hyperion.ng
Public Member Functions | Static Public Member Functions | List of all members
LedDeviceLpd8806 Class Reference

Implementation of the LedDevice interface for writing to LPD8806 led device. More...

#include <LedDeviceLpd8806.h>

Inheritance diagram for LedDeviceLpd8806:
ProviderSpi LedDevice

Public Member Functions

 LedDeviceLpd8806 (const QJsonObject &deviceConfig)
 Constructs specific LedDevice. More...
 
virtual bool init (const QJsonObject &deviceConfig)
 Sets configuration. More...
 
- Public Member Functions inherited from ProviderSpi
 ProviderSpi ()
 Constructs specific LedDevice.
 
virtual ~ProviderSpi ()
 Destructor of the LedDevice; closes the output device if it is open.
 
int open ()
 Opens and configures the output device. More...
 
- Public Member Functions inherited from LedDevice
 LedDevice (const QJsonObject &config=QJsonObject(), QObject *parent=nullptr)
 
virtual int switchOff ()
 Switch the leds off (led hardware disable)
 
virtual int switchOn ()
 Switch the leds on (led hardware enable), used if reinitialization is required for the device implementation.
 
virtual int setLedValues (const std::vector< ColorRgb > &ledValues)
 
const QString & getColorOrder ()
 Get color order of device. More...
 
void setActiveDevice (QString dev)
 
const QString & getActiveDevice ()
 
void setLedCount (int ledCount)
 
int getLedCount ()
 
void setEnable (bool enable)
 
bool enabled ()
 
int getLatchTime ()
 
bool componentState ()
 

Static Public Member Functions

static LedDeviceconstruct (const QJsonObject &deviceConfig)
 constructs leddevice
 

Additional Inherited Members

- Public Slots inherited from LedDevice
virtual void start ()
 Is called on thread start, all construction tasks and init should run here.
 
- Signals inherited from LedDevice
void enableStateChanged (bool newState)
 Emits whenever the led device switches between on/off. More...
 
void visiblePriorityChanged (const quint8 &priority)
 PIPER signal for Priority Muxer -> LedDevice. More...
 
- Protected Slots inherited from LedDevice
int rewriteLeds ()
 Write the last data to the leds again.
 
- Protected Member Functions inherited from ProviderSpi
int writeBytes (const unsigned size, const uint8_t *data)
 Writes the given bytes/bits to the SPI-device and sleeps the latch time to ensure that the values are latched. More...
 
- Protected Attributes inherited from ProviderSpi
QString _deviceName
 The name of the output device.
 
int _baudRate_Hz
 The used baudrate of the output device.
 
int _fid
 The File Identifier of the opened output device (or -1 if not opened)
 
int _spiMode
 which spi clock mode do we use? (0..3)
 
bool _spiDataInvert
 1=>invert the data pattern
 
spi_ioc_transfer _spi
 The transfer structure for writing to the spi-device.
 
- Protected Attributes inherited from LedDevice
QJsonObject _devConfig
 
Logger_log
 The common Logger instance for all LedDevices.
 
std::vector< uint8_t > _ledBuffer
 The buffer containing the packed RGB values.
 
bool _deviceReady
 
QString _activeDevice
 
int _ledCount
 
int _ledRGBCount
 
int _ledRGBWCount
 
QTimer _refresh_timer
 Timer object which makes sure that led data is written at a minimum rate e.g. More...
 
unsigned int _refresh_timer_interval
 
qint64 _last_write_time
 
unsigned int _latchTime_ms
 

Detailed Description

Implementation of the LedDevice interface for writing to LPD8806 led device.

The following description is copied from 'adafruit' (github.com/adafruit/LPD8806)

Clearing up some misconceptions about how the LPD8806 drivers work:

The LPD8806 is not a FIFO shift register. The first data out controls the LED closest to the processor (unlike a typical shift register, where the first data out winds up at the furthest LED). Each LED driver 'fills up' with data and then passes through all subsequent bytes until a latch condition takes place. This is actually pretty common among LED drivers.

All color data bytes have the high bit (128) set, with the remaining seven bits containing a brightness value (0-127). A byte with the high bit clear has special meaning (explained later).

The rest gets bizarre...

The LPD8806 does not perform an in-unison latch (which would display the newly-transmitted data all at once). Rather, each individual byte (even the separate G, R, B components of each LED) is latched AS IT ARRIVES... or more accurately, as the first bit of the subsequent byte arrives and is passed through. So the strip actually refreshes at the speed the data is issued, not instantaneously (this can be observed by greatly reducing the data rate). This has implications for POV displays and light painting applications. The 'subsequent' rule also means that at least one extra byte must follow the last pixel, in order for the final blue LED to latch.

To reset the pass-through behavior and begin sending new data to the start of the strip, a number of zero bytes must be issued (remember, all color data bytes have the high bit set, thus are in the range 128 to 255, so the zero is 'special'). This should be done before each full payload of color values to the strip. Curiously, zero bytes can only travel one meter (32 LEDs) down the line before needing backup; the next meter requires an extra zero byte, and so forth. Longer strips will require progressively more zeros. *(see note below)

In the interest of efficiency, it's possible to combine the former EOD extra latch byte and the latter zero reset...the same data can do double duty, latching the last blue LED while also resetting the strip for the next payload.

So: reset byte(s) of suitable length are issued once at startup to 'prime' the strip to a known ready state. After each subsequent LED color payload, these reset byte(s) are then issued at the END of each payload, both to latch the last LED and to prep the strip for the start of the next payload (even if that data does not arrive immediately). This avoids a tiny bit of latency as the new color payload can begin issuing immediately on some signal, such as a timer or GPIO trigger.

Technically these zero byte(s) are not a latch, as the color data (save for the last byte) is already latched. It's a start-of-data marker, or an indicator to clear the thing-that's-not-a-shift-register. But for conversational consistency with other LED drivers, we'll refer to it as a 'latch' anyway.

This has been validated independently with multiple customers' hardware. Please do not report as a bug or issue pull requests for this. Fewer zeros sometimes gives the illusion of working, the first payload will correctly load and latch, but subsequent frames will drop data at the end. The data shortfall won't always be visually apparent depending on the color data loaded on the prior and subsequent frames. Tested. Confirmed. Fact.

The summary of the story is that the following needs to be writen on the spi-device: 1RRRRRRR 1GGGGGGG 1BBBBBBB 1RRRRRRR 1GGGGGGG ... ... 1GGGGGGG 1BBBBBBB 00000000 00000000 ... |------—led_1-------—| |------—led_2– -led_n-------—| |-—clear data–

The number of zeroes in the 'clear data' is (#led/32 + 1)bytes (or *8 for bits)

Constructor & Destructor Documentation

§ LedDeviceLpd8806()

LedDeviceLpd8806::LedDeviceLpd8806 ( const QJsonObject &  deviceConfig)

Constructs specific LedDevice.

Parameters
deviceConfigjson device config

Member Function Documentation

§ init()

bool LedDeviceLpd8806::init ( const QJsonObject &  deviceConfig)
virtual

Sets configuration.

Parameters
deviceConfigthe json device config
Returns
true if success

Reimplemented from ProviderSpi.


The documentation for this class was generated from the following files: