TigerSHARC processor and evaluation board

1
TigerSHARCprocessor and evaluation board

• Different capabilitiesDifferent functionality

2
Concepts tackled

• Differences between processor and evaluation
  board
• Functionality present on TigerSHARC evaluation
  board
• Using the A/D and D/A during Lab. 1
• Developing the code for a C Moving sound
  source and a C FIR filter using a test driven
  development approach (TDD)
• Assembly code and optimized assembly code
  versions of this code will be developed during
  Lab. 2 and Lab. 3 and validated using the tests
  developed during Lab. 1.
• A dual core version of the FIR filter will be
  attempted during Lab. 4. This is a very new
  laboratory to me we will see.

3
TS201S Evaluation board contains 2 processors P0
(DSPA) and P1 (DSPB)
TWO TS201S processors Error messageswhen
linking about P0 and P1 mean problems with code
on specific processor
4
Evaluation board Link Ports
Designed for star Configuration Up to 8
processorsworking together P0 to P7
(selectableby switches on backof evaluation
board(Never tried Dontdestroy link ports)
5
Audio connectors.Max input voltage around 1 V
p-pwith no DC. Too great a voltage destroys the
board Use MPEG or other players and keep Volume
down Damage and distortion issues

• JTAG ONLY AVALIABLE IN 3rd floor labs
• Higher speed communication rates than USB
  (factor of 20 - 100)
• ICT318 TigerSHARC (some stations).
• ICT320 Blackfin (some stations).

6
Buttons and LEDs splitbetween processors Much
stuff on the boardshared between
processors(but NOT all)
7
Evaluation BoardArchitecture
P0 P1
8
Two processor operations in ENCM515

• Lab. 1, 2 and 3
• During TESTS only 1 processor being used as we
  check the functionality of various functions
  written in C (lab1), partially optimized
  assembly (Lab. 2), highly optimized but not
  ultimate (Lab. 3). Other processor is halted
• WARNING Make sure that you run the processor
  that you loaded the code into
• When running the code, one processor is idle
• Lab. 4
• Plans to have the work load split between two
  processors.

9
Processorscan access each othersINTERNALmemor
y and registers
Processorscan shareEXTERNALmemory Memory alloc
ationdepends on ID
10
Note labeled processor A and Brather than
processor 0 and 1
11
Note labeled processor A and Brather than
processor 0 and 1
12
How to use the evaluation board capability

• Read a lot in the hardware manual
• Modify some of the example programs provided in
• C\Program Files\Analog Devices\VisualDSP
  4.5\TS\Examples\ADSP-TS201 EZ-Kit Lite
• and then Read a lot in the hardware manual

13
Laboratory 1

• This makes use of the analog to digital converter
  and the digital to analogue converter to provide
  audio sound signals that we can process with the
  high speed algorithms we develop during the
  course
• By the end of the course we will need to learn
  about
• Interrupts which, how many, why, how?
• A2D, D2A which, how many, why, how?
• We need to learn about all this stuff before the
  end of the course but not now Now we just
  want to use it
• The C programming environment provides
  functions (interrupt( ), interruptf( ),
  interrupts( ) that handle all the interrupt
  overhead, and then call our functions as a
  subroutine.

14
ECE-ADI-Project

• Was asked to develop (as part of Analog Devices
  University Ambassadorship) to develop
  demonstration programs and place on the web
• Have now developed
• SHARC ADSP-21061 (Previous 515)
• Audio project utilities
• Blackfin ADSP-BF533 (Current ENCM415 , ENCM417
  in 2007)
• Audio project utilities, Video Project utilities
• Blackfin ADSP-BF533 (Current ENCM491 Nygren)
• Real time operating systems
• TigerSHARC ADSP-TS201 (Current ENCM515,
  ENCM619.23)
• Embedded Test Driven Development Environment
  (E-TDD)
• Blackfin, TigerSHARC, SHARC, Microsoft studio
• Find at http//www.enel.ucalgary.ca/People/Smith/
  ECE-ADI-Project/Index/index.htm

