 is the MSB.
5
FN4450.4
May 1, 2007
AGCGNSEL
TO OUTPUT FORMATTER
AND MICROPROCESSOR
INTERFACE
PROCCLK
AGCOUT
LOOP
FILTER
LIMIT
ERROR
DETECT
A
CLKIN
GAINADJ(2:0)
(CO = 1;
Cn = 0)
INTRRP
0 TO 5 HALFBAND FILTER;
DECIMATION UP TO 32
255-TAP
PROGRAMMABLE
FIR FILTER
(DECIMATE UP TO 16)
AGC
SHIFT
COS
SIN
(CO = 1;
Cn = 0)
COF
RE-SAMPLER
2
INTERPOLATE
BY 2/4
HALFBAND
FILTERS
I +Q
Q
DISCRIMINATOR
63-TAP
dθ PROGRAMMABLE
FIR FILTER
dt
(SYMBOL TRACKING)
BOUT(15:0)
Q
atan ⎛ ----⎞
⎝ I⎠
NCO
(CARRIER TRACKING)
AOUT(15:0)
2
OEAH
OEAL
OEBH
OEBL
INTRRP
SEL(2:0)
SEROUTA
SOF
SEROUTB
NCO
SOFSYNC
DIFFERENCE
REFCLK
TIMING ERROR
A
MICROPROCESSOR
READ/WRITE
RD
WR
A(2:0)
C(7:0)
CONTROL
SECTION
OUTPUT SECTION
DISCRIMINATOR SECTION
INPUT SECTION
LEVEL DETECT SECTION
SYNCHRONIZATION SECTION
CARRIER NCO SECTIONS
CIC, HALFBAND FILTER, AND FIR SECTIONS
DIGITAL AGC SECTION
RE-SAMPLER/INTERPOLATION HALFBAND SECTION
TIMING NCO
CLKIN
PROCCLK
FN4450.4
May 1, 2007
FIGURE 1. FUNCTIONAL BLOCK DIAGRAM OF THE HSP50214B PROGRAMMABLE DOWNCONVERTER
AGCOUT
CHIP
SYNCHRONIZATION
CIRCUITRY
SERCLK
SERSYNC
SEROE
MSYNCI
SYNCOUT
MSYNCO
BACK END
SYNCHRONIZATION
CIRCUITRY
SYNCIN2
FRONT END
SYNCHRONIZATION
CIRCUITRY
SYNCIN1
HSP50214B
TO
μPROCESSOR
INTERFACE
COFSYNC
DATARDY
OUTPUT FORMATTER
MIXER
5TH ORDER
CIC
DECIMATE
FROM 4-32
I
POLYPHASE
FILTER
LEVEL
DETECT
CARTESIAN
TO
POLAR
POLYPHASE
FILTER
SHIFT
6
IN(13:0)
INPUT
SECTION
ENI
HSP50214B
Functional Description
The HSP50214B Programmable Downconverter (PDC) is an
agile digital tuner designed to meet the requirements of a
wide variety of communications industry standards. The
PDC contains the processing functions needed to convert
sampled IF signals to baseband digital samples. These
functions include LO generation/mixing, decimation filtering,
programmable FIR shaping/bandlimiting filtering,
resampling, Automatic Gain Control (AGC), frequency
discrimination and detection as well as multi-chip
synchronization. The HSP50214B interfaces directly with a
DSP microprocessor to pass baseband and status data.
A top level functional block diagram of the HSP50214B is
shown in Figure 1. The diagram shows the major blocks and
multiplexers used to reconfigure the data path for various
architectures. The HSP50214B can be broken into 13
sections: Synchronization, Input, Input Level Detector,
Carrier Mixer/Numerically Control Oscillator (NCO), CIC
Decimating Filter, Halfband Decimating Filter, 255-Tap
Programmable FIR Filter, Automatic Gain Control (AGC),
Re-sampler/Halfband Filter, Timing NCO, Cartesian to Polar
Converter, Discriminator, and Output Sections. All of these
sections are configured through a microprocessor interface.
The HSP50214B has three clock inputs; two are required and
one is optional. The input level detector, carrier NCO, and CIC
decimating filter sections operate on the rising edge of the
input clock, CLKIN. The halfband filter, programmable FIR
filter, AGC, Re-Sampler / Halfband filters, timing NCO,
discriminator, and output sections operate on the rising edge
of PROCCLK. The third clock, REFCLK, is used to generate
timing error information.
NOTE: All of the clocks may be asynchronous.
PDC Applications Overview
This section highlights the motivation behind the key
programmable features from a communications system level
perspective. These motivations will be defined in terms of ability
to provide DSP processing capability for specific modulation
formats and communication applications. The versatility of the
Programmable Downconverter can be intimidating because of
the many Control Words required for chip configuration. This
section provides system level insight to help allay reservations
about this versatile DSP product. It should help the designer
capitalize on the greatest feature of the PDC - VERSATILITY
THROUGH PROGRAMMABILITY. It is this feature, when fully
understood, that brings the greatest return on design
investment by offering a single receiver design that can process
the many waveforms required in the communications
marketplace.
FDM Based Standards and Applications
Table 1 provides an overview of some common frequency
division multiplex (FDM) base station applications to which the
PDC can be applied. The PDC provides excellent selectivity
7
for frequency division multiple access (FDMA) signals. This
high selectivity is achieved with 0.012Hz resolution frequency
control of the NCO and the sharp filter responses capable with
a 255-tap, 22-bit coefficient FIR filter. The 16-bit resolution out
of the Cartesian to Polar Coordinate Converter are routed to
the frequency detector, which is followed by a 63-tap, 22-bit
coefficient FIR filter structure for facilitating FM and FSK
detection. The 14-bit input resolution is the smallest bit
resolution found throughout the conversion and filtering
sections, providing excellent dynamic range in the DSP
processing. A unique input gain scaler adds an additional
42dB of range to the input level variation, to compensate for
changes in the analog RF front end receive equipment.
Synchronization circuitry allows precise timing control of the
base station reconfiguration for all receive channels
simultaneously. Portions of this table were corroborated with
reference [2].
TABLE 1. CELLULAR PHONE BASE STATION
APPLICATIONS USING FDMA
STANDARD
AMPS
(IS-91)
RX BAND 824-849
(MHz)
MCS-L1 NMT-400
MCS-L2 NMT-900
C450
ETACS
NTACS
925-940
453-458
890-915
451-456
871-904
915-925
CHANNEL
BW (kHz)
30
25.0
12.5
25
12.5
20.0
10.0
25.0
12.5
# TRAFFIC
CHANNELS
832
600
1200
200
1999
222
444
1240
800
VOICE
MODULATION
FM
FM
FM
FM
FM
PEAK
DEVIATION
(kHz)
12
5
5
4
9.5
CONTROL
MODULATION
FSK
FSK
FSK
FSK
FSK
PEAK
DEVIATION
(kHz)
8
4.5
3.5
2.5
6.4
CONTROL
CHANNEL
RATE
(Kbps)
10
0.3
1.2
5.3
8
TDM Based Standards and Applications
Table 2 provides an overview of some common Time Division
Multiplexed (TDM) base station applications to which the PDC
can be applied. For time division multiple access (TDMA)
applications, such as North American TDMA (IS136), where
30kHz is the received band of interest for the PCS
basestation, the PDC offers 0.012Hz frequency resolution in
downconversion in addition to α = 0.35 matched
(programmable) filtering capability. The π/4 DPSK modulation
can be processed using the PDC Cartesian to Polar
coordinate converter and dφ/dt detector circuitry or by
processing the I/Q samples in the DSP μP. The PDC provides
FN4450.4
May 1, 2007
HSP50214B
the ability to change the received signal gain and frequency,
synchronous with burst timing. The synchronous gain
adjustment allows the user to measure the power of the signal
at the A/D at the end of a burst, and synchronously reload that
same gain value at the arrival of the next user burst.
For applications other than cellular phones (where the
preambles are not changed), the PDC frequency
discriminator output can be used to obtain correlation on the
preamble pattern to aid in burst acquisition.
TABLE 2. CELLULAR BASESTATION APPLICATIONS USING
TDMA
STANDARD
GSM
PCN
IS-54
TYPE
Cellular
Cellular
Cellular
BASESTATION RX
BAND (MHz)
935-960
1805-1880
824-849
200
200
30
CHANNEL BW (kHz)
# TRAFFIC CHANNELS
VOICE MODULATION
8
16
3
GMSK
GMSK
π/4
DQPSK
CHANNEL RATE (Kbps)
270.8
270.8
48.6
CONTROL
MODULATION
GMSK
GMSK
π/4
DQPSK
CHANNEL RATE (Kbps)
270.8
270.8
48.6
Several applications are combinations of frequency and time
domain multiple access schemes. For example, GSM is a
TDMA signal that is frequency hopped. The individual
channels contain Gaussian MSK modulated signals. The
PDC again offers the 0.012Hz tuning resolution for dehopping the received signal. The combination of halfband
and 256-tap programmable, 22-bit coefficient FIR filters
readily performs the necessary matched filtering for
demodulation and optimum detection of the GMSK signals.
CDMA Based Standards and Applications
For Code Division Multiple Access (CDMA) type signals, the
PDC offers the ability to have a single wideband RF front
end, from which it can select a single spread channel of
interest. The synchronization circuitry provides for easy
control of multiple PDC for applications where multiple
received signals are required, such as base-stations.
In IS-95 CDMA, the receive signal bandwidth is
approximately 1.2288MHz wide with many spread spectrum
channel in the band. The PDC supplies the downconversion
and filtering required to receive a single RF channel in the
presence of strong adjacent interference. Multiple PDC’s
would be sourced from a single receive RF chain, each
processing a different receive frequency channel. The
despreader would usually follow the PDC. In some very
specific applications, with short, fixed codes, the filtering and
despreading may be possible with innovative use of the
programmable, 22-bit coefficient FIR filter. The PDC offers
0.012Hz resolution on tuning to the desired receive channel
and excellent rejection of the portions of the band not being
8
processed, via the halfband and 255-tap programmable, 22bit coefficient FIR filter.
Traditional Modulation Formats
AM, ASK, FM AND FSK
The PDC has the capability to fully demodulate AM and FM
modulated waveforms. The PDC outputs 15-bits of amplitude
or 16-bits of frequency for these modulation formats. The FM
discriminator has a 63-tap programmable, 22-bit coefficient FIR
filter for additional signal conditioning of the FM signal. Digital
versions of these formats, ASK and FSK are also readily
processed using the PDC. Just as in the AM modulated case,
ASK signals will use 15-bit magnitude output of the Cartesian to
Polar Coordinate converter. Multi-tone FSK can be processed
several ways. The frequency information out of the
discriminator can be used to identify the received tone, or the
filter can be used to identify and power detect a specific tone of
the received signal. AMPS is an example of an FM application.
PM AND PSK
The PDC provides the downconversion, demodulation,
matched filtering and coordinate conversion required for
demodulation of PM and PSK modulated waveforms. These
modulation formats will require external carrier and symbol
timing recovery loop filters to complete the receiver design.
The PDC was designed to interface with the HSP50210
Digital Costas Loop to implement the carrier phase and
symbol timing recovery loop filters (for continuous PSK
signals-not burst).
Digital modulation formats that combine amplitude and
phase for symbol mapping, such as m-ary QAM, can also be
downconverted, demodulated, and matched filtered. The
received symbol information is provided with 16-bits of
resolution in either Cartesian or Polar coordinates to
facilitate remapping into bits and to recover the carrier
phase. External Symbol mapping and Carrier Recovery
Loop Filtering is required for this waveform.
Resampling and Interpolation Filters
Two key features of the resampling FIR filter are that the resampler filter allows the output sample rate to be
programmed with millihertz resolution and that the output
sample rate can be phase locked to an independent
separate clock. The re-sampler frees the front end sampling
clocks from having to be synchronous or integrally related in
rate to the baseband output. The asynchronous relationship
between front end and back end clocks is critical in
applications where ISDN interfaces drive the baseband
interfaces, but the channel sample rates are not related in
any way. The interpolation halfband filters can increase the
rate of the output when narrow frequency bands are being
processed. The increase in output rate allows maximum use
of the programmable FIR while preserving time resolution in
the baseband data.
FN4450.4
May 1, 2007
HSP50214B
14-Bit Input and Processing Resolution
The PDC maintains a minimum of 14-bits of processing
resolution through to the output, providing over 84dB of
dynamic range. The 18-bits of resolution on the internal
references provide a spurious floor that is better than 98dBc.
Furthermore, the PDC provides up to 42dB of gain scaling to
compensate for any change in gain in the RF front end as
well as up to 96dB of gain in the internal PDC AGC. This
gain maximizes the output resolution for small signals and
compensates for changes in the RF front end gain, to handle
changes in the incoming signal.
Summary
The greatest feature of the PDC is its ability to be
reconfigured to process many common standards in the
communications industry. Thus, a single hardware element
can receive and process a wide variety of signals from PCS
to traditional cellular, from wireless local loop to SATCOM.
The high resolution frequency tuning and narrowband
filtering are instrumental in almost all of the applications.
Multiple Chip Synchronization
Multiple PDCs are synchronized using a MASTER/SLAVE
configuration. One part is responsible for synchronizing the
front end internal circuitry using CLKIN while another part is
responsible for synchronizing the backend internal circuitry
using PROCCLK.
The PDC is synchronized with other PDCs using five control
lines: SYNCOUT, SYNCIN1, SYNCIN2, MSYNCO, and
MSYNCI. Figure 2 shows the interconnection of these five
signals for multiple chip synchronization where different
sources are used for CLKIN and PROCCLK.
PDC A is the Master sync through MSO.
PDC B configures the CLKIN sync through SYNCIN1.
PDC A configures the PROCCLK sync through SYNCIN2.
A
B
HSP50214B
MSYNCO
MSYNCI
MSYNCI
(MASTER
SYNCOUT
SYNCIN2)
SYNCOUT
SYNCIN2
SYNCIN2
SYNCIN1
SYNCIN1
(MASTER
SYNCIN1)
ALL OTHER SYNCIN1
ALL OTHER SYNCIN2
ALL OTHER MSI
FIGURE 2. SYNCHRONIZATION CIRCUIT
SYNCOUT for PDC B should be set to be synchronous with
CLKIN (Control Word 0, Bit 3 = 0. See the Microprocessor
Write Section). SYNCOUT for PDC B is tied to the SYNCIN1
of all the PDCs. The SYNCIN1 can be programmed so that
9
SYNCOUT for one of the PDC’s other than PDC B, should
be set for PROCCLK (bit 3 = 1 in Control Word 0). This
output signal is tied to the SYNCIN2 of all PDCs. The
SYNCIN2 can be programmed so that the AGC updates its
accumulator with the contents in the master registers
(Control Word 8, Bit 29 in the Microprocessor Write Section).
SYNCIN2 is also used to load or reset the timing NCO using
bit 5, Control Word 11. The halfband and FIR filters can be
reset on SYNCIN2 using Control Word 7, Bit 21. The
MSYNCO of one of the PDCs is then used to drive the
MSYNCI of all the PDCs (including its own).
For application configurations where CLKIN and PROCCLK
have the same source, SYNCIN1 and SYNCIN2 can be tied
together. However, if different enabling is desired for the
front end and backend processing of the PDC’s, these
signals can still be controlled independently.
In the HSP50214B, the Control Word 25 reset signal has
been extended so that the front end reset is 10 CLKIN
periods wide and the back end reset is 10 PROCCLK
periods wide. This guarantees that no enables will be caught
in the pipelines. In addition, the SYNCIN1 internal reset
signal, which is enabled by setting Control Word 7,
Bit 21 = 1, has been extended to 10 cycles.
In summary, SYNCIN1 is used to update carrier phase
offset, update carrier center frequency, reset CIC decimation
counters and reset the carrier NCO (clear the feedback in
the NCO). SYNCIN2 is used to reset the HB filter, FIR filter,
re-sampler / HB state machines and the output FIFO, load a
new gain into the AGC and load a new re-sampler NCO
center frequency and phase offset.
Input Section
HSP50214B
MSYNCO
the carrier NCO and/or the 5th order CIC filter of all PDCs can
be synchronously loaded / updated using SYNCIN1. See
Control Word 0, Bits 19 and 20 in the Microprocessor Write
Section for details.
The block diagram of the input controller is provided in
Figure 3. The input can support offset binary or two’s
complement data and can be operated in gated or
interpolated mode (see Control Word 0 from the
Microprocessor Write Section). The gated mode takes one
sample per clock when the input enable (ENI) is asserted.
The gated mode allows the user to synchronize a low speed
sampling clock to a high speed CLKIN.
The interpolated mode allows the user to input data at a low
sample rate and to zero-stuff the data prior to filtering. This
zero stuffing effectively interpolates the input signal up to the
rate of the input clock (CLKIN). This interpolated mode
allows the part to be used at rates where the sampling
frequency is above the maximum input rate range of the
halfband filter section, and where the desired output
bandwidth is too wide to use a Cascaded Integrator Comb
(CIC) filter without significantly reducing the dynamic range.
FN4450.4
May 1, 2007
HSP50214B
programmed via the microprocessor interface, as shown in
Figure 9. The bit weighting of the data path through the
input threshold detector is shown in Figure 10. The
threshold is a signed number, so it should be set to the
inverse of the desired input level. The threshold can be set
to zero if the average input level is desired instead of the
error. The sum of the threshold and the absolute value of
the input is accumulated in a 32-bit accumulator. The
accumulator can handle up to 218 samples without
overflow. The integration time is controlled by an 18-bit
counter. The integration counter preload (ICPrel) is
programmed via the microprocessor interface through
Control Word 1. Only the upper 16-bits are programmable.
The 2 LSBs are always zero. Control Word 1, Bits 29-14
are programmed to:
See Figures 4 to 7 for an interpolated input example,
detailing the associated spectral results.
Interpolation Example:
The specifications for the interpolated input example are:
CLKIN = 40MHz
Input Sample Rate = 5MSPS
PROCCLK = 28MHz
Interpolate by 8, Decimate by 10
Desired 85dB dynamic range output bandwidth = 500kHz
Input Level Detector
The Input Level Detector Section measures the average
magnitude error at the PDC input for the microprocessor by
comparing the input level against a programmable
threshold and then integrating the result. It is intended to
provide a gain error for use in an AGC loop with either the
RF/IF or A/D converter stages (see Figure 8). The AGC
loop includes Input Level Detector, the microprocessor and
an external gain control amplifier (or attenuator). The input
samples are rectified and added to a threshold
ICPrel = ( N ) ⁄ 4 + 1
(EQ. 1)
where N is the desired integration period, defined as the
number of input samples to be integrated. N must be a
multiple of 4: [0, 4, 8, 12, 16 .... , 218].
INPUT LEVEL DETECTOR †
STATUS (0) †
INPUT_THRESH †
INTG_MODE
†
15
REG
14
15
18
†
SHIFT
14
REG
REG
IN(13:0)
REG
INPUT FORMAT
INTG_INTEVAL
MUX
LEVEL
DETECT
18
NCO ††
LIMIT
EN
3
GAINADJ(2:0)
BYPASS
4
†
CIC
INPUT
FORMAT †
∑
DELAY 3
4
ENI
INTERP
DELAY 3
†
CONTROL WORD 0
CONTROL
LOGIC
CONTROL WORD 1
CLKIN
INPUT_MODE †
INPUT_FMT †
INPUT_THRESH †
INTG_MODE †
INTG_INTEVAL †
† Controlled via microprocessor interface.
†† See NCO Section for more details.
FIGURE 3. BLOCK DIAGRAM OF THE INPUT SECTION
BYPASS
CIC
FILTER
MIN. R = 4
CLKIN = 5MHz
MUX
5MHz
PROCCLK = 28MHz
HB/FIR FILTER
MAX. fS = 4MHz
(EXCEEDED IN
BYPASS PATH)
500kHz = 85dB
BANDWIDTH
(NOT ACHIEVED
WITH CIC FILTER
PATH)
Without Interpolation, the CIC bypass path exceeds the HB/FIR filter
input sample rate and the CIC filter path will not yield the desired
85dB dynamic range band width of 500kHz.
FIGURE 4. STATEMENT OF THE PROBLEM
10
↑8 (0 STUFF) = 40MHz
5MHz
CIC FILTER
R = ↓10
4MHz
HB/FIR FILTER
500kHz = 85dB
BANDWIDTH
CLKIN = 40MHz
FIGURE 5. BLOCK DIAGRAM OF THE INTERPOLATION
APPROACH
FN4450.4
May 1, 2007
HSP50214B
fS
5MHz
2fS
10MHz
3fS
4fS
5fS
6fS
7fS
8fS
15MHz
20MHz
25MHz
30MHz
35MHz
40MHz
THE INPUT DATA SPECTRUM SAMPLED AT RATE R = fS
f’S /8
5MHz
f’S/4
10MHz
3f’S/8
15MHz
f’S/2
20MHz
5f’S /8
25MHz
3f’S /4
30MHz
7f’S /8
35MHz
9fS
45MHz
10fS
50MHz
f’S
40MHz
FIGURE 6. INTERPOLATION SPECTRUM: INTERPOLATE BY 8 THE INPUT DATA WITH ZERO STUFFING; SAMPLE AT RATE R = f’s
4MHz
8MHz
12MHz 16MHz 20MHz 24MHz
28MHz 32MHz 36MHz 40MHz
DECIMATE BY 10 AND CIC FILTER; SAMPLE AT RATE R = f’s/10
85dB DYNAMIC RANGE BANDWIDTH
CIC FILTER
FREQUENCY
RESPONSE
CIC FILTER ALIAS PROFILE
O.5MHz
1MHz
2MHz
3MHz
4MHz
FIGURE 7. ALIAS PROFILE AND THE 85dB DYNAMIC RANGE BANDWIDTH
μPROC
INPUT LEVEL
DETECTOR (24-BIT
ERROR VALUE)
DAC
THRESH
IF
INPUT
A/D
PDC
GCA
FIGURE 8. PROCESSOR BASED EXTERNAL IF AGC
11
FN4450.4
May 1, 2007
HSP50214B
ACCUMULATOR
ADDR(2:0)
32
INPUT
GATING
LOGIC
IN(13:0)
+
|X|
+
CLKIN
R
E
G
R
E
G
24
M
U
X
8
TO
μPROC
R
E
INPUT_THRESHOLD †
G
INTEGRATION_INTERVAL†
16
START †
INTEGRATION_MODE †
“0”
COUNTER
CLKIN
† Controlled via microprocessor interface.
CONTINUOUS
SINGLE
010
001
0
218
217
216
215
214
213
212
211
210
29
28
27
26
25
24
23
22
21
20
2-1
2-2
2-3
2-4
000
-20
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
218
217
216
215
214
213
212
211
210
29
28
27
26
25
24
23
22
21
20
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
μPROC READ
PORTS
READ CODE A(2:0)
MAGNITUDE
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
ACCUMULATOR
20
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
THESHOLD
f
0
S 2
-6dB 2-1
-12dB 2-2
-18dB 2-3
-24dB 2-4
-30dB 2-5
-36dB 2-6
-42dB 2-7
-48dB 2-8
-54dB 2-9
-60dB 2-10
-66dB 2-11
-72dB 2-12
-78dB 2-13
INPUT
A/D
OUTPUT
FIGURE 9.
FIGURE 10. INPUT THRESHOLD DETECTOR BIT WEIGHTING
12
The integration period counter can be set up to run
continuously or to count down and stop. Continuous integration
counter operation lets the counter run, with sampling occurring
every time the counter reaches zero. Because the processor
samples the detector read port asynchronous to the CLKIN,
data can be missed unless the status bit is monitored by the
processor to ensure that a sample is taken for every integration
count down sequence.
Additionally, in the HSP50214B, the ability to align the
start/restart of the input level detector integration period with
an external event is provided. This allows the sync signals,
which are synchronized to external events, to be used to align
all of the gain adjustments or measurements. If Control Word
27, Bit 17 is set to a logic one, the SYNCIN1 signal will cause
the input level detector to start/restart its integration period. If
Control Word 27, Bit 17 is set to a logic zero, control of the
start/restart of the input level detector integration period does
not respond to SYNCIN1.
In the count down and stop mode, the microprocessor read
commands can be synchronized to system events, such as the
start of a burst for a TDMA application. The integration counter
can be started at any time by writing to Control Word 2. At the
end of the integration period (counter = 0000), the upper 23-bits
of the accumulator are transferred to a holding register for
reading by the microprocessor. Note that it is not the restarting
of the counter (by writing to Control Word 2) that latches the
current value, but the end of the integration count. When the
accumulator results are latched, a bit is set in the Status
Register to notify the processor. Reading the most significant
byte of the 23-bits clears the status bit. See the Microprocessor
Read Section. Figure 11 illustrates a typical AGC detection
process.
FN4450.4
May 1, 2007
HSP50214B
Typically, the average input error is read from the Input Level
Detector port for use in AGC Applications. By setting the
threshold to 0, however, the average value of the input signal
can be read directly. The calculation is:
modulo 232. The output frequency of the NCO is computed
as:
dBFS RMS = ( 20 ) log [ ( 1.111 ) ( level ) ⁄ ( ( N ) ( 16 ) ) ]
or in terms of the programmed value:
(EQ. 2)
where “level” is the 24-bit value read from the 3 level
Detector Registers and “N” is the number of samples to be
integrated. Note that to get the RMS value of a sinusoid,
multiply the average value of the rectified sinusoid by 1.111.
For a full scale input sinusoid, this yields an RMS value of
approximately 3dBfS.
fC = fS * N ⁄ ( 2
32
N = INT [ f C × 2
(EQ. 3)
),
32
where N is the 32-bit sum of the center and offset frequency
terms, fC is the frequency of the carrier NCO sinusoids, fS is
the input sampling frequency, and INT is the integer of the
computation. See the Microprocessor Write Section on
instructions for writing Control Word 3.
NOTE: 1.111 scales the rectified sinusoid average (2/π) to 1/ √2
TO MIXERS
COS
SIN
18 18
.
B) RECTIFIED SIGNAL
AMPLITUDE
AMPLITUDE
A) INPUT SIGNAL
(EQ. 3A)
⁄ f S ] HEX ,
REG
SIN/COS
ROM
18
AMPLITUDE
E) DETECTOR OUTPUT
+
AMPLITUDE
D) ACCUMULATOR INPUTS
F) CLOSED LOOP STEADY STATE
(CONSTANT INPUT)
AMPLITUDE
AMPLITUDE
C) THRESHOLD
In the HSP50214B, the polarity of the LSB’s of the
integration period pre-load is selectable. If Control Word 27,
Bit 23 is set to a logic one, the two LSB’s of the integration
period preload are set to logic ones. This allows a power of
two to be set for the integration period, for easy
normalization in the processor. If Control Word 27, Bit 23 is
set to a logic zero, then the two LSB’s of the integration
period preload are set to zeros as in the HSP50214.
Carrier Synthesizer/Mixer
The Carrier Synthesizer/Mixer Section of the HSP50214B is
shown in Figure 12. The NCO has a 32-bit phase
accumulator, a 10-bit phase offset adder, and a sine/cosine
ROM. The frequency of the NCO is the sum of a center
frequency Control Word, loaded via the microprocessor
interface (Control Word 3, Bits 0 to 31), and an offset
frequency, loaded serially via the COF and COFSYNC pins.
The offset frequency can be zeroed in Control Word 0, Bit 1.
Both frequency control terms are 32-bits and the addition is
13
10 R
E
G
R
E
G
PHASE
ACCUMULATOR
CARRIER
PHASE
OFFSET †
CLEAR
PHASE
ACCUM †
0
REG
ENI
MUX
R
E
G
+
COF
ENABLE †
MUX
32
COF
0
32
CF
REG
COFSYNC
FIGURE 11. SIGNAL PROCESSING WITHIN LEVEL DETECTOR
CARRIER
PHASE
STROBE †
REG
SYNC
COF
SHIFT REG
SYNCIN1
SYNC
CIRCUITRY
CARRIER
FREQUENCY
STROBE †
REG
REG
CARRIER
FREQUENCY†
(fC)
CARRIER
LOAD ON
UPDATE†
† Controlled via microprocessor interface.
FIGURE 12. BLOCK DIAGRAM OF NCO SECTION
For example, if N is 3267 (decimal), and fS is 65MHz, then fC
is 49.44Hz. If received data is modulated at a carrier
frequency of 10MHz, then the synthesizer/mixer should be
programmed for N = 27627627 (hex) or D89D89D8 (hex).
Because the input enable, ENI, controls the operation of the
phase accumulator, the NCO output frequency is computed
relative to the input sample rate, fS, not to fCLKIN. The
frequency control, N, is interpreted as two’s complement
because the output of the NCO is quadrature. Negative
frequency L.O.s select the upper sideband; positive frequency
L.O.s select the lower sideband. The range of the NCO is
-fS /2 to +fS /2. The frequency resolution of the NCO is fS /(232)
or approximately 0.015Hz when CLKIN is 65MSPS and ENI is
tied low.
FN4450.4
May 1, 2007
HSP50214B
φ OFF = 2π × ( PO ⁄ 2
10
9
9
) ;( – ( 2 ) ≤ PO ≤ ( 2 – 1 ) )
(EQ. 4)
( – 512 to 511 )
or, in terms of the parameter to be programmed:
PO = INT [ ( 2
10
φ OFF ) ⁄ 2π ] HEX ;( – π < φ OFF < π )
(EQ. 4A)
where PO is the 10-bit two’s complement value loaded into the
Phase Offset Register (Control Word 4, Bits 9-0). For example,
a value of 32 (decimal) loaded into the Phase Offset Register
would produce a phase offset of 11.25o and a value of -512
would produce an offset of 180o. The phase offset is loaded via
the microprocessor interface. See the Microprocessor Write
Section on instructions for writing Control Word 4.
The most significant 18-bits from the phase adder are used
as the address a sin/cos lookup table. This lookup table
maps phase into sinusoidal amplitude. The sine and cosine
values have 18-bits of amplitude resolution. The spurious
components in the sine/cosine generation are at least
-102dBc. The sine and cosine samples are routed to the
mixer section where they are multiplied with the input
samples to translate the signal of interest to baseband.
The mixer multiplies the 14-bit input by the 18-bit quadrature
sinusoids. The mixer equations are:
I OUT = I IN × cos ( ω c )
(EQ. 5)
Q OUT = I IN × sin ( ω c )
(EQ. 5A)
The mixer output is rounded symmetrically to 15-bits.
To allow the frequency and phase of multiple parts to be
updated synchronously, two sets of registers are used for
latching the center frequency and phase offset words. The
offset phase and center frequency Control Words are first
loaded into holding registers. The contents of the holding
registers are transferred to active registers in one of two ways.
The first technique involves writing to a specific Control Word
Address. A processor write to Control Word 5, transfers the
center frequency value to the active register while a processor
write to Control Word 6 transfers the phase offset value to the
active register.
The second technique, designed for synchronizing updates to
multiple parts, uses the SYNCIN1 pin to update the active
registers. When Control Word 1, Bit 20 is set to 1, the SYNCIN1
pin causes both the center frequency and Phase Offset Holding
Registers to be transferred to active registers. Additionally,
when Control Word 0, Bit 0 is set to 1, the feedback in the
phase accumulator is zeroed when the transfer from the
holding to active register occurs. This feature provides
14
synchronization of the phase accumulator starting phase of
multiple parts. It can also be used to reset the phase of the
NCO synchronous with a specific event.
The carrier offset frequency is loaded using the COF and
COFSYNC pins. Figure 13 details the timing relationship
between COF, COFSYNC and CLKIN. The offset frequency
word can be zeroed if it is not needed. Similarly, the
Sample Offset Frequency Register controlling the ReSampler NCO is loaded via the SOF and SOFSYNC pins.
The procedure for loading data through the two pin NCO
interfaces is identical except that the timing of SOF and
SOFSYNC is relative to PROCCLK.
CLKIN
COFSYNC/
SOFSYNC
COF/
SOF
NOTE:
MSB
LSB
MSB
Data must be loaded MSB first.
FIGURE 13. SERIAL INPUT TIMING FOR COF AND SOF INPUTS
Each serial word has a programmable word width of either 8,
16, 24, or 32-bits (See Control Word 0, Bits 4 and 5, for the
Carrier NCO programming and Control Word 11, Bits 3 and
4, for Timing NCO programming). On the rising edge of the
clock, data on COF or SOF is clocked into an input shift
register. The beginning of a serial word is designated by
asserting either COFSYNC or SOFSYNC “high” one CLK
period prior to the first data bit.
SHIFT COUNTER VALUE
The phase of the Carrier NCO can be shifted by adding a 10bit phase offset to the MSB’s (modulo 360o) of the output of
the phase accumulator. This phase offset control has a
resolution of 0.35o and can be interpreted as two’s
complement from -180o to 180o (-π to π) or as binary from 0
to 360o (0 to 2π). The phase offset is given by:
32 †
30
28
26
24 †
22
20
18
16 †
14
12
10
8†
6
4
2
0
ASSERTION OF
COFSYNC, SOFSYNC
DATA TRANSFERRED
TO HOLDING REGISTER
(8)
(24) (32)
(16)
2
6
10 14 18 22 26 30 34 38 42 46 50 54
CLK TIMES
TD ††
TD ††
TD ††
TD ††
† Serial word width can be: 8, 16, 24, 32 bits wide.
†† TD is determined by the COFSYNC, COFSYNC rate.
FIGURE 14. HOLDING REGISTERS LOAD SEQUENCE FOR
COF AND SOF SERIAL OFFSET FREQUENCY
DATA
NOTE: Serial Data must be loaded MSB first, and COFSYNC or
SOFSYNC should not be asserted for more than one
CLK cycle.
FN4450.4
May 1, 2007
NOTE: COF loading and timing is relative to CLKIN while SOF
loading and timing is relative to PROCCLK.
NOTE: TD can be 0, and the fastest rate is with 8-bit word width.
The assertion of the COFSYNC (or SOFSYNC) starts a count
down from the programmed word width. On following CLKs,
data is shifted into the register until the specified number of
bits have been input. At this point the contents of the register
are transferred from the Shift Register to the respective 32-bit
Holding Register. The Shift Register can accept new data on
the following CLK. If the serial input word is defined to be less
than 32-bits, it will be transferred to the MSBs of the 32-bit
Holding Register and the LSBs of the Holding Register will be
zeroed. See Figure 14 for details.
CIC SHIFT GAIN (FROM PROCESSOR)
HSP50214B
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
8-BIT INPUT
10-BIT INPUT
12-BIT INPUT
14-BIT INPUT
4
8
CIC Decimation Filter
The mixer output may be filtered with the CIC filter or it may
be routed directly to the halfband filters. The CIC filter is used
to reduce the sample rate of a wideband signal to a rate that
the halfbands and programmable filters can process, given
the maximum computation speed of PROCCLK. (See
Halfband and FIR Filter Sections for techniques to calculate
this value).
Prior to the CIC filter, the output of the mixer goes through a
barrel shifter. The shifter is used to adjust the gain in 6dB
steps to compensate for the variation in CIC filter gain with
decimation. (See Equation 6). Fine gain adjustments must
be done in the AGC Section. The shifter is controlled by the
sum of a 4-bit CIC Shift Gain word from the microprocessor
and a 3-bit gain word from the GAINADJ(2:0) pins. The three
bit value is pipelined to match the delay of the input samples.
The sum of the 3 and 4-bit shift gain words saturates at a
value of 15. Table 1 details the permissible values for the
GAINADJ(2:0) barrel shifter control, while Figure 15 shows
the permissible CIC Shift Gain values.
The CIC filter structure for the HSP50214B is fifth order; that
is it has five integrator/comb pairs. A fifth order CIC has
84dB of alias attenuation for output frequencies below 1/8
the CIC output sample rate.
ALLOWABLE CIC SHIFT
GAINS ARE BELOW THE
CURVES
12 16 20 24 28 32 36 40 44 48 52 56 60 64
DECIMATION (R)
FIGURE 15. CIC SHIFT GAIN VALUES
The decimation factor of the CIC filter is programmed in
Control Word 0, Bits 12 - 7. The CIC Shift Gain is programmed
in Control Word 0, Bits 16-13. The CIC Bypass is set in
Control Word 0, Bit 6. When bypassing the CIC filter, the ENI
signal must be de-asserted between samples, i.e., the CLKIN
rate must be ≥ 2 • fS.
TABLE 3. GAIN ADJUST CONTROL AND CIC DECIMATION
ΔGAIN VALUE
(dB)
GAIN ADJ(2:0)
MAX. CIC
DECIMATION
0
000
32
6
001
27
12
010
24
18
011
21
24
100
18
30
101
16
36
110
12
42
111
10
CIC Gain Calculations
The gain through the CIC filter increases with increased
decimation. The programmable barrel shifter that precedes
the first integrator in the CIC is used to offset this variation.
Gain variations due to decimation should be offset using the
4-bit CIC Shift Gain word. This allows the input signal level to
be adjusted in 6dB steps to control the CIC output level.
The gain at each stage of the CIC is:
N
k = R ,
(EQ. 6)
where R is the decimation factor and N is the number of stages.
The input to the CIC from the mixer is 15-bits, and the bit widths
of the accumulators for the five stages in the HSP50214B are
40, 36, 32, 32, and 32, as shown in Figure 16. This limits the
maximum decimation in the CIC to 32 for a full scale input.
If R is 32, the gain through all five integrator stages is 325 = 225.
(The gain through the last four CIC stages is 220, through the
15
FN4450.4
May 1, 2007
HSP50214B
last 3 it is 215, etc.). The sum of the input bits and the growth
bits cannot exceed the accumulator size. This means that for a
decimation of 32 and 15 input bits, the first accumulator must
be 15 + 25 = 40-bits.
The CIC filter decimation counter can be loaded synchronous
with other PDC chips, using the SYNCIN1 signal and the CIC
External Sync Enable bit. The CIC external Sync Enable is set
via Control Word 0, Bit 19.
Thus, the value of the CIC Shift Gain word can be
calculated:
Halfband Decimating Filters
5
SG = FLOOR [ 39 - ( IIN ) - log 2 (R) for 4<R<32
= 15
for R = 4
(EQ. 7)
NOTE: The number of input bits is IIN. (If the number of bits into
the CIC filter is used, the value 40 replaces 39).
For 14-bits, Equation 7 becomes:
5
SG = FLOOR [ 25 – log 2 ( R ) ]for 4 < R < 32
= 15
(EQ. 8A)
for R = 4
For 12-bits, Equation 7 becomes:
5
SG = FLOOR [ 27 – log 2 ( R ) ]for 5 < R < 40
for 4 ≤ R ≤ 5
= 15
(EQ. 8B)
The Programmable Down Converter has five halfband filter
stages, as shown in Figure 17. Each stage decimates by 2
and filters out half of the available bandwidth. The first
halfband, or HB1, has 7 taps. The remaining halfbands;
HB2, HB3, HB4, and HB5; have 11, 15, 19, and 23 taps
respectively. The coefficients for these halfbands are given
in Table 4. Figure 18 shows the frequency response of each
of the halfband filters with respect to normalized frequency,
FN. Frequency normalization is with respect to the input
sampling frequency of each filter section. Each stage is
activated by their respective bit location (15-20) in Control
Word 7. Any combination of halfband filters may be used, or
all may be bypassed.
For 10-bits, Equation 7 becomes:
5
SG = FLOOR [ 29 – log 2 ( R ) ]for 6 < R < 52
for 4 ≤ R ≤ 6
= 15
(EQ. 8C)
For 8-bits, Equation 7 becomes:
5
SG = FLOOR [ 31 – log 2 ( R ) ]for 9 < R < 64
= 15
(EQ. 8D)
for 4 ≤ R ≤ 9
Figure 15 is a plot of Equations 8A through 8D. The 4-bit CIC
Shift Gain word has a range from 0 to 15. The 6-bit
Decimation Factor counter preload field, (R-1), has a range
from 0 to 63, limited by the input resolution as cited above.
Using the Input Gain Adjust Control Signals
The input gain offset control GAINADJ(2:0)) is provided to
offset the signal gain through the part, i.e., to keep the CIC
filter output level constant as the analog front end
attenuation is changed. The gain adjust offset is 6dB per
code, so the gain adjust range is 0 to 42dB. For example, if
12dB of attenuation is switched in at the receiver RF front
end, a code of 2 would increase the gain at the input to the
CIC filter up 12dB so that the CIC filter output would not drop
by 12dB. This fixed gain adjust eliminates the need for the
software to continually normalize.
One must exercise care when using this function as it can
cause overflow in the CIC filter. Each gain adjust in the
shifter from the gain adjust control signals is the equivalent
of an extra bit of input. The maximum decimation in the CIC
is reduced accordingly. With a decimation of 32, all 40-bits of
the CIC are needed, so no input offset gain is allowed. As
the decimation is reduced, the allowable offset gain
increases. Table 3 shows the decimation range versus
desired offset gain range. Table 3 assumes that the CIC Shift
Gain has been programmed per Equation 7 or 8A.
16
FN4450.4
May 1, 2007
ACC4
ACC5
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
2-31
2-32
2-33
2-34
2-35
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
2-31
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
2-31
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
2-31
HALFBAND
FILTER INPUT
CIC
OUTPUT
ACC3
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
ACC1
INPUT
(SHFT=15)
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
2-31
2-32
2-33
2-34
2-35
2-36
2-37
2-38
2-39
ACC2
OUTPUT SHIFTER BITS TAKEN WHEN CIC IS BYPASSED
INPUT
(SHFT=0)
HSP50214B
fIN = fS
F N = fS
HALFBAND FILTER 1
0
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
0
CONTROL WORD 7, BIT 15
1
FHB1 = fS OR fS /2
FN = FHB1
HALFBAND FILTER 2
0
CONTROL WORD 7, BIT 16
†
1
FHB2 = FHB1 OR FHB1/2
FN = FHB2
HALFBAND FILTER 3
0
CONTROL WORD 7, BIT 17
†
1
FHB3 = FHB2 OR FHB2/2
FN = FHB3
HALFBAND FILTER 4
0
CONTROL WORD 7, BIT 18
†
1
FHB4 = FHB3 OR FHB3/2
FN = FHB4
HALFBAND FILTER 5
CONTROL WORD 7, BIT 19
†
0
†
1
F5 = FHB4 OR FHB4/2
HALFBAND
FILTER OUTPUT
† Each halfband section decimates by 2.
FIGURE 17. BLOCK DIAGRAM OF HALFBAND FILTER
SECTION
NOTE: If 14 input bits are not needed, the gain adjust can be increased by one for each bit that the input is shifted down
at the input. For example, if only 12-bits are needed, an
offset range of 24dB is possible for a decimation of 24.
FIGURE 16. CIC FILTER BIT WEIGHTING
Since each halfband filter section decimates by 2, the total
decimation through the halfband filter is given by:
DEC HB = 2
N
(EQ. 9)
where N = Number of Halfband Filters Selected (1 - 5).
17
FN4450.4
May 1, 2007
HSP50214B
0
-6dB BANDWIDTH
MAGNITUDE (dB)
-20
-40
HALFBAND FILTER 5
HALFBAND FILTER 4
HALFBAND FILTER 3
HALFBAND FILTER 2
HALFBAND FILTER 1
-60
-80
-100
-120
0.125
0.25
0.375
0.5
NORMALIZED FREQUENCY (FN)
FIGURE 18. HALFBAND FILTER FREQUENCY RESPONSE
0
ALIAS
PROFILES
-6dB BANDWIDTH
MAGNITUDE (dB)
-20
-40
-60
HALFBAND FILTER 5
-80
HALFBAND FILTER 4
HALFBAND FILTER 3
HALFBAND FILTER 2
HALFBAND FILTER 1
-100
-120
0.125
0.25
0.375
0.5
NORMALIZED FREQUENCY (FN)
FIGURE 19. HALFBAND FILTER ALIAS CONSIDERATIONS
18
FN4450.4
May 1, 2007
HSP50214B
TABLE 4. HALFBAND FILTER COEFFICIENTS
COEFFICIENTS
HALFBAND #1
HALFBAND #2
HALFBAND #3
-0.00130558
HALFBAND #4
HALFBAND #5
C0
- 0.031303406
0.005929947
0.000378609
-0.000347137
C1
0.000000000
0.000000000
0.000000000
0.000000000
0.000000000
C2
0.281280518
-0.049036026
0.012379646
-0.003810883
0.00251317
C3
0.499954224
0.000000000
0.000000000
0.000000000
0.000000000
C4
0.281280518
0.29309082
0.019245148
-0.010158539
C5
0.000000000
0.499969482
0.000000000
0.000000000
0.000000000
C6
- 0.031303406
-0.06055069
0.29309082
0.299453735
-0.069904327
0.03055191
C7
0.000000000
0.499954224
0.000000000
0.000000000
C8
-0.049036026
0.299453735
0.304092407
-0.081981659
C9
0.000000000
0.000000000
0.500000000
0.000000000
C10
0.005929947
0.304092407
0.309417725
C11
-0.06055069
0.000000000
0.000000000
0.500000000
C12
0.012379646
-0.069904327
0.309417725
C13
0.000000000
0.000000000
0.000000000
0.019245148
-0.081981659
C15
-0.00130558
0.000000000
0.000000000
C16
-0.003810883
0.03055191
C17
0.000000000
0.000000000
C18
0.000378609
-0.010158539
C14
C19
0.000000000
C20
0.00251317
C21
0.000000000
C22
-0.000347137
NOTE: While Halfband filters are typically selected starting with the last stage in the filter chain to give the maximum alias free bandwidth,
a higher throughput rate may be obtained using other filter combinations. See Application Note 9720, “Calculating Maximum
Processing Rates of the PDC”.
Depending on the number of halfbands used, PROCCLK
must operate at a minimum rate above the input sample
rate, fS, to the halfband. This relationship depends on the
number of multiplies for each of the halfband filter stages.
The filter calculations take 3, 4, 5, 6, and 7 multiplies per
input for HB1, HB2, HB3, HB4, and HB5 respectively. If we
keep the assumption that fS is the input sampling frequency,
then Equation 10 shows the minimum ratio needed.
fPROCCLK/fS ≥ ([(7)(HB5)(2HB5)+
(6)(HB4)(2(HB4 + HB5))+
(5)(HB3)(2(HB3+HB4+HB5))+
(4)(HB2)(2(HB2+HB3+ HB4+HB5))+
(3)(HB1)(2(HB1+HB2+HB3+HB4+HB5))]/2T
Suppose we enable HB1, HB3, and HB5. Using Figure 16,
HB1= 1, HB3 = 1, and HB5 = 1. Since stage 2 and stage 4
are not used, HB2 and HB4 = 0. PROCCLK must operate
faster than (7x2+5x4+3x8)/8 = 7.25 times faster than fS .
If all five halfbands are used, then PROCCLK must operate at
(7x2+6x4+5x8+4x16+3x32)/32 = 7.4375 times faster than fS .
255-Tap Programmable FIR Filter
(EQ. 10)
where
HB1 = 1 if this section is selected and 0 if it is bypassed;
HB2 = 1 if this section is selected and 0 if it is bypassed;
HB3 = 1 if this section is selected and 0 if it is bypassed;
HB4 = 1 if this section is selected and 0 if it is bypassed;
HB5 = 1 if this section is selected and 0 if it is bypassed;
T = number of Halfband Filters Selected. The range for T is
from 0 to 5.
