Categories
arduino C/C++ circuits Coding Embedded esp32 esp8266 expressif Internet of Things microcontrollers

ESP-32 : How to write multi-threaded application with priority, CPU core affinity, asynchronous non-blocking event driven loop.

Espressif Systems Shanghai launched the game changing low cost ESP-8266 microcontroller in 2014 – a key enabler for embedded Internet of Things (IOT) development.

Internet of Things (IOT)

Adding WiFi 802.11 and Bluetooth LE wireless connectivity to system on a chip (SoC) product costing roughly price of a cup of coffee meant innovators and micro electronics DIY enthusiasts could easily interface edge smart sensor or legacy hardware systems with cloud or mobile devices.

The successor chip ESP-32 (launched 2016) introduced Xtensa LX6 processor and FreeRTOS (a real time operating system (OS) for embedded) – enabling multi-threaded applications running on multi core CPU architecture.

To add context, many even relatively complex tasks especially automation, can be run on tiny embedded 8 bit processor such as Arduino. More complex applications, video, audio signal processing, image recognition or AI require much more compute power.

ESP-32 in the eco-system sits between Arduino and more powerful systems Raspberry Pi, Windows or Embedded Linux.

Sounds great, but how does it work in practice?

ESP8266

In this sketch ( https://github.com/steveio/arduino/tree/master/ESP32GPSMultiTask ) we assemble a UBLOX GPS data logger with an SD card internal storage, LoRaWAN wireless relay and an OLED display.


The goal is to demonstrate running 4 separate non-blocking tasks concurrently using FreeRTOS to schedule tasks, suspend, interrupt, queue and share data in thread-safe way with mutex semaphore locks.

Let’s take a look at the code…. concentrating on FreeRTOS multi-core multi-thread potential, rather than peripherals / sensors.

Data Structures, Multi-thread Semantics & Setup()

First, a struct to encapsulate core data model message (GPS data):

// GPS position data
struct XPosit
{
float Lat;
float Lon;
float Alt;
float Course;
float Speed;
} xPosit;

Next two semaphores to serialise reading and writing tasks to ensure data consistency:

// Semaphores to lock / serialise data structure IO
SemaphoreHandle_t sema_GPS_Gate;
SemaphoreHandle_t sema_Posit;

Then a pointer queue for passing messages between threads / tasks:

// GPS position data queue
QueueHandle_t xQ_Posit;

Here is the related setup():

sema_GPS_Gate = xSemaphoreCreateMutex();
sema_Posit = xSemaphoreCreateMutex();

xSemaphoreGive( sema_GPS_Gate );
xSemaphoreGive( sema_Posit );

Now let’s define 4 tasks:

// task handles
static TaskHandle_t xGPSTask;
static TaskHandle_t xLoRATask;
static TaskHandle_t xSDWriteTask;
static TaskHandle_t xOLEDTask;

// hexadecimal notification code
define GPS_READ_BIT 0x01
define LORA_TX_BIT 0x02
define LORA_RX_BIT 0x04
define SD_WRITE_BIT 0x06
define OLED_BIT 0x08

ISR / Interrupt for Asynchronous Non-Blocking Event Loop

The system will be driven by a periodic ISR timer raising an interrupt calling a handler routine running & managing tasks.

Keeping loop() empty results in non-blocking program execution, no waiting on calls to delay() in main routine – and we only need to call ISR once every ten seconds to read GPS (rather than polling loop()) – which is better for low power consumption.

Under the hood FreeRTOS works in the same way:

> FreeRTOS implements multiple threads by having the host program call a thread tick method at regular short intervals. The thread tick method switches tasks depending on priority and a round-robin scheduling scheme.
( https://en.wikipedia.org/wiki/FreeRTOS#References )

With a task based modular event driven architecture, failure of one task – if write to SD card task blocks or fails because there is no storage card present), other tasks – reading GPS messages, LoRaWAN radio transmit continue.

// ISR timer
hw_timer_t * timer = NULL;
portMUX_TYPE timerMux = portMUX_INITIALIZER_UNLOCKED;
unsigned long isrCounter = 0;

ISR Handler Function (Main control routine, replaces loop()):

// ISR Interupt Handler
void IRAM_ATTR fLoRASendISR( void )
{
portENTER_CRITICAL_ISR(&timerMux);

// Main program routine here ..

portEXIT_CRITICAL_ISR(&timerMux);
}

In setup() we schedule ISR timer:

// Configure Prescaler to 80, as our timer runs @ 80Mhz
// Giving an output of 80,000,000 / 80 = 1,000,000 ticks / second
timer = timerBegin(0, 80, true);
timerAttachInterrupt(timer, &fLoRASendISR, true);

// Fire Interrupt every 10s (10 * 1 million ticks)
timerAlarmWrite(timer, 10000000, true);
timerAlarmEnable(timer);

FreeRTOS Task (Thread) Setup

Next in setup() we describe & initialise tasks, specify which CPU core they should run on, their priority and a few other details

// create a pointer queue to pass position data
xQ_Posit = xQueueCreate( 15, sizeof( &xPosit ) );

Serial.print("Start Task fGPS_Parse() priority 0 on core 0");
xTaskCreatePinnedToCore( fGPS_Parse, "fGPS_Parse", 1000, NULL, 0, &xGPSTask, taskCore0 );
configASSERT( xGPSTask );


Serial.println("Start Task fSD_Write() priority 4 on core 1");
xTaskCreatePinnedToCore( fSD_Write, "fSD_Write", 1000, NULL, 4, &xSDWriteTask, taskCore1 ); // assigned to core 1
configASSERT( xSDWriteTask );

Serial.println("Start Task fLoRA_Send() priority 3 on core 1");
xTaskCreatePinnedToCore( fLoRA_Send, "fLoRA_Send", 1000, NULL, 3, &xLoRATask, taskCore1 ); // assigned to core 1
configASSERT( xLoRATask );

Here’s how to have a task start and wait suspended pending a notification. LoRaWAN transmit task starts and waits until read GPS notifies of new message to send; no need to constantly run radio draining battery or poll the message queue.

for( ;; )
{
/* block until task notification */
xResult = xTaskNotifyWait( LORA_TX_BIT,
                     ULONG_MAX,        /* Clear all bits on exit. */
                     &ulNotifiedValue, /* Stores the notified value. */
                     portMAX_DELAY );  /* Block indefinately */

if( xResult == pdPASS ) 
{

Executing FreeRTOS threads : ISR Main Routine

In main ISR program loop, here’s how to notify tasks:

BaseType_t xHigherPriorityTaskWoken = pdFALSE;

/* Notify (trigger) Read GPS
xTaskNotifyFromISR( xGPSTask,
                   GPS_READ_BIT,
                   eSetBits,
                   &xHigherPriorityTaskWoken );

/* Notify LoRA send task to transmit by setting the TX_BIT */
xTaskNotifyFromISR( xLoRATask,
                   LORA_WRITE_BIT,
                   eSetBits,
                   &xHigherPriorityTaskWoken );

Inter process Communication (IPC): Semaphore Mutex Locks

Concurrency & data consistency in real time and multi threaded systems requires care to ensure serial ops – a write by one thread to a data struct for example must not corrupted by another concurrent thread.

The classic answer to this is locks – mutex, semaphores. A thread requests obtains & takes a lock and other tasks block (wait, retry). The lock is revoked only when lock holding task completes or rolls back.

Here is how it’s done in FreeRTOS:

   if ( xSemaphoreTake( sema_GPS_Gate, xTicksToWait0 ) == pdTRUE )
{
 if ( xSemaphoreTake( sema_Posit, xTicksToWait1000 ) == pdTRUE )
    {
      xPosit.Lat = gps.location.lat();
      xPosit.Lon = gps.location.lng();
      xPosit.Alt = gps.altitude.meters();
      xPosit.Course = gps.course.deg();
      xPosit.Speed = gps.speed.kmph();

      xSemaphoreGive( sema_Posit );
    }


    if ( xSemaphoreTake( sema_Posit, xTicksToWait1000 ) == pdTRUE )
    {
      Serial.println("xQueueSend()");
      pxPosit = &xPosit;
      xQueueSend( xQ_Posit, ( void * ) &pxPosit, ( TickType_t ) 0 );  
      xSemaphoreGive( sema_Posit );
    }

    xSemaphoreGive( sema_GPS_Gate );
  }

FreeRTOS : Message Queue

Those interested in using Queue’s (a sized FIFO pointer linked buffer) to pass data between tasks should consult the API reference:

https://www.freertos.org/a00018.html

// Examples:

// Writing
if ( xSemaphoreTake( sema_Posit, xTicksToWait1000 ) == pdTRUE )
        {
          pxPosit = &xPosit;
          xQueueSend( xQ_Posit, ( void * ) &pxPosit, ( TickType_t ) 0 );  
          xSemaphoreGive( sema_Posit );
        }

// Reading..
struct XPosit xPosit, *pxPosit;
  
 if( xQueueReceive( xQ_Posit,
                         &( pxPosit ),
                         ( TickType_t ) 10 ) == pdPASS )
      {

Conclusion

Espressif changed the game in 2014 with ESP8266 – bringing WiFi & Bluetooth to Arduino’s eco-system of low cost interoperable sensors & components – the internet of things (IOT) long promised since at least 1980s as “smart home” concept finally came of age.

Modern mobile devices and laptops are now so incredibly complex as to be virtually indecipherable, especially at hardware / OS level – to most people, even those within IT industry.

Arduino made it possible for students, DIY enthusiasts, makers & researchers to work with Microcontrollers in a way that is relatively simple, comprehensible and fun. No longer do you necessarily need a PhD, work for tech giant or own a CPU manufacturer to build a micro-electronics project.

By adding WiFi, Bluetooth & LoRaWAN wireless, Espressif opened door to new information age of cloud connected smart sensor & control devices. IOT smart home devices & solutions from Amazon AWS and Google (Matter) are already available in marketplace.

But the option endures – those who wish to can make their own or projects or use a component approach to Internet connect existing machines.

It remains to be seen in decades to come whether people use this tech for good, to benefit people, to make things better; or for nefarious activities – stalking, tracking, surveillance; or worse AI enabled weapons…

(my plan for this prototype was to add 6 axis accelerometer, attach prototype to a stunt kite, aquire some flight input data, create some cool real time web visualisations and explore some more Python SciPi, Mathplotlib & Pandas libraries (alas the prototype proved too fragile to get off the ground).

Anyway hope this tutorial helps you, it took me 30 years to even vaguely understand micro electronics hardware, multi-threading, c++…

Categories
arduino C/C++ circuits Coding Embedded esp32 expressif Internet of Things microcontrollers MQTT sensors weather station WebSockets

SparkFun Weather Sensor Kit

Wind and Rain sensor kit newly arrived from SparkFun Electronics to upgrade an Arduino Weather Station project.

SparkFun Weather Sensor Kit, DIY prototypes, Arduino Weather Station

Also pictured are earlier DIY prototypes – a childrens bee wind spinner with hall effect sensor to count rotations, an anemometer made from recycled plastic packaging utilising a IR Led optical rotary encoder and a wind vane with eight fixed directional magnetic switches.

( more here: http://www.steveio.com/2020/07/21/weather-station-wind-vane-history-science/ and http://www.steveio.com/2020/07/21/weather-vane-hall-sensor-magnetic-rotary-encoder/ ).

Bee Windmill Anemometer with ESP32 LoRa Transmitter running on single 3.3v Li-Ion cell.
8 Durection WInd Vane with magnetic hall sensor array and WebSocket TCP web browser interface.
ESP8266 Anemometer with optical IR Led sensor, wifi connectivity and D3.js websocket provisioned UI.

( Code for these projects can be found on GitHib. )

Weather station projects are a popular accessible introduction to microelectronics; a microcontroller and sensors can be found at low cost, modular hardware design results in easy assembly and open software platforms like Arduino IDE streamline packaging and deployment of code to devices.

Analysing real time or historical time series data, from weather sensors is a lot of fun. Frameworks like R Project for Math Stats: https://www.r-project.org/ ) and Python, Pandas, Numpy & Mathplotlib provide implementations of most alogirithms and convenient data structures for importing & manipulating data.

Techniques and methods are transferable and can be applied to other domains or ontologies – finanicial, accounting data for example.

SparkFun offer an OEM Wind & Rain sensor kit manufactured by Shenzen Fine Offset Electronics, China.

With advent of 3d modelling & printing it is also feasible for an enthusiast to design and fabricate via a 3d printer custom sensor components, perhaps using template models downloaded from repos like ThingiVerse.

In competition marine OpenWind are defining what smart network connected sensors can achieve utilising Bluetooth LE to make near real time wind data available on smartphone.

Assembled SparkFun Weather Sensor Kit

Ideal for enthusiast or educator SparkFun Weather kit comes wihout circuitry,  microcontroller or software.  An add-on PCB designed for use with  Arduino / ESP32 can be purchased or Datasheet Technical Specs provide reference sensor circuit designs, not significantly complex due to use of magnetic reed switch and variable resistance technology.

MCU Sensor Control & Relay Unit – IP67 Weather Proof Enclosure, ESP32 TTGO LoRa microcontroller, light, temperature and air pressure sensors.

Traditionally 433MHz RF has been used for base station to transmitter devices. A popular project is to use Arduino, a cheap 433Mhz receiver and a library to read data from a commercial weather station designed for use with manufacturers display, enabling this data to be provisioned to the cloud.

For data transmission non GPRS (cellular) options include Bluetooth LE (range ~100 metres) or LoRa (Long Range Low Power Network – range between 300 – 10km depending on antenae) offering cableless wireless connectivity allowing remote sensor situation with no associated network costs.

At data layer WebSockets and MQTT for IOT devices are challenging serial protocols as defacto lightweight, reliable & easy to implement transport relays.

Apart from range and connectivity goals of low power consumption for efficient and long battery running time combined with solar charging enable devices to run standalone for long periods.

Is a single 3.3v Li-Ion Battery Cell Sufficient? TP405 Charging Module & Solar Panel

Weather Stations have applications beyond meteorology in smart agriculture, industrial, safety monitoring and for wind or wave based leisure pursuits. 

Assembling DIY Arduino Mega Weather Station v1.0

More generally Internet of things wireless networked smart sensor platforms can be used for many purposes and combined with AI and Machine Learning algorithms useful insight and patterns within data can be analysed, classified and predicted. 

SparkFun Smart ETextiles & Conductive Thread Kit

Personally, I really enjoyed SparkFun Arduino LilyPad e-textile, smart fabrics and conductive thread kit, so looking forward to now spinning up the Weather Station sensors!

Categories
arduino C/C++ Coding Embedded esp32 esp8266 Internet of Things microcontrollers Software

Timed Device – Simple & lightweight scheduling for on/off devices

Timed Device is a library for Arduino / ESP 8266/32 embedded platforms written to emulate a simple plug in timer for an electrical device, where pins in a ring are set to define hour/minute status. Designed to be simple and lightweight optimisation is for low memory.

Arduino ATMega328 and related embedded microcontrollers provide internal high precision clock based timers suitable for events with second, millisecond and fine granularity (to clock speed 8 / 16 mhz).

When lower precision is sufficient, recurring timers for example where an event should occur on a specific minute, hour or day of week, a lightweight implementation can be achieved and storage for internal data structures can be optimised, a significant win factor on embedded systems where memory use is constrained.

The library follows an object orientated design pattern resulting in an extensible, scale-able architecture suitable for controlling from one to a large number of devices sharing a common timing core with each timed device class able to implement specific instructions for switching on/off.

Example use cases:

  • Switch lights or any electrical relay device on/off
  • Activate a pump or valve
  • Open/Close blinds, curtains or shutters
  • Control a fan, heat source or air conditioning

How to Define Time

In C time.h the tm structure has the following definition −

struct tm {
   int tm_sec;         /* seconds,  range 0 to 59          */
   int tm_min;         /* minutes, range 0 to 59           */
   int tm_hour;        /* hours, range 0 to 23             */
   int tm_mday;        /* day of the month, range 1 to 31  */
   int tm_mon;         /* month, range 0 to 11             */
   int tm_year;        /* The number of years since 1900   */
   int tm_wday;        /* day of the week, range 0 to 6    */
   int tm_yday;        /* day in the year, range 0 to 365  */
   int tm_isdst;       /* daylight saving time             */
};

While this structure is useful for point in time defined events, in case of a recurring timers this can be simplified.

If hour precision is sufficient, a single 32 bit mask can be used:

// Bitmask defines hours (from 24h hours) device is on (1) or off (0)
// 0b 00000000 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
long hourTimerBitmask =0b00000000001111111111111111100000;

Day of week (bit index 0 = Sunday – 6 Saturday) can be defined similarly:

long dayOfWeekTimerBitmask = 0b0101010;

With bitmasks we can determine if an event is scheduled using bit shifts:

// Check if bit at pos n is set in 32bit long l
bool Timer::_checkBitSet(int n, long * l)
{
  bool b = (*l >> (long) n) & 1U;
  return b;
}

To check if an hour timer bitmask is on/off for 7am:

bool isSet = _checkBitSet(7, hourTimerBitmask)

For a more detailed guide to Bitmasks, shifts and bitwise operations see this article.

An on / off time occuring at specific (hour:minute) intervals can be defined as:

typedef struct tmElements_t {
  uint8_t Min;
  uint8_t Hour;
};

We can use a struct as container for an array of 1..n pairs of on/off times occurring on specified weekdays:

#define SZ_TIME_ELEMENT 2

typedef struct tmElementArray_t {
  unsigned long Wday;   // bitmap - days of week (bit index 0 = Sun - 6 Sat)
  struct tmElements_t onTime[SZ_TIME_ELEMENT];
  struct tmElements_t offTime[SZ_TIME_ELEMENT];
};

With these data structures we can setup up a pair of on/off times (08:00 -> 08:10 and 19:00 -> 19:30) to occur on Sun, Mon, Thur, Fri:

// create variables to define on/off time pairs
struct tmElements_t t1_on, t1_off, t2_on, t2_off, t3_on, t3_off;
struct tmElementArray_t timeArray;

t1_on.Hour = 8;
t1_on.Min = 0;
t1_off.Hour = 8;
t1_off.Min = 10;

t2_on.Hour = 19;
t2_on.Min = 0;
t2_off.Hour = 19;
t2_off.Min = 30;

timeArray.n = 2; // number of time pairs
timeArray.Wday = 0b00110011; // define days of week timer is active on

timeArray.onTime[0] = t1_on;
timeArray.offTime[0] = t1_off;

Check if a recurring timer is set

How we check a timer (whether bitmask or time based) depends on the type of time source we have.

Most familiar time source on Arduino is millis() function providing elapsed time in milliseconds since device reset / startup.

This is useful for timing events in a non-blocking way at recurring intervals:

if (millis() >= sampleTimer + sampleInterval)
{
  ... do something
  sampleTimer = millis();
}

But what if we want to schedule in terms of hour, minute or day of week based on current time?

To achieve this a micrcontroller is combined with a Real Time Clock module, providing a battery backed time source that once synced (for example to an NTP time source) maintains current time even when device is switched off.

Arduino Nano in pin breakout board with DS3231 Real Time Clock Module

Arduino DS1307/3231 RTC modules have supporting libraries (for example RTCLib from Adafruit) to obtain current time usually in form of:

DateTime now = rtc.now();

Where DateTime object provides an API to return specific time elements:

Serial.print(now.hour());
Serial.print(':');
Serial.println(now.minute());
Serial.print(now.dayOfTheWeek());

Is an event scheduled (at a specific point in time)?

In C function overloading can be used to provide multiple interfaces to an isScheduled() method according to timer precision / time source type :

bool isScheduled(int h);
bool isScheduled(int h, int d);
bool isScheduled(int m, int h, int d);
bool isScheduled(unsigned long ts);

(Where h = hour, m = minute, d = week day)

Most users of Unix based systems are familiar with “timestamp” a 32bit unsigned long representing elapsed seconds since a fixed point in time, the Unix Epoch which occurred 1970-01-01 00:00:00 UTC.

Unix is not unique in using a reference Epoch, they have been used in calender systems worldwide since ancient times.

We can obtain current timestamp on Linux via command line with date command:

stevee@ideapad-530S:~$ date +%s
1620155520

A timestamp as a time source can be used with any timer definition, whether point in time or recurring. It has advantage (compared to using numeric representations of individual time elements) that a single long number can be used for computation and conversion and we don’t need to worry about problems like variable number of days in month or leap years.

Timestamps and time and date conversion tricks

For a timer defining specific days of week, the first question might be how to obtain weekday from a unix timestamp (elapsed seconds since epoch)?

int weekday = (floor((ts / 86400)) + 4) % 7;

1970-01-01 was a Thursday, dividing timestamp by 86400 (number of seconds in a day: 24 * 60 * 60) gives number of days since epoch, adding 4 shifts start day to Sun and modulo 7 returns day of week.

