QL6325E Eclipse-E Data Sheet
••••••
FPGA Combining Performance, Density, and Embedded RAM
Device Highlights
Flexible Programmable Logic
• 0.18 µm six layer metal CMOS process
• 1.8/2.5/3.3 V drive capable I/O
• 1536 logic cells
• Up to 4,002 flip-flops
• Up to 310 I/O pins
• Up to 335 user-available pins
• 320,460 maximum system gates
Advanced Clock Network
• Nine global clock networks:
One dedicated
Eight programmable
• 20 quad-net networks—five per quadrant
• 16 I/O controls—two per I/O bank
• Four phase locked loops
Embedded Computational Units
Twelve ECUs provide integrated Multiply, Add, and
Accumulate functions.
Embedded Dual Port SRAM
• Twenty-four 2,304-bit dual port high
performance SRAM blocks
• 55,296 RAM bits
• RAM/ROM/FIFO wizard for automatic
configuration
• Configurable and cascadable
Security Features
The QuickLogic products come with secure
ViaLink technology that protects intellectual
property from design theft and reverse engineering.
No external configuration memory needed;
instant-on at power-up.
Figure 1: QL6325E Eclipse-E Block Diagram
PLL
Embedded RAM Blocks
Embeded Computational Units
PLL
Programmable I/O
• High performance I/O cell
• Programmable slew rate control
• Programmable I/O standards:
LVTTL, LVCMOS, LVCMOS18, PCI, GTL+,
SSTL2, and SSTL3
Eight independent I/O banks
Three register configurations: Input, Output,
and Output Enable
Fabric
PLL
Embedded RAM Blocks
PLL
© 2005 QuickLogic Corporation
www.quicklogic.com
•
•
•
•
•
•
1
QL6325E Eclipse-E Data Sheet Rev. F
QuickWorks Design Software
The QuickWorks package provides the most complete ESP and FPGA software solution from design entry
to logic synthesis, to place and route, to power calculation, and simulation. The package provides a solution
for designers who use third-party tools from Cadence, Mentor, OrCAD, Synopsys, Viewlogic, and other third-
party tools for design entry, synthesis, or simulation.
Programmable Logic Architectural Overview
The Eclipse-E logic cell structure is presented in
Figure 2.
This architectural feature addresses today's register-
intensive designs.
Table 1: Performance Standards
Function
Multiplexer
Parity Tree
Counter
Description
16:1
24
36
16 bit
32 bit
128 x 32
FIFO
Clock-to-Out
System clock
128 x 64
256 x 16
Slowest Speed Grade
2.8 ns
3.4 ns
4.6 ns
275 MHz
250 MHz
197 MHz
188 MHz
208 MHz
4 ns
200 MHz
Fastest Speed Grade
2.4 ns
2.9 ns
3.9 ns
328 MHz
300 MHz
235 MHz
266 MHz
248 MHz
3.3 ns
300 MHz
The Eclipse-E logic cell structure presented in
Figure 2
is a dual register, multiplexor-based logic cell. It is
designed for wide fan-in and multiple, simultaneous output functions. Both registers share CLK, SET, and
RESET inputs. The second register has a two-to-one multiplexer controlling its input. The register can be
loaded from the NZ output or directly from a dedicated input.
NOTE:
The input PP is not an “input” in the classical sense. It is a static input to the logic cell and selects
which path (NZ or PS) is used as an input to the Q2Z register. All other inputs are dynamic and can
be connected to multiple routing channels.
The complete logic cell consists of two six-input AND gates, four two-input AND gates, seven two-to-one
multiplexers, and two D flip-flops with asynchronous SET and RESET controls. The cell has a fan-in of 30
(including register control lines), fits a wide range of functions with up to 17 simultaneous inputs, and has six
outputs (four combinatorial and two registered). The high logic capacity and fan-in of the logic cell
accommodates many user functions with a single level of logic delay while other architectures require two or
more levels of delay.