19
Examples of PROCCLK Rate Calculations
The Programmable FIR filter can be used to implement real
filters with even or odd symmetry, using up to 255 filter taps,
or complex filters with up to 64 taps. The FIR filter takes
advantage of symmetry in coefficients by summing data
samples that share a common coefficient, prior to
multiplication. In this manner, two filter taps are calculated
per multiply accumulate cycle. Asymmetric filters cannot
share common coefficients, so only one tap per multiply
accumulate cycle is calculated. The filter can be effectively
bypassed by setting the coefficient C0 = 1 and all other
coefficients, CN = 0.
FN4450.4
May 1, 2007
(EQ. 11A)
SYM) – [ ( SYM ) ( ODD# ) ]
for real filters, and
TAPS = floor [ (PROCCLK ⁄ ( F SAMP ⁄ R ) – R ) ⁄ 2]
(EQ. 11B)
EVEN SYMMETRIC
EVEN TAP FILTER
CN-1
C0
COEFFICIENT
NUMBER
EVEN SYMMETRIC
ODD TAP FILTER
C0
CN-1
COEFFICIENT
NUMBER
ASYMMETRIC
EVEN TAP FILTER
COEFFICIENT VALUE
CN-1
COEFFICIENT
NUMBER
COEFFICIENT VALUE
C0
C0
CN-1
COEFFICIENT
NUMBER
ODD SYMMETRIC
EVEN TAP FILTER
CN
C0
COEFFICIENT VALUE
TAPS = ( floor [ PROCCLK ⁄ ( F SAMP ⁄ R ) – R ] ) ( 1 +
COEFFICIENT VALUE
For real filter configurations, use Equations 11A and 11B to
calculate the number of taps available at a given input filter
sample rate.
COEFFICIENT VALUE
Additionally, the Programmable FIR filter provides for
decimation factors, R, from 1 to 16. The processing rate of
the Filter Compute Engine is PROCCLK. As a result, the
frequency of PROCCLK must exceed a minimum value to
ensure that a filter calculation is complete before the result is
required for output. In configurations which do not use
decimation, one input sample period is available for filter
calculation before an output is required. For configurations
which employ decimation, up to 16 input sample periods
may be available for filter calculation.
COEFFICIENT VALUE
HSP50214B
COEFFICIENT
NUMBER
ODD SYMMETRIC
ODD TAP FILTER
C0
CN-1
COEFFICIENT
NUMBER
ASYMMETRIC
ODD TAP FILTER
REAL FILTERS
Use Equation 12A to calculate the maximum input rate.
F SAMP = ( PROCCLK ) ( R ) ⁄ [ R + [ floor [ ( Taps ) +
CQ
IMAGINARY
COEFFICIENT
VALUE
for complex filters, where floor is defined as the integer
portion of a number; PROCCLK is the compute clock; fS AMP
= the FIR input sample rate; R = Decimation Factor; SYM =
1 for symmetrical filter, 0 for asymmetrical filter; ODD# = 1
for an odd number of filter taps, 0 = an even number of taps.
CQ(N-1)
CQ(0)
(EQ. 12A)
( SYM ) ( ODD# ) ] ⁄ ( 1 + SYM ) ] ]
RE AL C
for real filters, and
COEFFICIENT
NUMBER
CI(N-1)
CI(0)
OEFFIC
IENT VA
CI
L UE
COMPLEX FILTERS
F SAMP = [ ( PROCCLK ) ( R ) ] ⁄ [ R + floor [ ( Taps ) ( 2 ) ] ] (EQ. 12B)
for complex filters, where floor[x], PROCCLK, fS AMP, R =
Decimation Factor, SYM, and ODD# are defined as in
Equation 11A.
Definitions:
Even Symmetric: h(n) = h(N-n-1) for n = 0 to N-1
Odd Symmetric: h(n) = -h(N-n-1) for n = 0 to N-1
Asymmetric:
A filter with no coefficient symmetry.
Even Tap filter:
A filter where N is an even number.
Use Equation 13 to calculate the maximum output sample
rate for both real and complex filters.
Odd Tap filter:
A filter where N is an odd number.
Real Filter:
A filter implemented with real coefficients.
F FIR OUT = ( F SAMP ) ⁄ R
Complex Filters: A filter with quadrature coefficients.
(EQ. 13)
The coefficients are 22-bits and are loaded using writes to
Control Words 128 through 255 (see Microprocessor Write
Section). For real filters, the same coefficients are used by I
and Q paths. If the filter is configured as a symmetric filter
using Control Word 17, Bit 9, then coefficients are loaded
starting with the center coefficient in Control Word 128 and
proceeding to last coefficient in Control Word 128+n. The
filter symmetry type can be set to even or odd symmetric,
and the number of filter coefficients can be even or odd, as
illustrated in Figure 20. Note that complex filters can also be
realized but are only allowed to be asymmetric. Only the
coefficients that are used need to be loaded.
20
FIGURE 20. DEMONSTRATION OF DIFFERENT TYPES OF
DIGITAL FIR FILTERS CONFIGURED IN THE
PROGRAMMABLE DOWNCONVERTER
Automatic Gain Control (AGC)
The AGC Section provides gain to small signals, after the
large signals and out-of-band noise have been filtered out, to
ensure that small signals have sufficient bit resolution in the
Resampling/Interpolating Halfband filters and the Output
Formatter. The AGC can also be used to manually set the
gain. The AGC optimizes the bit resolution for a variety of
input amplitude signal levels. The AGC loop automatically
adds gain to bring small signals from the lower bits of the 26bit programmable FIR filter output into the 16-bit range of the
FN4450.4
May 1, 2007
HSP50214B
The AGC Multiplier/Shifter portion of the AGC is identified in
Figure 23. The gain control from the AGC loop filter is
sampled when new data enters the Multiplier/Shifter. The
limit detector detects overflow in the shifter or the multiplier
and saturates the output of I and Q data paths
independently. The shifter has a gain from 0 to 90.31dB in
6.021dB steps, where 90.31dB = 20log10(2N), when N = 15.
The mantissa provides an additional 6dB of gain in
0.0338dB steps where 6.0204dB = 20log10[1+(X)2-15],
where X = 215-1. Thus, the AGC multiplier/shifter transfer
function is expressed as:
N
AGC Mult/Shift Gain = 2 [ 1 + ( X )2
– 15
],
(EQ. 14)
where N, the shifter exponent, has a range of 0<N<15 and X,
the mantissa, has a range of 0<X<(215-1).
Equation 14 can be expressed in dB,
N
(AGC Mult/Shift Gain)dB = 20log 10 ( 2 [ 1 + ( X )2
5
GAIN - LINEAR AND dB
Figure 23 shows the Block Diagram for the AGC Section.
The FIR filter data output is routed to the Resampling and
Halfband filters after passing through the AGC multipliers
and Shift Registers. The outputs of the Interpolating
Halfband filters are routed to the Cartesian to Polar
coordinate converter. The magnitude output of the
coordinate converter is routed through the AGC error
detector, the AGC error scaler and into the AGC loop filter.
This filtered error term is used to drive the AGC multiplier
and shifters, completing the AGC control loop.
6
4
G (dB)
3
G (LINEAR)
2
1
0
16 32 48 64 80 96 112 128 144 160 176 192 208 224 240
0
AGC CONTROL MANTISSA VALUES (TIMES 256)
FIGURE 21. AGC MULTIPLIER LINEAR AND dB TRANSFER
FUNCTION
100
N = 15
N = 14
90
N = 13
80
N = 12
N = 11
70
N = 10
N=9
60
GAIN (dB)
output section. Without gain control, a signal at -72dBFS =
20log10(2-12) at the input would have only 4-bits of
resolution at the output (12-bits less than the full scale
16-bits). The potential increase in the bit resolution due to
processing gain of the filters can be lost without the use of
the AGC.
N=8
50
N=7
N=6
40
N=5
N=4
30
N=3
20
N=2
N=1
10
-15
])
N=0
(EQ. 14A)
0
0
The full AGC range of the Multiplier/Shifter is from 0 to
96.331dB (20log10[1+(215-1)2-15] + 20log10[215] = 96.331).
Figure 21 illustrates the transfer function of the AGC
multiplier versus mantissa control for N = 0. Figure 22
illustrates the complete AGC Multiplier/Shifter Transfer
function for all values of exponent and mantissa control.
The resolution of the mantissa was increased to 16-bits in
the A Version, to provide a theoretical AM modulation level
of -96dBc (depending on loop gain, settling mode and SNR).
This effectively eliminates AM spurious caused by the AGC
resolution.
For fixed gains, either set the upper and lower AGC limits to
the same value, or set the limits to minimum and maximum
gains and set the AGC loop gain to zero.
64
128
192
AGC CONTROL WORD (MANTISSA x 256)
FIGURE 22. AGC GAIN CONTROL TRANSFER FUNCTION
The Cartesian to Polar Coordinate converter accepts I and Q
data and generates magnitude and phase data. The
magnitude output is determined by the Equation 15:
2
(EQ. 15)
2
r = 1.64676 I + Q .
where the magnitude limits are determined by the maximum
I and Q signal levels into the Cartesian to Polar converter.
Taking fractional 2’s complement representation, magnitude
ranges from 0 to 2.329, where the maximum output is
2
r = 1.64676 ( 1.0 ) + ( 1.0 )
2
= 1.64676 × 1.414 = 2.329 .
The AGC loop feedback path consists of an error detector, error
scaling, and an AGC loop filter. The error detector subtracts the
magnitude output of the coordinate converter from the
programmable AGC THRESHOLD value. The bit weighting of
21
FN4450.4
May 1, 2007
HSP50214B
the AGC THRESHOLD value (Control Word 8, Bits 16-28) is
shown in Table 5. Note that the MSB is always zero. The range
of the AGC THRESHOLD value is 0 to +3.9995. The AGC
Error Detector output has the identical range.
TABLE 5. AGC THRESHOLD (CONTROL WORD 8) BIT
WEIGHTING
28
27
26
25
24
23
22
21
20
19
18
17
16
22
21
20. 2-1 2-2 2-3 2-4 2-5 2-6 2-7 2-8 2-9 2-10
The loop gain is set in the AGC Error Scaling circuitry, using
the two programmable mantissas and exponents. The
mantissa, M, is a 4-bit value which weights the loop filter
input from 0.0 to 0.9375. The exponent, E, defines a shift
factor that provides additional weighting from 20 to 2-15.
Together the mantissa and exponent define the loop gain as
given by,
AGC Loop Gain = M LG 2
–4
2
– ( 15 – E LG )
(EQ. 16)
where MLG is a 4-bit binary value ranging from 0 to 15, and
ELG is a 4-bit binary value ranging from 0 to 15. Table 7 and
8 detail the binary values and the resulting scaling effects of
the AGC scaling mantissa and exponent. The composite
(shifter and multiplier) AGC scaling Gain range is from
0.0000 to 2.329(0.9375)20 = 0.0000 to 2.18344. The scaled
gain error can range (depending on threshold) from 0 to
2.18344, which maps to a “gain change per sample” range of
0 to 3.275dB/sample.
The AGC Gain mantissa and exponent values are
programmed into Control Word 8, Bits 0-15. The PDC
provides for the storing of two values of AGC Scaling Gain
(both exponent and mantissa). This allows for quick
adjustment of the loop gain by simply asserting the external
control line AGCGNSEL. When AGCGNSEL = 0, then AGC
GAIN 0 is selected, and when AGCGNSEL = 1, AGC Loop
Gain 1 is selected. Possible applications include
acquisition/tracking, no burst present/burst present, strong
signal/weak signal, track/hold, or fast/slow AGC values.
The AGC loop filter consists of an accumulator with a built in
limiting function. The maximum and minimum AGC gain
limits are provided to keep the gain within a specified range
and are programed by 12-bit Control Words using Equation 17:
AGC Gain Limit = ( 1 + m AGC 2
–9
)2
e
(EQ. 17)
( AGC Gain Limit )dB = ( 6.02 ) ( eeee ) + 20 log ( 1.0 + 0.mmmmmmmm )
(EQ. 17A)
where m is an 8-bit mantissa value between 0 and 255, and e
is the 4-bit exponent ranging from 0 to 15. Control Word 9,
Bits 16-27 are used for programming the upper limit, while bits
0-11 are used to program the lower threshold. The ranges and
format for these limit values are shown in Tables 6A through
6C. The bit weightings for the AGC Loop Feedback elements
are detailed in Table 9A.
22
TABLE 6A. AGC LIMIT EXPONENT vs GAIN
GAIN(dB)
EXPONENT
MANTISSA
96.332
15
255
90.309
15
0
84.288
14
0
78.268
13
0
72.247
12
0
66.227
11
0
60.206
10
0
54.185
9
0
48.165
8
0
42.144
7
0
36.124
6
0
30.103
5
0
24.082
4
0
18.062
3
0
12.041
2
0
6.021
1
0
0.000
0
0
TABLE 6B. AGC LIMIT MANTISSA vs GAIN
GAIN(dB)
EXPONENT
MANTISSA
6.000
0
255
5.750
0
240
5.500
0
226
5.250
0
212
5.000
0
199
4.750
0
185
4.500
0
173
4.250
0
161
4.000
0
149
3.750
0
138
3.500
0
127
3.250
0
116
3.000
0
105
2.750
0
95
2.500
0
85
2.250
0
75
2.000
0
66
1.750
0
57
1.500
0
48
1.250
0
39
1.000
0
31
0.750
0
23
0.500
0
15
0.250
0
7
0.020
0
1
FN4450.4
May 1, 2007
HSP50214B
TABLE 6C. AGC LIMIT DATA FORMAT
CONTROL WORD 9 BIT:
27
26
25
24
23
22
21
20
19
18
17
16
FORMAT
e
e
e
e
m
m
m
m
m
m
m
m
AGC
ERROR
DETECTOR
AGC ERROR SCALING
μP
16
μP
(11 MANTISSA
4 EXPONENT)
M
U
X
SHIFT
MSB = 0
LIMITER
16
13
+
4
MSB = 0
LIMIT
DET
EN
16 MANTISSA =
01.XXXXXXXXXXXXXX
EXP †
MAN †
LOOP GAIN 1
AGC REGISTER 1
AGC REGISTER 0
EXP †
LOOP GAIN 0
20
MAN †
UPPER LIMIT †
LOWER LIMIT †
EXP=2NNNN
Δ
AGCGNSEL
AGC
LOAD
4
13
MANTISSA
4
EXP
REGISTER
SERIAL
OUT
REGISTER
(RANGE = -2.18344 TO 2.18344)
UNSIGNED †
THRESHOLD
STT.TTTTTTTTTT
(S = 0)
AGC LOOP FILTER
MAGNITUDE
(RANGE = 0 TO 2.3)
LIMIT
DET
LIMITER
26
IFIR
SHIFTER
(RANGE = 0 TO 1)
18
18
IAGC
26
18
LIMITER
QFIR
SHIFTER
LIMIT
DET
RESAMPLING
FIR FILTERS
AND
INTERPOLATING
HALFBAND
FILTERS
18
CARTESIAN
TO
POLAR
COORDINATE
CONVERTER
(G = 1.64676)
QAGC
AGC MULTIPLIER/SHIFTER
† Controlled via microprocessor interface.
FIGURE 23. AGC BLOCK DIAGRAM
Using AGC loop gain, the AGC range, and expected error
detector output, the gain adjustments per output sample for
the Loop Filter Section of the Digital AGC can be given by:
AGC Slew Rate = 1.5dB ( THRESH – ( MAG*1.64676 ) ) ×
( M LG ) ( 2
–4 ⎛
) 2
– ( 15 – E
⎝
23
)
LG ⎞
⎠
(EQ. 18)
The loop gain determines the growth rate of the sum in the
loop accumulator which, in turn, determines how quickly the
AGC gain traces the transfer function given in Figures 21
and 22. Since the log of the gain response is roughly linear,
the loop response can be approximated by multiplying the
maximum AGC gain error by the loop gain. The expected
FN4450.4
May 1, 2007
HSP50214B
range for the AGC rate is ~ 0.000106 to 3.275dB/output
sample time for a threshold of 1/2 scale. See the notes at the
bottom of Table 9A for calculation of the AGC response
times. The maximum AGC Response is given by:
AGC Response Max = Input(Cart/Polar Gain)(Error Det Gain) ( AGC
Loop Gain)(AGC Output Weighting)
(EQ. 19)
Since the AGC error is scaled to adjust the gain, the loop
settles asymptotically to its final value. The loop settles to
the mean of the signal.
TABLE 7. AGC LOOP GAIN BINARY MANTISSA TO GAIN
SCALE FACTOR MAPPING
BINARY
CODE
(MMMM)
SCALE
FACTOR
BINARY
CODE
(MMMM)
SCALE
FACTOR
0000
0.0000
1000
0.5000
0001
0.0625
1001
0.5625
0010
0.1250
1010
0.6250
0011
0.1875
1011
0.6875
0100
0.2500
1100
0.7500
0101
0.3125
1101
0.8125
0110
0.3750
1110
0.8750
0111
0.4375
1111
0.9375
TABLE 8. AGC LOOP GAIN BINARY EXPONENT TO GAIN
SCALE FACTOR MAPPING
BINARY
CODE
(EEEE)
SCALE
FACTOR
BINARY
CODE
(EEEE)
SCALE
FACTOR
0000
215
1000
27
0001
214
1001
26
0010
213
1010
25
0011
212
1011
24
0100
211
1100
23
0101
210-
1101
22
0110
29
1110
21
0111
28
1111
20
when AGCGNSEL = 0, and selecting Gain 1 when
AGCGNSEL = 1.
In the HSP50214, a reset event (caused by SYNCIN2 or
CW25) would clear the AGC loop filter accumulator. In the
HSP50214B, if Control Word 27, Bit 15 is set to zero, the
AGC loop filter accumulator will clear as in the original
HSP50214. If Control Word 27, Bit 15 is set to a one, the
backend reset (from CW25) will not clear the AGC loop filter
accumulator.
In the HSP50214, the settling mode of the AGC forces the
mean of the signal magnitude error to zero. The gain error is
scaled and used to adjust the gain up or down. This
proportional scaling mode causes the AGC to settle to the
final gain value asymptotically. This AGC settling mode is
preferred in many applications because the loop gain
adjustments get smaller and smaller as the loop settles,
reducing any AM distortion caused by the AGC.
With this AGC settling mode, the proportional gain error
causes the loop to settle more slowly if the threshold is
small. This is because the maximum value of the threshold
minus the magnitude is smaller. Also, the settling can be
asymmetric, where the loop may settle faster for “over
range” signals than for “under range” signals (or vice versa).
In some applications, such as burst signals or TDMA signals,
a very fast settling time and/or a more predictable settling
time is desired. The AGC may be turned off or slowed down
after an initial AGC settling period.
To minimize the settling time, a median AGC settling mode
has been added to the HSP50214B. This mode uses a fixed
gain adjustment with only the direction of the adjustment
controlled by the gain error. This makes the settling time
independent of the signal level.
For example, if the loop is set to adjust 0.5dB per output
sample, the loop gain can slew up or down by 16dB in 16
symbol times, assuming a 2 samples per symbol output
sample rate. This is called a median settling mode because
the loop settles to where there is an equal number of
magnitude samples above and below the threshold. The
disadvantage of this mode is that the loop will have a wander
(dither) equal to the programmed step size. For this reason,
it is advisable to set one loop gain for fast settling at the
beginning of the burst and the second loop gain for small
adjustments during tracking.
For example, if MLG = 0101 and ELG = 1100, the AGC Loop
Gain = 0.3125*2-7. The loop gain mantissas and exponents
are set in the AGC Loop Parameter Control Register (Control
Word 8, Bits 0-15).
The median settling mode is enabled by setting Control
Word 27, Bit 16 to a logic one. If Control Word 27, Bit 16 is
zero, the mean loop settling mode is selected and the loop
works identically to the HSP50214.
Two AGC loop gains are provided in the Programmable Down
Converter, for quick adjustment of the AGC loop. The AGC
Gain select is a control input to the device, selecting Gain 0
In the median mode, the loop works as follows:
24
The sign of the true gain error selects a fixed gain error of
0010000000000b or 1110000000000b.
FN4450.4
May 1, 2007
HSP50214B
These gain error values are scaled by the programmable
AGC loop gains to adjust the data path gain.
The maximum slew rate is ~1.5dB per output sample. See
Equation 18.
In order to fully evaluate the dynamic range of the PDC,
Table 9B is provided, which details the bit weighting from the
input to the AGC Multiplier.
Re-Sampler/Halfband Filter
The re-sampler is an NCO controlled polyphase filter that
allows the output sample rate to have a non-integer
relationship to the input sample rate. The filter engine can be
viewed conceptually as a fixed interpolation filter, followed by
an NCO controlled decimator.
The prototype polyphase filter has 192 taps designed at 32
times the input sample rate. Each of the 32 phases has 6
filter taps (6)(32) = 192. The stopband attenuation of the
prototype filter is greater than 60dB, as shown in Figure 24.
The signal to total image power ratio is approximately 55dB,
due to the aliasing of the interpolation images. The filter is
capable of decimation factors from 1 to 4. If the output is at
least 2x the baud rate, the 32 interpolation phases yield an
effective sample rate of 64x the baud rate or approximately
1.5%, (1/64), maximum timing error.
Following the Re-sampler are two interpolation halfband
filters. The halfband filters allow the user to up-sample by 2
or 4 to recover time resolution lost by decimating.
Interpolating by 2 or 4 gives 1/4 or 1/8 baud time resolution
(assuming 2x baud at the re-sampler output). The halfband
filters use the same coefficients as HB3 and HB5 from the
Halfband Filters Section. If one halfband is used, the 23-tap
filter is chosen. If two are used, the 23-tap filter runs first
followed by the 15-tap filter operating at twice the first
halfband’s rate. The 23-tap filter requires 7 multiplies, and
the 15-tap filter requires 5 multiplies to complete a filter
calculation.
Using the interpolation halfband filters allows for reduction in
the FIR filter sample rate. This optimizes the use of the
programmable FIR filter by allowing the FIR output sample
rate to be closer to the Nyquist rate of the desired bandwidth.
Optimizing the FIR filter performance provides better use of
the programmable FIR taps. Table 10 details the maximum
clocking rates for the possible resampling and interpolation
halfband filter configurations of this section of the PDC.
Control Word 16, Bits 2-0 identify the filter configuration.
Control Word 16, Bit 3 is used to bypass the polyphase resampler filter.
For proper data output from the interpolation filters, the data
ready signal must account for the interpolation process.
Figure 25 illustrates the insertion of additional data ready
pulses to provide sufficient pulses for the new output sample
rate. The Re-sampler Output Pulse Delay parameter is set in
Control Word 16, Bits 4-11. These bits set the delay between
the output samples when interpolation is utilized. Program
this distance between pulses using
(EQ. 20)
N = [ ( f PROCCLK /f OUT ) – 1 ]
A value of at least 5 is required to have sufficient time to
update the Output Buffer Register. (Writing 5 samples
requires 5 clock cycles) A value of at least 16 is required for
proper serial output from the part. (Conversion from 16-bit
parallel to serial). The value is programmed in numbers of
PROCCLK’s.
