Nextion Touchscreen with ESP32

Nextion has a nice series of touch displays that make a very good human machine interface (HMI) solution for embedded products. The interface combines an onboard processor and memory with a touch display. Nextion has developed a software editor to support HMI GUI development.

The Nextion HMI editor has drag and drop components to create your user interface. The display is connected via TTL serial (3V3) RX/TX and requires a 5V/GND power connection with enough current to run the display.

The device I used was part of the Basic Series, NX3224T024, which is a 2.4″, 320×240, resistive touch panel. The document is not clear on which wires are the TX and RX. I found the Nextion TX is the blue wire and the RX the yellow wire. Since the original post I have found documentation showing the interface is 3V3 compatible and will accept 5V on the RX.

For my application I needed a method for a user to configure the embedded system. Instead of using a serial port I selected the Nextion touchscreen. The application was hosted on an ESP32 and code was developed using the Arduino IDE. I found the itead library, which implemented all of the objects I needed in version v0.9.0. I had to modify some of the files in the library to work properly with the ESP32. In NexUpload.cpp I moved the include software serial #include to be within the define block USE_SOFTWARE_SERIAL (makes sense). I also had to modify the NexHardware.cpp NexInit function to assigned the hardware serial pins I used (had to move Serial1 pins due to conflict with flash control).

// NexUpload.cpp
//#define USE_SOFTWARE_SERIAL
#ifdef USE_SOFTWARE_SERIAL
#include <SoftwareSerial.h>
SoftwareSerial dbSerial(3, 2); /* RX:D3, TX:D2 */
#define DEBUG_SERIAL_ENABLE
#endif
// NexHardware.cpp
bool nexInit(void)
{
    bool ret1 = false;
    bool ret2 = false;
    
    dbSerialBegin(9600);
    nexSerial.begin(9600,SERIAL_8N1, 32, 33);	// modified for serial1 I/O
    sendCommand("");
    sendCommand("bkcmd=1");
    ret1 = recvRetCommandFinished();
    sendCommand("page 0");
    ret2 = recvRetCommandFinished();
    return ret1 && ret2;
}

To build the ESP32 Nextion interface there are basically 5 steps after you create the HMI using the Nextion editor.

  • Define objects
  • Create an object listen list
  • Create object callbacks (what to do on touch screen events)
  • Attach callbacks to objects
  • Use nexLoop to monitor the Nextion device

Defining the object requires information from the Nextion HMI editor. To define an object you need the object name, page, and id number. Watch out when editing Nextion pages, ids can change. The listen list is a NexTouch type array of pointers to the objects you created that have the events you are interested in. The callbacks are actions to take based on Nextion events (button push, release, etc.). Attaching callbacks associates the callback routine with an object and event. The object method attachPop() is used when attaching a callback for a button release.

So putting it all together. For a MIDI project I had multiple parameters that could be modified, e.g. MIDI channel number. The parameters were all numerical values so I create a change page that allowed a user to increase or decrease the value by pushing a plus or minus button. Once finished, the value was updated.

Nextion Parameter Change Page

The parameter page had five objects, a title (text), value (numeric), and buttons for plus, minus, and done. The title was static and loaded with the page. The parameter value was also loaded showing the current parameter value. All objects except the title and value were setup for “Touch Release Event” to “Send Component ID”, which means when the button is pushed and then released, a serial string is transmitted with the Nextion object page, id, and name. The protocol is handled by the itead Nextion library.

Defining the objects for the parameter change page used the itead library definitions. These objects were on my Nextion page 6. The library parameters are <Object>(<page>,<id>,<name>).

// Define Nextion Objects
// page 6
NexButton minusChange         = NexButton(6,4, "b1");
NexButton plusChange          = NexButton(6,5, "b2");
NexNumber parameterValue      = NexNumber(6,2, "n0");
NexButton setParameter        = NexButton(6,3, "b0");
NexText parameterName         = NexText(6,1, "t0");

Next these objects were added to the listen list. The reduced version showing the page 6 objects is shown below. Remember to put a NULL at the end of the list. Note that I didn’t setup events for the parameter name or value. Only the plus and minus button that changed the parameter and a done button to set the new value were setup for events.

