Load Balancing for Spatial-grid-based Parallel Numeric Simulations on Clusters
of SMPs
Huaien Gao Andreas Schmidt Amitava Gupta Peter Luksch
Technical University, Munich, Germany
Emails : {gao, schmiand, gupta, Peter.Luksch}@in.tum.de
Abstract
Load distribution is an essential factor to parallel efficiency of numerical simulations that are based on spatial grids, especially on clusters of symmetric multiprocessors (SMPs). This paper presents a method of mapping spatial grid nodes to processors that combines two load balancing methodologies, graph partitioning and graph matching, to achieve maximum parallel efficiency on SMP clusters. The method has been successfully applied to load distribution in a parallel Computational Fluid Dynamics (CFD) simulation. Test runs on a PC cluster prove the effectiveness of our method.
1. Introduction
Scientific simulations in many domains, e.g., CFD, Computational Structural Mechanics (CSM), VLSI simulations, are based on the concept of a spatial grid. A natural way to execute this type of applications in parallel is to follow the Single Program Multiple Data (SPMD) approach, i.e., to distribute the spatial grid onto multiple processes, each of which is assigned a partition of the grid.
Load balancing is a key factor in achieving high parallel efficiency, especially on platforms with a large number of nodes. The amount of work assigned to each processor has to be determined such that the turnaround time is minimized. For a parallel application running on a large number of processors, the turnaround time is defined as the maximum of all the times taken by the individual processors to complete the task. For a given problem, there also is an optimal number of processors for which the turnaround time is minimized. Knowing this number is important in order to maximize throughput. For instance, crash simulations usually involve parameter studies, where a large number of simulation jobs are executed with the same model, each with different parameter settings. Using more than the optimal number of processors for a single job would result in a waste of resources, because faster execution would be possible with fewer processors. With fewer processors per job, more jobs can run in parallel on a large cluster.
Approaches to task assignment in distributed systems can be classified into three broad categories, namely graph-theoretic ([1], [2] as examples), mathematical programming ([3], [4]) and heuristic [5]. Graph-theoretic algorithms view the task as a graph representing the inter-modular dependencies and apply graph partitioning methodologies to obtained equal partitions of the task with the inter-node communication or the volume of such communication minimized. A concise tool for this purpose is Metis [6]. With the mathematical programming approach the problem is viewed as an optimisation problem and solved using mathematical programming techniques. Heuristic methods provide fast but often sub-optimal solutions within a finite time, where an optimal solution cannot be obtained within a finite time.
Shen and Tsai propose a method in [7], where the problem of load balancing by optimal task assignment is viewed as a graph matching problem. The task is represented as a task graph with each module represented by a vertex and the communication between these modules represented by edges. The weights associated with the vertices represent the computation cost associated with each module and the weights associated with each edge represents the communication cost for the interaction between two adjacent vertices of the task graph. Both these costs are expressed in terms of time. The connectivity between the processors is viewed as a
processor-graph where each vertex represents a node and each edge represents a communication channel between two adjacent nodes. Each graph match corresponds to a specific task assignment. Assuming that each module can require different processing time on different nodes, and the communication links between the different nodes have different speeds, optimal task assignment for a task comprising of N modules on M processors would have to explore M N possibilities, in the worst case. The work done by Shen and Tsai uses a heuristic approach based on A* algorithm to reduce this search to obtain an optimal solution. The method can be used for any topology, a cluster of SMPs as an example, as it can tackle heterogeneity. This method has a complexity that is exponential in the number of nodes. Hence it cannot be applied directly to real-world applications where the number of interacting modules is huge. In addition, SMP clusters have strongly connected processor graphs, which results in a huge number of searches.
There are also problems to use graph partitioning tools like Metis to obtain optimal task assignment on a cluster of SMPs. Metis, for example, can only be applied for homogeneous platform and cannot be applied to a cluster of SMPs directly, as the heterogeneity in terms of difference in link speeds across two processors located in the same node and two processors located on two different nodes cannot be represented.
This paper puts forward a methodology by which an optimal task assignment on a cluster of SMPs is obtained in two steps using a hybrid methodology comprising of first graph partitioning and then graph matching. Graph partitioning is used to assign a set of modules to each SMP node of the cluster. Subsequently, graph matching is used to assign these modules to individual processors of each SMP node. The results have been tested using a representative application comprising of sixty-four modules on a cluster comprising of four two way SMPs. The results establish this methodology as an efficient means of obtaining optimal task assignment on a cluster of SMPs for applications comprising of large number of modules on homogeneous clusters comprising of large number of nodes, each node consisting of finite number of processors.
2. A test case with graph matching and
graph partitioning
To study the effectiveness of a graph matching based method, it was first applied to a task comprising of 10 modules defined by the set T = {0,1,2…9} on a set of identical nodes (each comprising of one processor) defined by the set P = {a, b, c, d, e} such that the turnaround time is minimized. The computation time associated with these modules on any node is defined by the set TP= {0.26,0.4,0.4,0.4,0.4,0.4,0.4,0.4,0.4,0.26} with the i th element in TP denoting the computation time associated with the i th element in T. The inter module communication is defined by the matrix.
For the graph matching based optimal task assignment problem, the methodology used by Shen and Tsai in [7] was used.
The A* based search algorithm starts with an initial partial mapping 0a meaning module 0 is mapped on node a. The algorithm then generated successive partial maps with successive tasks mapped on any of the nodes and applied the A* based search on the state space so generated. The n th element in the state space is associated with a cost f(n) = g(n) + h(n) where the cost functions g(n) and h(n) are as defined in [7] . Typically g(n) represents the minimum path cost from the start element to the element n in the state-space and h(n) is the lower-bound estimate, using any heuristic information available of the minimum path cost from the element n to the goal element in the state-space.
Since all nodes are identical, it is wise to start from an initial element with any of the tasks mapped on any of the processors, so as to reduce the number of elements generated in the state space. For a perfectly homogeneous cluster, comprising of M nodes, there would be M different optimal mappings, all with the same turnaround time. The optimal task assignment for the present case, found out by this method, is 0a 1d 2e 3b 4b 5c 6c 7e 8a 9d with an estimated turnaround time of 0.98 sec. The algorithm expanded 4870 elements and generated more than 19,000 elements.
This same methodology is then repeated to obtain an optimal assignment for a task comprising of 64 modules on a cluster of four identical single processor nodes, each module assumed to be taking the same amount of processing time and the communication between the modules assumed to be in the form of a ring such that module 0 communicates with module 1 and module 63, module 63 communicates with module 62 and module 0, and any intermediate module i communicates with module i-1 and i+1. It was found that the graph matching based methodology failed to find an optimal task
assignment even after generating as many as 90,000 elements in the state space. This is because of the NP complete nature of the problem . Accordingly, the graph matching algorithm tends to have an exponential complexity when the number of modules increases[7].
Following this, graph partitioning is applied. For this purpose ready-made graph partitioning routines provided by Metis is used. Metis produces the mapping presented in fig. 1
Fig. 1 An optimal task assignment using Metis
In fig. 1, the edge (a,b) maps edges (16,17) and (17,16), the edge (b,c) maps edges (32,33) and (33,32), the edge (c,d) maps edges (48,49) and (49,48) while the edge (d,a) maps the edges (0,1) and (1,0). As with graph matching, the computation weight associated with each module was assumed to be equal (i.e. 0.4 sec.) and the communication weight associated with each edge of the task graph was assumed to be 0.03 sec. The results are in agreement with the expected values.
While using Metis, various partitioning objectives are used. A partitioning with edge-cut minimization objective aims to minimize the inter-node communication by minimizing the number of cut-edges. Thus for the task assignment shown in fig. 1, the sum of the weights of the edges (a, b), (b, c), (c, d) and (d, a) is minimum. This requires that weights be assigned to each edge in the task graph. An alternative partitioning objective is minimisation of the volume of communication, and this requires two weights to be associated with each vertex of the task graph viz. a computation weight denoting the processing time required by the module and the communication weight denoting the sum of the weight of edges originating from the vertex. This objective partitions the task graph in such a manner such that the sum of the communication weights of the straddle vertices in the partition (i.e. vertices which are associated with one or more edges between two partitions). Either way, the result is the same for the present case. However, for a
general case, for most graphs, partitioning with the objective to minimize the total volume of communication yields better results [6].