10
0
-20
-10
-40
-20
MAGNITUDE (dB)
MAGNITUDE (dB)
0
-60
-80
-100
-30
-40
-50
-60
-70
1
0.875
0.9375
0.75
0
0.8125
-80
0.625
10 11 12 13 14 15 16
0.6875
9
0.5625
8
0.5
7
0.375
6
0.4375
5
FREQUENCY (RELATIVE TO fS)
0.3125
4
0.25
3
0.125
2
0.1875
1
0.0625
-120
FREQUENCY (RELATIVE TO fS)
FIGURE 24A. POLYPHASE RESAMPLER FILTER BROADBAND
FREQUENCY RESPONSE
FIGURE 24B. POLYPHASE RESAMPLER FILTER PASS BAND
FREQUENCY RESPONSE
There is a 65dB limitation in SNR using the Re-Sampler Filter. When only the Interpolation FIRs are used, the full SNR range is passed.
FIGURE 24.
25
FN4450.4
May 1, 2007
HSP50214B
.
TABLE 9A. BIT WEIGHTING FOR AGC LOOP FEEDBACK PATH
AGC
ACCUM
BIT
POSITION
GAIN
ERROR
INPUT
GAIN
ERROR
BIT
WEIGHT
AGC LOOP
FILTER GAIN
(MANTISSA)
AGC LOOP
FILTER
GAIN
MULTIPLIER
(OUTPUT)
AGC
OUTPUT
AND AGC
LIMITS BIT
WEIGHT
AGC GAIN
RESOLUTION
(dB)
SHIFT
=0
SHIFT
=4
SHIFT
=8
SHIFT
= 15
31
2
2
2
2
0
30
2
2
2
2
E
3
48
29
2
2
2
2
E
2
24
28
2
2
2
2
E
1
12
27
12
=2
2
2
2
2
2
E
0
6
26
11
=1
1
2
2
2
1
M
-1
3
25
10
= 0.
0.
0.
2
2
2
0.
M
-2
1.5
24
9
=1
x
1
2
2
2
1
M
-3
0.75
23
8
=2
x
2
2
2
2
2
M
-4
0.375
22
7
=3
x
3
2
2
2
3
M
-5
0.1875
21
6
=4
x
4
2
2
2
4
M
-6
0.09375
20
5
=5
5
2
2
2
5
M
-7
0.04688
19
4
=6
6
2
2
1
6
X
-8
0.02344
18
3
=7
7
2
2
0.
7
-9
0.01172
17
2
=8
8
2
2
1
8
-10
0.00586
16
1
=9
9
2
2
2
9
-11
0.00293
15
0
= 10
10
2
1
3
10
-12
0.00146
14
11
2
0.
4
11
-13
0.000732
13
12
2
1
5
12
-14
0.000366
12
13
2
2
6
13
-15
0.000183
11
14
1
3
7
14
-16
0.0000916
10
0.
4
8
G
-17
0.0000458
9
1
5
9
G
-18
0.0000229
8
2
6
10
G
-19
0.0000114
7
3
7
11
G
-20
0.00000572
6
4
8
12
G
-21
0.00000286
5
5
9
13
G
4
6
10
14
G
3
7
11
G
G
2
8
12
G
G
1
9
13
G
G
0
10
14
G
G
AGC ResponseMax = Input(Cart/PolarGain)(Error Det Gain)(AGC Loop GainMax)(AGC Output Weighting).G = ground = 0.
15
AGC ResponseMax = ( 2 ) ( 1.64676 ) ⎛ ------⎞ ( 1 ) ( 1.5dB ) ∼ 3.275dB/output sample time .
⎝ 16⎠
AGC ResponseMin = ( 2 ) ( 1.64676 ) ( 2
– 15
) ( 1 ) ( 1.5dB ) ∼ 0.000106dB/output sample time .
Thus, the expected range for the AGC rate is ~ 0.000106 to 3.275dB/output sample time.
26
FN4450.4
May 1, 2007
HSP50214B
TABLE 9B. PDC BIT WEIGHTING
BIT
WEIGHT
INPUT
SIN/COS
0
0
1
2
CIC BIT
WEIGHTS
IIIIICCCCC
FIR
COEF
FIR
MULTI/
ACC
0
0
1
1
1
1
0
0
2
2
2
1
1
3
3
3
3
2
2
4
4
4
4
3
3
xxxxxxxxxx
5
5
5
5
4
4
xxxxxxxxxx
6
6
6
6
5
5
S
xxxxxxxxxx
7
7
7
7
6
6
S
xxxxxxxxxx
8
8
8
8
7
7
MIX OUT
CIC IN
SHIFT = 0
CIC IN
SHIFT = 15
0
0
S
S
xxxxxxxxxx
0
0
1
1
1
S
S
xxxxxxxxxx
1
2
2
2
S
S
xxxxxxxxxx
2
3
3
3
3
S
S
xxxxxxxxxx
4
4
4
4
S
S
xxxxxxxxxx
5
5
5
5
S
S
6
6
6
6
S
S
7
7
7
7
S
8
8
8
8
S
CIC OUT
HB DATA
HB DATA IN OUT/FIR IN
FIR
OUT
9
9
9
9
S
S
xxxxxxxxxx
9
9
9
9
8
8
10
10
10
10
S
10(S)
xxxxxxxxxx
10
10
10
10
9
9
11
11
11
11
S
11
xxxxxxxxxx
11
11
11
11
10
10
12
12
12
12
S
12
xxxxxxxxxx
12
12
12
12
11
11
13
13
13
13
S
13
xxxxxxxxxx
13
13
13
13
12
12
14
14
14
S
14
xxxxxxxxxx
14
14
14
14
13
13
15
15
SRnd
S
15
xxxxxxxxxx
15
15
15
15
14
14
16
16
S
16
xxxxxxxxxx
16
16
16
16
15
15
17
17
S
17
xxxxxxxxxx
17
17
17
17
16
16
18
S
18
xxxxxxxxxx
18
18
18
18
17
17
19
S
19
xxxxxxxxxx
19
19
19
19
18
18
20
S
20
xxxxxxxxxx
20
20
20
20
19
19
21
S
21
xxxxxxxxxx
21
21
21
21
20
20
22
S
22
xxxxxxxxxx
22
22
22
21
21
23
S
23
xxxxxxxxxx
23
23
23
22
22
24
S
24
xxxxx
Rnd
Rnd
Rnd
23
23
25
25(S)
25
xxxxx
SAT
SAT
SAT
24
24
26
26
26
xxxxx
25
Rnd
27
27
27
xxxxx
26
SAT
28
28
28
xxxxx
27
29
29
29
xxxxx
28
30
30
30
xxxxx
29
31
31
31
xxxxx
30
32
32
32
xx
31
33
33
33
xx
32
34
34
34
xx
(Rnd Out
of Mult.)
35
35
35
xx
36
36
36
x
37
37
37
x
38
38
38
x
39
39
39
x
NOTES:
1. SRnd = Symmetric Round; Rnd = Round; SAT = Saturation.
2. The NBW out of the CIC filter is 0.5 x fSOUT. If the NBWIN = fS / 4 and NBWOUT = fSOUT/2, then the processing gain for a decimation x 16 CIC
should be ~ 8 (9dB or 1.5 bits) versus A/D noise, the processing gain should be 10log(BWIN / BWOUT).
27
FN4450.4
May 1, 2007
HSP50214B
0
-2
-3
-4
-5
-6
-7
-8
-9
0.5
0.4375
0.375
0.3125
0.25
0.1875
0.125
0.0625
0
-10
FREQUENCY (RELATIVE TO fS)
FIGURE 24C. POLYPHASE RESAMPLER FILTER EXPANDED
RESOLUTION PASSBAND FREQUENCY
RESPONSE
TABLE 10. POLYPHASE AND INTERPOLATING HALFBAND
FILTER MAXIMUM CLOCKING RATES
MODE
RE-SAMPLER
INPUT
INTERCLOCK
RATE
POLATION
CYCLES
(MHz)
RATE
POLYPHASE
RESAMPLER
FILTER
OUTPUT
RATE
(MHz)
Bypass
0
55.00
-
55.00
Polyphase Filter
6
55/6 = 9.17
-
9.17
(Note 3)
Polyphase and
1 Halfband
Filter
13
55/13 = 4.23
2
8.46
(Note 3)
Polyphase and
2 Halfband
Filters
23
55/23 = 2.39
4
9.56
(Note 3)
1 Halfband
Filter
7
55/7 = 7.86
2
15.72
2 Halfband
Filters
17
55/17 = 3.24
4
12.94
RESAMPLER
NCO
PROCCLK
PROCCLK/N
3. This frequency is set by the Resampler NCO.
In burst systems (such as TDMA), time resolution is needed
for quickly identifying the optimum sample point. The timing
is adjusted by shifting the decimation in the DSP μP to the
closest sample. Use of timing error in this way may yield a
faster acquisition than a phase-locked loop coherent bit
synchronization. Finding the optimum sample point
minimizes intersymbol interference.
Fine time resolution is needed in CDMA systems to resolve
different multipath rays. In CDMA systems, the demands on
the programmable FIR can only be relieved by the
resampler/interpolation halfband filters. Assume the chip rate
for a baseband CDMA system is 1.2288MHz and PROCCLK
28
HALFBAND
FILTER #2
PULSE
DELAY
PULSE DELAY
COUNTER
THIS BLOCK GENERATES EXTRA
DATA READY PULSES FOR THE
NEW OUTPUTS FROM THE
INTERPOLATION PROCESS.
NV = INVALID MODE
NOTE:
HALFBAND
FILTER #1
HB2
MAGNITUDE (dB)
-1
MUX
1
is limited to 55MHz. Using the symmetric filter pre-sum
approach, PROCCLK limits the programmable FIR to
110MIPS (millions of instructions per second) effective due to
symmetry. If the CDMA filter (loaded into the programmable
FIR Section) requires an impulse response with a span of 12
chips, the filter at 2x the chip-rate would need 24 taps. The 24
taps would translate into 59MIPS = (1.2288MHz)(2)(24). To
get the same filtering at 8x the chip rate would require
944MIPS = (1.2288MHz)(8)(96). Direct 8x filtering can not be
accomplished with the programmable filter alone because
944MIPS are much greater that the 60MIPs effective limit set
by PROCCLK. It is necessary to decimate down to 2x the chip
rate to get a realistic number of filter taps. Both interpolation
halfband filters are then used to obtain the 8x CDMA output.
944MIPS is a lot of MIPS. The HSP50214B gets the
equivalent processing by decimating down and interpolating
backup.
0
0
0
0
1
1
1
1
HB1
RSMPLR
2
# EXTRA
PULSES
0
0
1
1
0
0
1
1
0 BYPASS
0
1
1
3 (NV)
3 (NV)
3
3
0
1
0
1
0
1
0
1
FIGURE 25. GENERATING DATA READY PULSES FOR
OUTPUT DATA
Timing NCO
The Timing NCO is very similar to the carrier NCO Phase
Accumulator Section. It provides the NCO driven sample
pulse and associated phase information to the resampling
filter process described in the Re-sampler Filter Section. The
Timing NCO does not include the SIN/COS Section found in
the Carrier NCO. The top level block diagram is shown in
Figure 26.
FN4450.4
May 1, 2007
HSP50214B
REG
TIMING NCO
PHASE OFFSET †
5
8
CARRY OUT = RUN
FILTER STROBE
+
PHASE
ACCUMULATOR
CLEAR
PHASE
ACC †
0
REG
NCO DIVIDE†
(NCO DIVIDE)/2 †
PROGRAMMABLE
DIVIDER
12
+
4
MUX
REFERENCE
DIVIDE †
EN
+
ENABLE SOF †
MUX
32
SOF
SOFSYNC
SOF
TIMING NCO
PH ACC
LOAD ON
UPDATE†
32
SCF
0
REG
REG
REG
SYNC
SHIFT REG
NUMBER OF SOF BITS
†
PROGRAMMABLE
DIVIDER
REFCLK
† Controlled via microprocessor interface.
FIGURE 27. TIMING ERROR GENERATION
TIMING FREQ
STROBE †
SYNC
TE(15:0)
Δ
†
REG
TIMING PHASE STROBE
TIMING
NCO
ACC.
REG
FILTER PHASE
SELECT
SYNC
SYNCIN2
CARRY
†
PHASE(31:28)
EN EXT TIMING NCO SYNC
TIMING NCO CENTER
FREQUENCY †
† Controlled via microprocessor interface.
Figure 27A illustrates an application where the Timing Error
Generator is used to lock the receiver samples with a
transmit data rate. In this example, the receive samples are
at four times the transmit data rate. An external loop filter is
required, whose frequency error output is fed into the Timing
NCO. This allows the loop to track out the long term drift
between the receive sample rate and the transmit data clock.
FIGURE 26. TIMING NCO BLOCK DIAGRAM
A timing error detector is provided for measuring the phase
difference between the timing NCO and a external clock input,
REFCLK. Timing Error is generated by comparing the values
of two programmable counters. One counter is clocked with
the Timing NCO carry out and the other is clocked by the
REFCLK. The 12-bit NCO Divide parameter is set in Control
Word 18, Bits 16-27. The NCO Divide parameter is the
preload to the counter that is clocked by the Timing NCO carry
out. The 12-bit Reference Divide parameter is set in Control
Word 18, Bits 0-11, and is the preload for the counter that is
clocked by the Reference clock. Figure 26 details the block
diagram of the timing error generation circuit. The 16-bits of
timing error are available both as a PDC serial output and as a
processor read parameter. See the Processor Read Section
for more details on accessing this value.
29
LOOP
FILTER
μP
CLKIN/RT
(NCO DIVIDE)/2 †
NCO DIVIDE = 4N†
PROGRAMMABLE
DIVIDER
+
4
REFERENCE
DIVIDE = N †
Tx DATA CLK
(REFCLK)
TE(15:0)
Δ
12
REG
CARRY
TIMING
NCO
ACC.
PHASE(31:28)
The programmable parameters for the Timing NCO include
an Enable External Timing NCO Sync (Control Word 11, Bit
5), the serial word width, Number of Offset Frequency Bits
(Control Word 11, Bits 3-4), an Enable Offset Frequency
control (Control Word 11, Bit 2), a Clear NCO Accumulator
control (Control Word 11, Bit 1), a Timing NCO Phase
Accumulator Load On Update control (Control Word 11, Bit
0), the Timing NCO Center Frequency (Control Word 12), a
Timing Phase Offset (Control Word 13, Bits 0-7), a Timing
Frequency Strobe (Control Word 14) and a Timing Phase
Strobe (Control Word 15). Refer to the Carrier Synthesizer
Mixer Section for a detailed discussion of the serial interface
for the Timing NCO offset frequency word.
EN
PROGRAMMABLE
DIVIDER
TO Tx BLOCK
(MODULATOR)
RT = TOTAL DECIMATION (CIC, HB FILTERS AND FIR)
† Controlled via microprocessor interface.
FIGURE 27A. TIMING ERROR APPLICATION
FN4450.4
May 1, 2007
HSP50214B
Cartesian to Polar Converter
The Cartesian to Polar converter computes the magnitude
and phase of the I/Q vector. The I and Q inputs are 18 bits.
The converter phase output is 18-bits (truncated) with the 16
MSB’s routed to the output formatter and all 18-bits routed to
the frequency discriminator. The 16-bit output phase can be
interpreted either as two’s complement (-0.5 to
approximately 0.5) or unsigned (0.0 to approximately 1.0),
as shown in Figure 28. The phase conversion gain is 1/2π.
The phase resolution is 16-bits. The 16-bit magnitude is
unsigned binary format with a range from 0 to 2.32. The
magnitude conversion gain is 1.64676. The magnitude
resolution is 16 bits. The MSB is always zero.
Table 11 details the phase and magnitude weighting for the
16-bits output from the PDC.
TABLE 11. MAG/PHASE BIT WEIGHTING
PHASE (o)
The magnitude and phase computation requires 17 clocks
for full precision. At the end of the 17 clocks, the magnitude
and phase are latched into a register to be held for the next
stage, either the output formatter or frequency discriminator.
If a new input sample arrives before the end of the 17 cycles,
the results of the computations up until that time, are
latched. This latching means that an increase in speed
causes only a decrease in resolution. Table 12 details the
exact resolution that can be obtained with a fixed number of
clock cycles up to the required 17. The input magnitude and
phase errors induced by normal SNR values will almost
always be worse than the Cartesian to Polar conversion.
TABLE 12. MAG/PHASE ACCURACY vs CLOCK CYCLES
CLOCKS
MAGNITUDE
ERROR
(% fS)
PHASE
ERROR
(DEG.)†
PHASE
ERROR
(% fS)
6
0.065
3.5
2
7
0.016
1.8
1
8
0.004
0.9
0.5
9
<0.004
0.45
0.25
10
<0.004
0.22
0.12
BIT
MAGNITUDE
15 (MSB)
22 (Always 0)
180
14
21
90
13
20
45
12
2-1
22.5
11
<0.004
0.11
0.062
11
2-2
11.25
12
<0.004
0.056
0.03
10
2-3
5.625
13
<0.004
0.028
0.016
9
2-4
2.8125
14
<0.004
0.014
0.008
8
2-5
1.40625
15
<0.004
0.007
0.004
7
2-6
0.703125
16
<0.004
0.0035
0.002
6
2-7
0.3515625
17
<0.004
0.00175
0.001
5
2-8
0.17578125
4
2-9
0.087890625
3
2-10
0.043945312
2
2-11
0.021972656
1
2-12
0.010986328
0 (LSB)
2-13
0.005483164
+π/2
4000
π/2
4000 3fff
3ff f
Q
7fff
±π
8000
Q
7fff
π
8000
I 0000
0
ffff
bfff
† Assumes ±180o = fS.
I
c000
bfff
3π/2
c000
-π/2
FIGURE 28. PHASE BIT MAPPING OF COORDINATE
CONVERTER OUTPUT
30
0000
0
ffff
In the HSP50214, the input to the coordinate converter I/Q to
|r|/θ) block is 18-bits. If the signal range is large and the AGC
is not used, the quantization noise can become a
contributing factor in the phase and frequency computations.
For example, if the signal range is 84dB and the maximum
signal is set at full scale, the minimum signal would have
only 4-bits each for I and Q.
In the HSP50214B, an additional data path option was
added that allows the output of the 255 tap programmable
FIR filter to be routed directly to the coordinate converter.
Rather than having to select only 18-bits out of the available
26 bit output, all 26-bits of the FIR output are routed to the
coordinate converter. This change eliminates quantization
effects to give more accuracy in the phase and frequency
discriminator outputs. The AGC settling time is not a factor
because the AGC is effectively bypassed for the magnitude,
phase, and frequency computations.
NOTE: The most significant 18-bits of the computed phase are
still routed to the discriminator.
FN4450.4
May 1, 2007
HSP50214B
One caveat to selecting the FIR outputs to be routed directly
to the coordinate converter is that because the I/Q samples
for the coordinate conversion are chosen from before the
resampler, the magnitude and phase samples will not align
with the I/Q samples, if the resampler or interpolation
halfband filters are used.
This optional signal routing mode was intended for FM or for
burst PSK where a fixed decimation can be used. It is also
applicable when resampling or timing adjustments on the
demodulated samples are done in a processor following
PDC.
The magnitude resolution may suffer because there is no
gain adjustment before computing the magnitude. If the
signal is < - 90dBFS, it will be below the LSB of the
magnitude output.
where D is the discriminator delay defined in Equation 21
(1 < D < 8), fSAMPOUT is the Discriminator FIR filter output
sample rate and CW is the desired center frequency. When
the phase multiplier is set to a value other than 20, the
discriminator range is reduced proportionally. The phase
multiplier can be 1, 2, 4 or 8 (20 to 23). Thus, a multiply of 21
reduces the range by 2, a multiply of 22 reduces the range
by 4, and a multiply of 23 reduces the range by 8.
The FIR filter can be configured with up to 63 symmetric taps
and up to 32 asymmetric taps. In the symmetric mode, the
FIR can be configured for even or odd symmetry, as well as
with an even or odd number of filter taps. Decimation is
provided to allow more processing time for longer (i.e., more
taps) filter structures.
PHASE INPUT
The enable signal for gating data into the coordinate
converter is either the AGC data ready signal or the
resampler data ready signal. If the resampler is bypassed,
the AGC data ready signal is used and there is a delay of 6
clock cycles between the FIR data being ready and the
coordinate converter block sampling it. If the resampler is
enabled, its data ready signal will be delayed by 6 clocks (for
the AGC) plus the compute delay of the resampler block.
This may cause the I/Q to |r|/θ output sample alignment to
shift with decimation. For this reason, it is recommended that
the resampler/halfband filter block be bypassed when using
this new data path.
To select the output of the 255 tap programmable FIR filter to
be routed to the coordinate converter, set Control Word 27,
Bit 13 to a logic one. For routing as in the HSP50214, set
Control Word 27, Bit 13 to a logic zero.
PHASE MULTIPLIER
†
DISCRIMINATOR DELAY
DISCRIMINATOR EN
†
†
DELAY
(1-8)
+
FIR COEFFICIENTS
DISC. FIR DECIMATION
FIR SYMMETRY TYPE
FIR SYMMETRY
FIR TAPS
†
†
†
†
†
+
63-TAP
FIR
FILTER
FREQ(15:0)
(2’s COMPLEMENT)
† Controlled via microprocessor interface.
FIGURE 29. FREQUENCY DISCRIMINATOR BLOCK DIAGRAM
Frequency Discriminator
The discriminator block delays phase from the Cartesian to
Polar Section and subtracts it from the latest sample. This
delay and subtract can be modeled as a programmable
delay comb filter. The output of the filter is dθ/dt, or
frequency. The transfer function of the discriminator is
set by
H(z)= 1 – Z
–D
(EQ. 21)
where D is the programmable discriminator delay expressed
in number of sample clock delays. The discriminator output
frequency is then filtered with a programmable FIR filter. The
Block Diagram of the Frequency Discriminator is shown in
Figure 29.
The range of delay in the discriminator is from 1 to 8
samples. Modulo 2π subtraction eliminates rollover problems
in the subtraction at 2π. The alias free discriminator
frequency range is given by:
Range FREQDISC = CW ± FSAMPOUT ⁄ ( D + 1 ) ;
31
(EQ. 22)
The HSP50214B offers an expanded choice of signals to be
filtered by the discriminator FIR. The choices are:
1) 18-bits of delayed, and subtracted (and optionally shifted)
phase. This is the Discriminator FIR filter input found in the
HSP50214.
2) 18-bits of magnitude from the coordinate converter block.
This was added to provide for post-detection filtering of AM
signals.
3) 18-bits from the I output of the resampler/interpolation
halfband filter block. This was added to provide for
processing of SSB signals.
The shift, delay, and subtract functions are bypassed for
items (2) and (3).
In addition to the FIR input selections, the Q input to the
coordinate converter block can be zeroed so that the
magnitude output is the magnitude of I only. Again this was
added to provide for processing SSB signals.
FN4450.4
May 1, 2007
HSP50214B
Control Word 27, Bit 14 is used to control the Q input to the
coordinate converter. The bit definitions is:
0
I and Q enabled to the I/Q to R/Theta block.
1
The Q input to the I/Q to R/Theta block is zeroed.
The enable signals associated with the various input
selections to the Discriminator FIR filter are:
1
The data ready strobe from the coordinate converter block.
2
The data ready strobe from the coordinate converter block.
The enable signals associated with the various input
selections to the coordinate converter are:
3a
The data ready signal to the coordinate converter
block when the resampler is bypassed. This is the
AGC output data ready signal.
3b
The data ready to the coordinate converter block
when the resampler/halfband filters are enabled.
This is the resampler halfband filter block output
data ready signal.