To check if a timestamp is within range of one or more on/off time pairs (specified in hh:mm format as uint8_t Min; uint8_t Hour;) first we convert all time elements to elapsed seconds:

// convert fully qualified timestamp to elapsed secs from previous midnight
unsigned long elapsedTime = ts % SECS_PER_DAY;

unsigned long s1, s2, onTime, offTime;

// check each timeArray on/off pair
for (int i = 0; i < _timeArray->n; i++)
{
  s1 = _timeArray->onTime[i].Min * SECS_PER_MIN;
  s2 = _timeArray->onTime[i].Hour * SECS_PER_HOUR;
  onTime = s1 + s2;
  s1 = _timeArray->offTime[i].Min * SECS_PER_MIN;
  s2 = _timeArray->offTime[i].Hour * SECS_PER_HOUR;
  offTime = s1 + s2;

  if (elapsedTime >= onTime &amp;&amp; elapsedTime <= offTime)
  {
    return true;
  }
}

Conclusion

While TimedDevice library is non-blocking (it does not technique like delay()) it is not truely asynchronous or event driven as each device implementing a timer must poll for an event on each iteration of loop().

This could constrained to checking once per second/minute/hour but perhaps a better architecture would be to use either RTC DS3231 squarewave or Arduino internal timer to register and trigger an interrupt with an associated handler only when an event is due to occur.

Similarly if either a very large number of events are scheduled, or a large set of timed devices created, it would be sensible to register these with a scheduler (akin to a CPU scheduler) tasked with sorting, prioritising and actioning event queue in an efficient way, which might utilise a multi-threaded approach on multi-core CPU architectures.

Source code for Timed Device library can be found on GitHub including a full Arduino sketch example for defining a solenoid valve timer with an RTC timesource.

Categories
arduino C/C++ Coding microcontrollers Software weather station

Sunrise / Sunset – Is Arduino ahead of Its Time?

Recently I added Sun Moon Library to my DIY Arduino Weather Station.

Given a location specified as Latitude / Longitude coordinates and a date / time this clever algorithm uses astronomical math routines to provide timing of sunrise, sunset and moon age.

https://www.flickr.com/photos/jannerboy62/31589567761/

In spite of Arduino ATMega328p chipset being only an 8 bit architecture and float data type being low precision ( 6 – 7 decimal digits ) timings output by sun moon are said to be accurate to second scale.

Comparing results with UK Met Office weather forecast during testing I found Arduino sunrise / sunset timings to be out, with a margin or error of around 15 minutes.

Being suspicious of data, one sunny afternoon I found a vantage point and watched sun disappear below horizon, timing the event.

Sure enough Arduino data was indeed inaccurate.

Arduino DIY Weather Station and sensors.

What could be the problem?

While my knowledge of maths is not sufficiently advanced to properly understand each line of Sun Moon library algorithm, to eliminate possibility of faulty code I decided to try another implementation, Dusk to Dawn.

After running a test for same location, result with new library continued to show an inaccuracy of 15 minutes.

Weather Station timing (date / time) is provided by a Real Time Clock (DS3132 module) which is synced from internet via Network Time Protocol (NTP). A quick check showed date and time to be correct, although manual correction for daylight savings time (DST) seemed sub-optimal.

This pointed to a coordinate error.

Google gives my location Bournemouth, UK as Longitude / Latitude 50.7192° N, 1.8808° W.

Google results for Latitude / Longitude Bournemouth, UK

These coordinates are defined in Weather Station code as:

// Lat/Long: Bournemouth 50.7192° N, 1.8808° W
#define LOC_latitude    50.7192
#define LOC_longtitude  1.8808

Here was the problem.

Checking cooordinates for another location, London, UK ( 51.5074° N, 0.1278° W ) Arduino gave sunrise / sunset with only a minor ( < 1 minute) difference in timing.

Bournemouth is West of Greenwich Meridian (zero line for Longitude) by approx 107 miles. Each degree of latitude is approximately 69 miles (111 kilometers) apart.

Google gives coordinates (50.7192° N, 1.8808° W) as decimal degrees and “N/S/E/W”, a human friendly representation.

Arduino code expects decimal degrees and Plus/minus symbol.

ISO 6709 International Standard defines Longitude as a number preceded by a sign character. A plus sign (+) denotes east longitude or the prime meridian, and a minus sign (-) denotes west longitude or 180° meridian (opposite of the prime meridian)

Similarly, according to Microsoft, “The latitude is preceded by a minus sign ( – ) if it is south of the equator (a positive number implies north)”.

Updating Weather Station to include latitude minus symbol, code now provided accurate timings –

// Lat/Long: Bournemouth 50.7192° N, 1.8808° W
#define LOC_latitude    50.7192
#define LOC_longtitude  -1.8808

In navigation, the 1 in 60 rule states that for each degree off (or displacement) over a distance of 60 nautical miles (NM), it will result in 1 NM off course

A trans Atlantic journey by boat from Southampton to New York ( 2974.5 nautical miles ), given a similar error in bearing of ~2 degrees might arrive in Boston or possibly Washington.

Sunrise / Sunset times on Arduino LCD Weather Station display

We were in agreement at last, today sunrise would occur at 06:25 and sunset at 19:46.

References:

https://github.com/steveio/arduino/blob/master/WeatherStation/WeatherStation.ino

https://en.wikipedia.org/wiki/ISO_6709#Longitude

https://docs.microsoft.com/en-us/previous-versions/mappoint/aa578799(v=msdn.10)?redirectedfrom=MSDN

https://stackoverflow.com/questions/51626412/getting-negative-values-of-latitude-and-longitude

https://www.metoffice.gov.uk/weather/forecast

Categories
C/C++ Coding Embedded microcontrollers Software

Binary – Bits, Bitwise Operators & Bitmasks

First we define bits, bit fields and bit masks, relating these to data structures, addressable memory storage, registers and binary protocol frames.

Bitwise logic is introduced with simple practical examples – bitwise operators and bit shifts – setting individual bits on or off, extracting bits for reading, manipulating bits or checking bit field values.

Bitwise Logic Gates

In Assembler, C and other programming languages bit manipulation is common especially when implementing low level hardware interfaces, embedded microcontroller based IOT devices, processing sensor, machine, network and peripheral data.

Graphics programming and image manipulation also utilises bitwise operations and bitmaps to perform animation (shifting and moving bits), compositing and pixel value modification.

Bits, Bytes & Bitmasks Explained

Bits are smallest unit of storage representing either a binary 1 or 0.

A byte is a group of binary digits or bits (typically eight) operated on as a unit.

Bit fields are a series or array of bits in adjacent memory.

Bit field values might represent a set of individual attributes (boolean on/off “flags” or status fields), a register value (storage address / instruction), or encapsulate binary encoded data.

AND bitmask

A bitmask (or mask) is an ordered set of bits of defined length used in bitwise logic to set, invert or manipulate another bit field.

Registers – Bit Arrays

Registers encapsulate a set of ordered bits in storage with a defined bit length typically expressed as power of base 2 (commonly 8 / 16 / 32 / 64 / 128 bit).

76543210
8 bit register where least significant bit (LSB) starts with (2º = 1)

An 8 bit register can hold unsigned (positive) numbers from 0 to 255 or signed (includes negative) values from -128 to +127 (where bit 7 is a sign bit).