// Listen List
// object list for touch screen
NexTouch *nex_listen_list[] = {
  // page 6
  &minusChange,
  &plusChange,
  &setParameter,
  NULL
};

The call backs are the actions to perform when the event occurs for that object. For the plus and minus the parameter value is read from the Nextion object, checked against a max/min value, and updated accordingly. The Done button updates the parameter value to the current page value and returns to the calling page. Global variables were used for the min, max, and return information.

// Callbacks
// page 6
void minusPopCallback(void *ptr) {
  uint32_t number;
  parameterValue.getValue(&number);
  if (number > minParameterValue)
    parameterValue.setValue(number - 1);
}
void plusPopCallback(void *ptr) {
  uint32_t number;
  
  parameterValue.getValue(&number);
  if (number < maxParameterValue)
    parameterValue.setValue(number + 1);
}
void setParameterPopCallback(void *prt) {
  uint32_t getValue;
  // update variable that is changing
  parameterValue.getValue(&getValue);
  *modifyParameter = (int)getValue;
  // go back to last page
  returnPage->show();
  // more stuff based on returning page
}

Attaching the callbacks for the release event uses the attachPop() object method. The parameters are the callback function and object. If I had setup the touch event for a push, then the method is attachPush().

// Attach Callbacks
void setupNextion (void) {
  // page 6
  minusChange.attachPop(minusPopCallback, &minusChange);
  plusChange.attachPop(plusPopCallback, &plusChange);
  setParameter.attachPop(setParameterPopCallback, &setParameter);
}

Finally, once everything is setup and initialized, use nexLoop with the listen list to handle the interface to the Nextion touch screen. The nexLoop polls the serial interface and is non-blocking, so my listener was setup as a task.

// Monitor Nextion Interface
// task
void getControl(void * parameter) {
  for(;;) {
    nexLoop(nex_listen_list);
    vTaskDelay(50 / portTICK_PERIOD_MS);
  }
}

Overall the interface was very successful and I was happy with the results. The project had 7 different pages that required almost 900 lines of code. I’m sure there are more efficient ways of coding this interface, which I will discover as I continue to work with Nextion touchscreens.

Bin Packing Problem

I saw an interesting project on Guru.com to look at pallet packing efficiency. The client input had job numbers/quantities and they wanted an optimized pallet loading. The loading was based on a minimum/maximum pallet quantities with a criteria that a job quantity can’t be split across pallets (a job must ship together). The client also had a “desired” quantity, which really didn’t provide additional information. The goal was to develop an Excel VBA to provide the optimized pallet loading for what I assumed to be their daily output.

Although I haven’t worked on bin packing algorithms, this sounded a lot like memory allocation schemes, which I have implemented; first, best, and worst fit algorithms. I performed some basic research and found that bin packing solutions are similar to memory management allocation schemes. The bin packing algorithms include next-fit, first-fit, and best-fit algorithms. There are also variations of these algorithms that pre-sorts the data in decreasing quantities sizes before apply the algorithm.

The next-fit algorithm checks to see if the the current bin can hold the quantity. If so, then place that quantity in that bin, else place the quantity in a new bin. With the next-fit algorithm you never go back to an earlier bin.

The first-fit algorithm scans open bins in order and places the quantity in the first bin that will hold it. If the quantity doesn’t fit in any bin, then start a new bin. The best-fit scans all bins for the best fit, if it doesn’t fit in an existing bin, then start a new one. The decreasing algorithm versions have the quantities sorted by size, largest to smallest, and then the algorithm is run.

Although not seen in the literature, I also included a worst fit algorithm which is the opposite of the best-fit algorithm where bins are scanned for the worst fit (most space left after adding the quantity to the bin). If the quantity doesn’t fit in an existing bin, start a new one.

I created an Excel VBA to run some trials where I could vary different parameters and create constrained random values. Random parameters includes the number of jobs and quantities per job. Fixed parameters are the min/max quantity on a pallet (in a bin) and the number of trials. After each trial the most efficient solution was selected (least number of pallets/bins) that met the minimum pallet quantity (i.e. a solution with the minimum number of pallets wasn’t selected if any pallet quantity was below the minimum threshold).

After varying a number of parameters and using uniform random numbers, generally the best fit or best fit decreasing algorithm was the most favorable. This experiment did not account for processor speed. One would expect that best and worst fit algorithms to require the most processing power since every open bin needs to be examined prior to placing a job in a bin.

DAQFactory Introduction

Recently I had an assignment to interface and display sensor data on an old Windows XP laptop. The client was reusing some existing infrastructure, sensors and serial communication radios, to display real-time sensor information at a remote office location. The original effort was to interface Excel serially to existing sensors. The client had already selected a 4-20mA DAT3015-I sensor interface that used Modbus protocol.