The discriminator input is 18-bits, and the output is rounded
asymmetrically to 16-bits. The phase into the discriminator
can be multiplied by 20, 21, 22, or 23 (modulo 2π) to remove
PSK data modulation. All programmable parameters for the
Frequency Discriminator are set in Control Word 17. Bits 15
and 16 are the phase multiplier which represents the shift
applied to the input phase. For CW, the multiply should equal
20, (00). For BPSK, QPSK, and 8PSK, the multiply should
equal 21, (01); 22, (10); or 23, (11); respectively. Bit 14 is
used to enable or disable the discriminator. Bits 11-13 set the
decimation in the programmable FIR filter. Bit 10 sets the
filter symmetry type as either odd or even, bit 9 sets whether
the filter is asymmetric or symmetric, and bits 3-8 set the
number of FIR filter taps. Bits 0-2 set the number of delays in
the frequency discriminator.
Output Section
The Output Section routes the 7 types of processed signals to
output pins in three basic modes. These basic modes are:
Parallel Direct Output, Serial Direct Output, and the Buffer
RAM Output. The Serial and Parallel Direct Output modes
were designed to output data strobes and “real time”
continuous streams of data. The Buffer RAM Output mode
outputs data upon receipt of an asynchronous request from an
external DSP processor or other baseband processing
engine. The use of the interrupt signal from the
Programmable Down Converter in conjunction with the
32
Parallel Direct Output Port Mode
The Parallel Direct Output Port Mode outputs two 16-bit words,
AOUT and BOUT, of “real time” data. Figure 30 details the
parallel output circuitry. Selection of the data source for the
AOUT and BOUT parallel outputs is done via Control Word 20,
Bits 22-23, and 20-21, respectively. The AOUT port can output
I, Magnitude, or Frequency data. The BOUT port can output Q,
Phase or Magnitude data. The upper bytes of AOUT and
BOUT are always in the parallel direct mode. The 16-bit parallel
direct mode is selected by setting Control Word 20, Bit 25, to
zero.
The DATARDY output is asserted during the first clock cycle
of the new data on the AOUT bus. The rate at which the data
out of the HSP50214 transitions and the rate at which
DATARDY is asserted can be different.
Data Transitions:
The transition rate of the parallel output data is dependent on
which of the three types of data is selected for the AOUT
Output channel: I (real symbols), |r| (magnitude), or f
(frequency). Q (quadrature symbols), ø (phase), or |r|
(magnitude) are available on the BOUT output. When selected
as an output, the I Q, |r|, and ø outputs transition at the symbol
rate. The f (frequency) output transitions at the discriminator
FIR filter output rate.
AOUT DIRECT PAR
OUTPUT MODE
DATA SOURCE †
DATARDY
16
I
(2’s COMPLEMENT)
16
MAG
(UNSIGNED BINARY)
FREQ
(2’s COMPLEMENT)
A(15:8)
16
AOUT(15:8)
A(7:0)
RAM(15:8)
MUX
1X Item (3) described above.
AOUT(7:0)
BOUT DIRECT PAR
OUTPUT MODE
DATA SOURCE †
Q
(2’s COMPLEMENT)
PHAS
(2’s COMPLEMENT)
MAG
(UNSIGNED BINARY)
16
B(15:8)
16
16
BOUT(15:8)
B(7:0)
RAM (7:0)
MUX
01 Item (2) described above.
MUX
00 Item (1) described above.
request strobes from the controller ensures that data is
transferred only when both the controller and the
Programmable Down Converter are ready. The Buffer RAM
output can be operated in a First In First Out (FIFO) or
SNAPSHOT mode with the data output either via the 8-bit
processor interface or a 16-bit processor interface.
MUX
The Discriminator FIR filter input selections are made in
Control Word 27, Bits 18 and 19. The bit definitions are:
BOUT(7:0)
RAM (15:0)
DATA SOURCE FOR LSB †
† Controlled via microprocessor interface.
FIGURE 30. PARALLEL OUTPUT BLOCK DIAGRAM
FN4450.4
May 1, 2007
HSP50214B
Data Ready Signal Assertion Rate:
Note that the BOUT data word may be at a different rate and
skewed in time with respect to DATARDY, depending on the
type of data selected for output. This is because of the timing
relationships defined above, and because the DATARDY is
driven by the AOUT signal. Figure 32 details such a
configuration.
The assertion rate of the DATARDY signal Is the data
transition rate of the AOUT output data either [I, |r| or f]. The
time alignment of parallel data words available for output are
as follows:
• I and Q are aligned in time,
When the f (frequency) word is selected for output on AOUT,
the DATARDY signal is asserted at the discriminator FIR
filter output rate, which will be a reduced rate when
decimation is engaged in the filter. The f (frequency)S output
is delayed from the associated I and Q samples one sample
time plus, the discriminator FIR filter impulse response time.
Figure 33 details the timing of this configuration for a FIR
filter that decimates by 4.
• |r| and ø are time aligned, but one sample clock delayed
from the associated I and Q samples.
DATARDY Is asserted time aligned with and at the same
rate as the data type selected for the AOUT output.
Figure 31 details the timing of the AOUT and DATARDY for
an AOUT = I data selection.
I0
AOUT
I1
I2
I3
DR2
DR3
I4
I5
I6
DR5
DR6
PROCCLK
DATARDY
DR0
DR1
DR4
NOTE: The number of PROCCLKS per output symbol is not representative, but shown to be small for clarity of establishing timing with respect to
the DATARDY signal. For each application, the relationship of the output symbol rate to PROCCLK must be properly illustrated to determine the
exact nature of the timing.
FIGURE 31. DATARDY WAVEFORMS WHEN I (READ DATA) IS SELECTED AS AOUT
BOUT
Q0
Q1
|r|0
AOUT
Q2
|r|1
Q3
Q4
|r|2
|r|3
Q5
Q6
|r|4
|r|5
|r|6
PROCCLK
DATARDY
DR1
DR0
DR2
DR3
DR4
DR5
DR6
FIGURE 32. DATARDY WAVEFORMS WHEN |r| (MAGNITUDE) IS SELECTED AS AOUT
1 + FIR Delay
BOUT
Q0
Q1
Q2
Q3
Q4
Qn
Qn+1
Qn+2
f0 (R = 4)
AOUT
PROCCLK
DATARDY
DR-1
DR0
NOTE: I and Q are sample aligned in time. |r| and φ are sample aligned in time, but one sample delayed from I or Q. The frequency sample
is delayed in time from I or Q by 1 sample time + 63 tap FIR impulse response. If the FIR is set to decimate and frequency is selected
for AOUT, the DATARDY signal will be at the discriminator FIR output (decimated) rate.
FIGURE 33. DATARDY WAVEFORMS WHEN f (FREQUENCY) IS SELECTED AS AOUT
33
FN4450.4
May 1, 2007
HSP50214B
Serial Direct Output Port Mode
The Serial Direct Output Port Mode offers the ability to
construct two serial output data streams, SEROUTA AND
SEROUTB, from 16-bit I, Q, magnitude, phase, frequency,
timing error, and AGC level data words. The total number of
data words (1 to 8) for serial output, and the sequential order of
these data word components of the serial output are
programmable. Each data word may be used once in either the
SEROUTA or SEROUTB data streams. Figure 34 illustrates the
conceptual implementation of the Serial Direct Output Port
Mode.
In the Serial Direct Mode, the output data is loaded into
Serial Shift Registers and routed to two serial output pins,
SEROUTA and SEROUTB. The serial output shift clock,
SERCLK, is PROCCLK divided by 1, 2, 4, 8, or 16. The
divide down ratio is programmed using Control Word 20,
Bits 14-16. The data is shifted out on the rising edge of the
internal SERCLK . The external clock polarity of SERCLK is
programmable via Control Word 20, Bit 18. A sync signal is
provided for detection of the start or end of each word in
the serial sequence. Control Word 20, Bit 17, sets the
SERSYNC signal location as either preceding the MSB
(typical for interfacing with microprocessors) or following
the LSB (typical for interfacing to D/A converters). Control
Word 20, Bit 19, sets the SERSYNC polarity as active low
or high. The LSB of each data word can be configured as
either the true LSB data, or set at a fixed logic “1” or “0” for
use as a tag bit. Control Word 20, Bits 0-13 set the LSB of
each of the 7 types of data words that can be configured in
the serial output stream. Control Word 19, Bits 21-24 set
the number of serial data words that will be linked to form
the serial outputs. Up to 7 data words can be linked to form
the serial output. SEROUTA and SEROUTB will have an
identical number of words in the serial output streams.
process for identifying the next word is to select a three bit
data type identifier which represents the data type to follow
the source data type. Program these bits into the Control
Word 19 field representing the “Link following X data”,
where X = the source data type, defines the second word in
the sequence. Likewise, the third data word is linked by
selecting the Control Word 19-bits that identify the “Link
following X data”, where X = the data type of the second
word in the serial chain. The process continues until all the
desired data words have been linked.
NOTE: I and Q are sample aligned in time. |r| and φ are sample
aligned in time, but one sample delayed from I or Q. The
frequency sample is delayed in time from I or Q by 1
sample time + 63 tap FIR impulse response. If the FIR is
set to decimate, the FIR output will be repeated every
sample time until a new value appears at the filter output.
(i.e., the frequency samples are clocked out at the I, Q
sample rate regardless of decimation.)
TABLE 13. LINKING CONTROL WORDS FOR SERIAL OUTPUT
DATA TYPE
IDENTIFIER
DATA TYPE
000
I Data
001
Q Data
010
Magnitude (MAG) Data
011
Phase (PHAS) Data
100
Frequency (FREQ) Data
101
Timing Error (TIMER) Data
110
AGC Gain
111
Zeros
Two examples will illustrate the process of configuring a serial
output using the Serial Output mode.
The 16-bit I, Q, magnitude, phase, frequency, timing error,
AGC level, and “zeros” data words are loaded into their
respective shift registers. The Magnitude and AGC Level
data word are unsigned binary format with a leading zero,
while the remaining signals are 2’s complement format.
The serial data stream looks like:
I data word >
SEROUTA source data = 000
Any of the eight data sources can be selected as the first
serial word for SEROUTA or SEROUTB. Control Word 19,
Bits 25-30 set the data type for the first serial word for
SEROUTA and SEROUTB. The three bit data type
identifier is shown both in Table 13 and in Figure 34, to the
right of the controls for the cross matrix switch. Serial
output data word sequences are formed by linking data
words by programming the data source for each shift
requester’s shift input signal. This programming links the
Shift Registers together in one or two serial chains. Thus,
the Control Word 19 term “Link follows X data”, where X is
one of the seven data types. Once the data source data
word is selected (by programming a three bit word
representing one of the data types into Control Word 19,
Bits 25-27 (SEROUTA), and 28-30 (SEROUTB)), the
Q data word >
Link following I data = 001
φ data word >
Link following Q data = 011
Zero data word >
Link following φ data = 111
34
SEROUTA:
CONTROL WORD 19 FIELD
start
end >
SEROUTB:
CONTROL WORD 19 FIELD
start
|r| data word >
SEROUTB source data = 010
f data word >
Link following |r| data = 100
TE data word>
Link following f data = 101
AGC data word >
Link following TE data = 110
FN4450.4
May 1, 2007
HSP50214B
SEROUTB:
CONTROL WORD 19 FIELD
end >
AGC DATA SERIAL OUTPUT TAG BIT †
TIMING ERROR DATA SERIAL OUTPUT TAG BIT †
FREQUENCY DATA SERIAL OUTPUT TAG BIT †
PHASE DATA SERIAL OUTPUT TAG BIT †
MAGNITUDE DATA SERIAL OUTPUT TAG BIT †
Q DATA SERIAL OUTPUT TAG BIT †
I DATA SERIAL OUTPUT TAG BIT †
REG
Q (15:0)
(2’s COMP)
REG
f (15:0)
AGC
(15:0)
(O; UNSIGNED BINARY)
ZERO
FOLLOWS Q
SHIFT REG
FOLLOWS |r|
SHIFT REG
CROSS
MATRIX
SWITCH
CROSS
MATRIX
SWITCH
SHIFT REG
SHIFT REG
FOLLOWS φ
SHIFT REG
FOLLOWS f
SHIFT REG
SHIFT REG
FOLLOWS TE
SHIFT REG
SHIFT REG
FOLLOWS AGC
REG
TE
(15:0)
(2’s COMP)
SHIFT REG
SHIFT REG
SHIFT REG
REG
(2’s COMP)
REG
φ (15:0)
(2’s COMP)
REG
|r| (15:0)
(O; UNSIGNED BINARY)
SOURCE
I
Q
MAG
PHASE
FREQUENCY
TIMING ERROR
AGC
ZERO
FOLLOWS I
SHIFT REG
REG
I (15:0)
(2’s COMP)
REG
DATA SOURCE FOR SEROUTA †
LINK FOLLOWING I DATA †
LINK FOLLOWING Q DATA †
LINK FOLLOWING MAG DATA †
LINK FOLLOWING PHASE DATA †
LINK FOLLOWING FREQ DATA †
LINK FOLLOWING TIMING DATA †
LINK FOLLOWING AGC DATA †
XXX
000
001
010
011
100
101
110
111
SHIFT REG
SHIFT REG
SEROUTA
SOURCE
MUX
6
5
4
3
2
1
0
SERIAL OUTPUT SHIFT REGISTER
DATA SOURCE FOR SEROUTB †
SEROUTA
SEROUTB
SOURCE
6
PROGRAMMABLE
DIVIDER
PROCCLK
NUM OF SER WORD LINKS IN A CHAIN †
SERIAL OUT CLOCK DIVIDER †
SERIAL OUTPUT SYNC POSITION †
SERIAL OUTPUT CLOCK POLARITY †
SERIAL OUTPUT SYNC POLARITY †
5
4
3
2
1
0
SERIAL OUTPUT SHIFT REGISTER
SEROUTB
‡
SERCLK
‡
SERSYNC
† Controlled via microprocessor interface
‡ Polarity is programmable
FIGURE 34. SERIAL OUTPUT FORMATTER BLOCK DIAGRAM
35
FN4450.4
May 1, 2007
HSP50214B
Serial Output Configuration Example 1:
It is desired to output the I data word, followed by the Q data
word, followed by the Phase data word on the SEROUTA
output. Similarly, it is desired to output the Magnitude data
word followed by the Frequency data word, followed by the
Timing Error data word, followed by the AGC Level data word
on the SEROUTB output. Table 14 illustrates how Control
Word 19 should be programmed.
TABLE 15. EXAMPLE 2 SERIAL OUTPUT CONTROL SETTINGS
CONTROL
WORD 19
BIT POSITION
TABLE 14. EXAMPLE 1 SERIAL OUTPUT CONTROL SETTINGS
CONTROL
WORD 19
BIT POSITION
FUNCTION
BIT
VALUE
FUNCTION
BIT
VALUE
RESULT
30-28
SEROUTA Data Source
000
(I)
27-25
SEROUTB Data Source
001
(Q)
24-21
Number of Serial Word
Links in a Chain
011
(3)
20-18
Link following I data
001
(Q)
17-15
Link following Q data
010
(|r|)
14-12
Link following |r| data
TBD
TBD
11-9
Link following φ data
XXX
(N/A)
8-6
Link following f data
XXX
(N/A)
5-3
Link following AGC data
XXX
(N/A)
2-0
Link following Timing
Error data
XXX
(N/A)
RESULT
30-28
SEROUTA Data Source
000
(I)
27-25
SEROUTB Data Source
010
(|r|)
24-21
Number of Serial Word
Links in a Chain
100
(4)
20-18
Link following I data
001
(Q)
17-15
Link following Q data
011
(φ)
start
14-12
Link following |r| data
100
(f)
I data word >
SEROUTA source data = 000
Q data word >
Link following I data = 001
|r|data word >
Link following Q data = 010
The serial data stream looks like:
11-9
Link following φ data
111
(Zeros)
8-6
Link following f data
101
(Timing)
5-3
Link following AGC data
XXX
(N/A)
2-0
Link following Timing
Error data
110
(AGC)
NOTE: Because all but the first data word in the serial output is
identified by the data type that it follows, SEROUTB can only
be fully independent of the sequence in SEROUTA if it does
not use any of the same data word types. This implies a
partition as described in Example 1. Once a data word that is
used in SEROUTA is called out in SEROUTB, the remaining
sequence in SEROUTB will be identical to that portion of
SEROUTA sequence that follows the duplicate data type.
This follows from using the “Link follows ‘data type’ data”
for word linkage.
NOTE: Each type of data word should be used only once in each
data stream. If the “Link following I data” is programmed
with the data type identifier for I, then the part will repeat the
I data word until all of the data word locations are filled. In
Example 1, if bits 20-18 were erroneously programmed to
000 (I data) then the SEROUTA would be four sequential
repeats of the I data word.
Serial Output Configuration Example 2:
It is desired to output only three data words on each serial
output. The I data word, followed by the Q data word, followed
by the Magnitude data word is to be output on SEROUTA.
The Q data word followed by the Magnitude data word,
followed by the one other data word to be output on
SEROUTB. The choices for the remaining data word in the
SEROUTB signal are: phase, frequency, AGC level and
timing error. Table 15 illustrates how Control Word 19 should
be programmed.
36
SEROUTA:
CONTROL WORD 19 FIELD
end >
SEROUTB:
start
Q data word >
|r|data word >
TBD data word>
end >
CONTROL WORD 19 FIELD
SEROUTB source data = 001
Link following Q data = 010
Link following | r | data = TBD
As shown by this example, once Q was linked to |r| in the
SEROUTA chain, the SEROUTB chain must have |r|
following Q, if Q is selected. Figure 35 illustrates the
construction of the serial output streams. If the serial data
stream was changed to be a length of four data words, then,
by default, the SEROUTA would be whatever is selected for
SEROUTB data word 3. SEROUTB would need to identify
the fourth data word. Thus, SEROUTA and SEROUTB are
not fully independent because they share the Q data
word (and by default, the MAGNITUDE follows Q data link
and whatever is selected for data word 3 to follow
MAGNITUDE data in SEROUTB).
The other signals provided with the SEROUTA and
SEROUTB are the SERSYNC and the SERCLK. The
SERSYNC signal can be programmed in either early or late
sync mode. The sync signal is pulsed active low or active
high for each information word link of the chain of data
created using Control Word 19. Figure 36 shows the four
possible configurations of SERSYNC as programmed using
Control Word 20.
As previously discussed, Control Word 20, Bits 17 and 19,
set the functionality of the LSB of each data word. These bits
may be programmed to be either a logic “0”, logic “1” or as
normal data. The fixed states are designed to allow the
microprocessor to synchronize to the serial data stream.
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 19, BITS 24-21 = 011
(3 DATA WORDS IN EACH SERIAL OUTPUT)
DATA WORD 3
DATA WORD 2
DATA WORD 1
MAGNITUDE
Q
I
DATA WORD 3
DATA WORD 2
DATA WORD 1
TBD
MAGNITUDE
Q
SEROUTA
SEROUTB
NOTE: Once magnitude is identified to follow Q,
it must be that way on both serial outputs.
THE REMAINING CHOICES FOR THE THIRD LINK ON SEROUTB ARE:
PHASE, FREQUENCY, AGC LEVEL, AND TIMING ERROR
FIGURE 35. EXAMPLE 2 SERIAL OUTPUT DATA STREAM
“NORMAL”
0
1
2
1
2
LATE
SERSYNC
MODE
SERSYNC FOLLOWS LSB
0
“INVERTED”
“NORMAL”
1
2
3
EARLY
SERSYNC
MODE
SERSYNC PRECEDES MSB
“INVERTED”
MSB WORD1
LSB WORD0
2
1
3
2
1
0
15
14
•••
DATA SHIFT MSB FIRST
MSB WORD2
2
1
0
15
14
•••
MSB WORD3
2
1
0
15
14
•••
2
LSB WORD2
LSB WORD1
FIGURE 36. VALID SERSYNC CONFIGURATION OPTIONS
The serial direct output can be programmed to output less
than 16-bits. New output data preempts old output data, so if
SERSYNC is programmed to precede the MSB, then data
will shift out until new data comes along. Note that if
SERSYNC is programmed to follow the LSB, then a sync will
never occur.
Buffer RAM Output Port
The Buffer RAM parallel output mode utilizes a RAM to store
output data for future retrieval by either the 8-bit
microprocessor that is configuring the PDC or by a 16-bit
baseband processing engine (which could also be a
microprocessor). Data is output from the RAM only on request
and can be obtained from either the 8-bit μP interface or from
a 16-bit interface that uses the two LSBytes of AOUT and
BOUT. The RAM holds up to eight 80-bit sample sets. Each
sample set includes 16-bits of each I, Q, magnitude, phase,
and frequency data. The RAM samples are mapped as shown
in Table 16. The Buffer RAM controller supports both FIFO
and Snapshot modes.
37
TABLE 16. RAM DATA STORAGE MAP
RAM
SAMPLE
SET
I
DATA
(000)
Q
DATA
(001)
|r|
DATA
(010)
Φ
DATA
(011)
F
DATA
(100)
0
I0(15:0)
Q0(15:0)
|r|0(15:0)
φ0(15:0)
f0(15:0)
1
I1(15:0)
Q1(15:0)
|r|1(15:0)
φ1(15:0)
f1(15:0)
2
I2(15:0)
Q2(15:0)
|r|2(15:0)
φ2(15:0)
f2(15:0)
3
I3(15:0)
Q3(15:0)
|r|3(15:0)
φ3(15:0)
f3(15:0)
4
I4(15:0)
Q4(15:0)
|r|4(15:0)
φ4(15:0)
f4(15:0)
5
I5(15:0)
Q5(15:0)
|r|5(15:0)
φ5(15:0)
f5(15:0)
6
I6(15:0)
Q6(15:0)
|r|6(15:0)
φ6(15:0)
f6(15:0)
7
I7(15:0)
Q7(15:0)
|r|7(15:0)
φ7(15:0)
f7(15:0)
NOTE: I and Q are sample aligned in time. |r| and φ are sample
aligned in time, but one sample delayed from I or Q. The
frequency sample is delayed in time from I or Q by 1
sample time + 63 tap FIR impulse response. If the FIR is
set to decimate, the FIR output will be repeated every
sample time until a new value appears at the filter output.
(i.e., the frequency samples are clocked out at the I, Q
sample rate regardless of decimation.)
FN4450.4
May 1, 2007
The INTRRP output signal goes low for 8 PROCCLK cycles
when the number of samples in the Buffer RAM (depth)
reaches the programmed depth. The depth of the RAM is
calculated using Equation 23. A DSP microprocessor or the
data processing engine can use the INTRRP signal to know
that the RAM is ready to be read.
D RAM = [ ( ADDR WRITE – ADDR READ ) – 1 ] MOD8
(EQ. 23)
FIFO Operation Via 16-Bit μProcessor
Interface
Figure 37 shows the conceptual configuration of the 16-bit
μProcessor interface. This interface looks like a 16-bit
μProcessor read-only microprocessor interface. The
SEL(2:0) lines are the address bus and the OEAL and OEBL
lines are the read lines. The address is decoded as shown in
Table 17.
Use of the 16-bit interface for Buffer RAM output requires
Control Word 20, Bit 25, to be set to a logic “0” and Control
Word 20, Bit 24, to be set to a logic “1”. Once the Control
Word 20 has been set to route data to AOUT(7:0) and
BOUT(7:0), then the microprocessor must place a value on
the PDC input pins SEL(2:0), to choose which data type will
be output on AOUT(7:0) and 6BOUT(7:0). Table 17 defines
the data types in terms of SEL(2:0). With the control lines
set, the selected data is read MSByte on AOUT(7:0) and
LSByte on BOUT(7:0) when OEAL and OEBL (are low).
New data only read when OEBL goes low, so use μP for 8bit modes. Programming SEL(2:0) = 110 outputs a 16-bit
status signal on AOUT and BOUT. The FIFO status includes
FULL, EMPTY, FIFO Depth, and READYB. These status
signals are defined in Table 18.