Therefore, the methodology to obtain an optimal task assignment with N number of modules on a cluster of M number of P way SMPs essentially starts with the
application of graph partitioning to obtain N equal partitions with the volume of communication minimised. Once this is done, it is possible to obtain N balanced
partitions of the original task each of which comprises of smaller number of modules ( << N ) and on each of these
sub-graphs graph matching can be applied to obtain an optimal task assignment for each node, which will produce an optimal turnaround time for the entire task. 3. Using the SMP advantage
Taking the specific example of the task with 64 modules discussed in the previous example, to obtain the
optimal task assignment on four processors, for example, for each of the nodes a, b, c and d, the next step would be to treat each of these sub-graphs as a task comprising
of a finite number of modules and find an optimal task assignment on four identical processors in each case. A single step partitioning into 4P number of processors
with Metis, for example, does not yield the optimal task assignment for the cluster of SMPs, as the communication costs within a node is much less compared to the
communication costs between two nodes.
Of the two approaches available, graph matching is
first explored and then the results are compared to graph partitioning. For graph matching, basically, the methodology proposed by Shen and Tsai is followed with
some modifications. As in [7], a partial mapping S is defined as a set of tuples, s1, s2, s3,……s n , each being of the form {x,y} , x ∈ T′, y = m(x) , y ∈ [0,1,2,…P-1]. T′ is the set of modules assigned to the node in question and m(x) is the function which maps a module x on to a
processor. The goal is the mapping in which all elements of T′ are mapped on one of the P processors with the minimum turnaround time for the node in question. For each partial mapping, described in [7], the values of g(n) and h(n) are computed as follows:-
(i) Computation of g(n)
For a partial mapping S, let
X be the ordered set of processors , ordered in the
ascending order of computation times;
TP i be the computation time spent by the processor corresponding to the i th element in X ;
TC i be the time spent by the same processor in communication;
Then, the turnaround time TA i for the i th processor in X is defined as
TP i + TC i , if i =0;
TA i =
max((TP i + TC i),( TA i-1 + TC i)),i≠0; .(3.1) g(n)= TA p-1, for a p processor system;(3.2)
Since X is an ordered set, the turnaround time for the mapping S , being the maximum of the turnaround times of all processors in X, is the turnaround time of the processor which finishes its computation last, i.e. the processor corresponding to the last element in X. This is the value of g(n).
Equation 3.2 takes into account the fact that in a SMP node comprising of P processors, the communication executed by the processors with modules on other nodes is essentially serialized ( since each SMP node has only one network interface), while the computations proceed in parallel. Thus, the maximum time, a process can take to complete its computation and communication is represented by the second term of equation 3.1 assuming that the sum of its computation and communication times is less than this.
(ii) Computation of h(n)
)
(=
n
h
(3.3)
The result represented by equation (3.3) follows from the analysis presented in [7] and hence, this is a specific case, where the lower bound of h(n) has a value zero. Experiments with various configurations have shown that this methodology is capable of finding an optimal distribution up to a maximum of about 20 modules per node. The modification in the computation of g(n) is necessary to calculate the cost of mapping taking into account the fact that even though the processors compute in parallel, they can only communicate serially with modules located on other nodes and therefore equation 3.2 represents the exact calculation of the turnaround time of a processor.
For the task described in Appendix, vector A stores the adjacency information and communication weights associated with the edges in tuples representation. Each edge between a vertex i and vertex j with a transfer of w bytes is represented by the tuple {i,j,w} .The vector B stores the computation weights associated with each module in terms of the number of units of computation. The actual weights used for partitioning are computed using representative cluster specific data for computation and communication defined by COMP_WORTH and COMM_WORTH respectively in Appendix. Table 1 shows the distribution of the modules on each of the eight
nodes designated as a, b, c, d, e, f, g & h.
Table.1 Assignment on 8 single nodes for the
case depicted in Appendix using Metis
Node Modules
Turnaround
Time (secs.) a
22,25, 26, 30, 48, 51, 52, 53,
55, 56, 57, 58, 59
20.82
b
20, 21, 23, 24, 44, 45, 46, 47,
54, 60
22.76
c 0, 13, 41, 42, 43, 63 23.28
d 5, 12, 14, 15 24.72
e 3, 27, 29, 32, 35, 37, 38, 40 20.38
f
4, 19, 28, 31, 33, 34, 36, 39,
50
22.55
g 2,10, 11, 16, 49, 61, 62 24.23
h 1, 6, 7, 8, 9, 17, 18 25.12
For the test case shown, the turnaround times for each individual node, assumed to be comprised of 1 processor
is denoted against each processor in table 1.
Assuming that each of these nodes is comprised of
two processors each, the detail of the optimal distribution
of a cluster of eight two-way SMPs is as shown in table 2.
In table 2, and also in table 3, each node is assumed to consist of two processors viz. Processor 0 and Processor
1. The turnaround time for each node is also indicated against each distribution taking into consideration the fact
that the communication to modules located on other nodes
is serialized for all modules on the same SMP node.
Table 2. Optimal distribution on a cluster of eight two way SMPs using graph matching
Node Processor 0 Processor 1
Node
turnaround
time (sec.)
a 22,25,55,56,58
26,30,48,51,52,53,
57,59
12.05
b 20,21,23
24,44,45,46,47,54,
60
13.82
c 0,63 13,41,42,43 14.33
d 5 12,14,15
15.43
e 3,35,38 27,29,32,37,40 12.00
f 4,19,28 31,33,34,36,39,50 13.39
g 2,10,11,61 16,49,62 14.73
h 1,7,8,9,18 6,17 15.44
To understand the effectiveness of the assignment, it is
now attempted to obtain an optimal task assignment on
the same cluster using graph partitioning using a standard
graph partitioning tool like Metis. Table 3 shows the details of the distribution. For this distribution, each module is assumed to be represented by a vertex with a weight, which is equal to the sum of the computation time
of the module for one iteration plus the time spent by the module in communicating with other modules located in other nodes. The idea is to find out partitions with equal turnaround times .
Table 3. Optimal distribution on a cluster of eight
two way SMPs using graph partitioning
Node Processor 0 Processor 1 Node
turnaround
time (sec.)
a 22,30,51,52 25,26,48,53,55,
56,57,58,59
13.16 b
20,21,23,54,64 24,44,45,46,47 15.37 c
0 13,41,42,43,63 14.89 d
5 12,14,15 15.43 e
3,27,29,32 35,37,38,40 13.41 f
4,28,36,50 19,31,33,34,39 14.80 g
2,10,11,16,62 49,61 16.41 h
17,18 1,6,7,8,9 16.02
Tables 2 and 3 clearly establish the effectiveness of
the hybrid methodology comprising of graph partitioning and graph matching as compared to graph partitioning for obtaining an optimal task assignment on a cluster of SMPs. It is clearly seen that using the hybrid partitioning methodology, the turnaround time for the entire task is 15.44 sec. as compared to 16.41 sec. obtained by a two-step graph partitioning. This is true for most cases as the hybrid methodology takes into account the interleaving of computation and communication amongst the processors of a SMP node.
4. Results and discussion
To test the efficiency of the proposed methodology, a sample parallel CFD code was used. The computation and communication times associated with the various modules of this code and the interaction between the modules is shown in Appendix.
The test bed comprised of a 6 node SMP cluster of which 4 nodes were used. Every node is equipped with 2 Intel Xeon Processor clocked at 450 MHz. Each node has 512 MB memory resulting in a total of 3 GB RAM for the whole cluster. All 6 nodes are connected using fast-Ethernet PCI-boards and a dedicated fast-Ethernet switch. For comparison of the results two cases were considered. In the first case, a non-optimised assignment of modules to processors was considered. This meant that modules were loaded consecutively on the processors 0,1,2,….8, considering on 4 nodes of the cluster. In the second case, an optimal task assignment was obtained using the methodology presented in sections 2.0 and 3.0. Table 4 represents the results.
