Wednesday, February 29, 2012

Artcfox's awesome TLC5940 lib and the Arduino IDE

Artcfox has released an absolutely brilliant library and book (!) on how to get an ATMega328 talking to a TLC5940 chip: Demystifying the TLC5940

If you want to control lots of LEDS (RGB or otherwise) the TLC5940 is your chip, unfortunately it has a somewhat complex interface. The TLC5940 is pretty ubiquitous in the Arduino world, but Arctfox (being the hardcore kind of guy he is), wrote his library in pure AVR C.

This does not play well with the Arduino IDE. So for those of you out there that would like to have the ease of use of the IDE and get access to this awesome code, I've done the hard work for you. Here is a breakdown of what is required.

First off some background. I am using the example code from chapter 9 of the source zip.
This is a full example including the library and a test program.

To convert the code we need to understand some things about compiling in the IDE:

  • The example contains a main.c file and in this file is defined the program entry point:
    int main(void)

    We need to expose this function to the IDE so that we can call the example code. In order to avoid conflicts, we need to rename it to something other than main (the Arduino sketch has a hidden main function that calls setup() and loop()). So we name it something creative like:
    int themain(void)

  • Now we need to define this function in a header so it can be shared with the sketch. To accomplish this, create a main.h file with these contents:
    #ifndef __MAIN_H
    #define __MAIN_H

    int themain(void);


  • To include and correctly link to the C function, declare this at the top of your sketch:
    extern "C"
    #include "themain.h"
    This tells the compiler to include the header as a C header and to expect function names defined in the header to be in a C linker format.

  • The IDE does not support C99 mode, this means that code like this:
    for( int i = 0; i < 10; i++ )

    will not compile. The declaration of the variable in the for loop is not allowed and needs to be changed to:
    int i;
    for( i = 0; i < 10; i++ )
    There are a few places in the code where this occurs, fixing this is left as an exercise for the reader... :) The IDE should quickly point out these issues!

  • Now all we have to do is get the #defines correct so that the code will compile and be configured correctly. This gets interesting. Artcfox has used some makefile ninja wizardry to accomplish this which I have converted into a helper header called "ifdefs.h". This header needs to be included at the top of the "tlc5940.h" file after the #include <avr/io.h> line:
    #pragma once

    #include <stdint.h>
    #include <avr/io.h>

    #include "ifdefs.h"
    This file (which is basically a port of the makefile to header definitions) looks like this:
    #define PB1 PINB1
    #define PB2 PINB2
    #define PB3 PINB3
    #define PB5 PINB5

    #define PC0 PINC0
    #define PC1 PINC1
    #define PC2 PINC3

    #define PD1 PIND1
    #define PD3 PIND3
    #define PD4 PIND4
    #define PD5 PIND5
    #define PD6 PIND6
    #define PD7 PIND7

    // Defines the number of TLC5940 chips that are connected in series
    #define TLC5940_N 4

    // Flag for including functions for manually setting the dot correction
    // 0 = Do not include dot correction features (generates smaller code)
    // 1 = Include dot correction features (will still read from EEPROM by default)
    #define TLC5940_INCLUDE_DC_FUNCS 1

    // Flag for including efficient functions for setting the grayscale
    // and possibly dot correction values of four channels at once.
    // 0 = Do not include functions for ganging outputs in groups of four
    // 1 = Include functions for ganging outputs in groups of four
    // Note: Any number of outputs can be ganged together at any time by simply
    // connecting them together. These function only provide a more efficient
    // way of setting the values if outputs 0-3, 4-7, 8-11, 12-15, ... are
    // connected together
    #define TLC5940_INCLUDE_SET4_FUNCS 0

    // Flag for including a default implementation of the TIMER0_COMPA_vect ISR
    // 0 = For advanced users only! Only choose this if you want to override the
    // default implementation of the ISR(TIMER0_COMPA_vect) with your own custom
    // implemetation inside main.c
    // 1 = Most users should use this setting. Use the default implementation of the
    // TIMER0_COMPA_vect ISR as defined in tlc5940.c
    #define TLC5940_INCLUDE_DEFAULT_ISR 1