38
WRITE
SEQUENCER
NEW
DATA
DUAL
PORT
RAM
I
Q
|r|
0
1
2
3
φ
ƒ
4
STATUS 6
MUX
φ
ƒ
16
16
16
16
16
DATA OUTPUT
I
Q
|r|
MUX
The FIFO mode allows the processor to service the interface
only when enough samples are present in the RAM. This
mode is provided so that the μProcessor does not have to
service the PDC every output sample. An interrupt,
INTRRPT, is asserted when the desired number of samples
are available. The PDC can be programmed to assert the
interrupt when up to 7 samples are available. Control Word
21, Bit 15 is used to set the Buffer RAM controller to the
FIFO mode, while Control Word 21, Bits 12-14 set the
number of RAM samples to be stored (0 to 7) before the
interrupt (INTRRPT) is asserted. Control Word 20, Bit 24
determines whether the RAM output interface is the 8-bit
microprocessor interface or the 16-bit processor interface. In
the 16-bit interface the MSByte is sent to AOUT(7:0) while
the LSByte is sent to BOUT(7:0).
DATA INPUT
HSP50214B
OUTPUT
DATA
OEBL
“SET OF WORDS”
ADDRESS
SEQUENCER
INCR
INCR
RD
WR
SEL(2:0)
PROCCLK
FIGURE 37. 16-BIT MICROPROCESSOR INTERFACE BUFFER
RAM MODE BLOCK DIAGRAM
TABLE 17. BUFFER RAM OUTPUT SELECT DEFINITIONS
SEL(2:0)
OUTPUT DATA TYPE
000
I Data
001
Q Data
010
Magnitude
011
Phase
100
Frequency
101
Unused
110
Memory Status
111
Reading this address increments to the next
sample set
TABLE 18. STATUS BIT DEFINITIONS
AOUT BIT
LOCATION
INFORMATION
(7:5)
FIFO depth - When in FIFO mode, these bits are
the current depth of the FIFO.
4
EMPTY - When in FIFO mode, the FIFO is
empty, and the read pointer cannot be
advanced. Active High.
3
FULL - When in FIFO mode, the FIFO is full, and
new samples will not be written.
Active High.
2
READYB - When in FIFO mode, the output
buffer has reached the programmed threshold.
In the snapshot mode, the programmed number
of samples have been taken. Active Low.
1-0
GND
NOTE: In the Status output, BOUT(7:0) are all GND.
FN4450.4
May 1, 2007
HSP50214B
Figure 37 shows the interface between a 16-bit
microprocessor (or other baseband processing engine) and
the Buffer RAM Output Section of the Programmable Down
Converter, configured for data output via the parallel outputs
AOUT and BOUT. In the 16-bit microprocessor interface
configuration, the Buffer RAM pointer is incremented when
the μProcessor reads address SEL(2:0) = 7 and OEBL = 0.
After reset, the FIFO must be incremented to read the first
sample set. This is because the RAM read and write pointers
cannot point to the same address. Thus, the FIFO pointer
must move to the next address before reading the next set of
data (I, Q, |r|, φ, and f) samples. 4 PROCCLK cycles are
required after an increment before reading can resume. The
FIFO write pointer is reset to zero (the first data sample) when
Control Word 22 is written to via the 8-bit microprocessor
interface. See the Microprocessor Read Section for more
detail on how to obtain the Buffer RAM output with this
technique. Figure 39 shows the timing diagram required for
parallel output operations. In this diagram, only the I, Q and
Frequency data are taken from each sample before
incrementing to the next sample. Figure 39 assumes that the
pointer has already been incremented into a sample.
NOTE: For the very first sample read, the pointer must be
incremented first and 4 PROCCLKs must pass before
this sample can be read.
Figure 38 shows INTRRP going low before the FIFO is read.
The FIFO can be read before the number of samples
reaches the INTRRP pointer. The number of samples in the
FIFO must be monitored by the user via a status read.
INT
PDC
OEAL
HSP50214B
INTRRP
AOUT(7:0)
D(15:8)
OEBL
16-BIT
μP
BOUT(7:0)
SEL(2:0)
RD
D(7:0)
INTRRP
8 CLKS
> 4 CLKS
OEAL,
OEBL
SEL(0:2)
AOUT(7:0),
BOUT(7:0)
0
1
4
I
Q
FR
1 2 3 4 5 6 7 8
7
0
1
I
Q
1 2 3 4
PROCCLK
FIGURE 39. TIMING DIAGRAM FOR PDC IN FIFO MODE WITH
OUTPUTS I, Q, AND FREQUENCY SENT TO
AOUT(7:0) AND BOUT(7:0)
Suppose the depth of the Buffer RAM Output Section is
programmed for an INTRRP pointer depth of 4. If the output
is at 4 times the baud rate, the processing routine for the
microprocessor may only need to read the buffer when the
Buffer RAM had 4 samples since processing is usually on a
baud by baud basis.
Figure 39 illustrates the conceptual view of the FIFO as a
circular buffer, with the Write address one step ahead of the
Read Address.
Figure 40A deals with clockwise read and write address
incrementing. The FIFO depth is the difference between the
Write and Read pointers, modulo 8. Figure 40B illustrates a
FIFO status of Full, while Figure 40C illustrates a FIFO
empty status condition. Figure 40D illustrates a programmed
FIFO depth of 3 and the INTRRP signal indicating that the
buffer has sufficient data to be read.
Following some simple rules for operating the FIFO will
eliminate most operational errors:
Rule #1: The Read and Write Pointers cannot point at the
same address (the circuitry will not allow this).
Rule #2: The FIFO is full when the Write Address = Read
Address -1 (no more data will be written until some samples
are read or the FIFO is reset).
A(2:0)
FIGURE 38. INTERFACE BETWEEN A 16-BIT
MICROPROCESSOR AND PDC IN FIFO BUFFER
Rule #3: The FIFO is empty when the Read Address =
(Write Address -1) (the circuitry will not allow the read
pointer to be incremented).
Rule #4: You cannot write over what you have not read.
Rule #5: RESET places the Write address pointer = 000 and
Read address pointer = 111.
Rule #6: The best addressing scheme is to read the FIFO
until it is empty. This avoids erroneous INTRRP assertions
and provides for simple FIFO depth monitoring. The interrupt
is generated when the depth increments past the threshold.
39
FN4450.4
May 1, 2007
HSP50214B
FIFO Operation Via 8-Bit μProcessor
Interface
6
The Buffer RAM Output may also be accessed via the 8-bit
microprocessor interface C(7:0). Figure 41 shows the
conceptual configuration of the 8-bit μprocessor interface.
Control Word 20, Bit 24 must be set to 0 in order to obtain
Buffer RAM data to this output. The Microprocessor Read
Section describes how to read the data from each sample
out of the C(7:0) interface.
7
5
0
4
1
3
FIFO
DEPTH
2
WRITE
FIGURE 40A. FIFO DEPTH IS (WRITE - READ)
WRITE
6
7
READ
5
0
4
1
3
Recall that INTRRP stays low for 8 PROCCLK cycles. The
FIFO can be read before the INTRRP signal goes low; the
number of samples in the FIFO must be monitored by the user.
The timing relationship of the INTRRP to the snapshot data is
shown in Figure 42.
The read pointer of the FIFO is incremented when Control
Word 23 is written to. The data cannot be read from the
next sample until 4 PROCCLKs after the Buffer RAM
pointer has been incremented. Control Word 22 is used to
reset the Read and Write pointers of the Buffer RAM output
to the first sample to 000 and 007 for write and read
respectively.
2
FIGURE 40B. FIFO FULL IS WHEN (WRITE - READ) = 7
READ
6
7
5
WRITE
0
4
1
3
2
FIGURE 40C. FIFO EMPTY IS WHEN (WRITE - READ) = 1
READ
6
7
5
0
4
WRITE
1
3
2
READY
FIGURE 40D. FIFO READY IS WHEN (WRITE - READ) > DEPTH
FIGURE 40.
40
FN4450.4
May 1, 2007
3
R2 R1 R0 A2 A1 A0 SELECTION
0
4
MUX
φ
ƒ
0
1
2
MUX
I
Q
|r|
LSByte
0
R1 R0 A1
“SET OF WORDS”
ADDRESS
STATUS
1
SEQUENCER
R2
1
INCR
INCR
RD
WR
MSByte
A0
0
MUX
MUX
WRITE
SEQUENCER
DUAL
PORT
RAM
DATA OUTPUT
φ
ƒ
16
16
16
16
16
MUX
I
Q
|r|
DATA INPUT
HSP50214B
OUTPUT
DATA
1
NEW
DATA
A(2:0)
WRITE
ADDRESS “5”
INT(15:0)
INT(22:16)
R2, R1, R0
AGC
A2, A1, A0
0: I;Q (2’s COMP)
TIMING
1: |r|; φ (O; UNSIGNED BINARY; 2’s COMP)
2: ƒ (2’s COMPLEMENT)
4: INPUT AGC (O; UNSIGNED BINARY)
5: AGC; TIMING (O; UNSIGNED BINARY;
2’s COMP)
0
1
2
MUX
CONTROL
WORD 23
A2 A1 A0
3
R0 A1
0
0
0
0
0
0 RAM I LSB
0
0
0
0
0
1 RAM I MSB
0
0
0
0
1
0 RAM Q LSB
0
0
0
0
1
1 RAM Q MSB
0
0
1
0
0
0 RAM |r| LSB
0
0
1
0
0
1 RAM |r| MSB
0
0
1
0
1
0 RAM φ LSB
0
0
1
0
1
1 RAM φ MSB
0
1
0
0
0
0 RAM ƒ LSB
0
1
0
0
0
1 RAM ƒ MSB
0
1
1
X
X
X NOT USED
1
0
0
0
0
0 INPUT INTEG LSB
1
0
0
0
0
1 INPUT INTEG NMSB
1
0
0
0
1
0 INPUT INTEG MSB
1
0
1
0
0
0 AGC LSB
1
0
1
0
0
1 AGC MSB
1
0
1
0
1
0 TIMING LSB
1
0
1
0
1
1 TIMING MSB
1
1
X
X
X
X NOT USED
X
X
X
1
1
1 STATUS
RD
FIGURE 41. 8-BIT MICROPROCESSOR INTERFACE BUFFER RAM MODE BLOCK DIAGRAM
PROCCLK
I/Q
DATARDY
(I/Q SELECTED)
R/φ
DELAY TO DATARDY DEPENDS ON LENGTH OF FIR IF FREQ CHOSEN
DATARDY
(R/φ SELECTED)
INTRRP
WRITES TO
SNAPSHOT
RAM
I
Q
R
φ
ƒ
FIGURE 42. RAM LOAD SEQUENCE
Snap Shot Operation
The snapshot mode takes sets of adjacent samples at
programmed intervals. It is provided for tracking algorithms
that do not require processing of every sample, but do
require sets of adjacent samples. For example, bit sync
algorithms have narrow loop bandwidths that may not need
to be updated every sample. Computing the bit phase may
require 4 adjacent samples at 2 times the baud rate. The
snapshot mode allows the processor to implement the
tracking algorithms for high speed data without having to
handle every data sample.
The interval from the start of one snapshot to the start of a
second snapshot is programmed into bits 11-4 (where bit 11
is the MSB) of Control Word 21. The actual interval is the
41
value programmed plus 1. If bits 11-4 = 11111111, then the
interval is set to 256. If sample sets are to be taken every 4
samples, then bits 11-4 = 00000011.
Figure 43 shows the relationship between the snapshot
samples and the snapshot interval.
ADJACENT
SAMPLES
0
1
2
3
4
62
63
64
65
# SAMPLES = 4
INTERVAL = 64
FIGURE 43. SNAP SHOT SAMPLING
FN4450.4
May 1, 2007
HSP50214B
The PDC begins to fill the buffer each time an interval
number of samples have passed. The number of sample
sets the PDC writes into the buffer and is programmed into
bits 3-0 of Control Word 21. The number of samples stored
is the programmed value and may be from 1 to 8 sample
sets. A sample set consists of I, Q, | r | , φ and ƒ.
In snap shot operations, the buffer is read the same as for
FIFO operations. Figures 37 and 39 describe the Design
Blocks and Timing required to output data on AOUT(7:0) and
BOUT(7:0). Table 17 summarizes the selectable output
signals. The method for reading data through the
Microprocessor Section in snap shot mode is identical to the
method described in the FIFO mode subsection and the
Microprocessor Read Section.
Avoiding Timing Pitfalls When Using the Buffer
RAM Output Port
In snapshot mode, the whole buffer is written whenever the
interval counter has timed-out. After time-out, old data can
be written over. Thus, the data contained within the buffer
must be retrieved before time-out to avoid data loss.
It may be desirable to disable the INTRRPT into the
controlling microprocessor during read cycles to avoid the
generating extra interrupts. Figure 44 details how the WRITE
address can trigger extra interrupts. Care must be taken to
either read sufficient data out of memory or RESET the
addressing to ensure that a complete set of data is the cause
of the interrupt.
INTRRP
INTRRP
INTRRP
Microprocessor Write Section
The Microprocessor Write Section uses an indirect addressing
scheme where a 32-bit data word is first loaded in a four 8-bit
byte master registers using four writes via C(7:0). The desired
destination register address is then written to another address
using C(7:0). Writing this address triggers a circuit that
generates a pulse, synchronous to clock, that loads the
Destination Register. The sync circuits and data words are
synchronized to different clocks, CLKIN or PROCCLK,
depending on the Destination Registers.
A(2:0) determines the destination for the data on bus, C(7:0).
Table 19 shows the address map for microprocessor
interface. Figure 45 shows the Control Register loading
sequence. The data in C(7:0) and address map in A(2:0) is
loaded into the PDC on the rising edge of WR and is latched
into the Master Register on the rising edge of WR and A(2:0)
= 100. Four clocks must pass before loading the next Control
Word to guarantee that the data has been transferred.
Some registers can be loaded (i.e., transferred from the
Master Register to a Configuration Register or from a Holding
Register to an active register) by initiating a sync. For
example, to load the AGC Gain, the value of the AGC gain is
first loaded into the Holding Registers, then a transfer is
initiated by SYNCIN2 if Control Word 8, Bit 29 = 1. This allows
the AGC gain to be loaded by detecting a system event, such
as a start of a new burst. Bit 20 of Control Word 0 has the
same effect on the Carrier NCO center frequency for assertion
of SYNCIN1, except it transfers from a dedicated holding
register - not the Master Register.
WRITE
ADDRESS
TABLE 19. DEFINITION OF ADDRESS MAP
A2-0
REGISTER DESCRIPTION
0
Holding Register 0. Transfers to bits 7-0 of the 32-bit
Destination Register. Bit 0 is the LSB of the 32-bit register.
1
Holding Register 1. Transfers to bits 15-8 of a 32-bit
Destination Register.
2
Holding Register 2. Transfers to bits 23-16 of a 32-bit
Destination Register.
3
Holding Register 3. Transfers to bits 31-24 of a 32-bit
Destination Register. Bit 31 is the MSB of the 32-bit
register.
4
This is the Destination Address Register. On the fourth CLK
following a write to this register, the contents of the Holding
Registers are transferred to the Destination Register. All 8
bits written to this register are decoded into the Destination
Register Address. The configuration destination address
map is given in the tables in the Control Word Section.
5
Selects data source for reading. See Microprocessor Read
Section.
WR
RD
REEST
TIME
WR
WR
RD
A COMPLETE SET OF 3 DATA SAMPLES IS IN MEMORY AT INTRRP
A: NORMAL READ/WRITE SEQUENCE
INTRRP
INTRRP
WRITE
ADDRESS
INTRRP
WR
WR
RD
RD
TIME
WR
RD
THE THIRD INTERRUPT HAS ONLY 1 NEW DATA ENTRY
(INSTEAD OF 3) AT INTRRP
B: FALSE TRIGGERED INTERRUPT READ/WRITE SEQUENCE
Suppose a (0018D038)H needs to be loaded into Control Word
0, then Table 20 details the steps to be taken.
FIGURE 44. AVOIDING FALSE INTRRP ASSERTIONS
42
FN4450.4
May 1, 2007
HSP50214B
TABLE 20. EXAMPLE PROCESSOR WRITE SEQUENCE
PROCLK
STEP
A(2:0)
1
000
C(7:0)
COMMENT
0011 1000 Loads 38 into Master Register
(7:0) on rising edge of WR.
2
001
1101 0000 Loads D0 into Master Register
(15:8) on rising edge of WR.
3
010
0001 1000 Loads 18 into Master Register
(23:16) on rising edge of WR.
4
011
0000 0000 Loads 00 into Master Register
(31:24) on rising edge of WR.
5
100
0000 0000 Load
“0018D038”
into
Configuration Control Register 0.
6
WR
RD
A2-0
5
READ ADDRESS
C7-0
OUTPUT DATA C(7:0)
READ CODE C(2:0)
LOAD ADDRESS
OF TARGET
CONTROL REGISTER
THREE-STATE
INPUT BUS
ASSERT RD
TO ENABLE DATA
OUTPUT ON C0-7
FIGURE 46. READING THE CONTROL REGISTERS USING A
LATCH CODE EQUAL TO A 5, A READ ADDRESS
AND A READ CODE
Wait 4 CLKS.
TABLE 21. PROCESSOR READ SEQUENCE (INPUT LEVEL
SELECTOR)
1
CLK =
(PROCCLK,
CLKIN)
2
3
4
WR
A2-0
0
C7-0
LSB
1
2
3
4
0
LOAD ADDRESS OF
TARGET CONTROL
REGISTER AND
WAIT 4 CLKs
A(2:0)
C(7:0)
COMMENT
1
101
100
Write Read Code, 100 to
Address 5, WR pulled high to
generate rising edge.
2
000
1111 1000
(F4)H
3
001
0001 1010 Pull RD high, then drop low,
(1A)H
Read AGC NLSB.
4
010
0011 0010 Pull RD high, then drop low,
(32)H
Read AGC MSB.
2
MSB ADD
LOAD
CONFIGURATION
DATA
STEP
LOAD NEXT
CONFIGURATION
REGISTER
FIGURE 45. LOADING THE CONTROL REGISTERS WITH
32-BIT CONTROL WORDS
Microprocessor Read Section
The microprocessor read uses both read and write
procedures to obtain data from the PDC. A write must be
done to location 5 to select the source of data to be read.
The read source is determined by the value placed on the
lower three bits of C(7:0). The output from a particular read
code is selected using a read address placed on A(2:0). The
output is sent to C(7:0) on the falling edge of RD.
If the Read Address is equal to 111, the Read Code is
ignored, and the status bits shown in Table 22 in the Output
Section is sent to C(7:0). This state was provided so that the
user could obtain the status bits quickly.
Refer to the Timing Diagram in Figure 46. Suppose the input
level detector has a hex value of (321AF5)H, then Table 21
details the steps to be taken.
43
Drop RD low, Read AGC LSB.
TABLE 22. DEFINITION OF ADDRESS MAP
READ
CODE C(2:0)
STATUS
TYPE
READ ADDRESS A(2:0)
000
Buffer RAM 000- I LSB.
I and Q
001- I MSB.
010- Q LSB.
011- Q MSB.
See Output Section.
001
Buffer RAM 000- MAG LSB (7-0).
Output
001- MAG MSB (15-8).
(|r| and φ) 010- PHASE LSB (7-0).
011- PHASE MSB (15-8).
See Output Section.
010
Buffered
000- FREQ LSB.
Frequency 001- FREQ MSB.
See Output Section.
011
Not Used
100
Input Level Input AGC
Detector
000- input AGC LSB (0-7).
001- input AGC NLSB (8-15).
010- input AGC MSB (16-23).
FN4450.4
May 1, 2007
HSP50214B
TABLE 22. DEFINITION OF ADDRESS MAP (Continued)
READ
CODE C(2:0)
101
STATUS
TYPE
Not Used
111
Not Used
Don’t Care
READ ADDRESS A(2:0)
AGC Data AGC (must write to location 10 to
and Timing sample)
Error
000- AGC LSB (lower 8-bits of linear
Control Word 3 used by multiplier)
mmmmmmmm LSB.
001- AGC MSB (4 shift control bits
and first three bits of linear) Control
Word oeeeemmm MSB. This yields
11-bits of the linear control mantissa.
010- Timing error LSB, not stabilized.
011- Timing error MSB, not
stabilized.
110
Status
124 CHANNELS
111- Status (6:0) consisting of
(6:4)-FIFO depth when output is in
FIFO Buffer RAM Output Mode.
(3)-EMPTY signalling the FIFO is
empty and the read pointer cannot be
advanced (Active High).
(2)-FULL signalling the FIFO is full
and new samples will not be written
(Active High).
(1)-READYB Output buffer has
reached the programmed threshold
in FIFO mode or the programmed
number of samples have been taken
in snapshot mode. (Active Low).
(0)-INTEGRATION has been
completed in the input level detector
and is ready to be read. (Active
High).
Applications
Composite Filter Response Example
For this example consider a total receive band roughly
25MHz wide containing 124 200kHz wide FDM channels as
shown in Figure 47. The design goal for the PDC is to tune
to and filter out a single 200kHz FDM channel from the FDM
band, passing only baseband samples onto the baseband
processor at a multiple of the 270.8 KBPS bit rate.
44
• • •
FREQUENCY
200kHz
CHANNEL
FREQUENCY
FIGURE 47. RECEIVE SIGNAL FREQUENCY SPECTRUM
RF/IF Considerations
The input frequency to the PDC is dependent on the A/D
converter selected, the RF/IF frequency, the bandwidth of
interest and the sample rate of the converter. If the A/D
converter has sufficient bandwidth, then undersampling
techniques can be used to downconvert IF/RF frequencies
as part of the digitizing process, using the PDC to process a
lower frequency alias of the input signal.
For example, a 70MHz IF can be sampled at 40MHz and the
resulting 10MHz signal alias can be processed by the PDC
to perform the desired downconversion/tuning and filtering. If
the IF signal is less than 1/2 the sample frequency then
standard oversampling techniques can be used to process
the signal. Of the two techniques, only undersampling allows
part of the down conversion function to be brought into the
digital domain just through sampling, assuming that a
sampling frequency can be found that keeps the alias
signals low and that the A/D converter has the bandwidth to
accept the unconverted analog signal.
FN4450.4
May 1, 2007
HSP50214B
PDC Configuration
References
For this example, the PDC is configured as follows:
For Intersil documents available on the web, see
www.intersil.com/
CLKIN: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39MHz
Mode: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gated
[1] HSP50210 Data Sheet, Intersil Corporation, FN3652.
Input Format: . . . . . . . . . . . . . As required by Digital Source
[2] Cellular Radio and Personal Communications: A Book
of Selected Readings, Theodore S. Rappaport, 1995 by
IEEE, Inc.
Carrier NCO Fc: . . . . . . . . As determined by Channel Freq.
Carrier NCO Phase Offset: . . . . . . . . . . . . . . . . . . . . . . . . .0
Carrier NCO Offset Frequency: . . . . . . . . . . . . . . . .Disabled
CIC Filter: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Enabled
Decimation: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18
PROCCLK: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28MHz
[3] AN9720 Application Note, Intersil Corporation,
“Calculating Maximum Processing Rates of the PDC
(HSP50214B)”.
[4] FO-007 Block Diagram of HSP50214.
Half Band Filters: . . . . . . . . . . . . . . . . . . HB3 and 5 Enabled
FIR Filter: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . gsmtemp file
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . fS = 541.667kHz
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Decimation = 1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Passband: 90kHz
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Transition Band: 25kHz
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Passband Atten: 3dB
. . . . . . . . . . . . . . . . . . . . . . . . . Stop Band Atten: 111.25713
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .FIR Order: 90
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FIR Symmetry: Even
Resampling Filter: . . . . . . . . . . . . . . . . . . . . . . HB1 Enabled
The basis for this configuration is:
Sampling Rate: Select a high rate PROCCLK
Output Rate: 1.083MHz (4x Bit Rate; 8x Baud Rate)
CIC Filtering: Primarily Rate Reduction (39/18 = 2.166MHz).