I hadn’t planned on the Modbus protocol. Although I have worked with data communications with sensors for more than 35 years, I had never worked with Modbus. Modbus is an application layer messaging protocol that provides client/server communication over different types of busses and networks. Modbus has been a serial de facto standard since 1979 and has a request/replay PDU structure.

Although there are Excel VBA code to support Modbus, I assumed that if the standard has been around since 1979 with companies are still building Modbus protocol hardware, there must be commercially available software for Windows XP that would work for this application.

I performed an industry survey and found about 10 software packages that met our requirements. I selected DAQFactory Express since it has custom screens that contain multiple real-time indicators and graphs, it appeared to be a robust environment with flexible compatibilities including custom coding, and the Express version is free. The free version limits the number of pages (2), I/O (8 channels), and screen components (11).

The display used a single screen and had indictors for

  • 2D trending graph showing real-time sensor data
  • Instantaneous sensor values in mA and GPM (converted data)
  • Signal status using colors and blink text when an error was detected
  • Real-time operating controls (stop/continue), reset, test mode and mode indicators
  • Error feedback
  • Display control (graph data and error thresholds)
DAQFactory Finished Product

DAQFactory has multiple setup screens for device setup (physical interface and protocol), channel setup, conversion formulas, and others. DAQFactory provides a lot of flexibility in setting up your device interface and how data are handled within the environment. DAQFactory is not limited to Modbus protocol and supports even custom interfaces.

I chose to use DAQFactory sequences, which are code segments similar to C in syntax and is object oriented. For this application the primary sequence is GetData that is used to get sensor current readings, the device name, and reset coil status (a Modbus thing). All other sequences work with runtime operation such as stop/continue and entering/exiting test mode.

Sequences have a nice try/catch structure to handle events like data timeouts. With the Modbus protocol and server requests data from the client. Sometimes the message is missed and no data are received from the client. The try/catch is perfect to handle this type of error. To read data from the DAT3015-I, a simple method was used that specified the data type, address, and number of values.

inputData = Device.DAT3015.ReadHoldingS16(1,40015,2)  // read registers (15 & 16)

This call created the properly formatted Modbus string that was sent to the client and received the response and placed the 16-bit signed data into an 2-D array, inputData. To get the instantaneous values from the most recent data a simple array access was used.

Outfall_uA = inputData[0][0]     // outfall uA value reg 40015
Overflow_uA = inputData[0][1]    // overflow uA value reg 40016 

The most recent data are then converted and added to the real-time graphical display using the AddValue method.

 // convert data algorithm, data in uA
 // if < 4mA GPM = 0
 // GPM = uA * 1.25 - 5000
       
 // Outfall sensor
 if(Outfall_uA < 4000)
    OF01_Outfall.AddValue(0)
 else
    OF01_Outfall.AddValue(Outfall_uA * 1.25 - 5000)
 endif
  
 // Overflow sensor
 if(Overflow_uA < 4000)
    OF01_Overflow.AddValue(0)
 else
    OF01_Overflow.AddValue(Overflow_uA * 1.25 - 5000)
 endif 

A 2D trending graph was used for each sensor and displays GPM verses time. The graphs auto scales in the y-axis (GPM) utilizing the Min/Max expressions on the channel data (e.g. Min(OF01_Outfall)). The time scale is user adjustable using the runtime display control at the button of the display. This setting uses a registry variable and is retained between DAQFactory restarts.

Instantaneous values for mA and GPM are displayed for each sensor using Variable Value Components. The value is set using an expression to the most recent reading and conversion (e.g for mA Overflow_uA/1000, for GPM OF01_Overflow[0]). The [0] index is the most recent value. No color changes or actions are associated with these objects.

Each sensor has a prominent display using a Descriptive Text Component (OutfallStatus, OverflowStatus). An inline if expression on the sensor uA variable is used to generate either a 1 (good) or 0 (problem).

iif(abs(Outfall_uA) < Registry.signalThreshold,0,1)

Overall the project was a success. The develop time was short and helped by the client sending the DAT3015-I for integration testing prior to installation. The client was able to install and add the DAQFactory project to the Windows XP laptop. Once the client found a problem with an interface cable, the system ran without an issues. I highly recommend DAQFactory.