Hello,

I have an unreliable ModbusTCP device I want to read via this library which sometimes just doesn't reply.

It is unfortunate that the ModbusTCPClient waits 30 seconds for a reply. Even more bad: I cannot change this timeout (excerpt of ModbusTCPClient.cpp):

int ModbusTCPClient::begin(IPAddress ip, uint16_t port)
{
  modbus_t* mb = modbus_new_tcp(_client, ip, port);

  // Here is the timeout set to 30 Seconds
  modbus_set_response_timeout(mb, 30, 0);

  return ModbusClient::begin(mb, MODBUS_TCP_SLAVE);
}

An overload would be great here! You can exchange the whole upper part with:

int ModbusTCPClient::begin(IPAddress ip, uint16_t port, uint32_t usecTimeout)
{
  modbus_t* mb = modbus_new_tcp(_client, ip, port);

  // Set Timeout to value provided
  modbus_set_response_timeout(mb, 0, usecTimeout);

  return ModbusClient::begin(mb, MODBUS_TCP_SLAVE);
}
int ModbusTCPClient::begin(IPAddress ip, uint16_t port)
{
  // Keep standard timeout of 30 seconds
  return begin(ip, port, 30000000);
}

Do you think you can make this official so I can revert my local changes in your library.

Thanks Soko

This is an alternative to pull request #11

Instead of having a parameter to begin(...) a new ModbusRTUClientClass or ModbusRTUServerClass instance can be declared with an alternative RS485Class instance.

Removing Arduino.h causes compile issues on stm32duino so adding back just the needed dependency.

Here is the definition of delayMicroseconds in stm32duino https://github.com/stm32duino/Arduino_Core_STM32/blob/a7283739a4d8bafe95c9f9b9156e6f82a1ebbd75/cores/arduino/wiring_time.h#L64

this breaking change was introduced on https://github.com/arduino-libraries/ArduinoModbus/pull/58 pull request

Compiling .pio/build/universal/lib640/ArduinoModbus/libmodbus/modbus.c.o
.pio/libdeps/universal/ArduinoModbus/src/libmodbus/modbus.c: In function '_sleep_response_timeout':
.pio/libdeps/universal/ArduinoModbus/src/libmodbus/modbus.c:156:5: warning: implicit declaration of function 'delay' [-Wimplicit-function-declaration]
  156 |     delay(ctx->response_timeout.tv_sec * 1000);
      |     ^~~~~
.pio/libdeps/universal/ArduinoModbus/src/libmodbus/modbus.c:157:5: warning: implicit declaration of function 'delayMicroseconds' [-Wimplicit-function-declaration]
  157 |     delayMicroseconds(ctx->response_timeout.tv_usec);
      |     ^~~~~~~~~~~~~~~~~
Archiving .pio/build/universal/lib640/libArduinoModbus.a
Indexing .pio/build/universal/lib640/libArduinoModbus.a
Linking .pio/build/universal/firmware.elf
/home/user/.platformio/packages/toolchain-gccarmnoneeabi/bin/../lib/gcc/arm-none-eabi/9.2.1/../../../../arm-none-eabi/bin/ld: .pio/build/universal/lib640/libArduinoModbus.a(modbus.c.o): in function `_sleep_response_timeout':
modbus.c:(.text._sleep_response_timeout+0x1e): undefined reference to `delayMicroseconds'
collect2: error: ld returned 1 exit status
*** [.pio/build/universal/firmware.elf] Error 1

With memory being tight on most micro controllers, I felt like I shouldn't have to allocate memory for modbus to use, but instead point the server at existing data structures. If the structures are not of the same type, the user can cast it as such.

This proposed change makes it possible to use an RS485Class object other than the default RS485 with the contained libmodbus. This is useful with boards with multiple serial ports or when the user wants to specify different DE/RE pins.

I'm using the latest version of the library. Via the rs485 library, I can get my counter to answer the request correctly. Implementing the sketch with this lib, it gives me Invalid CRC error and Response not from requested slave. So there is a problem with the receiving buffer. Could it be timeout? I think it is splitting the msg in two.

HW: Arduino MKR WIFI1010 with MKR485 shield Finder 7E.64.8.230.0210 Meter

The #define ON and #define OFF in modbus.h creates a conflict with Arduino Menu 4 library. It is only used once in modbus.c and can be substituted for TRUE FALSE, also defined in modbus.h

This fix might be called "trivial" but it fixes a compiler warning, and with some boards that use inline code for delayMicroseconds, it is an error rather than just a warning.

Despite the last fix, I cannot run the exemple program on uno wifi rev 2. I used the code in the ArduinoModbus version 1.02 modified 8 days ago, and the IDE 1.8.13 under OSX. I get the following message :

Arduino : 1.8.13 (Mac OS X), Carte : "Arduino Uno WiFi Rev2, None (ATMEGA4809)"