15
Using the Lab. 1 code

• You should have already done the Familiarization
  exercise
• Part of the Familiarization exercise involved
  loading the ENCM515Lab directory onto your
  UDrive
• An audio talk-through program is provided at Lab.
  1
• Audio-In signal is sampled (A/D) automatically
  and placed in known input location in memory
• An interrupt is made to happen each time a sample
  occurs (44kHz)
• During the interrupt, the input value is
  processed and the processed value is placed in
  a known output location in memory.
• The output value is then automatically moved to
  the D/A for reconstruction.
• WARNING NEVER START THE PROCESSOR WITH YOUR
  EAR-PHONES ON YOUR EARS. IF YOU HAVE THE
  PROCESSING INCORRECT, YOU CAN GET AN INFINITELY
  LARGE SIGNAL COME OUT WHICH HURTS YOUR EARS
• Start the processor with the earphones on the
  desk, then bring the earphones near your ears
  ONLY when it is safe to do so.

16
Lab. 1 Task 1 conceptConcept ? idea exact,
details inexact

• Run the (modified) audio-talk-through program
• Each time an interrupt occurs
• ProcessData(INCBuffer ? OUTCBuffer) // Four
  circular buffers.// Lab. 1 Task 1 ProcessData(
  ) is just a copy
• The oldest processed sound sample is send to
  the D/A. There is a sound sample from both left
  and right channels (circular buffers) leftOUT
  leftOUTCBuffer leftOUTIndex - 128
  rightOUT rightOUTCBuffer rightOUTIndex - 128
  
• The newest input sound sample (from A/D)
  overwrites the oldest sound sample.
  leftINCBuffer leftINputIndex sample
  (leftIN rightIN) / 2 rightINCBuffer
  rightINputIndex sample (leftIN rightIN)
  / 2
• RESULT You hear a slightly delayed version of
  the sound in each ear (delay 128 / 44000
  seconds). It will sound normal

17
Lab. 1 Task 1 -- concept

• Why average the left and the right signals
  rather than putting each signal into each own
  array.
• The reason will be come obvious as we proceed
  through Labs. 1, 2 and 3
• Basically the ProcessData( ) function gets too
  complicated to handle.
• Current plan is, in Lab. 4, to finish with one
  TigerSHARC core doing ProcessData( ) for the left
  channel and the other TigerSHARC core doing all
  the processing on the right channel

18
Lab. 1 Task 2A In CLeft no delay, right
with delay

• Run the (modified) audio-talk-through program
• Each time an interrupt occurs
• ProcessData(INCBuffer ? OUTCBuffer) // Four
  circular buffers.// Lab. 1 Task 1 ProcessData(
  ) is just a copy
• The oldest processed sound sample is send to
  the D/A. There is a sound sample from both left
  and right channels (circular buffers) leftOUT
  leftOUTCBuffer leftOUTIndex - 0 rightOUT
  rightOUTCBuffer rightOUTIndex -256
• The newest input sound sample (from A/D)
  overwrites the oldest sound sample.
  leftINCBuffer leftINputIndex sample
  (leftIN rightIN) / 2 rightINCBuffer
  rightINputIndex sample (leftIN rightIN)
  / 2
• RESULT With no delay on the sound going to the
  left ear , and delay on the sound going to the
  right ear, the brain THINKS the sound comes
  from close to the left ear

19
Lab. 1 Task 2B In CLeft delay, right with
no delay

• Run the (modified) audio-talk-through program
• Each time an interrupt occurs
• ProcessData(INCBuffer ? OUTCBuffer) // Four
  circular buffers.// Lab. 1 Task 1 ProcessData(
  ) is just a copy
• The oldest processed sound sample is send to
  the D/A. There is a sound sample from both left
  and right channels (circular buffers) leftOUT
  leftOUTCBuffer leftOUTIndex - 256 rightOUT
  rightOUTCBuffer rightOUTIndex - 0
• The newest input sound sample (from A/D)
  overwrites the oldest sound sample.
  leftINCBuffer leftINputIndex sample
  (leftIN rightIN) / 2 rightINCBuffer
  rightINputIndex sample (leftIN rightIN)
  / 2
• RESULT With delay on the sound going to the
  leftear , and no delay on the sound going to the
  right ear, the brain THINKS the sound comes
  from close to the right ear