Table 4. Comparison of results: with and without optimal assignment of a cluster of four two way SMPs
OD # ND
Node No.
Time*
P 0 P 1 P 0 P 1 TP
246.51 342.65 577.18 296.43 TC 195.62 91.01 67.49 363.48 1 TA 465.19 465.13 685.46 685.42 TP 318.25 323.90 307.86 232.15 TC 110.74 107.87 349.37 426.17 2 TA 465.12 465.10 685.33 685.33 TP 346.14 285.71 510.43 216.58 TC 79.38 144.91 143.87 445.68 3 TA 465.11 465.10 685.39 685.39 TP 252.87 368.36 220.73 129.20 TC 169.93 51.24 445.85 542.12 4 TA 465.17 465.11 685.39 685.34 * Time for 50 iterations (sec.) # OD stands for Optimised Distribution, ND means Non-optimised Distribution. P0 and P1 refer to the processor number 0 and 1 respectively for each node.
Thus it is clearly seen that the hybrid methodology developed is extremely effective for obtaining the optimal task assignment on a cluster of SMPs and produces a 47 % speed up for the present application. Furthermore, the task assignment algorithm takes less than a second for the test case in question, which requires normally two to three thousand iterations to converge at the minimum. Hence the assignment module can easily be integrated as a pre-processor, without causing any appreciable increase in the run time of the application.
5. Conclusion
The methodology presented in this paper combines two diverse load-balancing strategies to obtain an optimal task assignment on a cluster of SMPs. It uses graph matching as a complement to graph partitioning to address the limitation of each. Graph partitioning tools suffer from the inability of being used directly to obtain an optimal task assignment on a cluster of SMPs as they cannot tackle heterogeneity presented even by a homogeneous cluster of SMPs. The graph matching algorithms, on the other hand can tackle this homogeneity, but cannot be used when the number of modules is large. The methodology of using graph matching on smaller size partitions produced by graph partitioning proves to be very effective. The methodology can be used for obtaining efficient assignments, which optimise turnaround times for fairly large parallel applications on a homogeneous cluster of SMPs within a finite time.
6. Acknowledgements
This work is funded by the Bavarian Science Foundation within the scope of Competence Network for Technical, Scientific High Performance Computing in Bavaria, KONWIHR.
7. References
[1] Stone, H. S., and Bokhari, S.H., “Control of Distributed Processes”, Computer, vol. 11, pp. 97-106, July 1978
[2] Chow, T. C. K., and Abraham, J. A., “Load Balancing in Distributed Systems”, IEEE Transactions on Software Engineering, vol. SE-8, July 1982
[3] Chu, W. W., Holloway, L.J., Lan, M.T., and Efe, K., “Task Allocation in Distributed Data Processing”, Computer, vol. 13, pp. 57-69, Nov. 1980
[4] Ma, P. R., et al, “A Task Allocation Model for Distributed Computing Systems”, IEEE Transactions on Computers, vol. C-31, pp. 41-47, Jan. 1982
[5] Efe, K., “Heuristic Models of Task Assignment Scheduling in Distributed Systems”, Computer, vol. 15, pp. 50-56, June 1982. [6] Metis: Serial Graph Partitioning, https://www.doczj.com/doc/fe9260523.html,/~karypis/metis/metis/
[7] Shen, Chien-Chung and Tsai, Wen-Hsiang, “A Graph Matching Approach to Optimal Task Assignment in Distributed Computing Systems using a Minimax Criterion”, IEEE Transactions on Computers, vol. C-34, No. 3, March 1985.
8. Appendix
A=
{
{0,5,278528} {0,44,69632} {0,22,139264} {0,41,34816} {0,63,104448} {0,43,69632} {0,47,69632} {0,48,139264} {0,45,69632} {0,42,69632} {0,46,69632} {0,0,34816} {0,0,34816} {1,5,69632} {1,8,69632} {1,7,69632} {1,6,43520} {1,61,26112} {1,17,69632} {2,7,69632} {2,12,69632} {2,18,69632} {2,17,69632} {2,16,69632} {3,8,69632} {3,20,69632} {3,29,69632} {3,28,69632} {3,27,69632} {3,35,69632} {4,18,69632} {4,23,69632} {4,28,69632} {4,36,69632} {4,35,69632} {4,34,69632} {5,0,278528} {5,44,52224} {5,41,13056} {5,45,52224} {5,63,39168} {5,42,52224} {5,1,69632} {5,8,69632} {5,7,69632} {5,9,69632} {5,12,69632} {5,14,69632} {5,20,69632} {5,21,69632} {5,13,69632} {5,15,69632} {5,41,13056} {5,43,52224} {5,63,39168} {6,1,43520} {6,10,32640} {6,61,52224} {6,49,43520} {6,17,52224} {7,1,69632} {7,5,69632} {7,9,69632} {7,61,17408} {7,2,69632} {7,62,52224} {8,1,69632} {8,5,69632} {8,9,69632} {8,10,69632} {8,3,69632} {8,17,69632} {9,5,69632} {9,7,69632} {9,8,69632} {9,11,69632} {9,18,69632} {9,28,69632} {10,6,32640} {10,8,69632} {10,11,52224} {10,61,19584} {10,27,52224} {10,49,69632} {10,17,52224} {11,9,69632} {11,10,52224} {11,61,13056} {11,19,52224} {11,62,39168} {11,31,52224} {11,49,69632} {12,2,69632} {12,5,69632} {12,14,69632} {12,13,69632} {12,41,17408} {12,63,52224} {13,12,69632} {13,17,69632} {13,15,69632} {13,41,17408} {13,63,52224} {13,5,69632} {14,5,69632} {14,12,69632} {14,18,69632} {14,15,69632} {14,23,69632} {14,42,69632} {15,13,69632} {15,14,69632} {15,17,69632} {15,60,34816} {15,43,69632} {15,24,34816} {15,5,69632}