•
2
•
www.quicklogic.com
•
•
•
•
© 2005 QuickLogic Corporation
QL6325E Eclipse-E Data Sheet Rev. F
Figure 2: Eclipse-E Logic Cell
QS
A1
A2
A3
A4
A5
A6
OS
OP
B1
B2
C1
C2
MP
MS
D1
D2
E1
E2
NP
NS
F1
F2
F3
F4
F5
F6
PS
PP
QC
QR
AZ
OZ
QZ
NZ
Q2Z
FZ
RAM Modules
The QL6325E includes 24 dual-port 2,304-bit RAM modules for implementing RAM, ROM, and FIFO
functions. Each module is user-configurable into four different block organizations and can be cascaded
horizontally to increase their effective width, or vertically to increase their effective depth as shown in
Figure 4.
Figure 3: 2,304-bit RAM Module
MODE[1:0]
WA[9:0]
WD[17:0]
WE
WCLK
ASYNCRD
RA[9:0]
RD[17:0]
RE
RCLK
Using the two “mode” pins, designers can configure each module into 128 x 18 and 256 x 9. The blocks are
also easily cascadable to increase their effective width and/or depth (see
Figure 4
)
.
© 2005 QuickLogic Corporation
www.quicklogic.com
•
•
•
•
•
•
3
QL6325E Eclipse-E Data Sheet Rev. F
Figure 4: Cascaded RAM Modules
WDATA
WADDR
RAM
Module
(2,304 bits)
RDATA
RADDR
WDATA
RAM
Module
(2,304 bits)
RDATA
The RAM modules are dual-port, with completely independent READ and WRITE ports and separate READ
and WRITE clocks. The READ ports support asynchronous and synchronous operation, while the WRITE
ports support synchronous operation. Each port has 18 data lines and 8 address lines, allowing word lengths
of up to 18 bits and address spaces of up to 256 words. Depending on the mode selected, however, some
higher order data or address lines may not be used.
The Write Enable (WE) line acts as a clock enable for synchronous write operation. The Read Enable (RE) acts
as a clock enable for synchronous READ operation (ASYNCRD input low), or as a flow-through enable for
asynchronous READ operation (ASYNCRD input high).
Designers can cascade multiple RAM modules to increase the depth or width allowed in single modules by
connecting corresponding address lines together and dividing the words between modules.
A similar technique can be used to create depths greater than 256 words. In this case address signals higher
than the MSB are encoded onto the write enable (WE) input for WRITE operations. The READ data outputs
are multiplexed together using encoded higher READ address bits for the multiplexer SELECT signals.
The RAM blocks can be loaded with data generated internally (typically for RAM or FIFO functions) or with
data from an external PROM (typically for ROM functions).
Embedded Computational Unit (ECU)
Traditional Programmable Logic architectures do not implement arithmetic functions efficiently or effectively—
these functions require high logic cell usage while garnering only moderate performance results.
The Eclipse-E architecture allows for functionality above and beyond that achievable using programmable logic
devices. By embedding a dynamically reconfigurable computational unit, the Eclipse-E device can address
various arithmetic functions efficiently. This approach offers greater performance and utilization than
traditional programmable logic implementations. The embedded block is implemented at the transistor level
as shown in
Figure 5.
•
4
•
www.quicklogic.com
•
•
•
•
© 2005 QuickLogic Corporation
QL6325E Eclipse-E Data Sheet Rev. F
Figure 5: ECU Block Diagram
RESET
D
S1
S2
S3
CIN
SIGN1
SIGN2
00
01 3-1
mux
10
Q[16:0]
3-4
decoder
C
B
A
A[7:0]
A[15:8]
8-bit
Multiplier
2-1
mux
16-bit
Adder
D
Q
17-bit
Register
A[0:15]
CLK
B[0:15]
2-1
mux
The Eclipse-E ECU block (Table
2)
is placed next to the SRAM circuitry for efficient memory/instruction fetch
and addressing for DSP algorithmic implementations.
Table 2: Eclipse-E ECU Block
Device
QL6325E
ECUs
12
Up to twelve 8-bit MAC functions can be implemented per cycle for a total of 1.2 billion MACs/s when clocked
at 100 MHz. Additional multiply-accumulate functions can be implemented in the programmable logic.
The modes for the ECU block are dynamically re-programmable through the programmable logic.
© 2005 QuickLogic Corporation
www.quicklogic.com
•
•
•
•
•
•
5
京公网安备 11010802033920号
EEWORLD
