EECS 461 Winter 2009 Lab 2: Quadrature Decoding using the eTPU
Contents
1. EECS 461 Winter 2009 Lab 2 Quadrature Decoding using the eT PU 1 Overview The purpose of this lab is to learn about quadrature decoding and apply knowledge of sampling To do so we will use the Quadrature Decoding function of the MPC5553 enhanced Time Processor Unit eTPU as an angular position sensor 1 1 Encoder When developing a controller for a mechanical device with moving parts it is often necessary to gather information about the motion of the device by using an encoder The haptic device we will be using in lab has an optical encoder for the purpose of obtaining position and velocity information Optical encoders typically use optical gratings to produce two square wave signals that are 90 degrees out of phase This is an example of a quadrature encoded signal The graph below shows how the two signals vary as the shaft rotates in one direction a I LAF LI Le Figure 1 Output for a 2 channel optical encoder rotating in one direction 1 2 Time Processor Unit eTPU The time processor unit is described in the MPC5553 User s Manual in Chapter 18 The enhanced time processing unit eTPU is a new timing unit featured on the MPC5553 MPC5554 microcontroller that operates in parallel with the MPC5553 MPC5554 core CPU The eTPU does the following e Executes programs independently from the host core e Detects and precisely records timing of input events e Generates complex output waveforms EECS 461 Embedded Control Sy2. ant to discuss possible reasons for such a failure Post Lab Assignment What are two reasons that you think the quadrature decode function might stop working at high speeds The MPC5553 eTPU can be programmed to switch automatically from slow to normal to fast mode based on encoder speed Why would the quadrature decode function work at a higher speed when in fast mode Hint Read section 3 1 of AN2842 concerning how fast modes work Using the information about the encoder wheel given to you in fqd c what type of encoder are we using in the haptic wheel e g 32 pulse 256 pulse etc We now analyze situations for which the QD function may fail to correctly update the position count a The system clock for our setup is 40MHz Given that we are running the QD function in slow mode determine the maximum rate the haptic wheel can reach revolutions sec before the eT PU fails See section 3 7 in Application Note AN2842 for eT PU performance b Assume that the haptic wheel is running at the maximum speed allowed by the QD function i e at the speed computed in part a What is the slowest possible periodic rate that the the CPU can read the TPU position counter register while still knowing whether the counter value is increasing or decreasing You may assume that the wheel will be spinning at this maximum rate in one direction Calculate this rate based on both a 16 bit what we used and 24 bit what is available in the hardwa
3. desired value Note that the quantity CURR_FQDPC PREV_FQDPC can be changed to PREV_FQDPC CURR_FQDPC This will only change the directions for turning the wheel to increment or decrement the counter Note that LAST_TOTAL and PREV_FQDPC will have to be static variables in order to maintain their values between calls to updateCounter Note also that PREV_FQDPC must have the same initial value as the position count register in fqd c or there will be an initial offset in the count 3 Pre Lab Assignment Pre lab questions must be done individually and handed in at the start of your lab section You must also with your partner design an initial version of your software before you come to your lab section Your EECS 461 Embedded Control Systems 3 Winter 2009 Lab 2 Quadrature Decoding using the eTPU software does not need to be completely debugged but obviously more debugging before coming to lab means less time you will spend in the lab 1 Fill in the truth table in Table 1 for quadrature decoding X _ and Y _ refer to the previous state while X and Y refer to the current state The valid actions are increment decrement and error A few have already been done for you 2 A 32 pulse encoder is an encoder that provides a square wave signal with 32 periods or 64 edges rising and falling on each channel per revolution What is the resolution smallest measurable amount in degrees of the quadrature decode function operating in
4. ing A This creates 16 possible transitions 4 increment the counter 4 decrement the counter and 8 are errors see the prelab below for more information When the function receives a sequence that is an error the counter is unchanged You can also program the function to switch to normal and fast modes based on the speed of the encoder wheel Normal mode applies when the speed is greater than a defined slow to normal threshold but less than a specified normal to fast threshold In normal mode the counter is updated by one for each valid transition on either channel as in slow mode but the direction is not calculated the last direction calculated in the slow mode is kept When in fast mode the function only monitors the signal on the primary channel and updates the counter by 4 on each rising edge transition 2 Design Specifications 2 1 Hardware On the interface board you will see a header bank labeled TPU A 0 7 next to the DIP switches These channels can interface with up to four quadrature encoders the chip itself supports up to 32 channels per TPU 8 channels are made available on the interface board and two channels are needed for each quadrature encoder The pins for TPU A channels 0 and 1 are connected to the haptic wheel over the parallel cable You may use the header bank to measure the voltage signals coming from the quadrature encoder channels Test loops for channels 0 and 1 are located next to the header bank f