Registers are used to address memory locations for computations within central processing unit, configuration parameters, IO from data storage, ports, or peripherals.

Bits in a Byte – Memory Address

Bitwise shift operators, along with logical AND can be used to access individual bits within a byte of memory for any variable in C.

To read value of bit at index 4 of an 8 bit length variable x

int x = 16; // 10000
printf("%d\n", (x >> 4) &amp; 0x01); // Prints 1

Network Protocols – Binary Bit Sequences

Bit sequences are used in network protocols where packetised messages are arranged into frames containing fields with defined offset and length.

A frame refers to the entire data packet which is being sent/received during a communication. Each protocol defines a specification of its own frame format.

An RS232 serial protocol frame defines a bit sequence –

Frame:StartDataParityStop
Size (bits):15-911-2
Bit sequence format of a USART serial protocol frame

Bitwise operators and shifts can be used to efficiently read, set or modify fields within a network protocol packet.

Bitwise Operators

Bitwise operations allow setting bit sequence values in a single operation and are more efficient than loops or maintaining individual bits.

SymbolOperator
&bitwise AND
|bitwise inclusive OR
^bitwise XOR (exclusive OR)
<<left shift
>>right shift
~bitwise NOT (one’s complement) (unary)
Bitwise and shift operators

Logical operators AND, OR, XOR (Exclusive OR) and NOT (Negation) implement bitwise manipulation, incurring a minimal number of processing instructions.

Comparing bit by bit –

  • OR sets a bit if one or both operators are true: 1110 OR 0000 = 1110
  • AND sets a bit if both operators are true: 1010 AND 1101 = 1000
  • AND with zero-check tests if any bit is set: 1010 AND 0010 = 0010 ≠ 0
  • XOR sets a bit if only one operator is true: 1010 XOR 0100 = 1110
  • NOT inverts all bits: NOT 1011 = 0100

Bitwise operators work on integer and character data types and do not modify value of their arguments so assignment (=, +=, -=, |=, &=) is typically used for example x |= (y & z).

Arithmetic Bit Shifts

Left Arithmetic Shift

Bit or Arithmetic shift operators << >> treat a value as a series of individual bits rather than a numerical quantity, shifting (or moving) bit positions left or right.

These operations are useful to move a bit or set of bits to a specific positional index.

In a left shift operation (<<) bits are moved by one position to the left and last digit is zero filled. This is equivalent to multiplication of a signed integer by a power of 2.

01101011
8 bit Binary encoding of decimal 107

Left shifting 2 bits ( << 2) results in

10101100
8 bit Binary encoding of decimal 172

A right shift operation (>>) moves bits right by a specified number of positions, dividing number by 2.

In more general form we can say –

// x multiplied by 2ⁿ
x << n
// x divided by 2ⁿ 
x >> N

Notes –

  • Bits shifted from end are lost due to overflow, they do not wrap around.
  • Right bitshift on signed types – gcc promises to always give the sane behavior (sign-bit-extension) but ISO C allows the implementation to zero-fill the upper bits.

Defining bit masks in C

Bitmasks are used in bit manipulation and combined with a logical operator (AND, OR XOR) define a pattern for bits to keep or discard.

In c++14 which supports binary literals bit masks are defined –

const uint8_t mask0 = 0b00000001 ; // represents bit 0
const uint8_t mask1 = 0b00000010 ; // represents bit 1
const uint8_t mask2 = 0b00000100 ; // represents bit 2
...  

Using Hexadecimal

const uint8_t mask0 = 0x01 ; // bit 0  0000 0001
const uint8_t mask1 = 0x02 ; // bit 1  0000 0010
const uint8_t mask2 = 0x03 ; // bit 2  0000 0100
...  

Or with bit shift operator

const uint8_t mask0 = 1 << 0 ; // 0000 0001
const uint8_t mask1 = 1 << 1 ; // 0000 0010
const uint8_t mask2 = 1 << 2 ; // 0000 0100
...

Set, Clear, Modify, Toggle & Check Bits

Set a bit
Set the nth bit ( zero up to n-1 ) of number

number |= 1UL << n

Set bit 0 of i to one

i |= 1 << 0;

// Example
000 i
001 1 << 0;
001 i |= 1 << 0;

  • the << operator is left “bit shift” operator which moves all bits to the left n times.
  • 1UL species numeric literal 1 with type Unsigned Long

Clear a bit
Set nth bit of number to zero

number &= ~(1UL << n);

Set bit 1 of i to zero

i &= ~(1UL << 1);

// Example
010 i
000 ~(1UL << 1)
000 i &= ~(1UL << 1);
  • ~ negates value of bit 1 from 1 to 0
  • AND evaulates false as both operands are not equal

Toggle (flip) a bit

Toggle nth bit of number

number ^= 1UL << n;

Toggle (flip) bit 0 of i

i ^= 1 << 0;

// Example
001 i
001 1 << 0
000 i ^= 1 << 0
  • ^= represents XOR exclusive OR with assignment, output is true when operand bits are different

Check a bit is set

True if nth bit of number equals 1

bit = (number >> n) & 1U;

Check bit 7 of i assign 1 or 0 to bit

int bit = (i >> 7) & 1U;

Modify – Changing a bit to x (1 or 0)

// 2s complement system with negation behaviour
number ^= (-x ^ number) & (1UL << n); 
// portable
number = (number & ~(1UL << n)) | (x << n);

Set bit 7 of i to x

i ^= (-x ^ i) & (1UL << 7);
i = (i & ~(1UL << 7)) | (x << 7);

Notes –

  • range / bounds checking is not applied, out of bounds shift index result in undefined behaviour
  • Use 1ULL if number is wider than unsigned long
  • endianess (position of most / least significant bit) and signing varies between platforms and compilers

Bitwise Functions as C Pre-Processor Macros

To minimise code duplication bit operator functions can be defined in a header file as pre-processor macros

/* a=number, b=bit index 0-n */
#define BIT_SET(a,b) ((a) |= (1UL<<(b)))
#define BIT_CLEAR(a,b) ((a) &amp;= ~(1UL<<(b)))
#define BIT_FLIP(a,b) ((a) ^= (1UL<<(b)))
#define BIT_CHECK(a,b) (!!((a) &amp; (1UL<<(b))))

Print Bits in C as Left Padded Binary String

In C printf() does not have a format specifier to print a string representation of binary, but we can write a function to achieve this –

const unsigned bits = 8;

// print integer value as a left padded binary string
void print_bits(unsigned value)
{
    unsigned mask = 1 << (bits-1);

    while (mask)
    {
        printf("%d", (mask &amp; value) != 0);
        mask >>= 1;
    }

    printf("\n");
}

int main()
{
  int i = 0x145;
  print_bits(i);
}

Result:
01000101

References and Further Reading:

[1] Methods for Bit Manipulation & Discussion:
https://stackoverflow.com/questions/47981/how-do-you-set-clear-and-toggle-a-single-bit/263738#263738

[2] Theory and examples of Bitwise Operation:
https://en.wikipedia.org/wiki/Bitwise_operation

[3] Bit Twiddling Hacks – Advanced Bitwise Algorithms http://graphics.stanford.edu/~seander/bithacks.html

Categories
arduino C/C++ esp8266 Internet of Things Python Software WebSockets

WebSocket Binary – ESP8266 to Web Browser

Binary wire protocols are long established for embedded machine to machine (M2M) communication, network applications and wireless radio data transmission.

Internet of Things (IOT) devices, real time sensors, robotics, smart home of industrial machine control data also demand efficient, low latency & lightweight data communications.

WebSockets ( RFC6455 ) protocol brings native support for binary framed messaging to web browser clients, offering a compact lightweight format for fast and efficient endpoint messaging.

Why use binary format data messaging?

Compared to serialisation of more complex text based wire formats, binary is lightweight and requires minimal storage / bandwidth and processing.

Taking an example key/value command data message:

// JSON encoding
{"cmd":101,"value":180}
23 * 2 = 46 bytes

// CSV plain text encoding
101,180\n
9 * 2 = 18 bytes 

// Binary
101 180
int (4 bytes) + int(4 bytes)
4+4 = 8 bytes

In case of high performance applications supporting a large number of clients or very high frequency of data exchange, minimising data size, bandwidth and processing becomes an important priority.

Binary wire protocols are long established for embedded M2M messaging

Taking as a simple example an embedded ESP8266 WiFi device, message gateway and web browser client, data serialisation and bidirectional binary framed WebSocket data exchange are demonstrated.

ESP8266 Byte Array Serialisation

Internally data is represented in embedded microcontrollers as ones and zeros, sequences of bits arranged in addressable memory.

Higher level programming language abstraction provides human readable textual labels and in case of C/C++ associated type information.

Lets define a mixed type data structure that could be some kind of sensor or message data payload –

    // define mixed type data struct
    struct Data
    {
        int id;
        float v1;
        float v2;
        unsigned long v3;
        char v4[20];
    };

    struct Data data;

    // populate data values
    data.id = 67;
    data.v1 = 3.14157;
    data.v2 = -7.123;

    unsigned long ts = millis();
    data.v3 = ts;

    char c[20] = "N NE E SE S SW W NW";
    strncpy(data.v4, c, 20);

To access underlying bytes, a pointer to data structure address is created –

    uint8_t * bytePtr = (uint8_t*) &data;    
    webSocket.sendBIN(bytePtr, sizeof(data));

Data pointer and length are passed to WebSocket send method “webSocket.sendBIN()”, byte range is read, packaged (framed) according to protocol specification and written to TCP/IP network socket.

Hexidecimal and Binary text representation of in memory data structure can also be displayed –

void printBytes(const void *object, size_t size)
{
    const uint8_t * byte;
    for ( byte = (uint8_t *) object; size--; ++byte )
    {
        Serial.print(*byte, HEX);
        Serial.print("\t");
        Serial.println(*byte, BIN);
    }
    Serial.println('\n');
}

Python WebSocket Server

A Python3 middleware hosts WebSocket server and acts as a message relay gateway.

Binary WebSocket messages can be decoded in Python, the struct module performs conversions between Python data types and C structs –

async def wsApi(websocket, path):
    try:
        async for message in websocket:
            print('User-Agent: '+ websocket.request_headers['User-Agent'])
            print('Sec-WebSocket-Key: '+websocket.request_headers['Sec-WebSocket-Key'])
            print('MessageType: '+str(type(message)))
            print(message);
            print('Hex: '+message.hex());

            if isinstance(message, (bytes, bytearray)):

                i = message[:4];
                print(i);
                tuple_of_data = struct.unpack("i", i)
                print(tuple_of_data)

                tuple_of_data = struct.unpack_from("f", message, 4)
                print(tuple_of_data)

                tuple_of_data = struct.unpack_from("f", message, 8)
                print(tuple_of_data)

                tuple_of_data = struct.unpack_from("i", message, 12)
                print(tuple_of_data)

                tuple_of_data = struct.unpack_from("20s", message, 16)
                print(tuple_of_data[0])

                ## forward message
                await asyncio.wait([user.send(message) for user in USERS])

To index into byte array and read a number of bytes according to data type being unpacked Python’s array slice method “i = message[:4]” can be used where [<from>:<to>] specifies start/end positions.

Method struct.unpack_from() is another approach, taking as parameters a format character specifying data type (“i” – integer, “f” – float), data buffer and an index (in bytes) to read from.

Here is decoded binary message output including some WebSocket headers –

User-Agent: arduino-WebSocket-Client
Sec-WebSocket-Key: zoJ0aR/5XunSvEKKcUkWfQ==
MessageType: <class 'bytes'>
b'C\x00\x00\x00|\x0fI@\x9e\xef\xe3\xc0\xb9\x17\x00\x00N NE E SE S SW W NW\x00'
Hex: 430000007c0f49409eefe3c0b91700004e204e45204520534520532053572057204e5700
b'C\x00\x00\x00'
(67,)
(3.1415700912475586,)
(-7.123000144958496,)
(6073,)
b'N NE E SE S SW W NW\x00'

Web Browser – Binary Encode/Decode in JavaScript

In web browser, JavaScript primitives Blob, ArrayBuffer and TypedArray perform a similar conversion.

Firstly, received WebSocket messages (event object) can be debugged to console –

 websocket.onmessage = function (event) {
    console.log(event);

Binary framed data payload is reported as type “Blob” (raw data) of length 36 bytes –

Chrome Developer Tools console log for WebSocket Binary message receieve event

To de-serialise message, raw data Blob is converted asynchronously using FileReader API to ArrayBuffer, a generic fixed length binary data buffer –

    if (event.data instanceof Blob)  // Binary Frame
    {
      // convert Blob to ArrayBuffer
      var arrayPromise = new Promise(function(resolve) {
          var reader = new FileReader();

          reader.onloadend = function() {
              resolve(reader.result);
          };

          reader.readAsArrayBuffer(event.data);
      });

When promise is fulfilled, ArrayBuffer can be read using typed views (Uint32Array, Uint32Array) for integer (including long) and float types, TextDecoder API is used to decode character array –

arrayPromise.then(function(buffer) {

          // Decoding Binary Packed Data

          // int (4 bytes)
          var arrInt = new Uint32Array(buffer);
          var id = arrInt[0];
          console.log("id:"+id);

          // 2x float (4 bytes)
          var arrFloat = new Uint32Array(buffer,4);
          var v1 = arrFloat[0];
          var v2 = arrFloat[1];
          console.log("v1: "+v1);
          console.log("v2: "+v2);

          // long (4 bytes)
          var v3 = arrInt[3];
          console.log("v3:"+v3);

          // character data (20 bytes)
          var uint8Array = new Uint8Array(buffer,16);
          var string = new TextDecoder("utf-8").decode(uint8Array);
          console.log(string);
      });

JavaScript Binary Data Encoding

Binary data can also be encoded from native JavaScript. TypedArrays created for each data type – integer, float, long and character array are populated and packed into an ArrayBuffer suitable for use as WebSocket data payload –

console.log("Binary Encode example");