    // Flag for including a gamma correction table stored in the flash memory. When
    // driving LEDs, it is helpful to use the full 12-bits of PWM the TLC5940 offers
    // to output a 12-bit gamma-corrected value derived from an 8-bit value, since
    // the human eye has a non-linear perception of brightness.
    // For example, calling:
    // TLC5940_SetGS(0, 2047);
    // will not make the LED appear half as bright as calling:
    // TLC5940_SetGS(0, 4095);
    // However, calling:
    // TLC5940_SetGS(0, pgm_read_word(&TLC5940_GammaCorrect[127]));
    // will make the LED appear half as bright as calling:
    // TLC5940_SetGS(0, pgm_read_word(&TLC5940_GammaCorrect[255]));
    // 0 = Do not store a gamma correction table in flash memory
    // 1 = Stores a gamma correction table in flash memory

    // Flag for forced inlining of the SetGS, SetAllGS, and Set4GS functions.
    // 0 = Do not force inline the calls to Set*GS family of functions.
    // 1 = Force all calls to the Set*GS family of functions to be inlined. Use this
    // option if execution speed is critical, possibly at the expense of program
    // size, although I have found that forcing these calls to be inlined often
    // results in both smaller and faster code.
    #define TLC5940_INLINE_SETGS_FUNCS 1

    // Flag to enable multiplexing. This can be used to drive both common cathode
    // (preferred), or common anode RGB LEDs, or even single-color LEDs. Use a
    // P-Channel MOSFET such as an IRF9520 for each row to be multiplexed.
    // 0 = Disable multiplexing; library functions as normal.
    // 1 = Enable multiplexing; The gsData array will become two-dimensional, and
    // functions in the Set*GS family require another argument which corresponds
    // to the multiplexed row they operate on.

    // The following option only applies if TLC5940_ENABLE_MULTIPLEXING = 1
    // Defines the number of rows to be multiplexed.
    // Note: Without writing a custom ISR, that can toggle pins from multiple PORT
    // registers, the maximum number of rows that can be multiplexed is eight.
    // This option is ignored if TLC5940_ENABLE_MULTIPLEXING = 0
    #define TLC5940_MULTIPLEX_N 3

    // Flag to use the USART in MSPIM mode, rather than use the SPI Master bus to
    // communicate with the TLC5940. One major advantage of using the USART in MSPIM
    // mode is that the transmit register is double-buffered, so you can send data
    // to the TLC5940 much faster. Refer to schematics ending in _usart_mspim for
    // details on how to connect the hardware before enabling this mode.
    // 0 = Use normal SPI Master mode to communicate with TLC5940 (slower)
    // 1 = Use the USART in double-buffered MSPIM mode to communicate with the
    // TLC5940 (faster, but requires the use of different hardware pins)
    // WARNING: Before you enable this option, you must wire the chip up differently!
    #define TLC5940_USART_MSPIM 1

    // Defines the number of bits used to define a single PWM cycle. The default
    // is 12, but it may be lowered to achieve faster refreshes, at the expense
    // of the ISR being called more frequently. If TLC5940_INCLUDE_GAMMA_CORRECT = 1
    // then changing TLC5940_PWM_BITS will automatically rescale the gamma correction
    // table to use the appropriate maximum value, at the expense of precision.
    // 12 = Normal 12-bit PWM mode. Possible output values between 0-4095
    // 11 = 11-bit PWM mode. Possible output values between 0-2047
    // 10 = 10-bit PWM mode. Possible output values between 0-1023
    // 9 = 9-bit PWM mode. Possible output values between 0-511
    // 8 = 8-bit PWM mode. Possible output values between 0-255
    // Note: Lowering this value will decrease the amount of time you have in the
    // ISR to send the TLC5940 updated values, potentially limiting the
    // number of devices you can connect in series, and it will decrease the
    // number of cycles available to main(), since the ISR will be called
    // more often. Lowering this value will however, reduce flickering and
    // will allow for much quicker updates.
    #define TLC5940_PWM_BITS 12

    // Determines whether or not GPIOR0 is used to store flags. This special-purpose
    // register is designed to store bit flags, as it can set, clear or test a
    // single bit in only 2 clock cycles.
    // Note: If enabled, you must make sure that the flag bits assigned below do not
    // conflict with any other GPIOR0 flag bits your application might use.
    #define TLC5940_USE_GPIOR0 1

    // GPIOR0 flag bits used
    #if TLC5940_USE_GPIOR0 == 1
    #define TLC5940_FLAG_GS_UPDATE 0
    #define TLC5940_FLAG_XLAT_NEEDS_PULSE 1

    // BLANK is only configurable if the TLC5940 is using the USART in MSPIM mode
    #if TLC5940_USART_MSPIM == 1
    #define BLANK_DDR DDRD
    #define BLANK_PORT PORTD
    #define BLANK_PIN PD6