{16,2,69632} {16,62,52224} {16,19,52224} {16,17,52224} {16,49,69632} {17,2,69632} {17,13,69632} {17,15,69632} {17,18,69632} {17,16,52224} {17,19,52224} {17,35,69632} {17,38,52224} {17,49,139264} {17,1,69632} {17,8,69632} {17,6,52224} {17,10,52224} {18,2,69632} {18,9,69632} {18,14,69632} {18,17,69632} {18,19,69632} {18,4,69632} {19,11,52224} {19,16,52224} {19,18,69632} {19,17,52224} {19,34,52224} {19,49,69632} {20,3,69632} {20,5,69632} {20,22,69632} {20,21,69632} {20,44,69632} {20,24,34816} {20,60,34816} {21,5,69632} {21,20,69632} {21,28,69632} {21,22,69632} {21,23,69632} {21,45,69632} {22,0,139264} {22,44,52224} {22,48,52224} {22,45,52224} {22,20,69632} {22,29,69632} {22,21,69632} {22,30,8704} {22,25,69632} {22,51,52224} {22,52,8704} {22,48,52224} {22,54,17408} {22,56,17408} {22,57,17408} {22,59,17408} {23,4,69632} {23,14,69632} {23,21,69632} {23,25,69632} {23,60,34816} {23,46,69632} {23,24,34816} {24,15,34816} {24,23,34816} {24,47,69632} {24,60,69632} {24,26,17408} {24,54,17408} {24,20,34816} {25,22,69632} {25,23,69632} {25,36,69632} {25,53,17408} {25,48,69632} {25,26,17408} {25,55,17408} {25,58,17408} {26,24,17408} {26,25,17408} {26,48,17408} {26,53,17408} {26,54,17408} {26,55,17408} {27,3,69632} {27,10,52224} {27,32,52224} {27,31,52224} {27,49,69632} {27,38,52224} {28,3,69632} {28,9,69632} {28,21,69632} {28,30,8704} {28,31,69632} {28,50,34816} {28,4,69632} {28,51,26112} {29,3,69632} {29,22,69632} {29,30,17408} {29,32,69632} {29,50,34816} {29,52,17408} {29,37,69632} {30,22,8704} {30,28,8704} {30,29,17408} {30,50,8704} {30,51,17408} {30,52,8704} {31,11,52224} {31,27,52224} {31,28,69632} {31,33,52224} {31,34,52224} {31,49,69632} {32,27,52224} {32,29,69632} {32,33,52224} {32,49,69632} {32,40,52224} {33,31,52224} {33,32,52224} {33,50,69632} {33,39,52224} {33,49,69632} {34,4,69632} {34,19,52224} {34,31,52224} {34,39,52224} {34,38,52224} {34,49,69632} {35,4,69632} {35,17,69632} {35,60,69632} {35,37,69632} {35,38,69632} {35,3,69632} {36,4,69632} {36,25,69632} {36,51,34816} {36,37,69632} {36,39,69632} {36,50,34816} {37,35,69632} {37,36,69632} {37,53,17408} {37,40,69632} {37,58,17408} {37,57,17408} {37,59,17408} {37,29,69632} {38,17,52224} {38,34,52224} {38,35,69632} {38,40,52224} {38,49,69632} {38,27,52224} {39,33,52224} {39,34,52224} {39,36,69632} {39,40,52224} {39,49,69632} {40,37,69632} {40,38,52224} {40,39,52224} {40,49,69632} {40,32,52224} {41,0,34816} {41,63,104448} {41,5,13056} {41,12,17408} {41,13,17408} {41,5,13056} {42,0,69632} {42,5,52224} {42,63,52224} {42,46,52224} {42,43,52224} {42,14,69632} {43,0,69632} {43,63,52224} {43,47,52224} {43,42,52224} {43,15,69632} {43,5,52224} {44,0,69632} {44,5,52224} {44,22,52224} {44,45,52224} {44,20,69632} {44,47,52224} {45,0,69632} {45,5,52224} {45,44,52224} {45,22,52224} {45,46,52224} {45,21,69632} {46,0,69632} {46,42,52224} {46,45,52224} {46,48,52224} {46,47,52224} {46,23,69632} {47,0,69632} {47,43,52224} {47,48,52224} {47,46,52224} {47,24,69632} {47,44,52224} {48,0,139264} {48,47,52224} {48,22,52224} {48,46,52224} {48,25,69632} {48,26,17408} {48,54,17408} {48,55,17408} {48,56,17408} {48,22,52224} {49,6,43520} {49,10,69632} {49,61,43520} {49,27,69632} {49,11,69632} {49,31,69632} {49,16,69632} {49,19,69632} {49,62,52224} {49,34,69632} {49,17,139264} {49,38,69632} {49,32,69632} {49,33,69632} {49,39,69632} {49,40,69632} {49,49,34816} {49,49,34816} {50,28,34816} {50,29,34816} {50,33,69632} {50,30,8704} {50,36,34816} {50,51,52224} {50,52,8704} {51,22,52224} {51,28,26112} {51,36,34816} {51,50,52224} {51,30,17408} {51,52,17408} {52,22,8704} {52,29,17408} {52,51,17408} {52,30,8704} {52,50,8704} {53,25,17408} {53,37,17408} {53,60,17408} {53,26,17408} {53,57,17408} {53,58,17408} {54,24,17408} {54,48,17408} {54,26,17408} {54,57,17408} {54,56,17408} {54,22,17408} {55,25,17408} {55,26,17408} {55,48,17408} {55,58,17408} {55,56,17408} {56,48,17408} {56,54,17408} {56,55,17408} {56,59,17408} {56,22,17408} {57,54,17408} {57,60,17408} {57,53,17408} {57,59,17408} {57,37,17408} {57,22,17408} {58,25,17408} {58,37,17408} {58,53,17408} {58,55,17408} {58,59,17408} {59,56,17408} {59,57,17408} {59,58,17408} {59,37,17408} {59,22,17408} {60,15,34816} {60,23,34816} {60,35,69632} {60,53,17408} {60,24,69632} {60,57,17408} {60,20,34816} {61,1,26112} {61,7,17408} {61,10,19584} {61,11,13056} {61,6,52224} {61,49,43520} {61,62,52224} {62,7,52224} {62,16,52224} {62,11,39168} {62,61,52224} {62,49,52224} {63,0,104448} {63,41,104448} {63,43,52224} {63,5,39168} {63,42,52224} {63,12,52224} {63,13,52224} {63,5,39168}
}
B=
{
27744, 8000, 8000, 8000, 8000, 41472, 4480, 8000, 8000, 8000, 6400, 6400, 8000, 8000, 8000, 8000, 6400, 23040, 8000, 6400, 8000, 8000, 23040, 8000, 4800, 8000, 1728, 6400, 8000, 8000, 1152, 6400, 6400, 6400, 6400, 8000, 8000, 8000, 6400, 6400, 6400, 4608, 6400, 6400, 6400, 6400, 6400, 6400, 11520, 27744, 4800, 3840, 1152, 1728, 1728, 1728, 1728, 1728, 1728, 1728, 4800, 4480, 5120, 9216
} COMP_WORTH = 0.0928;
COMM_WORTH = 0.00846;
The actual computation time of a module i equals B[i]*COMP_WORTH msec.
COMM _WORTH is the average time taken to transmit a single byte in msec. computed for the application in question.
一、选择题 1．I’d like to________the mall because it’s crowded and noisy. A．visit B．hang out C．walk D．go off 2．That path ________ directly to my house.You won't miss it. A．leads B．forms C．repairs D．controls 3．I don’t want to go. __________, I am too tired. A．However B．And C．Besides D．But 4．Some animals carry seeds from one place to another, ________ plants can spread to new places. A．so B．or C．but D．for 5．When I as well as my cousins __________ as a volunteer in Beijing, I saw the Water Cube twice. A．were treated B．treated C．was served D．served 6．He is wearing his sunglasses to himself from the strong sunlight. A．prevent B．stop C．keep D．protect 7．When you are________, you should listen to music to cheer you up. A．shy B．afraid C．strict D．down 8．Mr. Smith gave us some________on how to improve our speaking skills. A．advice B．news C．knowledge D．information 9．World Book Day takes place ________ April 23rd every year. A．at B．in C．on 10．More and more people have realized that clear waters and green mountains are as ________ as mountain of gold and silver. A．central B．harmful C．valuable D．careful 11．We loved the food so much, ________the fish dishes. A．special B．especial C．specially D．especially 12．—Oh, my God! I have ________ five pounds after the Spring Festival. —All of the girls want to lose weight, but easier said than done. A．given up B．put on C．got on D．grown up 13．—What do you think of the performance today? —Great! ________ but a musical genius could perform so successfully. A．All B．None C．Anybody D．Everybody 14．He ________ his homework________the morning of Sunday. A．doesn’t do; on B．doesn’t do; in C．doesn’t; on 15．Maria ________ speaks Chinese because she doesn’t know much Chinese. A．seldom B．always C．often D．usually 16．In 2018, trade between China and Hungary rose by 7.5 percent, and recently on Friday companies from China and Hungary________ several cooperation (合作) agreements under the
现在进行时 撰稿：王红艳审稿：白雪雁 【概念引入】 I. 什么是现在进行时？ 1）现在进行时表示说话时正在进行或发生的动作。 例如：I am reading a book. 我正在看书。 2）表示现阶段正在进行而说话时不一定在进行的动作。 例如：I am learning English hard these days. 这些日子我正在努力学习英语。 II. 现在进行时的标志词。 现在进行时常和now、at the moment、look、listen等连用。 【用法讲解】 I.现在进行时的结构。 现在进行时的结构是：助动词be（am，is，are）＋现在分词v-ing 现在分词的构成： 1）动词的后面直接加－ing。例如：work－working，study－studying 2）以不发音的字母e结尾的动词，先去掉字母e，再加－ing。例如：live－living 3）以重读闭音节结尾并且只有一个辅音字母的动词，先双写这个辅音字母，再加－ing。 例如：stop－stopping，swim－swimming，run－running II. 现在进行时的用法。 1）现在进行时表示说话的时候正在进行的动作，经常和now，right now，at the moment 等时间状语或者动词look，listen等连用。 例如：My father is watching TV now.我爸爸现在在看电视。 Look! My brother is playing basketball there. 看！我弟弟正在那里打篮球。 2）现在进行时可以表示目前一段时间内一直进行的动作，经常和these days，this week，at present等时间状语连用。 例如：My parents are working on a farm these days. 这些天我的父母在农场干活。 3）现在进行时还可以表示现在不断发展变化的事情，表示不断发展变化的动词有get，grow，turn，become等。 例如：The leaves are turning yellow. 树叶在变黄。 4）现在进行时还可以表示将要发生的动作，只限于动词arrive，begin，go，come，leave，fly等动词。 例如：I am coming soon. 我马上来。 Ⅲ. 现在进行时的句式变化。 肯定句式：主语+be( am, is, are)+现在分词+其它. 否定句式：主语+be(am, is, are) +not +现在分词+其它. 一般疑问句：Be(am, is, are) +主语+现在分词+其它？ 特殊疑问句：疑问词+be(am, is, are)+主语+现在分词+其它？ 对现在进行时的特殊疑问句的回答，它不可以用Yes或No直接作答，要根据实际情况回答。 Ⅳ. 现在进行时的特殊用法。 表示位置移动的动词，如：leave/ come/go/begin等用于现在进行时，表示按计划或安排近期将要进行的动作，常与表示将来的时间状语连用。 —Can you help me? 你能帮我吗？
职业技术学院学生课程设计报告 课程名称：高频电路课程设计 专业班级：信工102 姓名： 学号：20110311202 学期：大三第一学期
目录 1课程设计题目……………………………………………2课程设计目的…………………………………………3课程设计题目描述和要求……………………………4课程设计报告内容……………………………………… 4.1二极管包络检波电路的设计……………………… 4.2同步检波器的设计……………………………5结论……………………………………………………6结束语………………………………………………………7参考书目……………………………………………………8附录………………………………………………………
摘要 振幅调制信号的解调过程称为检波。有载波振幅调制信号的包络直接 反映调制信号的变化规律，可以用二极管包络检波的方法进行检波。而抑 制载波的双边带或单边带振幅调制信号的包络不能直接反映调制信号的变 换规律，无法用包络检波进行解调，所以要采用同步检波方法。 同步检波器主要是用于对DSB和SSB信号进行解调（当然也可以用于AM）。它的特点是必须加一个与载波同频同相的恢复载波信号。外加载波信 号电压加入同步检波器的方法有两种。利用模拟乘法器的相乘原理，实现 （t）,和输入的同步 同步检波是很简单的，利用抑制载波的双边带信号V s （t），经过乘法器相乘，可得输出信号，实现了双 信号（即载波信号）V c 边带信号解调 课程设计作为高频电子线路课程的重要组成部分，目的是一方面使我们能够进一步理解课程内容，基本掌握数字系统设计和调试的方法，增加集成电路应用知识，培养我们的实际动手能力以及分析、解决问题的能力。 另一方面也可使我们更好地巩固和加深对基础知识的理解，学会设计中小型高频电子线路的方法，独立完成调试过程，增强我们理论联系实际的能力，提高电路分析和设计能力。通过实践引导我们在理论指导下有所创新，为专业课的学习和日后工程实践奠定基础。 通过设计，一方面可以加深我们的理论知识，另一方面也可以提高我们考虑问题的全面性，将理论知识上升到一个实践的阶段。