// Binary Encode example
var buffer = new ArrayBuffer(36)
var arrInt =   new Uint32Array(buffer, 0, 1);
arrInt[0] = 67;
var arrFloat = new Float32Array(buffer, 4, 2);
arrFloat[0] = 3.14157;
arrFloat[1] = -7.123;

var arrInt2 =   new Uint32Array(buffer, 12, 1);
arrInt2[0] = Date.now();

var uint8Array = new Uint8Array(buffer,16);
var charBuffer = new TextEncoder("utf-8").encode("N NE E SE S SW W NW");

for(var i = 0; i<charBuffer.length; i++)
{
  uint8Array[i] = charBuffer[i];
}

// send binary data
websocket.send(buffer);

At message gateway, logs demonstrate parity between data packed by embedded device and those sent from web browser client –

User-Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/69.0.3497.100 Safari/537.36
Sec-WebSocket-Key: 1TD9Zp71cMTivUbj+QSx5w==
MessageType: <class 'bytes'>
b'C\x00\x00\x00|\x0fI@\x9e\xef\xe3\xc0\x07\x1c\xff\x91N NE E SE S SW W NW\x00'
Hex: 430000007c0f49409eefe3c0071cff914e204e45204520534520532053572057204e5700
b'C\x00\x00\x00'
(67,)
(3.1415700912475586,)
(-7.123000144958496,)
(-1845552121,)
b'N NE E SE S SW W NW\x00'

Limitations / Drawbacks

Compared to UTF-8 text formats (XML, JSON) packed binary data has significant disadvantages –

  • legibility – text based key/value formats are easy to read, manipulate and maintain
  • fixed frame boundaries – using positional byte sequence indexes means even small changes to message structure, size or field position require updates to consumer client code
  • endianess / alignment / padding must be maintained consistently, compiler and platform implementation differences may occur

Security

WebSockets Secure (WSS) offers transport layer security (TLS) to encrypt data streams. An authentication and authorisation strategy (challenge/response password, token or certificate based) for client identification should also be deployed. Cryptographic message digest signing or encryption might also be used as extra protection for critical data.

Categories
C/C++ esp8266 Internet of Things microcontrollers WebSockets

ESP8266 IOT Microservo WebSocket

Internet of things devices can be networked wirelessly over internet and integrate easily with cloud, mobile and web applications.

WebSockets ( RFC6455 ) run on a variety of platforms including embedded systems and web browsers enabling low latency bidirectional binary socket communication over TCP/IP.

ESP8266 is a low cost Wi-Fi micro-controller compatible with Arduino open-source electronics framework designed for networked sensor, robotics and micro-electronic control applications.

A microservo controlled by ESP8266 with web & mobile interface

WebSocket MicroServo

In a simple example a micro-servo reports position data and is controlled via a potentiometer (rotary encoder), web and mobile interfaces.

This tutorial demonstrates:

  • Microservo / Potentiometer GPIO control
  • ESP8266 WebSocket Client
  • Binary format Messaging
  • Python WebSocket Library to synchronise state between clients
  • Responsive HTML5 / Javascript Browser Control Interface

Circuit Wiring

  • 5v regulated DC power supply
  • Microservo connected to ESP8266 GPIO 15
  • 100k Potentiometer – CLK -> GPIO 12, DT -> GPIO 13

Microservo Setup

Microservo power is provided by an external regulated 5v DC power supply sharing a common ground connection to microcontroller.

Servo library initialisation and global variables –

#include <Servo.h>
int servoPin = 15;
int angle; // current angle (degrees)
int servoStartAngle = 90; // initial position
int limit = 90; // range of servo in degrees
Servo Servo1;

// in setup()
Servo1.attach(servoPin);
rotateServo(0);

Servo position is set in a function by passing angle in degrees as a parameter –

void rotateServo(int angle)
{
  Serial.print("RotateServo: ");
  Serial.println(angle);
  Servo1.write(angle);
}

Potentiometer / Rotary Encoder Setup

Potentiometer, a type of incremental rotary encoder based on variable resistance, allows relative rotary motion to be tracked, 2 out-of-phase output channels indicate direction of travel.

#define outputA 12 // Rotary Encoder #1 CLK
#define outputB 13 // Rotary Encoder #2 DT

int counter = 0; // rotary encoder incremental position
int aState; // rotary encoder state comparator
int aLastState;  

void setup() {
   pinMode(outputA,INPUT);
   pinMode(outputB,INPUT);
       
   aLastState = digitalRead(outputA);
}

Rotary encoder CLK and DT pins are compared to determine direction of rotation. Calling map() converts counter (relative position) to angle –

void readRotaryEncoder()
{  

  aState = digitalRead(outputA);

   if (aState != aLastState){     
     // If the outputB state is different to the outputA state, that means the encoder is rotating clockwise
    if (digitalRead(outputB) != aState) { 
      if (counter < limit)
      {
        counter ++;
        angle = map(counter, -90, 90, 0, 180);
        rotateServo(angle);
      }
    } else {
      if (counter + limit > 0)
      {
        counter --;
        angle = map(counter, -90, 90, 0, 180);
        rotateServo(angle);
      }
    }
  } 
  aLastState = aState;
}

ESP8266 microcontroller, Potentiometer (Rotary Encoder), Microservo, 5v DC power supply

Websocket Binary Message Framing

WebSocket protocol natively supports binary framed messaging, offering a compact lightweight format for fast and efficient endpoint messaging.

To report and update microservo position command messages are passed between control interfaces and microcontroller as binary data.

A Python service running on server creates WebSocket and maintains and synchronises shared application state between connected clients.

In web browser user interface, Javascript also has native support for (un)packing binary data.

Data Serialisation

A C struct data_t encapsulates two fields – “cmd” (unit8_t) and “value” (int) – two commands are defined, one to report servo position and another to set a new position, value represents an angle.

To assist serialisation data_t is wrapped in a packet union –

// data message
typedef struct data_t
{
  uint8_t cmd;
  int value;
};

// message packaging / envelope
typedef union packet_t {
 struct data_t data;
 uint8_t packet[sizeof(struct data_t)];
};

#define PACKET_SIZE sizeof(struct data_t)

#define CMD_SERVO_ANGLE 12 // command to report servo position
#define CMD_SERVO_ROTATE 13 // move servo to a specified position

Instances of data structs are created representing send and receive messages.

// send / receive msg data structures
union packet_t sendMsg;
union packet_t receiveMsg;

// messaging function prototypes
void readByteArray(uint8_t * byteArray);
void writeByteArray();
void printByteArray();

// buffer 
uint8_t byteArray[PACKET_SIZE];

Methods allow struct data to be written and read from byte array buffer –

// read bytes from buffer
void readByteArray(uint8_t * byteArray)
{
  for (int i=0; i < PACKET_SIZE; i++)
  {
    receiveMsg.packet[i] = byteArray[i];
  }
}

// write data to buffer
void writeByteArray()
{
  for(int i=0; i<PACKET_SIZE; i++)
  {
    // msg into byte array
    byteArray[i] = sendMsg.packet[i]; 
  }
}

ESP8266 Wifi / Websocket Setup

Details for setting up ESP8266 wifi can be found here.