20
Lab. 1 Task 3 ConceptLeft / right changing
delays

• The oldest processed sound sample is send to
  the D/A. There is a sound sample from both left
  and right channels (circular buffers) leftOUT
  leftOUTCBuffer leftOUTIndex - 128
  LEFTIndexrightOUT leftOUTCBuffer
  rightOUTIndex 128 RIGHTIndex
• LEFTIndex LEFTIndex 1
• RIGHTIndex RIGHTIndex 1
• RESULT
• Sound seems to come from in front of you to start
  with
• With changing delays, the left sound will get
  more delayed, and the right sound will get less
  delayed
• The sound will move from in front of you, over to
  your right ear.
• When the sound is at your right ear, the indexes
  change direction (right increases, left
  decreases) till the sound comes from left side,
  and so on

21
Lab. 1 Task 3. Use TDD approach to develop C
code

• Work with customer to check that the tests
  properly express what the customer wants done.
  Iterative process with customer heavily
  involved Agile methodology.

CUSTOMER DEVELOPER
22
Lab. 1 Task 3. Use TDD approach to develop C
code

• void AdjustDelayOffsetCPP (long IndexOffset,
  long IndexOffsetDirection, long
  DelayLength)
• You work in the Lab1Tests DEVELOPERS directory,
  with the code for AdjustDelayOffsetCPP( ) placed
  in the LAB1 CUSTOMERS directory.
• The customer provides you information about how
  this function works to adjust the index so that
  the sound will move correctly
• The information is in the form of a case study
  based on customer developer consultation for a
  simple example.
• The information is expressed as an executable
  specification document
• You get the code to work for the simple example,
  and add additional tests (part of your report) as
  needed to confirm that the code will work

23
Lab. 1 Task 4. -- 2 minutes maximumUse
AdjustDelayOffsetCPP( ) for real

• void AdjustDelayOffsetCPP (long IndexOffset,
  long IndexOffsetDirection, long
  DelayLength)
• If you did the testing correctly, then this task
  simple means
• Recompiling the code in the LAB1 CUSTOMERS
  directory 2 minutes
• Demonstrate to T. A. and ask (politely) please
  give me full marks.
• Note that the same tests will be used Lab. 2 for
  AdjustDelayOffsetASM andLab. 3 for
  AdjustDelayOffsetOPtimizedASM

24
Lab. 1 Tasks 5, 6 and 7float FIRDebugModeCPP(floa
t coeffs, float inputData, int
index, int coeff_length)

• Task 5 Work with the customer to set up some
  tests for the FIR operation with is essentially
  SUM (coeffs) (inputData)
• Task 6 Another 2 minute job recompile the
  code in the Lab 1 CUSTOMER DIRECTORY and
  demonstrate.
• Task 7 Another 2 minute job
• The code will work with the arrays small (N lt
  128) but not for larger arrays chronic sound
  distortion
• Recompile the code using the C optimizer and
  show your FIR filter works no sound distortion

25
Speed improvements when we use the C optimizer

• Later we will need to modify the timing tests
  from the Familiarization Laboratory so we can
  check the speed of the filters, and delay
  adjustment algorithms
• We may not HEAR a lot of distortion, but all
  the processing must be completed in 1 / 44000 s
• We must have efficient code as we increase the
  length of the data being processed
• As one exercise we will examine the code for the
  Analog Devices Reverb program which simulates the
  sound of an opera filter using discrete Fourier
  transforms on 64000 points at a time!
• We will look at that code and say oh my!
  rather than writing it

26
Concepts tackled

• Differences between processor and evaluation
  board
• Functionality present on TigerSHARC evaluation
  board
• Using the A/D and D/A during Lab. 1
• Developing the code for a C Moving sound
  source and a C FIR filter using a test driven
  development approach (TDD)
• Assembly code and optimized assembly code
  versions of this code will be developed during
  Lab. 2 and Lab. 3 and validated using the tests
  developed during Lab. 1.
• A dual core version of the FIR filter will be
  attempted during Lab. 4. This is a very new
  laboratory to me we will see.