HB Filtering: Flat passband with rate reduction by 4 - low
enough (541.66kHz) for sufficient FIR Taps to be used.
FIR Filtering: Primary shaping filter/set final out of band
suppression.
Polyphase/HalfBand Filtering: Interpolate by two to output
8x baud rate or 4x bit rate.
The CIC and halfband filter responses are shown in Figures
48A and 48B.
The composite filter response constrained primarily by
halfband filter 5 and the FIR filter, are shown in Figure 48C
and 48D.
For a more detailed discussion of design approaches and
trades when designing with the PDC, refer to AN9720 [3],
“Calculating the Maximum Processing Rates of the PDC”.
45
FN4450.4
May 1, 2007
10
10
-10
-10
-30
-30
MAGNITUDE (dB)
MAGNITUDE (dB)
HSP50214B
-50
-70
-50
-70
-90
-90
-110
-110
fS = CIC INPUT RATE
fS = CIC INPUT RATE
-130
-130
FREQUENCY
FIGURE 48A. CIC FILTER RESPONSE
FIGURE 48B. HB3 FILTER RESPONSE
10
-10
-10
-30
-30
MAGNITUDE (dB)
10
-50
-70
-90
fS = CIC INPUT RATE
-50
-70
-90
-110
-110
fS = CIC INPUT RATE
-130
-130
fS
R
FREQUENCY
fS
R
FREQUENCY
FIGURE 48C. HB5 FILTER RESPONSE
FIGURE 48D. 255 FIR TAP FILTER RESPONSE
10
fS = CIC INPUT RATE
-10
MAGNITUDE (dB)
MAGNITUDE (dB)
fS
R
FREQUENCY
fS
R
-30
-50
-70
-90
-110
-130
FREQUENCY
fS
R
FIGURE 48E. COMPOSITE FILTER RESPONSE
FIGURE 48. PDC FILTER FREQUENCY SPECTRUMS EXAMPLE (NORMALIZED TO SAME SCALE)
46
FN4450.4
May 1, 2007
HSP50214B
Configuration Control Word Definitions
Note that in the Configuration Control Register Tables, some
of the available 32-bits in a Control Word are not used.
Unused bits do not need to be written to the Master Register.
If the destination only has 16-bits, then only 2 bytes need to
be written to the Master Register. Figure 45 details the timing
for proper operation of the Microprocessor Write Section. Bits
identified as “Reserved” should be programmed to a zero.
NOTE: CLKIN or PROCCLK must be present to properly load
control words. Note in the header which is applicable.
CONTROL WORD 0: CHIP CONFIGURATION, INPUT SECTION, CIC GAIN (SYNCHRONOUS TO CLKIN)
BIT
POSITION
31-21
FUNCTION
DESCRIPTION
Reserved
Reserved.
20
Carrier NCO External
Sync Enable
0- The SYNCIN1 pin has no effect on the Carrier NCO.
1- When the SYNCIN1 pin is asserted, the carrier center frequency and phase are updated from the
holding registers to the active register. Also, if bit 0 of this word is active, the carrier phase accumulator
feedback will be zeroed to set the Carrier NCO to a known phase, allowing the NCOs of multiple parts
to be initialized and updated synchronously.
19
CIC External Sync
Enable
0- The SYNCIN1 pin has no effect on the CIC filter.
1- When the SYNCIN1 pin is asserted, the decimation counter is loaded, allowing the decimation
counters in multiple chips to be synchronized. When CW27 bit-22 is set to a 1, SYNCIN1 will reset both
front end and back end circuitry.
18
Input Format
0- Two’s Complement Input Format.
1- Offset Binary Input Format.
17
Input Mode
0- Input operates in Gated Mode.
1- Input operates in Interpolated Mode.
16-13
CIC Shift Gain
These bits control the barrel shifter at the input to the CIC filter. These bits are added to the
GAINADJ(2:0) pins to determine the total shift. The sum is saturated at 15. See the CIC Decimation Filter
Section for values to be programmed in this field based on CIC filter Decimation. Bit 16 is the MSB.
SG = Floor [39 - (number of input bits) - 5log2(R)] for 4 < R < 31
SG = 15 for R = 4.
SG = 0 for R = 32.
12-7
CIC Decimation
Counter Preload
These bits control the decimation in the CIC filter. Program this field to R-1, where R is the desired
decimation factor in the filter. The decimation factor range is 4-32. See CIC Filter Section for effective
decimation range relative to the CIC Shift Gain value. Bit 12 is the MSB.
While this field allows values from 0 - 63, the valid values are in the range from 4- 32.
CIC Bypassed
Active high, this bit routes the output of the input shifter to the output of the CIC with no filtering.
When the CIC filter is bypassed, CLKIN must be at least twice the input sample rate (ENI should be
toggled to achieve this). When the CIC filter is bypassed, the bottom 24-bits of the barrel shifter output
are routed to the halfband filters.
Number of Offset
Frequency Bits
00 - 8-bits.
01 - 16.
10 - 24.
11 - 32.
3
Syncout CLK Select
This bit selects whether the SYNCOUT signal is generated from CLKIN of from PROCCLK
0- CLKIN.
1- PROCLK.
2
Clear Phase Accum
0- Enable accumulator in Carrier NCO.
1- Zero feedback in accumulator.
1
Carrier NCO Offset
Frequency Enable
When set to 1, this bit enables the offset frequency word to be added to the center frequency Control
Word. The offset is loaded serially via the COF and COFSYNC pins.
0
Carrier NCO Load
Phase Accum On
Update
When this bit is set to 1, the μP update to the Carrier NCO frequency or an external carrier NCO load
using SYNCIN1 will zero the feedback of the phase accumulator, as well as update the phase or
frequency. This function can be used to set the NCO to a known phase synchronized to an external
event.
6
5-4
47
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 1: INPUT LEVEL DETECTOR (SYNCHRONOUS TO CLKIN)
BIT
POSITION
FUNCTION
DESCRIPTION
31
Reserved
Reserved.
30
Integration Mode
0- Integration of magnitude error stops when the interval counter times out.
1- Integration runs continuously. When the interval counter times out, the integrator reloads, and the
results of the integration is sent to a register for the processor to read.
29-14
Integration Interval
These are the top 16-bits of the 18-bit integration counter, ICPrel. ICPrel = (N)/4+1; where N is the
desired integration period in CLKIN cycles, defined as the number of input samples to be integrated. N
must be a multiple of 4: [0, 4, 8, 12, 16.... , 218]. Bit 29 is the MSB. If the input is interpolated, then the
zeros must be accounted for, as they will be added to the threshold! If the gated input mode is used, the
same input sample will be accumulated multiple times.
13-0
Input Threshold
Input Magnitude Threshold. Bits 12-0 correspond to input bits 12-0. The magnitude of the input is added
to this threshold, where the threshold is a signed number. Bit 13 is the MSB.
CONTROL WORD 2: INPUT LEVEL DETECTOR START STROBE (SYNCHRONIZED TO CLKIN)
BIT
POSITION
N/A
FUNCTION
Start Input Level
Detector AGC
Integrator
DESCRIPTION
Writing to this location starts/restarts the input AGC error integrator. The integrator will either restart or
stop when the integration interval counter times out depending on bit 30 of Control Register 1 (see
Microprocessor Write Section).
CONTROL WORD 3: CARRIER NCO CENTER FREQUENCY (SYNCHRONIZED TO CLKIN)
BIT
POSITION
31-0
FUNCTION
Carrier Center
Frequency
DESCRIPTION
These bits control the frequency of the Carrier NCO. The frequency range of the NCO is ± fS / 2 where
fS is the input sample rate. The bits are computed by the equation N = (FNCO / fS)*232. Bit 31 is the MSB.
This location is a holding register. After loading, a transfer to the active register is done by writing to
Control Word 5 or by generating a SYNCIN1 with Control Word 0, Bit 20 set to 1. The Carrier NCO only
updates ENI is active.
NOTE: In the HSP50214B, if the SYNCIN1 occurs when the NCO is not updating, the load signal is held internal to the part until the next NCO
update.
CONTROL WORD 4: CARRIER PHASE OFFSET (SYNCHRONIZED TO CLKIN)
BIT
POSITION
31-10
9-0
FUNCTION
DESCRIPTION
Reserved
Reserved.
Carrier Phase Offset
These bits, PO, are used to offset the phase of the carrier NCO. The bits are computed by the Equation
PO = INT[(210φoff)/ 2π]HEX; (-π <φoff< π) for 10-bit 2’s complement representation or from 0 to 2π for 10bit offset binary representation. Bit 9 is the MSB. This location is a holding register. After loading, a
transfer to the active register is done by writing to Control Word 6 or by generating a SYNCIN1 with
Control Word 0, Bit 20 set to 1. The carrier NCO only updates when ENI is active.
CONTROL WORD 5: CARRIER FREQUENCY STROBE (SYNCHRONIZED TO CLKIN)
BIT
POSITION
N/A
FUNCTION
Carrier Frequency
Strobe
DESCRIPTION
Writing to this address updates the carrier frequency Control Word from the Holding Register.
CONTROL WORD 6: CARRIER PHASE STROBE (SYNCHRONIZED TO CLKIN)
BIT
POSITION
N/A
FUNCTION
Carrier Phase Strobe
48
DESCRIPTION
Writing to this address updates the carrier phase offset Control Word with the value written to the phase
offset (PO) register.
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 7: HB, FIR CONFIGURATION (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-22
FUNCTION
DESCRIPTION
Reserved
Reserved.
21
Enable External
Filter Sync
0- The SYNCIN2 pin has no effect on the halfband and FIR filters.
1- When the SYNCIN2 pin is asserted, the filter control circuitry in the halfband filters, the FIR, the
resampler, and the discriminator are reset. SYNCIN2 can be used to synchronize the computations of
the filters in multiple parts for the alignment (see Synchronization Section).
20
Halfband (HB)
Bypass
1- Bypass Halfband Filters.
0- Enable HB Filters (at least one HB must be enabled).
19
HB5 Enable
0- Disables HB number 5 (the last in the cascade).
1- Enables HB filter number 5.
18
HB4 Enable
Setting this bit enables HB filter number 4.
17
HB3 Enable
Setting this bit enables HB filter number 3.
16
HB2 Enable
Setting this bit enables HB filter number 2.
15
HB1 Enable
Setting this bit enables HB filter number 1.
FIR Decimation
Load decimation from 1-16, where 0000 = 16. Bit 14 is the MSB.
0001 - 1
1001 - 9
0010 - 2
1010 - 10
0011 - 3
1011 - 11
0100 - 4
1100 - 12
0101 - 5
1101 - 13
0110 - 6
1110 - 14
0111 - 7
1111 - 15
1000 - 8
0000 - 16
10
FIR Real/Complex
0- Complex Filter.
1- Dual Real Filters.
9
FIR Sym Type
0- Odd Symmetry.
1- Even Symmetry.
8
FIR Symmetry
0- Symmetric Filters.
1- Asymmetric Filters.
FIR Taps
Number of taps in the FIR filter. Range is 1 to 255, where 0000000 is invalid.
14-11
7-0
CONTROL WORD 8: AGC CONFIGURATION 1 (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-30
FUNCTION
DESCRIPTION
Reserved
Reserved.
Sync AGC Updates to
SYNCIN2
When this bit is 1, the SYNCIN2 pin loads the contents of the master registers into the AGC accumulator.
28-16
Threshold
The magnitude measurement out of the cartesian to polar converter is subtracted from this value to get
the gain error. A gain of 1.647 in the cartesian to polar conversion that must be taken into account when
computing this threshold. These bits are weighted -22 down to 2-10. Bit 28 is the MSB.
15-12
Loop Gain 1
Mantissa
Selected when AGCGNSEL = 1. These bits, MMMM, together with the exponent bits, EEEE (11-8), set
the loop gain for the AGC loop. The gain adjustment per output sample is:
1.5dB (Threshold -[Magnitude * 1.6]) 0.MMMM * 2-(15 - EEEE) where magnitude ranges from 0 to 1.414
and the threshold is programmed in bits 28-16. The decimal value for the mantissa is calculated as
DEC(MMMM)/16. Bit 15 is the MSB.
11-8
Loop Gain 1
Exponent
Selected when AGCGNSEL = 1. These bits are EEEE. See description of bits 15-12. Bit 11 is the MSB.
7-4
Loop Gain 0 Mantissa
Selected when AGCGNSEL = 0. These bits are MMMM. See description for bits 15-12. Same equations
are used for Loop 0. Bit 7 is the MSB.
3-0
Loop Gain 0
Exponent
Selected when AGCGNSEL = 0. These bits are EEEE. See description for bits 15-12. Same equations
are used for Loop 0. Bit 3 is the MSB.
29
49
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 9: AGC CONFIGURATION 2 (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
FUNCTION
DESCRIPTION
31-28
Reserved
Reserved.
27-16
Upper Limit
Maximum Gain/Minimum Signal. The upper four bits are used for exponent; the remaining bits form the
mantissa in the fractional offset binary: [eeeemmmmmmmm]. See the AGC Section for details. Bit 27 is
the MSB. The gain is in dB. G = (6.02)(eeee) + 20log10(1.0 + 0.mmmmmmmm)
eeee = Floor [log2(10GAIN dB/20)]
mmmmmmmm = Floor [256(10GAIN dB/20/2eeee - 1)]
15-12
Reserved
Reserved.
Lower Limit
Minimum Gain/Maximum Signal. The upper four bits are used for exponent; the remaining bits form the
mantissa in the fractional offset binary: [eeeemmmmmmmm]. See the AGC Section for details. Bit 11 is
the MSB. The gain is in dB. G = (6.02)(eeee) + 20log10(1.0 + 0.mmmmmmmm)
eeee = Floor [log2(10GAIN dB/20)]
mmmmmmmm = Floor [256(10GAIN dB/20/2eeee - 1)]
11-0
CONTROL WORD 10: AGC SAMPLE GAIN CONTROL STROBE (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
N/A
FUNCTION
Sample AGC Gain
Level
DESCRIPTION
Writing to this location samples the output of the AGC loop filter to stabilize the value for μP reading.
CONTROL WORD 11: TIMING NCO CONFIGURATION (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-6
FUNCTION
DESCRIPTION
Reserved
Reserved.
5
Enable External
Timing NCO Sync
0- SYNCIN2 has no effect on the timing NCO.
1- When SYNCIN2 is asserted, the timing NCO center frequency and phase are updated with the value
loaded in their holding registers. If bit 0 of this word is set to 1, the phase accumulator feedback is also
zeroed.
4-3
Number of Offset
Frequency Bits
00 - 8 bits.
01 - 16.
10 - 24.
11 - 32.
2
Enable Offset
Frequency
0- Zero Offset Frequency to Adder.
1- Enable Offset Frequency.
1
Clear Phase
Accumulator
0- Enable Accumulator.
1- Zero Feedback in Accumulator.
0
Timing NCO Phase
Accumulator Load On
Update
When this bit is set to 1, the μP update to the timing NCO frequency or an external timing NCO load
using SYNCIN2 will zero the feedback of the phase accumulator as well as update the phase and
frequency. This function can be used to set the NCO to a known phase synchronized to an external
event.
CONTROL WORD 12: TIMING NCO CENTER FREQUENCY (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-0
FUNCTION
Timing NCO Center
Frequency
50
DESCRIPTION
These bits control the frequency of the timing NCO. The frequency range of the NCO is from 0 to
FRESAMP where FRESAMP is the input sample rate to the resampling filter. The bits are computed by
the equation: N =(fOUT / FRESAMP)*232. Bit 31 is the MSB. This location is a holding register. After
loading, a transfer to the Active Register is done by writing to Control Word 14 or by generating a
SYNCIN2 with Control Word 11, Bit 5 set to 1.
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 13: TIMING PHASE OFFSET (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
FUNCTION
DESCRIPTION
31-8
Reserved
Reserved.
7-0
Timing NCO Phase
Offset
These bits are used to offset the phase of the Timing NCO. The range is 0 to 1 times the resampler input
period interpreted either as ± T/2 (2’s complement) or 0 to T (offset binary). Bit 7 is the MSB. This location
is a holding register. After loading, a transfer to the Active Register is done by writing to Control Word 15
or by generating a SYNCIN2 with Control Word 11, Bit 5 set to 1.
CONTROL WORD 14: TIMING FREQUENCY STROBE (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
N/A
FUNCTION
Timing Frequency
Strobe
DESCRIPTION
Writing to this address updates the active timing NCO Frequency Register in the timing NCO (see Timing
NCO Section).
CONTROL WORD 15: TIMING PHASE STROBE (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
N/A
FUNCTION
Timing Phase Strobe
DESCRIPTION
Writing to this address updates the active timing NCO Phase Offset Register in the timing NCO (see
Timing NCO Section).
CONTROL WORD 16: RESAMPLING FILTER CONTROL (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-12
11-4
FUNCTION
DESCRIPTION
Reserved
Reserved.
Re-Sampler Output
Pulse Delay
NOTE: These bits program the delay between output samples when interpolating. The extra outputs can
be delayed from 2 to 255 clocks from the first output. A delay of 2 equals 255 clocks of delay. A
delay of 0 or 1 is an invalid mode. When interpolating by 2, one extra output is generated; when
interpolating by 4, 3 extra outputs are generated. Program by the equation (PROCCLK/fOUT) - 1.
Bit 11 is the MSB.
NOTE: If less than 5 is programmed, there will not be sufficient time to fully update the output
buffer. If less than 16 is programmed, the serial output may be preempted. This means that
it won’t finish and if the sync is programmed to follow the data, there may never be a sync.
3
Re-Sampler Bypass
0- Resampling Filter Enabled. A valid combination of bits 2-0 must also be selected.
1- Resampling Filter Section (including Interpolation halfband filters) is bypassed.
2-0
Filter Mode Select;
2- HB2 Enabled
1- HB1 Enabled
0- Re-Sampler
Enabled
000- Not Valid.
001- Re-Sampler Enabled.
010- Halfband 1 Enabled.
011- Re-Sampler and Halfband Filter 1 Enabled.
100- Not Valid.
101- Not Valid.
110- Both Halfband Filters Enabled.
111- Re-Sampler and Both Halfband Filters Enabled.
51
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 17: DISCRIMINATOR FILTER CONTROL, DISCRIMINATOR DELAY (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
FUNCTION
DESCRIPTION
31-17
Reserved
Reserved.
16-15
Phase Multiplier
These bits program allow the phase output of the cartesian to polar converter to be multiplied by 1, 2, 4, or
8 (modulo 2π) to remove phase modulation before the frequency is measured.
00- No Shift on Phase Input to frequency discriminator.
01- Shift Phase Input to frequency discriminator up 1 (one bit), discarding the MSB and zero filling the LSB.
10- Shift Phase Input to frequency discriminator up 2 (two) bits, discarding the MSB and zero filling the LSB.
11- Shift Phase Input to frequency discriminator up 3 (three) bits, discarding the MSB and zero filling the LSB.
Discriminator Enable
0- Disable Discriminator.
1- Enable Discriminator.
Discriminator FIR
Decimation
The decimation can be programmed from 1 to 8, where 000 = decimate by 8; 001 = decimate by 1; 010 =
decimate by 2; 011 = decimate by 3; 100 = decimate by 4; 101 = decimate by 5; 110 = decimate by 6; and
111 - decimate by 7.
10
FIR Symmetry Type
0- Odd Symmetry.
1- Even Symmetry.
9
FIR Symmetry
0- Symmetric.
1- Asymmetric.
8-3
Number of FIR Taps
Number of FIR taps from 1 to 63, where 00000 is not valid (00001 = 1 tap, 00010 = 2 taps, etc. up to 11111
= 63 taps). Bit 8 is the MSB.
2-0
Discriminator Delay
Sets the number of delays from 1 to 8 in the discriminator. Set delay ddd to delay minus 1, where 000
represents 1 delay; 001 represents 2 delays, 010 represents 3 delays, 011 represents 4 delays, 100
represents 5 delays, 101 represents 6 delays, 110 represents 7 delays, and 111 represents 8 delays. If ddd
the decimal representation bits 2-0, then the discriminator a transfer function H(Z) = 1-Z-(ddd + 1).
14
13-11
CONTROL WORD 18: TIMING ERROR PRELOADS (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
FUNCTION
DESCRIPTION
31-28
Reserved
Reserved.
27-16
NCO Divide
The Re-Sampler NCO output is divided down by the value loaded into this register plus 1. Load with a
value that is one less than the desired period. Bit 27 is the MSB.
Reference Divide
The reference clock is divided down by the value loaded into this register plus 1. Load with a value that
is one less than the desired period. Bit 27 is the MSB. A minimum preload of “I” is required.
11-0
CONTROL WORD 19: SERIAL OUTPUT ORDER (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31
FUNCTION
DESCRIPTION
Reserved
Reserved.
30-28
Data Source for
SEROUTA
Serial Output A Source. The serial data source is selected using Table 12 (see Output Section).
27-25
Data Source for
SEROUTB
Serial Output B Source. The serial data source is selected using Table 12 (see Output Section).
24-21
Number of Serial Word This parameter determines the number of SERSYNC pulses generated. It can be set from 1 to 7. If this
Links in a Chain
parameter matches the number of serial words that are linked together to form a serial output chain, then
there will be a sync pulse for every word in the serial output. In applications where a processor is
receiving the serial data, it may be desirable to have a single SERSYNC pulse for the whole serial output
chain, instead of a SERSYNC for each word in the data chain. The processor then parses out the various
data words. As an example, if the I and Q are chained together and a single SERSYNC pulse is
generated for this serial output chain, no ambiguity exists in the processor about which two data samples
(one from I and one from Q) are related.
20-18
Link Following I Data
52
The serial data word, or link, following the I data word is selected using Table 12
(see Output Section).
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 19: SERIAL OUTPUT ORDER (SYNCHRONIZED TO PROCCLK) (Continued)
BIT
POSITION
FUNCTION
17-15
Link Following Q Data
The serial data word, or link, following the Q data word is selected using Table 12
(see Output Section).
14-12
Link Following
Magnitude Data
The serial data word, or link, following the MAG data word is selected using Table 12
(see Output Section).
11-9
Link Following Phase
Data
The serial data word, or link, following the PHAS data word is selected using Table 12
(see Output Section).
8-6
Link Following
Frequency Data
The serial data word, or link, following the FREQ data word is selected using Table 12
(see Output Section).
5-3
Link Following AGC
Level Data
The serial data word, or link, following the AGC data word is selected using Table 12
(see Output Section).
2-0
Link Following Timing
Error Data
The serial data word, or link, following the TIMER data word is selected using Table 12
(see Output Section).
DESCRIPTION
CONTROL WORD 20: BUFFER RAM, DIRECT PARALLEL, AND DIRECT SERIAL OUTPUT CONFIGURATION
(SYNCHRONIZED WITH PROCCLK)
BIT
POSITION
31-26
FUNCTION
DESCRIPTION
Reserved
Reserved.
25
Data Source for Least
Significant Bytes of
AOUT and BOUT
Output LSBytes, bits (7:0), of AOUT and BOUT can provide:
0- Buffer RAM Mode Output or,
1- Parallel Direct Mode Output.
24
Buffered Output Mode
Interface
Buffered Mode Output interfaces to either:
0- 8-bit μP (address = μP ASEL(5:#); CLK = μP RAM read).
1- 16-bit μP (address = SEL(2:0); CLK = OEBL).