ESP8266 Websocket library is added by including headers, defining server IP address / port / path and creating a WebSocketsClient class instance.

#include <WebSocketsClient.h>

WebSocketsClient webSocket;

const char* ws_server = "192.168.1.127";
const uint16_t ws_port = 6789;
const char* ws_path = "/";

In setup WebSocket library is initialised –

void setup() {
    ...

    // server address, port and URL
    webSocket.begin(ws_server, ws_port, ws_path);
  
    // event handler
    webSocket.onEvent(webSocketEvent);
  
    // use HTTP Basic Authorization (optional)
    //webSocket.setAuthorization("user", "Password");
  
    // try again if connection has failed
    webSocket.setReconnectInterval(5000);

WebSocket – Message Receive

Event of type “WStype_BIN” defines handling when a binary format message is received, size is reported and hexdump() displays message contents –

void webSocketEvent(WStype_t type, uint8_t * payload, size_t length) {

  switch(type) {
    ...
    case WStype_BIN:
      Serial.printf("[WSc] get binary length: %u\n", length);
      hexdump(payload, length);
      setServoPosition(payload);
      break;

In setServoPosition() received byte array is de-serialised into message data structure. Angle field is used to update servo and potentiometer position.

void setServoPosition(uint8_t * byteArray)
{
  readByteArray(byteArray);
  printByteArray(receiveMsg);

  int angle = (int) receiveMsg.data.value;
  rotateServo(angle);

  counter = map(angle, 0, 180, -90, 90); // update rotary encoder
}

WebSocket – Message Send

Servo position is reported by populating sendMsg data structure, converting to byte array and calling webSocket.sendBIN() passing a pointer payload data and size.

// send servo position to Websocket server
void sendServoPosition()
{
  sendMsg.data.cmd = CMD_SERVO_ANGLE;
  sendMsg.data.value = angle;

  // write message to buffer
  writeByteArray();
  
  webSocket.sendBIN(byteArray, PACKET_SIZE);
}

Server Side – Python WebSocket (WS) Library

A simple WebSocket server implemented in Python v3x is tasked with text / binary message exchange, tracking connected clients and maintaining shared application state.

WebSocket server event loop is started by passing function/method name, IP address and port –

start_server = websockets.serve(wsApi, "192.168.1.127", 6789)
asyncio.get_event_loop().run_until_complete(start_server)
asyncio.get_event_loop().run_forever()

In event loop new clients are registered. Clients can be identified by checking protocol headers –

async def wsApi(websocket, path):
    # register(websocket) sends user_event() to websocket
    await register(websocket)

...
async def register(websocket):
    USERS.add(websocket)
    await notify_users()
    print(websocket.request_headers);
...

### debug output
Host: 192.168.1.127:6789
Connection: Upgrade
Upgrade: websocket
Sec-WebSocket-Version: 13
Sec-WebSocket-Key: LnOM1uj7n3gE4cFGqJ1yFg==
Sec-WebSocket-Protocol: arduino
Origin: file://
User-Agent: arduino-WebSocket-Client

Received message and headers are printed and handling is defined for binary (byte array) format messages.

### <class 'bytes'>
### b'\x0c\x00\x00\x00@\x00\x00\x00'
### (12, 64)

Python struct library allows packed binary bytes representing C structs to be unpacked as native python data types.

Format specifier “Ii” represents a message containing an int and unsigned int.

    try:
        async for message in websocket:
            print('Sec-WebSocket-Key: '+websocket.request_headers['Sec-WebSocket-Key'])
            print('MessageType: '+str(type(message)))
            print(message);
            if isinstance(message, (bytes, bytearray)):

                tuple_of_data = struct.unpack("Ii", message)
                cmd = tuple_of_data[0]
                value = tuple_of_data[1]
                STATE["value"] = value
                await notify_state()
    except Exception as e:
        print(e);
    finally:
        await unregister(websocket)

Connected WebSocket clients are push notified (synchronised) when position data (state) is updated –

async def notify_state():
    if USERS: 
        binary_data = struct.pack("Ii", 12, STATE['value'])
        await asyncio.wait([user.send(binary_data) for user in USERS])

WebSocket Browser Client – Binary Messaging in JavaScript

Web Interface is responsive and runs in mobile and web browser clients.

A radial D3.js radial gauge displays current angle. HTML5 buttons (divs) and slider control allow position to be updated.

Web / Mobile Interface provisioned by WebSocket (RFC6455)

JavaScript WebSocket onmessage function handles decoding of binary framed data.

A FileReader object is used to convert Blob to ByteArray in an asynchronous function, with flow control provided by a promise (future). TypedArray Uint8Array is used to extract 4 byte int and unsigned int command and value data fields –

websocket.onmessage = function (event) {
    if (event.data instanceof Blob)  // Binary Frame
    {
      async function readBinaryData(blob) {
          let promise = new Promise((res, rej) => {

            var fileReader = new FileReader();
            fileReader.onload = function(event) {
                var arrayBuffer = event.target.result;
                res(arrayBuffer);
            };
            fileReader.readAsArrayBuffer(blob);
          });

          // wait until the promise returns us a value
          let arrayBuffer = await promise;

          var v = new Uint8Array(arrayBuffer);
          // v[0] = cmd, v[4] = value
          //console.log(v[0] + " " + v[4]);

          // update UI elements
          value.textContent = v[4];
          angleSlider.value = v[4];
          gauges.forEach(function(g) {
              g.gauge.update(v[4]);
          });
      };

      readBinaryData(event.data);

Conclusion

WebSockets offer significant advantages over HTTP request/response polling techniques for real time data exchange, principally overhead of opening connection should occur only once per client.

Message push notify model, lightweight protocol framing, client libraries for embedded devices and native support in modern web browsers make this protocol well suited to real time Internet of Things data message exchange.

While underlying TCP provides message ordering and re-transmission (of failed packets), higher level application abstractions: guaranteed message delivery (at least once, at most once), message acknowledgements and queue / persist / forward (to offline clients) are not specified by WebSocket specification.

Similarly, TLS SSL can be used at network layer to encrypt data transmission (WebSocket Secure WSS), but client authentication / authorisation is not handled by WebSocket protocol, meaning a strategy for token or key based client identification (OAuth for example) must be considered for secure use cases.

Event driven implementations supporting asynchronous non-blocking IO result in efficient, well structured and modular code with handlers dedicated to specific tasks.

Factors influencing choice between binary framing and text format messages (field delimeted, JSON) include legibility, convenience, compactness and parsing overhead. Binary format introduces complexity due to differences in compiler / platform / network architectures and data type implementations between programming languages.

Full Source Code on GitHub –

Web UI –
https://github.com/steveio/arduino/blob/master/python/wsESP8266RotaryEncoderServo.html

Python WebSocket Server –
https://github.com/steveio/arduino/blob/master/python/wsESP8266RotaryEncoderServo.html

ESP8266 WebSocket MicroServo Sketch –
https://github.com/steveio/arduino/blob/master/ESP8266RotaryEncoderServo/ESP8266RotaryEncoderServo.ino