（工作分析）计数器工作原理的模式化分析
计数器工作原理的模式化分析 时序逻辑电路是《脉冲和数字电路》这门课程的重要组成部分，计数器是时序逻辑电路基础知识的实际应用，其应用领域非常广泛。计数器原理是技工学校电工电子专业学生必须重点掌握的内容，也是本课程的考核重点，更是设计计数器或其他电子器件的基础。 但近年来技校学生的文化理论基础和理解能力普遍较差，按照课件体系讲授计数器这个章节的知识，超过70%的学生听不懂。 我先后为四届学生讲授过这门课，于教学实践中摸索出壹套分析计数器的方法——模式化分析，即把分析步骤模式化，引导学生按部就班地分析计数器。用这种方法分析，我只要以其中壹种计数器（如异步二进制计数器）为例讲解，学生便能够自行分析其他计数器。 教学实践证明，用这种方法讲授计数器知识，学生比较感兴趣，觉得条理清晰，易于理解，掌握起来比较轻松。这种方法仍有壹个好处，不管是同步计数器仍是异步计数器，不管是二进制计数器仍是十进制计数器，不管是简单的计数器仍是复杂的计数器，只要套用这种方法，计数器工作原理迎刃而解。即使是平时基础很差的学生，只要记住几个步骤，依葫芦画瓢，也能把计数器原理分析出个大概来。 一、明确计数器概念 分析计数器当然要先清楚什么是计数器啦。书上的概念是：
计数器是数字系统中能累计输入脉冲个数的数字电路。我告诉学生，计数器就是这样壹种电子设备：把它放于教室门口，每个进入教室的同学均于壹个按钮上按壹下，它就能告诉你壹共有多少位同学进入教室。其中，每个同学按壹下按钮就是给这个设备壹个输入信号，N个同学就给了N个信号，这N个信号就构成计数器的输入CP脉冲，计数器要统计的就是这个CP脉冲系列的个数。当然，如果没有接译码器，计数器的输出端显示的是二进制数而非十进制数，比如有9位同学进入教室，它不显示“9”，而是显示“1001”。 随后，我简要介绍了计数器的构成和分类，且强调，计数器工作前必须先复位，即每个触发器的输出端均置零。 二、回顾基础知识 分析计数器要用到触发器的关联知识，其中JK触发器最常用，偶尔用到T触发器和D触发器。因此，介绍完计数器概念后，我不急于教学生分析其原理，而是先提问JK、T、D触发器的关联知识，包括触发器的逻辑符号、特性方程、特性表等。 由于计数器的控制单元由逻辑门电路构成，分析前仍要简要回顾壹下和、或、非等常用逻辑门电路的关联知识。另外，用模式化方法分析计数器仍要用到逻辑代数的运算方法、逻辑函数的化简方法等关联知识。 三、画出解题模板 准备工作做完了，下面进入核心部分——列出分析计数器的
boring 令人厌烦的，乏味的，无聊的 tedious 乏味的，单调的，冗长的 flat 单调的，沉闷的 dull 乏味的，单调的 troublesome 令人烦恼的，讨厌的，麻烦的 tired 疲劳的，累的 bored 无聊的，无趣的，烦人的 exhausted 极其疲倦的 weary 疲劳的 bright 聪敏的，机灵的 apt 聪明的，反应敏捷的 intelligent 聪明的，有才智的 shrewd 机灵的，敏锐的，精明的（表示生意上的精明） ingenious （人，头脑）灵巧的 alert 警觉的，留神的 cute 聪明伶俐的，精明的 acute/cute acute 指的是视力，感觉的敏锐 dull 愚钝的，笨的 awkward 笨拙的，不灵巧的 absurd 荒谬的 ridiculous 可笑的，荒谬的 idiotic 白痴般的 blunt 率直的，直言不讳的 clumsy 笨拙的，粗陋的 happy 快乐的，幸福的 cheerful 欢乐的，高兴的 content 满意的，满足的 merry 欢乐的，愉快的，快乐的 pleasure 高兴，愉快，满足 enjoyment 享乐，快乐，乐趣 cheer 喝彩 applause 鼓掌，掌声 optimism 乐观，乐观主义 delight 快乐，高兴 kick 极大的乐趣 paradise 天堂，乐园 instant 立即的，即刻的 instantaneous 瞬间的，即刻的 immediate 立即的，即刻的 simultaneous 同时发生的，同时存在的，同步的punctual 严守时刻的，准时的，正点的 pick 挑选，选择 select 选择，挑选 single 选出，挑出 elect 选举，推举 vote 投票，选举 appoint 任命，委派 nominate 提名，任命 propose 提名，推荐 recommend 推荐，举荐 designate 指派，委任 delegate 委派（或选举）…为代表 install(l) 使就职，任命 ballot 使投票表决 dub 把…称为 choice 选择（权） option 选择 selection 选择，挑选 alternative 取舍，供选择的东西 favorite 特别喜爱的人（或物） inclination 爱好 preference 喜爱，偏爱，优先 observe 注意到，察觉到 perceive 认识到，意识到，理解 detect 察觉，发现 appreciate （充分）意识到，领会，体会 alert 使认识到，使意识到 awake 意识到，醒，觉醒 scent 察觉 ancient 古代的，古老的 primitive 原始的 preliminary 预备的，初步的 preliminary trial初审 primary 最初的，初级的 initial 开始的，最初的 original 起初的 former 在前的，以前的 previous 先，前 prior 在前的，优先的 beforehand 预先，事先 medieval 中世纪的，中古（时代）的preceding 在先的，在前的，前面的 senior 资格较老的，地位较高的 following 接着的，下述的 attendant 伴随的 subsequent 随后的，后来的 succeeding 以后的，随后的 consequent 作为结果（或后果）的，随之发生的 resultant 作为结果的，因而发生的therefore 因此，所以 consequently 所以，因此 then 那么，因而 thus 因此，从而 hence 因此，所以 accordingly 因此，所以，于是 thereby 因此，从而
现在进行时特殊用法展现 现在进行时表示说话时正在进行的动作，这是我们平时接触最多的。然而除此以外，现在进行时还有以下几种用法。 1．表示“在做某事的过程中”，此时动作不一定正在发生。例如： Next I'll give you a few minutes to read the article．When you are reading，make a mark where there is a new word．现在我给你们几分钟时间读一下这篇文章。在读的过程中，在有生词的地方作以标记。 2．表示按计划、方案或安排而进行的将来的动作。在这种情况下谓语动词多为非延续性动词，如come，go， leave，move，die，start，stop，arrive等，及少数延续性动词，如spend，stay等。例如： She is leaving for Guangzhou next week．她下星期就要去广州了。 We are spending the whole summer holiday inBeijing soon．不久我们要在北京度过整个暑假。 Where are you staying in Guangzhou？在广州你打算住什么地方？ 3．与副词forever，always，constantly等连用，表示赞成、厌烦、生气等情绪。例如： She is always talking loudly in our class．她总是在我们班上吵吵嚷嚷的。（表示厌烦） She's constantly changing her mind．她老是改变主意。（表示不以为然） He is forever complaining about his job．他总是对他的工作提出抱怨。（表示厌烦） 4．teach，work，live，study等表示状态的动词使用现在进行时可表状况，与一般现在时区别不大。例如： I'm studying in No．1Middle School．我在一中学习。（相当于：I study in No．1Middle School．） My brother is working in a big factory．我哥哥在一家大工厂工作。（相当于：My brother works in a big factory．） 5．表示目前经常发生的动作，然而此时动作不一定正在进行之中。例如：
1.侧向测井（电流聚焦测井）采用电屏蔽方法，使主电流聚焦后水平流入地层，减小井眼和围岩影响。主电流线沿井轴径向成饼状流入地层。 2.理想的侧向测井组合是双侧向加微球形聚焦，可较准确地确定地层电阻率、冲洗带电阻率和侵入带直径，是计算地层含油饱和度、判断地层含油性的重要参数。 3.侧向测井电极系的主电极A0位于电极系中心，两端有屏蔽电极A1、A2，呈对称排列。 七侧向电极系主电极A0，屏蔽电极A1、A2，两对监督电极M1N1和M2N2；Um1＝Un1或Um2＝Un2，使主电流沿水平方向流入地层。 七侧向四个参数：①电极系长度: 210A A L =影响侧向测井的径向探测深度。电极系长度越大，探测越深；②电极距：21O O L =影响纵向分辨率。L 越小纵向分层能力越强。③分布比：L L s /0=影响电流层的形状，一般取s 为3左右较适宜。④聚焦系数：L L L q /)0(-= 1-=s q 影响电流层的形状。 双侧向电极系由9个电极组成，第二屏蔽电极A1’、A2’有着双重的作用。 4. 如何保证屏流和主电流同极性？ 用同一电流源供给屏流和主电流。屏流大于主电流，在测井过程中屏流是浮动的。所以，屏流要由平衡放大电路输出的信号加以调制后通过功率放大后加到屏蔽电极上；二是用跟踪主电流来产生屏流，或用跟踪屏流来产生主电流，这种方式用在双侧向仪器中。 5.双侧向测井仪器中，增加屏蔽电极的长度可以加大聚焦能力，而增加仪器探测深度。相反，在屏蔽电极两端设置回流电极，可使主电极和屏流流入地层的深度变浅，降低探测深度。 6.侧向测井仪器工作方式：恒流式（高阻地层），恒压式（低阻地层），自由式（1229、JSC801）和恒功率式（DLT-E ）。 恒流式：保持主电流恒定，测量主电极（通常用监督电极M1或M2代替）至远处电极N 之间的电位差U 。地层的电阻率越高测量电压信号越大，测量误差越小。 恒压式：保持主电极电位恒定，测量主电流。地层的电阻率越低测量电流信号越大，测量误差越小。 自由式：电流和电压按一定规律浮动，同时测量电流、电压两个量，可以得到较宽的测量动态范围。 恒功率式或可控功率式：测量过程中使最高和最低电阻率的两个极点保持功率（IU 乘积）不变，让测量电压和电流保持在仪器可测量的范围之内（不被限幅）。比自由式仪器有更宽的测量动态范围。 7.1229双侧向测井仪采用屏流主动式供电，即先有屏流后又主电流，用屏流来激励产生主电流。工作方式为自由式，为提高仪器测量动态范围用U2D 来控制深、浅屏流、屏压的变化幅度在于此。 频分双侧向供电式，fS = 4fD ，深、浅侧向供电频率分别为32Hz 和128Hz 。使深、浅侧向两个系统相对独立地控制和测量。
计数器原理分析及应用实例 除了计数功能外，计数器产品还有一些附加功能，如异步复位、预置数（注意，有同步预置数和异步预置数两种。前者受时钟脉冲控制，后者不受时钟脉冲控制）、保持（注意，有保持进位和不保持进位两种）。虽然计数器产品一般只有二进制和十进制两种，有了这些附加功能，我们就可以方便地用我们可以得到的计数器来构成任意进制的计数器。下面我们举两个例子。在这两个例子中，我们分别用同步十进制加法计数器74LS160构成一个六进制计数器和一个一百进制计数器。 因为六进制计数器的有效状态有六个，而十进制计数器的有效状态有十个，所以用十进制计数器构成六进制计数器时，我们只需保留十进制计数器的六个状态即可。74LS160的十个有效状态是BCD编码的，即0000、0001、0010、0011、0100、0101、0110、0111、1000、1001[图5-1]。 图5-1 我们保留哪六个状态呢？理论上，我们保留哪六个状态都行。然而，为了使电路最简单，保留哪六个状态还是有一点讲究的。一般情况下，我们总是保留0000和1001两个状态。因为74LS160从1001变化到0000时，将在进位输出端产生一个进位脉冲，所以我们保留了0000和1001这两个状态后，我们就可以利用74LS160的进位输出端作为六进制计数器的进位输出端了。于是，六进制计数器的状态循环可以是0000、0001、0010、0011、0100和1001，也可以是0000、0101、0110、0111、1000和1001。我们不妨采用0000、0001、0010、0011、0100
和1001这六个状态。 如何让74LS160从0100状态跳到1001状态呢？我们用一个混合逻辑与非门构成一个译码器[图5.3.37b]，当74LS160的状态为0100时，与非门输出低电平，这个低电平使74LS160工作在预置数状态，当下一个时钟脉冲到来时，由于等于1001，74LS160就会预置成1001，从而我们实现了状态跳跃。 图5.3.37b用置数法将74160接成六进制计数器（置入1001） 比这个方案稍微繁琐一点的是利用74LS160的异步复位端。下面这个电路中[图5.3.34]，也有一个由混合逻辑与非门构成的译码器。 图5.3.34用置零法将74LS160接成六进制计数器
板框压滤机的工作原理 板框压滤机是由交替排列的滤板和滤框共同构成一组滤室。在滤板的表面有沟槽构造，它凸出部位是用来支撑滤布的。滤框和滤板的边角上各有通孔，组装以后可以构成一个完整的通道，能够通入洗涤水、悬浮液和引出滤液来。板和框的两侧各有把手支托在横梁的上面，由压紧装置压紧板、框。板、框之间的滤布起到密封垫片的作用。由供料泵将悬浮液压入滤室，在滤布的上面形成滤渣，直至充满了滤室。 滤液穿过滤布并沿滤板沟槽流至板框边角通道，集中排出。过滤完毕之后，可以通入清洗涤水洗涤滤渣。洗涤后，有时还通入压缩空气，除去剩余的洗涤液。随后打开压滤机卸除滤渣，清洗滤布，重新压紧板、框，开始下一工作循环。 板框压滤机主要由压紧板(活动滤板)、止推板(固定滤板)、过滤介质(滤布或滤纸等)、滤板和滤框、横梁(扁铁架)、压紧装置、集液槽等组成(参见附图)，其中的过滤介质和集液槽上由用户自备，当然也可以由上海大张过滤设备代配。 板框压滤机对于滤渣压缩性大或近于不可压缩的悬浮液都能适用。适合的悬浮液的固体颗粒浓度一般为10%以下，操作压力一般为0.3～0.6兆帕，特殊的可达3兆帕或更高。过滤面积可以随所用的板框数目增减。板框通常为正方形，滤框的内边长为 320～2000毫米，框厚为16～80毫米，过滤面积为1～1200米2。板与框用手动螺旋、电动螺旋和液压等方式压紧。板和框用木材、铸铁、铸钢、不锈钢、聚丙烯和橡胶等材料制造。 板框压滤机共有手动压紧、机械压紧和液压压紧三种形式。 手动压紧是螺旋千斤顶推动压紧板压紧；机械压紧是电动机配H型减速箱，经机架传动部件推动压紧板压紧；液压压紧是有液压站经机架上的液压缸部件推动压紧板压紧。
不能片面说人做主语用ed，物做主语ing ing形式是修饰引起这种感觉的人或物；ed形式是描写人或物的感受。（当然物一般是动物） 翻译的话 ing形式的词译为“令人……的”；ed形式译为“……的” boring是令人感到厌烦的；bored是厌烦的。 a boring person 能够指一个了无情趣的人，让人觉得无趣的人 a bored person 则是说这个人自己感到很无趣 1.bore 1）vt.使厌烦；挖 e.g. I'm bored with this job. 这件工作厌烦了。 The oldier bore the sharp pain in the wound with great courage. 这士兵以巨大的勇气忍受着伤口的剧烈疼痛。 2）n.令人讨厌的人（或事） e.g. It's a bore having to go out again. 外出真是讨厌。 boredom n.厌倦，无趣 e.g. in infinite boredom 极其无趣 boring n. 钻(孔) adj. 令人厌烦的(事或物） e.g. The play was boring. 这部短剧很一点意思都没有。 bored adj. 无聊的, 无趣的, 烦人的 e.g. Jack is so bored. 杰克是个没有趣的人。 2.surprising 是针对事或物感到惊奇。 surprised 则是针对人。 3.pleasant adj. 愉快的, 快乐的, 舒适的, 合意的可爱的, 举止文雅的, 活泼的滑稽的, 有趣的 (天气)晴朗的, 美好的容易相处的, 友爱的 e.g. a pleasant voice 悦耳的声音 a pleasant companion 可爱的伴侣 a pleasant time 愉快地度过时光 pleasing adj. 舒适的, 使人愉快的; 满意的; 惹人喜欢的, 可爱的 e.g. a pleasing look 使人愉快的神情 a very well mannered and pleasing young man 彬彬有礼而令人喜爱的年轻人
现在进行时动词加ing的变化规律 1)一般情况下，直接加 -ing: 如：go—going answer—answering study—studying be—being see—seeing [注一] 和名词复数、一般现在时动词第三人称单数加-s(-es)不同， 动词末尾如为“辅音字母 + y”时，y不变，其后直接加ing。 如: study—studying fly—flying carry—carrying [注二] 动词结尾为辅音字母r时，加-ing，r在此必须发音。 如: water—watering answer—answering wear—wearing 2)以不发音的e结尾的动词，去掉e，再加ing 如：come—coming write—writing take—taking become—becoming 3)动词是闭音节的单音节词，或是以重读闭音节结尾的多音节词， 而末尾只有一个辅音字母时，这个辅音字母须双写，然后再加ing。 如：sit—sitting run—running stop—stopping begin—beginning admit—admitting forget—forgetting [注一] send，think，accept等动词虽是闭音节或以重读闭音节结尾， 但末尾有一个以上的辅音字母，因此，这个辅音字母不双写，应直接加ing。 如：sending thinking accepting 4)少数几个以-ie结尾的动词，须将ie变作y，再加ing。 如：die—dying tie—tying lie—lying躺，说谎 5)少数以-c结尾的动词变为现在分词时和过去式，须先将-c变为ck,然后再加-ing 或-ed 。 如：picnic—picnicking (picnicked) traffic—trafficking (trafficked)
目录 第1章二极管检波电路设计方案论证 (1) 1.1检波的定义 (1) 1.2二极管检波电路原理 (1) 1.3二极管检波电路设计的要求及技术指标 (1) 第2章对二极管检波电路各单元电路设计 (2) 2.1检波器电路设计检波器电路 (2) 2.1.1检波器电路原理及工作原理 (2) 2.1.2检波器质量指标 (3) 第3章二极管检波电路整体电路设计及仿真结果 (4) 3.1整体电路图及工作原理 (4) 3.3电路仿真图形 (4) 第4章总结 (5) 参考文献 (6) 元器件清单 (7)
第1章二极管检波电路设计方案论证 1.1检波的定义 广义的检波通常称为解调，是调制的逆过程，即从已调波提取调制信号的过程。对调幅波来说，是从它的振幅变化提取调制信号的过程；对调频波来说，是从它的频率变化提取调制信号的过程；对调相波来说，是从它的相位变化提取调制信号的过程。 狭义的检波是指从调幅波的包络提取调制信号的过程。因此，有时把这种检波称为包络检波或幅度检波。图1-20-21出了表示这种检波的原理：先让调幅波经过检波器（通常是晶体二极管），从而得到依调幅波包络变化的脉动电流，再经过一个低通滤波器滤去高频成分，就得到反映调幅波包络的调制信号 1.2二极管检波电路原理 调幅波信号是二极管检波电路的输入，由于二极管只允许单向导电，所以，如果使用的是硅管，则只有电压高于0.7V的部分可以通过二极管。 同时，由于二极管的输出端连接了一个电容，这个电容与电阻配合对二极管输出中的高频信号对地短路，使得输出信号基本上就是AM信号包络线。电容和电阻构成的这种电路功能叫做滤波。 1.3二极管检波电路设计的要求及技术指标 1．对常规调幅信号进行二极管检波解调并仿真，能够观察输入输出波形。 2．根据电路结果求出电压利用系数 3．判断设计的电路是否能够产生失真 参数：常规调幅信号调幅系数为0.5，输入信号载波频率10000HZ，载波电压100mV左右。
