or product names are trademarks or registered trademarks of their respective holders. The specifications and information herein are subject to change without notice.
1
ispMACH 4064ZE
64
4.7
2.5
3.2
241
1.8V
32+4
48+4
48+4
64+10
ispMACH 4128ZE
128
5.8
2.9
3.8
200
1.8V
ispMACH 4256ZE
256
5.8
2.9
3.8
200
1.8V
32
4.4
2.2
3.0
260
1.8V
32+4
32+4
64+10
96+4
96+4
64+10
96+14
108+4
64+10
96+4
www.latticesemi.com
1
DS1022_01.8
ispMACH 4000ZE Family Data Sheet
Introduction
The high performance ispMACH 4000ZE family from Lattice offers an ultra low power CPLD solution. The new fam-
ily is based on Lattice’s industry-leading ispMACH 4000 architecture. Retaining the best of the previous generation,
the ispMACH 4000ZE architecture focuses on significant innovations to combine high performance with low power
in a flexible CPLD family. For example, the family’s new Power Guard feature minimizes dynamic power consump-
tion by preventing internal logic toggling due to unnecessary I/O pin activity.
The ispMACH 4000ZE combines high speed and low power with the flexibility needed for ease of design. With its
robust Global Routing Pool and Output Routing Pool, this family delivers excellent First-Time-Fit, timing predictabil-
ity, routing, pin-out retention and density migration.
The ispMACH 4000ZE family offers densities ranging from 32 to 256 macrocells. There are multiple density-I/O
combinations in Thin Quad Flat Pack (TQFP), Chip Scale BGA (csBGA), and Ultra Chip Scale BGA (ucBGA) pack-
ages ranging from 32 to 144 pins/balls. Table 1 shows the macrocell, package and I/O options, along with other key
parameters.
A user programmable internal oscillator and a timer are included in the device for tasks like LED control, keyboard
scanner and similar housekeeping type state machines. This feature can be optionally disabled to save power.
The ispMACH 4000ZE family has enhanced system integration capabilities. It supports a 1.8V supply voltage and
3.3V, 2.5V, 1.8V and 1.5V interface voltages. Additionally, inputs can be safely driven up to 5.5V when an I/O bank
is configured for 3.3V operation, making this family 5V tolerant. The ispMACH 4000ZE also offers enhanced I/O
features such as slew rate control, PCI compatibility, bus-keeper latches, pull-up resistors, pull-down resistors,
open drain outputs and hot socketing. Pull-up, pull-down and bus-keeper features are controllable on a “per-pin”
basis. The ispMACH 4000ZE family members are 1.8V in-system programmable through the IEEE Standard 1532
interface. IEEE Standard 1149.1 boundary scan testing capability also allows product testing on automated test
equipment. The 1532 interface signals TCK, TMS, TDI and TDO are referenced to V
CC
(logic core).
Overview
The ispMACH 4000ZE devices consist of multiple 36-input, 16-macrocell Generic Logic Blocks (GLBs) intercon-
nected by a Global Routing Pool (GRP). Output Routing Pools (ORPs) connect the GLBs to the I/O Blocks (IOBs),
which contain multiple I/O cells. This architecture is shown in Figure 1.
Figure 1. Functional Block Diagram
CLK0/I
CLK1/I
CLK2/I
CLK3/I
V
CCO0
GND
V
CCO1
GND
I/O
Block
16
ORP
GOE0
GOE1
V
CC
GND
OSC
I/O
Block
ORP
16
Global Routing Pool
Generic
Logic
Block
16
36
16
36
Generic
Logic
Block
I/O Bank 0
TCK
TMS
TDI
TDO
I/O
Block
ORP
16
Generic
Logic
Block
16
36
16
36
Generic
Logic
Block
I/O
Block
16
ORP
2
I/O Bank 1
ispMACH 4000ZE Family Data Sheet
The I/Os in the ispMACH 4000ZE are split into two banks. Each bank has a separate I/O power supply. Inputs can
support a variety of standards independent of the chip or bank power supply. Outputs support the standards com-
patible with the power supply provided to the bank. Support for a variety of standards helps designers implement
designs in mixed voltage environments. In addition, 5V tolerant inputs are specified within an I/O bank that is con-
nected to a V
CCO
of 3.0V to 3.6V for LVCMOS 3.3, LVTTL and PCI interfaces.
Architecture
There are a total of two GLBs in the ispMACH 4032ZE, increasing to 16 GLBs in the ispMACH 4256ZE. Each GLB
has 36 inputs. All GLB inputs come from the GRP and all outputs from the GLB are brought back into the GRP to
be connected to the inputs of any other GLB on the device. Even if feedback signals return to the same GLB, they
still must go through the GRP. This mechanism ensures that GLBs communicate with each other with consistent
and predictable delays. The outputs from the GLB are also sent to the ORP. The ORP then sends them to the asso-
ciated I/O cells in the I/O block.
Generic Logic Block
The ispMACH 4000ZE GLB consists of a programmable AND array, logic allocator, 16 macrocells and a GLB clock
generator. Macrocells are decoupled from the product terms through the logic allocator and the I/O pins are decou-
pled from macrocells through the ORP. Figure 2 illustrates the GLB.