5. or easy connection of oscilloscope probes The digital I O interface from Lab 1 will be used again 2 2 Software fqd h and fqd c You need to make a C file called fqd c from the fqd_template c file that contains four functions named init_FQD ReadFQD_pc updateCounter and updateAngle see Notes on Casting below The prototypes and details of these functions can be found in the fqd h file It is very important that you follow these prototypes and the comments in the fqd_template c file closely Place fqd c file in your 1ib directory and the fqd h file in your include directory EECS 461 Embedded Control Systems 2 Winter 2009 Lab 2 Quadrature Decoding using the eTPU Makefile To simplify the compilation process a makefile has been provided to you Note that for the makefile to work you need to make sure that all of your files are in their proper directories as specified in the lab handouts The gmake command must be run from your lab2 directory Freescale header files used in this lab etpu_qd h etpu_set h 2 3 Notes on Casting In previous years students have sometimes observed a peculiar problem in which their updateCounter function failed to update the position correctly The specific behavior observed was that the top 8 bits of the bar LEDs would alternate between being all on 0xFFXX or all off 0x00XX In the following exercise we will see that this problem is due to the subtleties involved in typecasting correctl
6. re position counter size c Relate that periodic rate to the number of processor cycles that can be executed between checks Include well documented copies of your fqd c lab2 c and lab2angle c files If you have comments that you feel would help improve the clarity or direction of this assignment please include them following your postlab solutions EECS 461 Embedded Control Systems 6 Winter 2009 Lab 2 Quadrature Decoding using the eTPU 6 Figures for Pre Lab Questions These may be collected at the beginning of lab Previous State Current State Action X 1 Yi 1 Aaya Error Increment or Decrement Decrement Increment Error Table 1 Encoder Transition Table for pre lab problem 1 0x0000 Figure 2 Sample output for pre lab problem 3a The 16 bit counter is initialized to 0x0000 EECS 461 Embedded Control Systems T Winter 2009 Lab 2 Quadrature Decoding using the eT PU B 0x0000 Figure 3 Sample output for pre lab problem 3b The 16 bit counter is initialized to 0x0000 EECS 461 Embedded Control Systems 8 Winter 2009
7. slow mode with a 32 pulse encoder With a 512 pulse encoder 3 a On Figure 2 draw lines where the quadrature decode function would update the counter and label each line with the value the counter would hold Assume the function is operating in slow mode b Repeat Part a for Figure 3 but assume that the function is running in fast mode now 4 Briefly not more than 60 words summarize the differences between slow normal and fast decoding 5 The eTPU is a separate co processor that operates in parallel with the 5553 core What is the advantage of having a separate processor 6 Complete fqd c EECS 461 Embedded Control Systems A Winter 2009 Lab 2 Quadrature Decoding using the eTPU 4 In Lab Assignment 4 1 Reading the Counter 1 Write a simple C program called ab2 c that uses the init_FQD and ReadFQD_pc functions from fqd c Create a uint16_t variable counter in your program and store the values returned by ReadFQD_pc in this variable You will be able to examine the count value of the decoder by looking at this variable when debugging 2 Compile and debug your lab2 c program using the makefile by entering gmake at an DOS prompt Make sure that all your files are in their respective directories or the compilation will fail 3 Power up the board and use the P amp E Debugger to download the lab2 elf file to the board Make sure you have the encoder connected to the interface board correctly ask the GSI if you are uns
8. stems 1 Winter 2009 Lab 2 Quadrature Decoding using the eTPU e Is controlled by the core without a requirement for real time host processing Software for specific eT PU functions may be written by the user or for common functions obtained from Freescale Freescale Application Note AN2842 on the course website describes the quadrature decode function Read the application note and look over Chapter 18 of the user s manual before coming to the lab You can find the necessary information concerning TPU configuration registers in Section 18 4 1 3 Quadrature Decoding eT PU Function The MPC5553 eTPU operates in one of three modes slow normal or fast We will operate the eT PU in the slow quadrature decode mode The quadrature decode function uses two eTPU channels to decode quadrature signals into a 24 bit counter Note In this lab we will use only 16 bits of the count register When in slow mode the quadrature decode function is able to determine direction by looking at both the rising and falling edges of each signal If the primary channel is leading the secondary channel the 24 bit counter increments If the primary channel is trailing the secondary channel the counter is decremented The function does this by comparing the current state of the two channels with the state before the most recent rising falling edge For example when channel B goes from low to high and channel A remains high the counter is incremented since B is trail
9. ure 4 In the debugger add the variable counter to the Variables Window Right click in the Variables Window and select Add Variable then type the variable name counter into the menu Display the value in Decimal and use the Default data type Click OK and you should now see counter in the Variables Window 5 Run the lab2 c program by clicking the High Level Go button in the debugger Move the wheel of the haptic device slightly Then stop running lab2 c by clicking the Stop button in the debugger which is shown as a black square The value of counter in the Variables Window should have changed If not correct the problem before proceeding 6 Reset the target by clicking on the Reset lightning bolt button in the debugger You may have to press the button twice until you see code run by in the Status Window and the Source Window is cleared Download and run lab2 elf again and turn the wheel of the haptic device one full revolution How much does the FQD function increment during one full revolution of the haptic wheel 4 2 Overflow and Underflow 7 Use your knowledge from the Digital I O Lab to output the value of posCount to the 16 LEDs Recompile your code using the same command as above Download and run your code 8 With your new code running turn the wheel until the LED counter shows 0x8000 This is the center position for the counter Now turn the wheel until the LED counter either underflows or overflows Ho
10. w many revolutions of the wheel does it take to overflow a 16 bit register starting at this center value 4 3 Faulty Casting 9 Modify your code so that you use your updateCounter function to keep track of the position count Output the middle 16 bits of your position counter to the LEDs You can do this by using a bitwise shift Verify that your updateCounter function works correctly 10 Repeat step 9 but modify updateCounter to use the faulty casting method from the Notes on Casting section of this lab Verify that there is in fact a problem with this code and show that problem to the GSI 4 4 Angle Calculation 11 Write a simple C program called lab2angle c that outputs how much the haptic wheel turned coun terclockwise in degrees to the LEDs by using the updateAng1le function in fqd c Note that similar to what you did in updateCounter you will need a static float in updateAngle to remember the current angle Do not call the function updateCounter from updateAngle Compile your program by entering gmake angle at an DOS prompt Download the lab2angle elf file to the board and demon strate the functionality to the GSI EECS 461 Embedded Control Systems 5 Winter 2009 Lab 2 12 5 Quadrature Decoding using the eTPU The post lab asks questions about why the quadrature decode function may fail to correctly keep track of position when the wheel is running at sufficiently high speeds Before you leave the lab you might w
11. y The following code does not work correctly NEW_TOTAL and LAST_TOTAL are int32_ts and CURR_FQDPC and PREV_FQDPC are uint16_ts int32_t updateCounter setup return LAST_TOTAL CURR_FQDPC PREV_FQDPC This code will work int32_t updateCounter setup return LAST_TOTAL int16_t CURR_FQDPC PREV_FQDPC Note the difference between these two examples they are identical except that the quantity CURR FQDPC PREV_FQDPC of type uint16_t is cast to an int16_t before being added to total This is needed so that the quantity is sign extended when it is turned into a long so as to be added with total For more information on sign extension read section 4 2 of Computer Organization and Design The Hardware Sortware Interface By John L Hennessy and David A Patterson For more information of casting read section 2 7 of The C Programming Language by Brian W Kernighan and Dennis M Ritchie To make this clearer here is an example LAST_TOTAL O0x00007FFF CURR_FQDPC 0OxFFFF PREV_FQDPC 0x0000 CURR_FQDPC PREV_FQDPC yields the unsigned quantity OxFFFF which when changed to an int32_t a signed value becomes Ox0000FFFF This is not the desired result It now represents a large positive number instead of 1 If the quantity CURR_FQDPC PREV_FQDPC is first cast as an int16_t as in the second example then when it is converted to an int32_t it is sign extended to OxFFFFFFFF or 1 the
Download Pdf Manuals
Related Search