现在进行时 一、定义及用法： 1定义：（1）表示说话时正在进行的动作及行为。（2）表示现阶段正在进行的动作。 2基本用法： (1)现在进行时主要表示说话人的说话时刻正在进行的动作、不断重复的动作或目前这个阶段（不一定是说话时刻）正在进行的动作，如： We’re having a meeting. 我们在开会。（说话时正在进行的动作） Be quiet！The baby is sleeping.安静，孩子在睡觉。 He is teaching in a middle school. 他在一所中学教书。（目前阶段在进行的动作） (2)现在进行时表将来: 现在进行时表将来，主要表示按计划或安排要发生的动作： I’m leaving tomorrow. 我明天走。 They’re getting married next month. 他们下个月结婚。 注意：现在进行时与一般现在时均可表示将来，区别是：用现在进行时表示将来，其计划性较强，并往往暗示一种意图；而一般现在时表示将来，则其客观性较强，即通常被视为客观事实，多指按时刻表或规定要发生的情况： I’m not going out this evening. 今晚我不准备出去。 What time does the train leave？火车什么时候开？ (3)现在进行时表示感情色彩: 现在进行时有时可表示满意、称赞、惊讶、厌恶等感情色彩，通常与always，forever，constantly，continually等副词连用。比较： She’s always helping people. 她老是帮助别人。（表赞扬） She always helps others. 他总是帮助别人。（陈述一个事实） The boy is constantly lying. 这孩子老是撒谎。（表示厌恶） The boy often lies. 这孩子常撒谎。（指出缺点） 二、结构: 现在进行时常有三种句型： （1）肯定式：主语＋be＋v-ing＋其它。 如：He is mending his bike．他正在修自行车。 （2）否定式：主语＋be＋not＋v-ing＋其它。 如：He is not（isn't）mending his bike．他没在修自行车。 （3）疑问式：主要分一般疑问句和特殊疑问句两种句式。 一般疑问句：Be＋主语＋v-ing＋其它？ 如：—Is he mending his bike？他正在修自行车吗？—Yes，he is．（No，he isn't．）特殊疑问句：疑问词＋be＋主语＋v-ing＋其它？ 如：—What is he doing？他正在干什么？ 三、何时用现在进行时？ （1）以Look！或Listen！开头的句子提示我们说话时动作正在进行，应用现在进行时。 如：Look！The children are playing games over there． Listen！Who's singing in the classroom？ （2）当句子中有now（现在）时，常表示说话时动作正在进行，这时也常用现在进行时。 如：We are reading English now． （3）描述图片中的人物的动作时常用现在进行时，以示生动。 如：Look at the picture．The girl is swimming． （4）表示当前一段时间内的活动或现阶段正在进行的动作时常用现在进行时。这时常与时间状语these days，this week等连用。
各种过滤器的原理及结构 株洲海润公司郑胜春（摘录） 石英砂过滤器主要用于去除水中的悬浮物。该设备与其它水处理设备配合，广泛地应用在给水净化、循环水净化污水处理等各类水处理工程中。 活性碳过滤器主要用于吸附水中游离氯（吸附力达99％），对有机物和色度也有较高的去除率。是软化、除盐系统制纯水工艺的预处理设备。滤料为活性碳。设备主要材质为碳钢防腐、玻璃钢和不锈钢等。 活性碳过滤器技术参数 过滤速度：8-10m3/h 进水浊度：≤5mg/L； 工作温度：常温工作压力：≤0.6Mpa; 反洗压缩空气量：18-25L/m2.S 滤料层高：1000-1200mm 反洗强度：4-12L/m2.S; 反冲洗时间：4-6分钟 石英砂过滤器技术参数： 1、运行参数 2、水质参数 设计滤速：8-10米/时期终水头损失：1.7米进水浊度小于15度，出水浊度小于3度。反清洗强度：4-15升/秒·平方米进水浊度小于10度，出水浊度小于2度。 冲洗历时：5-7分钟滤料：石英砂3、水压 垫层厚度：200-400毫米滤层厚度：800毫米进水水压：≥0.04Mpa 反冲洗进水水压：≥0.15Mpa 盘式过滤器原理与应用分析 工厂制水的预处理系统以前采用的是纤维过滤法，在近几年的运行过程中，这种方法暴露出许多问题：过滤效率明显下降，运行阻力增加，树脂破碎率升高，制水成本逐年上升；出现纤维扭曲，发生“粘连抱团”现象，纤维束不能垂直竖立，下移动不能复位；过滤器内部气囊破损严重，无法正常发挥松散纤维的作用。为了改善制水预处理系统的现状，转而采用盘式过滤器代替高效纤维过滤，取得了良好的效果。 一、盘式过滤器机理 1盘式过滤器的原理： 利用相邻盘片之间的沟槽纹交叉点实现对固体颗粒的拦截，运行时14组过滤头并联，在水和弹簧的压力作用下过滤滤芯的滤盘被压紧，水从盘片的端面进入，水中的颗粒杂质被压紧的盘片截留，从而起到过滤的作用。反洗时，其中一组进水阀关闭，排污阀打开，其他13组过滤单元的部分出水反向进入这组过滤单元，在反洗水压下促使碟片横向旋转和纵向颤动。滤芯盘片松开，同时反洗水沿管线方向冲向过滤盘片，导致盘片高速旋转，使截留在盘片上的杂物在离心力和水流冲洗的共同作用下脱离盘片，并经反洗水的作用排除。冲洗过程仅需十几秒钟，一个滤头反洗结束后，再对其它几组依次进行反洗。阿速德盘式过滤器又有其独到之处：旋转设计。水流进入到过滤器单元内时，沿外壳的切向进入，在过滤单元内高速旋转，没有真正进入盘式过滤器之前，系统已经将大部分的泥沙等杂质从水中分离，减轻了过滤器的负担，使其工作寿命是同类产品的10倍左右。
一、选择题 1．Is this a photo of your son? He looks________ in the blue T-shirt. A．lovely B．quietly C．beautiful D．happily 2．—Jerry looks so tired. He works too hard. —He has to ________ a family of four on his own. A．offer B．support C．provide D．remain 3．— Mr. Wilson, can I ask you some questions about your speech? — Certainly, feel __________ to ask me. A．good B．patient C．free D．happy 4．Some animals carry seeds from one place to another, ________ plants can spread to new places. A．so B．or C．but D．for 5．— Can you tell us about our new teacher? —Oh, I’m sorry. I know________ about him because I haven’t seen him before. A．something B．anything C．nothing D．everything 6．—Help yourselves! The drinks are ________ me. —Thank you. You’re always so generous. A．above B．in C．on D．over 7．Gina didn’t study medicine. ________, she decided to become an actor. A．Instead B．Again C．Anyway D．Also 8．—Have you got Kathy’s________ for her concert? —Yes, I’d like to go and enjoy it. A．interview B．information C．invitation D．introduction 9．More and more people have realized that clear waters and green mountains are as ________ as mountain of gold and silver. A．central B．harmful C．valuable D．careful 10．Kangkang usually does her homework ________ it is very late at night. A．until B．when C．before D．after 11．He ________all the “No Smoking” signs and lit up a cigarette. A．requested B．attacked C．protected D．ignored 12．一Where is Mr. Brown? 一I think he's _____________ the music hall. A．on B．in C．over D．from 13．— Is your home close to the school, Tom? — No, it's a long way, but I am________ late for school because I get up early daily. A．always B．usually C．never D．sometimes 14．—Mum, I don’t want the trousers. They’re too long.
英语现在进行时用法 初中英语――现在进行时 1】现在进行时的构成 现在进行时由"be+v-ing"构成。be应为助动词，初学者最容易漏掉，它应与主语的人称和数保持一致。 2】现在进行时的应用 在实际运用时，现在进行时常用以下几种情况： (1)当句子中有now时，常表示动作正在进行，这时要用现在进行时。如： They are playing basketball now.现在他们正在打篮球。 (2)以look, listen开头的句子，提示我们动作正进行，这时要用现在进行时。如： Listen！She is singing an English song.听，她正在唱英语歌。 (3)表示当前一段时间或现阶段正在进行的动作，且此时有this week, these days等时间状语，这时常用现在进行时。如： We are making model planes these days.这些天我们在做飞机模型。 (4)描述图片中的人物的动作，也为了表达更生动。此时也常用现在进行时。如： Look at the picture. The children are flying kites in the park.看这幅图，那些孩子正在公园放风筝。 3】现在进行时的变化 肯定句式：主语+be( am, is, are)+现在分词+其它. 否定句式：主语+be(am, is, are) +not +现在分词+其它. 一般疑问句：Be(am, is, are) +主语+现在分词+其它？ 特殊疑问句：疑问词+be(am, is, are)+主语+现在分词+其它？ 对现在进行时的特殊疑问句的回答，它不可以用Yes或No直接作答，要根据实际情况回答。 注意事项 1.在英语中，并不是所有的动词都要使用正在进行时。例如一些表示状态和感觉的动词，一般不用进行时态，而是用现在一般时表示。例如： I hear someone singing. 我正听见有人唱歌。