Figure 2. Generic Logic Block
CLK0
CLK1
CLK2
CLK3
To GRP
Clock
Generator
1+OE
16 MC Feedback Signals
1+OE
1+OE
1+OE
1+OE
1+OE
1+OE
1+OE
To Product Term
Output Enable Sharing.
Also, To Input Enable of
Power Guard on I/Os
in the block.
Logic Allocator
AND Array
36 Inputs,
83 Product Terms
16 Macrocells
36 Inputs
from GRP
AND Array
The programmable AND Array consists of 36 inputs and 83 output product terms. The 36 inputs from the GRP are
used to form 72 lines in the AND Array (true and complement of the inputs). Each line in the array can be con-
nected to any of the 83 output product terms via a wired-AND. Each of the 80 logic product terms feed the logic
allocator with the remaining three control product terms feeding the Shared PT Clock, Shared PT Initialization and
Shared PT OE. The Shared PT Clock and Shared PT Initialization signals can optionally be inverted before being
fed to the macrocells.
Every set of five product terms from the 80 logic product terms forms a product term cluster starting with PT0.
There is one product term cluster for every macrocell in the GLB. Figure 3 is a graphical representation of the AND
Array.
3
To ORP
ispMACH 4000ZE Family Data Sheet
Figure 3. AND Array
In[0]
In[34]
In[35]
PT0
PT1
PT2
PT3
PT4
Cluster 0
PT75
PT76
PT77
Cluster 15
PT78
PT79
PT80 Shared PT Clock
PT81 Shared PT Initialization
PT82 Shared PTOE/BIE
Note:
Indicates programmable fuse.
Enhanced Logic Allocator
Within the logic allocator, product terms are allocated to macrocells in product term clusters. Each product term
cluster is associated with a macrocell. The cluster size for the ispMACH 4000ZE family is 4+1 (total 5) product
terms. The software automatically considers the availability and distribution of product term clusters as it fits the
functions within a GLB. The logic allocator is designed to provide two speed paths: 20-PT Speed Locking path and
an up to 80-PT path. The availability of these two paths lets designers trade timing variability for increased perfor-
mance.
The enhanced Logic Allocator of the ispMACH 4000ZE family consists of the following blocks:
• Product Term Allocator
• Cluster Allocator
• Wide Steering Logic
Figure 4 shows a macrocell slice of the Logic Allocator. There are 16 such slices in the GLB.
Figure 4. Macrocell Slice
to to
n-1 n-2
from from
n-1 n-4
From
n-4
n
5-PT
1-80
PTs
To XOR (MC)
Cluster
to
n+1
Individual Product
Term Allocator
from from
n+2 n+1
Cluster
Allocator
To
n+4
SuperWIDE™
Steering Logic
4
ispMACH 4000ZE Family Data Sheet
Product Term Allocator
The product term allocator assigns product terms from a cluster to either logic or control applications as required
by the design being implemented. Product terms that are used as logic are steered into a 5-input OR gate associ-
ated with the cluster. Product terms that used for control are steered either to the macrocell or I/O cell associated
with the cluster. Table 2 shows the available functions for each of the five product terms in the cluster.
Table 2. Individual PT Steering
Product Term
PT
n
PT
n
+1
PT
n
+2
PT
n
+3
PT
n
+4
Logic
Logic PT
Logic PT
Logic PT
Logic PT
Logic PT
Single PT for XOR/OR
Individual Clock (PT Clock)
Individual Initialization or Individual Clock Enable (PT Initialization/CE)
Individual Initialization (PT Initialization)
Individual OE (PTOE)
Control
Cluster Allocator
The cluster allocator allows clusters to be steered to neighboring macrocells, thus allowing the creation of functions
with more product terms. Table 3 shows which clusters can be steered to which macrocells. Used in this manner,
the cluster allocator can be used to form functions of up to 20 product terms. Additionally, the cluster allocator
accepts inputs from the wide steering logic. Using these inputs, functions up to 80 product terms can be created.
Table 3. Available Clusters for Each Macrocell
Macrocell
M0
M1
M2
M3
M4
M5
M6
M7
M8
M9
M10
M11
M12
M13
M14
M15
—
C0
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
C14
C0
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
C14
C15
Available Clusters
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
C14
C15
—
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
C14
C15
—
—
Wide Steering Logic
The wide steering logic allows the output of the cluster allocator n to be connected to the input of the cluster alloca-
tor
n+4.
Thus, cluster chains can be formed with up to 80 product terms, supporting wide product term functions
and allowing performance to be increased through a single GLB implementation. Table 4 shows the product term
编译自semiengineering 业界越来越关注人工智能的功耗问题，但这个问题并没有简单的解决方案。这需要深入了解应用、半导体和系统层面的软件和硬件架构，以及所有这些的设计和实现方式。每个环节都会影响总功耗和提供的效用。这是最终必须做出的权衡。 但首先，必须解决效用问题。电力是否被浪费了？“我们将电力用于有价值的用途，”Ansys（现为新思科技旗下公司）产品营销总监 Marc Swi...[详细]
能源基础设施
京公网安备 11010802033920号
Copyright © 2005-2026 EEWORLD.com.cn, Inc. All rights reserved
电子工程世界