23-22
AOUT Direct Parallel
Output Mode Data
Source
The data word sent by the Direct Parallel Output Mode to AOUT is:
00- I Data. (2’s complement)
01- Magnitude. (O; unsigned binary)
1X- Frequency. (2’s complement)
21-20
BOUT Direct Parallel
Output Mode Data
Source
The data word sent by the Direct Parallel Output Mode to BOUT is:
00- Q Data (2’s complement).
01- Phase (2’s complement).
1X- Magnitude (O; unsigned binary).
19
Serial Output Sync
Polarity
0- Normal Sync Mode (active high).
1- Sync Inverted (active low).
18
Serial Output Clock
Polarity
0- Output Clock Inverted rising edge aligns with data transitions.
1- Output Clock Normal falling edge aligns with data transitions.
0
17
1
Serial Output Sync
Position
0- Sync is asserted one bit time after the last bit of the serial word (Late Mode).
1- Sync is asserted one bit time prior to the first bit of the serial word (Early Mode).
16-14
Serial Out Clock
Divider
000- Serial Output at PROCCLK/16.
001- Serial Output at PROCCLK/8.
010- Serial Output at PROCCLK/4.
011- Serial Output at PROCCLK/2.
1XX- Serial Output at PROCCLK rate.
13-12
I Data Serial Output
Tag Bit
00- No Tag Bit. LSB of word is passed.
01- 0 Tag Bit. LSB of word is set to zero.
1X- 1 Tag Bit. LSB of word is set to one.
53
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 20: BUFFER RAM, DIRECT PARALLEL, AND DIRECT SERIAL OUTPUT CONFIGURATION
(SYNCHRONIZED WITH PROCCLK) (Continued)
BIT
POSITION
11-10
FUNCTION
Q Data Serial Output
Tag Bit
DESCRIPTION
(See I Data Serial Output Tag selection above).
9-8
Magnitude Data Serial (See I Data Serial Output Tag selection above).
Output Tag Bit
7-6
Phase Data Serial
Output Tag Bit
5-4
Frequency Data Serial (See I Data Serial Output Tag selection above).
Output Tag Bit
3-2
AGC Data Serial
Output Tag Bit
(See I Data Serial Output Tag selection above).
1-0
Timing Error Data
Serial Output Tag Bit
(See I Data Serial Output Tag selection above).
(See I Data Serial Output Tag selection above).
CONTROL WORD 21: BUFFER RAM OUTPUT CONTROL REGISTER (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-16
FUNCTION
DESCRIPTION
Reserved
Reserved.
Output Buffer Mode
0- The output buffer operates in snapshot mode.
1- The output buffer operates in FIFO mode.
FIFO Mode Depth
Threshold
In FIFO mode, when the FIFO depth reaches this threshold, an interrupt is generated and the READY
flag is asserted. The threshold may be set from 0 to 7. Bit 14 is the MSB. The interrupt is generated when
the FIFO depth reaches the threshold, as the FIFO fills.
11-4
Snapshot Mode
Interval
In snapshot mode, the interval between snapshots in the output sample times is determined by this 8bit binary number, i.e. 256, (28), sample time counts between snapshot samples. Program this
parameter to 1 less than the desired interval. Bit 11 is the MSB.
3-0
Snapshot Mode
Number of Samples
In snapshot mode, the number of samples stored each time the snapshot interval counter times out is
equal to the decimal version of this 4-bit number. The range is 1- 8. Bit 3 is the MSB.
15
14-12
CONTROL WORD 22: BUFFER RAM OUTPUT FIFO RESET (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
N/A
FUNCTION
FIFO reset
DESCRIPTION
A write to this address increments the output FIFO RAM address pointers to READ = 111 and WRITE =
000.
CONTROL WORD 23: INCREMENT OUTPUT FIFO (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
N/A
FUNCTION
FIFO Strobe
DESCRIPTION
A write to this address increments the output FIFO/buffer to the next sample set.
CONTROL WORD 24: SYNCOUT STROBE OUTPUT PIN
(SYNCHRONIZED TO CLKIN OR PROCCLK DEPENDING ON PROGRAMMING IN CONTROL WORD 0)
BIT
POSITION
N/A
FUNCTION
SYNCOUT Strobe
54
DESCRIPTION
A write to this address generates a one clock period wide strobe on the SYNCOUT pin that is
synchronized to the clock. This strobe may be synchronized to CLKIN or PROCCLK based on the
programming of bit 3 of Control Word 0.
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 25: COUNTER AND ACCUMULATOR RESET (SYNCHRONIZED TO BOTH CLKIN AND PROCCLK)
BIT
POSITION
N/A
FUNCTION
Counter and
Accumulator Reset
DESCRIPTION
A write to this address initializes the counters and accumulators for testing. Items that are reset are:
Carrier NCO.
1. Loads phase offset <9:0> into register to be used for adding to accumulator.
2. Enables feedback on the accumulator.
CIC Filter
1.
2.
3.
4.
Resets the decimation counter.
Clears enables to CIC.
Clears accumulators in CIC.
Clears enable leaving CIC.
Halfband Filters
1.
2.
3.
4.
5.
Resets compute counter in Halfband control.
Resets read address for all Halfband Filters.
Resets write address for all Halfband Filters.
Clears input available strobe.
Resets Halfband control logic.
255 Tap FIR
1. Resets FIR read and write address pointers.
2. Zero’s coefficient read address.
AGC Loop
1. Clears accumulator in loop filter.
Re-Sampler and Interpolation Halfband Filters.
1. Resets counters for Halfband addresses for writing.
2. Resets output enable.
3. Reset controller for Re-Sampler.
Timing NCO
1. Initializes counters for inserting extra pulses when interpolating halfbands are enabled. In the
HSP50214B, a configuration control word bit determines if a Timing NCO reset is executed. If
Control Word 27, Bit 20 is set to a logic one, a reset will clear the feedback in the timing NCO phase
accumulator. If Control Word 27, Bit 20 is zero, a reset will not clear the timing NCO phase
accumulator feedback, which is how the HSP50214 operated.
Discriminator
1. Resets read and write address pointers.
2. Zero’s coefficient read address.
Cartesian to Polar Coordinate Counter
1. Resets Cordic counters (stops current computation).
FIFO Control
1.
2.
3.
4.
5.
6.
Resets decoder for controlling FIFO.
Resets write address for FIFO.
Clears RD and INTRRPT.
Resets “depth” and “full” flags.
Sets the empty flag.
Sets the read address to “7”, write address to “0”.
Snapshot Control
1. Zeros the group number.
2. Load interval counter.
3. Resets write address and read address for FIFO.
Output Serial Control
1.
2.
3.
4.
5.
55
Reloads shift counter.
Reloads “Number of Words” counter.
Reloads counter for sync (for early or late).
Reloads counter for dividing down SERCLK.
In the HSP50214B, the Control Word 25 reset signal is designed such that the front end reset is 10
CLKIN periods wide and the back end reset is 10 PROCCLK periods wide. This guarantees that no
enables will be caught in the pipelines.
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORD 26: LOAD AGC GAIN (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
FUNCTION
DESCRIPTION
(15:12)
(11:5)
(5:0)
eeee - AGC Exponent
mmmmmmm - AGC Mantissa
000000 - Not Used
AGC LOAD. Writing to this location generates a strobe to load the AGC loop accumulator with bits
(15:5) to the master registers. These bits are loaded into the MSBs of the AGC loop filter accumulator.
Bits 15:12 are the exponent associated with the AGC gain shifter, while bits 11:5 are the mantissa
associated with the AGC multiplier. The weighting of the AGC mantissa is 01.mmmmmmm. When
considering Figure 23, the AGC Block Diagram, note the mux between the Register and the Limiter
in the AGC Loop filter. The AGC LOAD controls the mux. Normally the mux would select the limiter
output. When the AGC LOAD is asserted via the write command, the mux selects the Write Master
Registers for data input See Table 20, Figure 45 and the associated text of the data sheet for an
explanation of how data is loaded into the Master Registers for use internal to the part. Note that for
the AGC LOAD only the lower 16-bits require data be valid to ensure a proper write of an AGC Value
that will be loaded on write to CONTROL WORD 26.
CONTROL WORD 27: TEST REGISTER (SYNCHRONIZED TO CLKIN)
BIT
POSITION
31-25
24
FUNCTION
DESCRIPTION
Reserved
A fixed value of 0000 000 is loaded here for normal operation.
RAM Test Enable
0 = Normal Operation; 1 = RAM Test Enabled. The B Version includes test circuitry for the ROM and RAM
blocks that was not present in the original release part. This circuitry must be disabled before loading the
coefficient RAM’s. This is done by setting bit 24 to zero.
Because the HSP50214 did not require a “write” to Control Word 27 and the HSP50214B does
require that Control Word 27, Bit 24 be set to zero for normal operation, software that was written
for the HSP50214 will require modification to work properly with the HSP50214B.
23
Input Level Detector
Counter Preload
Select
0 = The two LSB’s of the interpolation period preload are set to zero.
22
SYNCIN1 Reset
Control
0: SYNCIN1 causes only front end reset.
1: SYNCIN1 causes front end and back end resets.
21
Timing Error Input
Select
0 = Operates as HSP50214.
1 = Corrects an error in the 4 LSB’s.
20
Timing NCO Reset
Control Select
0 = Backend reset will not clear the timing NCO phase accumulator feedback.
Discriminator FIR
Input
00 = 18-bits of delayed and subtracted (optionally shifted) phase.
01 = 18-bits of magnitude from coordinate converter.
1X = 18-bits of resampler/halfband filer I output.
Input Level Detector
Integration Start
Select
0 = No external sync control of input end detector start/restart of integration period.
16
AGC Average
Control
0: AGC settles to mean.
1: AGC settles to median.
15
AGC Clear Inhibit
When set to zero, this bit will clear the AGC loop filter accumulator on a SYNCIN2 assertion or a WRITE
to CW25.
When set to a one, a WRITE to CW25 will not clear the AGC loop filter accumulator.
14
Q Input to Coordinate
Converter (see bits 19
- 15)
0 = I and Q enabled to coordinate converter.
I = Q input to coordinate converter is zeroed.
13
Coordinate Converter
Input
0 = The Resampler HB filter output is routed to coordinate converter.
1 = The output of 255 tap FIR is routed to coordinate converter.
Reserved
A fixed value 0 0010 0111 1000 [0278]hex is loaded here for normal operation.
A fixed value 0 0010 0111 1010 [027A]hex is loaded here for setting the Sin/Cos Generator outputs to
7FFF.
19 - 18
17
12-0
56
1 = The two LSB’s of the interpolation period preload are set to one.
1 = Backend reset clears the timing NCO phase accumulator.
1 = SYNCIN causes the input level detector to start/restart its integration period.
FN4450.4
May 1, 2007
HSP50214B
CONTROL WORDS 64-95: DISCRIMINATOR COEFFICIENT REGISTERS (SYNCHRONIZED TO PROCCLK)
BIT
POSITION
31-10
FUNCTION
Discriminator FIR
Coefficient
DESCRIPTION
The discriminator FIR coefficients are 22-bit-two’s complement. If the filter is symmetric, the coefficients
are loaded from the center coefficient at address 64 to the last coefficient. If the filter is asymmetric the
coefficients C0 to CN are loaded with C0 in address 64 up to 64+N, where N is number of asymmetric
coefficients.
CONTROL WORDS 128-255: 255 PROGRAMMABLE COEFFICIENT REGISTERS
BIT
POSITION
31-10
FUNCTION
Programmable FIR
Coefficient
DESCRIPTION
The programmable FIR coefficients are 22-bit-two’s complement. If the filter is symmetric, the
coefficients are loaded from the center coefficient at address 128 to the last coefficient. If the filter is
asymmetric the coefficients C0 to CN are loaded with C0 in address 128 up to 128+N, where N is number
of asymmetric coefficients.
Real Filters are computed as:
Xn-k+1 Ck1 + Xn-k+2 Ck-2 + ... XnC0),
where C0 is the coefficient in address 128 and Xo is the oldest data sample.
Complex filters outputs are computed as follows:
Xn is the most recent data sample.
k is the number of samples = number of (complex) taps.
C0_re is the coefficient loaded into CW128.
C0_im is the coefficient loaded into CW129.
The convolution starts with the oldest data, times the last complex coefficient, and ends with the newest
data, times the first complex coefficient loaded.
Iout
= (-Xn-k+1_q * Ck-1_im + Xn-k+1_i * Ck-1_re).
+ (-Xn-k+2_q * Ck-2_im + Xn-1+2_i * Ck-2_re).
+ ...
+ (-Xn_q
* C0_im + Xn_i
* C0_re).
Qout = (Xn-k+1_i * Ck-1_im + Xn-k+1_q * Ck-1_re).
+ (Xn-k+2_i * Ck-2_im + Xn-1+2_q * Ck-2_re).
+ ...
+ (Xn_i
57
* C0_im + Xn_q
* C0_re).
FN4450.4
May 1, 2007
HSP50214B
Absolute Maximum Ratings
Thermal Information
Supply Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +7.0V
Input, Output or I/O Voltage . . . . . . . . . . . . GND-0.5V to VCC +0.5V
ESD Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Class 2
Thermal Resistance (Typical, Note 4)
θJA (°C/W)
MQFP Package . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
Maximum Junction Temperature . . . . . . . . . . . . . . . . . . . . . . +150°C
Maximum Storage Temperature Range . . . . . . . . . .-65°C to +150°C
Pb-free reflow profile . . . . . . . . . . . . . . . . . . . . . . . . . .see link below
http://www.intersil.com/pbfree/Pb-FreeReflow.asp
Operating Conditions
Voltage Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . +4.75V to +5.25V
Temperature Range
Commercial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0°C to +70°C
Industrial. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .-40°C to +85°C
Input Low Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0V to +0.8V
Input High Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2V to VCC
Input Rise and Fall Time . . . . . . . . . . . . . . . . . . . . . . . . . 1V/ns Max
CAUTION: Stresses above those listed in “Absolute Maximum Ratings” may cause permanent damage to the device. This is a stress only rating and operation of the
device at these or any other conditions above those indicated in the operational sections of this specification is not implied.
NOTE:
4. θJA is measured with the component mounted on an evaluation PC board in free air.
VCC = 5 ±5%, TA = 0°C to +70°C, Commercial; -40°C to +85°C, Industrial
DC Electrical Specifications
MIN
MAX
UNITS
Logical One Input Voltage
PARAMETER
SYMBOL
VIH
VCC = 5.25V
2.0
-
V
Logical Zero Input Voltage
VIL
VCC = 4.75V
-
0.8
V
Clock Input High
VIHC
VCC = 5.25V
3.0
-
V
Clock Input Low
VILC
VCC = 4.75V
-
0.8
V
Output High Voltage
VOH
IOH = -400μA, VCC = 4.75V
2.6
-
V
Output Low Voltage
VOL
IOL = +2.0mA, VCC = 4.75V
-
0.4
V
Input Leakage Current
Standby Power Supply Current
Output Leakage Current
Operating Power Supply Current
II
VIN = VCC or GND, VCC = 5.25V
-10
+10
μA
ICCSB
VCC = 5.25V, Outputs Not Loaded
-
500
μA
IO
VIN = VCC or GND, VCC = 5.25V
-10
+10
μA
CLK = PROCCLK = 52MHz, VIN = VCC or
GND,
VCC = 5.25V, Outputs Not Loaded
-
420
mA
(Note 5)
Freq = 1MHz, VCC open, all measurements
are referenced to device ground
-
8
pF
(Note 6)
ICCOP
Input Capacitance
CIN
Output Capacitance
TEST CONDITIONS
COUT
NOTES:
5. Power Supply current is proportional to operation frequency. Typical rating for ICCOP is 7mA/MHz.
6. Capacitance TA = +25°C, controlled via design or process parameters and not directly tested. Characterized upon initial design and at major
process or design changes.
AC Electrical Specifications
VCC = 5 ±5%, TA = 0° to +70°C, Commercial (Note 7); -40°C to +85°C, Industrial (Note 7)
65MHz
PARAMETER
SYMBOL
MIN
MAX
UNITS
CLKIN Clock Period
tCP
15
-
ns
CLKIN High
tCH
6
-
ns
CLKIN Low
tCL
6
-
ns
PROCCLK Period
tPCP
18
-
ns
PROCCLK High
tPCH
7
-
ns
PROCCLK Low
tPCL
7
-
ns
REFCLK Clock Frequency
fRCP
-
PROCCLK/2
Hz
REFCLK High
tRCH
7
-
ns
REFCLK Low
tRCL
7
-
ns
Setup Time GAINADJ(2:0), IN(13:0), ENI, COF, COFSYNC, and
SYNCIN1 to CLKIN
tDS
7
-
ns
58
FN4450.4
May 1, 2007
HSP50214B
AC Electrical Specifications
VCC = 5 ±5%, TA = 0° to +70°C, Commercial (Note 7); -40°C to +85°C, Industrial (Note 7) (Continued)
65MHz
PARAMETER
SYMBOL
MIN
MAX
UNITS
Hold Time GAINADJ(2:0), IN(13:0), ENI, COF, COFSYNC, and SYNCIN1
from CLKIN
tDH
0
-
ns
Setup Time AGCGNSEL, SOF, MCSYNCI, SOFSYNC, and SYNCIN2 to
PROCCLK
tDSS
7
-
ns
Hold Time AGCGNSEL, SOF, MCSYNCI, SOFSYNC, and SYNCIN2
from PROCCLK
tDHS
0
-
ns
Setup Time, A(2:0) to Rising Edges of WR
tWSA
8
-
ns
Setup Time, A(2:0) C(7:0) to Rising Edges of WR
tWSC
10
-
ns
Hold Time, A(2:0) from Rising Edges of WR
tWHA
2
-
ns
Hold Time, A(2:0) C(7:0) from Rising Edges of WR
tWHC
0
-
ns
tWC
14
-
ns
(Note 9)
tDO_OUT
-
8
ns
PROCCLK to SYNCOUT
tDO_SYNCI
-
8
ns
PROCCK to MCSYNCO
WR to CLKIN
PROCCLK to AOUT(15:0), BOUT (15:0), DATARDY, SEROUTA,
SEROUTB, INTRRP
tDO_SYNCO
-
6
ns
PROCCLK to SERCLK, SERSYNC Valid
tDOS
-
12
ns
WR High
tWRH
15
-
ns
WR Low
tWRL
8
-
ns
RD Low
tRL
20
-
ns
Address Setup to Read Low
tAS
-
3
ns
RD LOW to Data Valid
tRDO
-
18
ns
RD HIGH to Output Disable
tROD
-
10
ns
(Note 8)
Output Enable Time
tOE
-
6
ns
tOEBL
-
15
ns
Output Disable Time
tOD
-
8
ns
(Note 8)
Output Rise, Fall Time
tRF
-
3
ns
(Note 8)
Output Enable Time - FIFO Read Mode
NOTES:
7. AC tests performed with CL = 40pF, IOL = 2mA, and IOH = -400μA. Input reference level for CLK is 2.0V, all other inputs 1.5V.
Test VIH = 3.0V, VIHC = 4.0V, VIL = 0V.
8. Controlled via design or process parameters and not directly tested. Characterized upon initial design and at major process or design changes.
9. Setup time required to ensure action initiated by WR will be seen by a particular CLKIN.
59
FN4450.4
May 1, 2007
HSP50214B
AC Test Load Circuit
S1
DUT
CL (NOTE)
±
SWITCH S1 OPEN FOR ICCSB AND ICCOP
IOH
1.5V
IOL
EQUIVALENT CIRCUIT
NOTE: Test head capacitance.
Waveforms
tRL
tWRL
RD
tWRH
tAS
WR
A(2-0)
tWSA tWHA
tWSC tWHC
C(0-7)
C(0-7), A(0-2)
tROD
tRDO
FIGURE 49. TIMING RELATIVE TO WR
FIGURE 50. TIMING RELATIVE TO RD
tCP
tCL
tCH
CLKIN
tDS
tRF
tRF
2.0V
0.8V
tDH
IN(13:0), COF
GAINADJ(2:0), ENI,
COFSYNC, SYNCIN1
WR
tWC
FIGURE 51. OUTPUT RISE AND FALL TIMES
60
FIGURE 52. TIMING RELATIVE TO CLKIN
FN4450.4
May 1, 2007
HSP50214B
Waveforms
tPCP
tPCL
tPCH
PROCCLK
AGCGNSEL,
MCSYNC1
SOF, SOFSYNC,
SYNCIN2
OEAH, OEAL,
OEBH, OEBL
1.5V
AOUT(15:0),
BOUT(15:0), DATARDY,
INTRRP, MCSYNC0,
SYNCOUT, SEROUTA,
SEROUTB
1.5V
tOE
tOEBL
tOD
OUTA(15:8), OUTA(7:0),
OUTB(15:8), OUTB(7:0)
tDSS
tDHS
tDO_ OUT, tDO _ MCSYNCO,
tDO_ SYNCO
1.7V
SERSYNC
1.3V
tDOS
FIGURE 53. OUTPUT ENABLE/DISABLE
FIGURE 54. TIMING RELATIVE TO PROCCLK
tRCP
fRCP =
I
tRCP
tRCP Š≥ 2 tRCP
tRCH
tRCL
FIGURE 55. REFCLK
61
FN4450.4
May 1, 2007
HSP50214B
Metric Plastic Quad Flatpack Packages (MQFP/PQFP)
Q120.28x28 (JEDEC MO-108DA-1 ISSUE A)
D
120 LEAD METRIC PLASTIC QUAD FLATPACK PACKAGE
D1
-D-
INCHES
-A-
-B-
E E1
e
PIN 1
SEATING
A PLANE
-H-
0.10
0.004
0.40
0.016 MIN
-C-
5o-16o
0.20
M
0.008
C A-B S
0o MIN
A2 A1
0o-7o
L
MIN
MAX
MIN
MAX
NOTES
A
-
0.160
-
4.07
-
A1
0.010
-
0.25
-
-
A2
0.125
0.144
3.17
3.67
-
B
0.012
0.018
0.30
0.45
6
B1
0.012
0.016
0.30
0.40
-
D
1.219
1.238
30.95
31.45
3
D1
1.098
1.106
27.90
28.10
4, 5
E
1.219
1.238
30.95
31.45
3
E1
1.098
1.106
27.90
28.10
4, 5
L
0.026
0.037
0.65
0.95
N
120
120
e
0.032 BSC
0.80 BSC
7
Rev. 0 1/94
NOTES:
1. Controlling dimension: MILLIMETER. Converted inch
dimensions are not necessarily exact.
2. All dimensions and tolerances per ANSI Y14.5M-1982.
3. Dimensions D and E to be determined at seating plane -C- .
B
4. Dimensions D1 and E1 to be determined at datum plane
-H- .
B1
5. Dimensions D1 and E1 do not include mold protrusion.
Allowable protrusion is 0.25mm (0.010 inch) per side.
6. Dimension B does not include dambar protrusion. Allowable
dambar protrusion shall be 0.08mm (0.003 inch) total.
7. “N” is the number of terminal positions.
BASE METAL
WITH PLATING
SYMBOL
D S
0.13/0.17
0.005/0.007
5o-16o
MILLIMETERS
0.13/0.23
0.005/0.009
All Intersil U.S. products are manufactured, assembled and tested utilizing ISO9000 quality systems.
Intersil Corporation’s quality certifications can be viewed at www.intersil.com/design/quality
Intersil products are sold by description only. Intersil Corporation reserves the right to make changes in circuit design, software and/or specifications at any time without
notice. Accordingly, the reader is cautioned to verify that data sheets are current before placing orders. Information furnished by Intersil is believed to be accurate and
reliable. However, no responsibility is assumed by Intersil or its subsidiaries 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 Intersil or its subsidiaries.
For information regarding Intersil Corporation and its products, see www.intersil.com
62
FN4450.4
May 1, 2007