/Users/xxx/Library/Arduino15/packages/arduino/hardware/megaavr/1.8.6/cores/arduino/main.cpp: In function 'main':
/Users/xxx/Library/Arduino15/packages/arduino/hardware/megaavr/1.8.6/cores/arduino/api/Stream.h:50:7: warning: 'MEM[(struct Stream &)&client]._startMillis' may be used uninitialized in this function [-Wmaybe-uninitialized]
 class Stream : public Print
       ^
/Users/xxx/Documents/Arduino/libraries/WiFiNINA/src/WiFiServer.cpp:69:20: note: 'MEM[(struct Stream &)&client]._startMillis' was declared here
         WiFiClient client(sock);
                    ^
/Users/xxx/Documents/Arduino/libraries/WiFiNINA/src/WiFiClient.h:28:7: warning: 'client._socket' may be used uninitialized in this function [-Wmaybe-uninitialized]
 class WiFiClient : public Client {
       ^
/Users/xxx/Documents/Arduino/libraries/WiFiNINA/src/WiFiServer.cpp:69:20: note: 'client._socket' was declared here
         WiFiClient client(sock);
                    ^
/var/folders/kf/5_pjskbd1yg3j6qlthc39v5m0000gn/T//ccpthV3p.ltrans0.ltrans.o: In function `begin':
/Users/xxx/Documents/Arduino/libraries/ArduinoModbus-master/src/ModbusTCPServer.cpp:40: undefined reference to `modbus_new_tcp'
/Users/xxx/Documents/Arduino/libraries/ArduinoModbus-master/src/ModbusTCPServer.cpp:46: undefined reference to `modbus_tcp_listen'
/var/folders/kf/5_pjskbd1yg3j6qlthc39v5m0000gn/T//ccpthV3p.ltrans0.ltrans.o: In function `accept':
/Users/xxx/Documents/Arduino/libraries/ArduinoModbus-master/src/ModbusTCPServer.cpp:55: undefined reference to `modbus_tcp_accept'
collect2: error: ld returned 1 exit status
exit status 1
Erreur de compilation pour la carte Arduino Uno WiFi Rev2

Did I missed something ?

Best Regards Frederic

Trying the modbus library with the following sketch:

#include <ArduinoModbus.h>

float temperature, humidity;

void setup() {
  Serial.begin(9600);
  while (!Serial);

  Serial.println("Modbus Temperature Humidity Sensor");

  if (!ModbusRTUClient.begin(9600)) {
    Serial.println("Failed to start Modbus RTU Client!");
    while (1);
  }
}
 int i=0;
void loop() {
  if (!ModbusRTUClient.requestFrom(1, HOLDING_REGISTERS, 0x00, 2)) {
    Serial.print("failed to read registers! ");
    Serial.println(ModbusRTUClient.lastError());
  } else {
    short rawtemperature = ModbusRTUClient.read();
    short rawhumidity = ModbusRTUClient.read();
    temperature = rawtemperature / 10.0;
    humidity = rawhumidity / 10.0;
    Serial.println(temperature);
    Serial.println(humidity);
  }
 Serial.println(i);
  delay(1000);
  i++;
  if(i>2){
    while(1);
  }
}

rise an issue related to reception of the packet from the RS485 channels, it alternate a good reception with a discard, as show in the in the image. I have sniffed the channel with the Saleae analyzer, and the packet are the same, the capture are in the following .zip.

modbus.zip

The problem could be located in the library libmodusb/modbus.c in modbus_read_registers(modbus_t *ctx, int addr, int nb, uint16_t *dest) or in the integrity check function of the class modbus_backend_type_t. In particular must be checked how the CRC is calculated in two consecutive reading from the RS485 . Following, also, a stamp of serial monitor output:

Bumps actions/checkout from 2 to 2.3.4.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

This is imho necessary since even one of our own examples does not link anymore, due to the required memory exceeding the available ressources of the ATMEGA328P based Arduino Nano.

Hi,

I'm trying to have an Arduino Uno serve as a Modbus RTU server to simulate a device I need for work. I'd like to setup multiple holding register blocks but so far I'm not having any luck as I get the "illegal data address" error.

What I mean is that if I would like to read from holding registers 6, 254, 867, and 1023 then I can't get it to work just by running

//Configure 1024 registers
result = ModbusRTUServer.configureHoldingRegisters(0, 1023);
if (result != 0) {
  Serial.println("Holding registers configured incorrectly!");
}

//initialize the 1024 registers with the number of the register
for (int i = 0; i < 1024; i++) {
  result = ModbusRTUServer.holdingRegisterWrite(i, (uint16_t)i);
  if (result != 1) {
    Serial.println("Failed to write to the holding registers!");
    break;
  }
}

What would be the correct way to go about doing this?

I know this was just opened and closed, but I've used a wrong account for that (https://github.com/arduino-libraries/ArduinoModbus/issues/101).

Thank you!

Some modbus client (e.g pymodbus) assert dtr before transmitting, this lead to reset arduino to bootloader ( i spent one day around the problem on arduino-nano used as modbus-slave) Resolved disabling reset with a 120 ohm resistor from 5v to rst to disable dtr reset. After that, my program works flawless Not an issue but i think would be nice to warn user if strange behaviour happen. Thank you for sharing your work!

The usual code for servicing a clients only works with one connected client at a time.

while (1) {
    
    ethClient = ethServer.available();
    
    if (ethClient) {

        modbusTCPServer.accept(ethClient);

        while (ethClient.connected()) {
          modbusTCPServer.poll();

Is there a correct way (or an example) for how to connect multiple clients to the server at one time.

Thanks

Got it to work on a Portenta H7 - only testing Holding Register reading (all required in my application).

I have a question about the software.

Is it possible to issue 2 configures as below

modbusTCPServer.configureHoldingRegisters(0, nSensors);
modbusTCPServer.configureHoldingRegisters(1000, nSensors*2);

I would like to have the first group be scaled integers, and the second group will be transmitted as floats. The user will decide which to access depending on their PLC etc

Thanks