    // DDR, PORT, and PIN connected to DCPRG
    #define DCPRG_DDR DDRD
    #define DCPRG_PORT PORTD
    // DCPRG is always configurable, but the default pin needs to change if
    // the TLC5940 is using USART MSPIM mode, because PD4 is needed for XCK
    #if TLC5940_USART_MSPIM == 1
    #define DCPRG_PIN PD3
    #define DCPRG_PIN PD4

    // DDR, PORT, and PIN connected to VPRG
    #define VPRG_DDR DDRD
    #define VPRG_PORT PORTD
    #define VPRG_PIN PD7

    // DDR, PORT, and PIN connected to XLAT
    #if TLC5940_USART_MSPIM == 1
    #define XLAT_DDR DDRD
    #define XLAT_PORT PORTD
    #define XLAT_PIN PD5
    #define XLAT_DDR DDRB
    #define XLAT_PORT PORTB
    #define XLAT_PIN PB1

    // The following options only apply if TLC5940_ENABLE_MULTIPLEXING = 1
    // DDR, PORT, and PIN registers used for driving the multiplexing IRF9520 MOSFETs
    // Note: All pins used for multiplexing must share the same DDR, PORT, and PIN
    // registers. These options are ignored if TLC5940_ENABLE_MULTIPLEXING = 0

    // List of PIN names of pins that are connected to the multiplexing IRF9520
    // MOSFETs. You can define up to eight unless you use a custom ISR that can
    // toggle PINs on multiple PORTs.
    // Note: All pins used for multiplexing must share the same DDR, PORT, and PIN
    // registers. These options are ignored if TLC5940_ENABLE_MULTIPLEXING = 0
    // Also: If you add any pins here, do not forget to add those variables to the
    // MULTIPLEXING_DEFINES flag below!
    #define R_PIN PC0
    #define G_PIN PC1
    #define B_PIN PC2


    // ---------- End TLC5940 Configuration Section ----------

  • Now all that is left to do is round out the sketch:
    extern "C"
    #include "main.h"

    void setup()

    void loop()

And there you have it! Using these techniques I have built an entire firmware and they can be applied to many C libraries out there. The ease of the Arduino IDE as a compiler and uploader can help lower complexity barriers in developing your code.

Hopes this help you out there in Arduino land!


  1. Did anybody get this working?? When I try to compile this in the Arduino IDE i get the error:

    In file included from main.cpp:15:
    main:6: error: previous declaration of 'int themain()' with 'C++' linkage
    themain.h:3: error: conflicts with new declaration with 'C' linkage

    Any ideas?

  2. Hi,

    Sorry about getting back to you so late, comment notifications seem to have disappeared into the ether!

    +Kev Griss where is the main.cpp file coming from? This should be main.c. Did you rename the file by any chance?


  3. Hi Peter,

    Thanks for replying... and yes it appear notifications are nolonger working. I don't have a main.cpp but I think the makefile is referencing it with

    276: .S.o:
    277: $(COMPILE) -x assembler-with-cpp -c $< -o $@

    I'll have another go when I have time to play


  4. Hello Peter, This looks interesting. I went over the schematic and everything looks like its there other than the capacitor which may not be necessary. Have you had any luck running the dot correction and grey scale modes separately as of yet???

    I have one of these working using another pinout but am looking forward to setting up a second device to see what can be done with this arrangement.

    Best M

    1. Hi,

      What schematic(s) are you looking at? IIRC Matt has some caps for the xtal on his standalone board design, whereas my articles have been specifically on porting to the Arduino (sorry bit confused here :)).

      If you look at some my later posts on this, I have the dot correction code (well Matt's) going and in the next instalment I am going to get his CH9 code (which is basically the first full version) running:

  5. Hello Peter - do you have any plans to release this as a ready-to-use Arduino library?

    1. Hi,

      I'm not sure. The changes are fairly easy to implement, but if there was a sufficient call to do so I guess I could be persuaded :) I'd need to discuss this with Matt first to see if he is keen on the idea.

      Ideally I'd like to finish up the dev so that this can run at the maximum possible level of capability and implementation (basically full CH9 compliance with multiplex LEDs but no accelerated UART stuff).

    2. This would be great! The changes are not sooo trivial, I think many people will like a ready-to-use library for Arduino.

      The existing library is under the LGLP license, so any changes / extensions etc. can again be published under the same license.