Documents

test bench for exponential function

Description
this document talk about how to write a HDL testbench for a verilog code.various data generation for input have presented
Categories
Published
of 4
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  Assignment 1  –   Design and writing a Testbench for an Exponentiation Module Mohammad Ebrahimi 810192234  Electrical and Computer Engineering School of Engineering Colleges, Campus 2 University of Tehran, 1450 North Amirabad, 14395-515 Tehran, Iran  Moh.ebrahimi@ece.ut.ac.ir Abstract  —   In this homework we are to design and implement a module named “Exp”  that calculates e x  using Tailor series. This module accepts a n bit input “x” in fixed point representation system. “x” is assumed  to be in range 0   . When the module detects a positive pulse on start signal, computation is started and output busy signal issues to be 1 meanwhile computation. At the end of computation busy signal becomes low indicating that outputs are valid. Two outputs are considered to showing the result. Two bit ”resultIntegralpart” to represent integral part of the result and 16 bit “result Fractional part”  to represent fractional part of the result. Array multiplier is the main portion of “Exp” in which multi plication is done. The design is flexible and user could determine bit width of “x. Keywords  —    Fixed point representation, Array multiplier. I. INTRODUCTION   Tailor series is one of well-known methods to compute mathematical functions such as sin(x), cos(x), e x , and etc. The following algorithm can be used to approximate e  x : As can be seen multiplication is the main function to calculate e  x . from the algorithm we can conclude that two multiplication and one adding are needed in each iteration. Moreover , values of term (1/i) for 1    have been computed before and stored in a table. All components will discussed in details in next sections. Sec. II introduces the the data path. Sec. III  presents the controller of the ―Exp‖ module. Testbench and data provision methods to apply inputs will describe in sec. IV. results will be shown in sec. V and finally sec. VI concludes the report. II.DATA PATH Figure 1 shows the circuit of data path. Data path component consist of several components including ― Multiplier  ‖ that gives 3 n-bit inputs and an n-  bit output. ― Multiplier  ‖ itself consists of two cascaded array multiplier. Array multiplier accepts two n-bit inputs and result is available on n-bit output.  Note that the array multiplier ignores n-bit least significant  bits. Multiplier has three inputs. One of inputs comes from ―x‖ input, another multiplier input(1/i) comes from look-up-table, and the other one is the previous output that comes from ―rega‖ output. An up -counting counter has been considered two address look-up-table entries sequentially. Up-counter counts while it’s start signal is high. Start signal come from controller and connected to controller ―busy‖ output. after each multiplication the intermediate results are accumulate using n+2 bit adder, n bit for fractional part and 2 bit for integral part of the result. Counter and registers are common and clear so we describe array multiplier and look up table details in the rest of this section: Array multiplier:  multiplication of binary operands in a digital system can be performed in a variety of ways that are different in speed and gate count. A combinational n*n multiplier requires an array on n*n multiplier cells each of which is responsible for multiplying a bit of multiplier with a  bit of multiplicand and adding the result with the product bit coming from a previous multiplication stage. Therefore an AND gate for 1*1 multiplication and a full-adder for the add operation constitute the required multiplier cell hardware, as shown in Figure 2. The cell shown in Figure 2 multiplies its x i  and y i  inputs using the AND gate and adds this result with its input partial  product p 0 , a carry output c 0 , and passes x i  and y i  inputs on to its outputs(x o  and y o ).    Figure 2: Multiplier Cell Figure 3 shows a 4 * 4 array multiplier that uses 16 of the multiplier cells of Figure 2. A 32 bit multiplier requires 1024 such cells. Figure 4 shows the Verilog code of a multiplier cell bit_multiplier corresponding to the diagram of Figure 2. Figure 3: 4*4 Multiplier The logical architecture of bit_multilier has a full adder, an AND gate and pass-through wires connecting inputd x i  and y i  to their corresponding outputs, x o  and y o . Figure 4: One-bit Multiplier ―array_mutiplier.v‖ includes  the Verilog code of an n*n unconstrained array multiplier. In the declarative part of this architecture xv,yv,cv, and pv n*n array are declared. These signals are used for connecting signals of the bit-multipliers of the array multiplier to their adjacent bit-multiplier. These arrays are for connecting intermediate x inputs, y inputs, carry, and partial products. For example xv(i,j) connects the x 0  of a bit-multiplier at location i, j to x i  of the bit-multiplier to its left.      Look up table: look up table   is a table in which (1/i) terms have been saved in fixed point representation system. We assume 1   . So the table has 16 16-bit entries. ―table.v‖ includes High level description Verilog code for generating table.   III.CONTROLLER Controller receives 3 signals as input ― CLK  ‖, ― start ‖,  and ―finish‖. First, controller is in ―Waiting‖ state and wait for  positiv e pulse on ― start ‖. After detecting positive pulse on ― start ‖, controller changes its state to Busy. The controller has  been written according to Huffman coding style so that sequential and combinational parts of controller is separated. Controller has 3 o utputs including ―  busy ‖, ― waiting ‖ and ― done ‖. Figure 7 shows the state machine of the controller. Waiting Busy(start,finish) (10) Done (X1)(0X)(x0)   Figure 7: State diagram of the controller   IV.TestBench ―tester.v‖ includes testbench for module ―Exp‖ . Data  provision methods including deterministic, arithmetic,  periodic, random and Text IO have been used to apply data to ―Exp‖. Data has been applied in random and equal time slots and results collected after falling edge of busy signal. V.Simulation Results Figure 8-a and 8-b show the result waveforms after running testbench. In addition to waveform, results have been displayed on screen using $diplay command in figure 9. VI.CONCLUSION In this homework we designed and implemented a module that calculates e x  using tailor series. In order to do multiplication we design a generic array multiplier. The ―Exp‖ module start calculating after active pulse detection on start signal on the rise edge of clk signal and issues busy signal to high simultaneously . After completion of calculation busy signal is deactivated and the module is ready to accept another input on x vector. module bit_mutiplier( xi,yi,pi,ci, xo,yo,po,co ); input xi,yi,pi,ci; output xo,yo,po,co; reg xy; always @(xi,yi) begin xy <= xi & yi; end assign co = (pi & xy) | (pi & ci) | (xy & ci); assign po = pi ^ xy ^ ci; assign xo = xi; assign yo = yi; endmodule  ControllerArray MultiplierArray MultiplierRega(16 bit)CLKSetCounter(4 bit)Look up tableAdderRegb(18 bit)Regb_outRegb_out[17:16]Regb_out[15:0]   MultiplierCLKSetCLKStartXSetCLKdonebusyStart Data path ResultIntegralpartResultFractionalpart   Finish   Figure 1:Data path and Controller Figure 8-a: result waveforms(x= 1001100110001101, result = 01.1101001001011001)      Figure 8-b: result waveforms(x= 0101001000010010, result = 01.0110000010111010)  Figure 9: result using $display command
Search
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks