HSP50415 查看數據表（PDF） - Intersil
零件编号
产品描述 (功能)
比赛名单
HSP50415 Datasheet PDF : 29 Pages
| |||
HSP50415
Iout<3:0>:Qout<3:0>. See Figure 7 for a constellation
mapping example. For bit widths less than 4-bits the data in
the RAM may simply be zero’s for the unused bit positions
and the unused addresses since the HSP50415 will discard
the unused bits. For example, if the user programs the
number of bits to be 1 and the upper bits of the DIN<15:0>
bus are tied to “0”, the user need only program addresses 0,
1, 16 and 17 since the other addresses will never be
selected. In this example, the only data that is used will be
memory address bits 4 and 0 since these map to I<0> and
Q<0> respectively. For data bit widths larger than 4 bits or if
mapping is not required, the constellation mapper may be
bypassed.
Shaping FIR Filters
Following the constellation mapping, the I/Q data pair is
input to the programmable FIR filters for the first stage of
interpolation. The interpolating FIR filters’ have
programmable coefficients and must be loaded via the
microprocessor interface. The I and Q filter stages are
identical and may be loaded simultaneously or separately
thus allowing for different gains and responses through the
FIR filter if desired. The loading options are programmable
including readback modes and will be discussed in detail in
the ‘Microprocessor Interface’ section. Since the hardware
for the I and Q filters is identical, further discussion will
pertain to a single channel.
The basic interpolation rates allowed through the FIR are x4,
x8 or x16. An optional decimate by 2 mode is available that
subsamples the output of the filter thus reducing the
interpolation rate by a factor of 2. Each filter multiplication is
implemented as a series of shifts and adds thus constraining
the maximum input symbol rate as follows:
symbolRateMax is the smaller of:
(CLK * 2 * 2^twoBitMode) / (#bits * interpolationRate)
and CLK/4
where CLK is the final sample rate clock (100MHz max),
#bits is the data bit width of a single channel and twoBitMode
is a special processing mode where 2-bits at a time are
computed. The gain through the filter is:
A = (sum of coefficients) / interpolation rate
The FIR filter contains saturation logic in the event that the
final output peaks over 1.0. Table 3 outlines the filter
characteristics for the various interpolation rates.
TABLE 3. FIR FILTER CHARACTERISTICS
2-BITMODE
INTERP.
RATE SYMBOL SPAN # FILTER TAPS
0
x4
24
96
0
x8
20
160
0
x16
16
256
1
x4
12
48
1
x8
10
80
1
x16
8
128
The programmable coefficients are stored in RAM as bit-
sliced sums of products.
The data exits the interpolating FIR filters as a parallel
I<15:0> and Q<15:0> data stream at the interpolated sample
rate. These filters may be totally bypassed if higher input
symbol rates are required. When bypassed, the RAMs may
be loaded with all zeros for power conservation.
Post FIR Gain Control
Following the FIR filter pair is a gain stage where I and Q are
scaled equally. The programmable gain consists of a 6-bit
mantissa and a 4-bit exponent stage. The equation for the
gain is as follows:
dataOut<15:0> =
(dataIn<15:0> * 1.MMMMMM) * 2 ^ (EEEE - 11)
where MMMMMM denotes the 6-bit gain value and EEEE is
the 4-bit shift value.
For a gain of 1.0 through this stage, program the mantissa to
0x00 and the exponent to 0xB. This stage is implemented
with a signed 16-bit by unsigned 7-bit multiplier with the
DATA FROM FIFO
16 “0x0008”
I/
16 “0x000A”
Q/
CONSTELLATION MAPPER
MAP
ADDRESS
Address
0x00
LOOK-UP 16
TABLE RAM /
8
“0x8A”
/
0x9B
16
/
Address 0x8A is formed
from the lower 4-bits of
I and Q data
0xFF
DATA TO SHAPING FILTERS
“0x0009”
I
“0x000B”
Q
Address 0x8A is previously
loaded with data 0x9B via
the control bus
FIGURE 7. CONSTELLATION MAPPING
10
FN4559.6
April 23, 2007
Share Link: 