ESP32 Forum

ffplus wrote:I have come across exact same thing with ModbusMaster library as well. Too many Modbus and Logic analyser tests lead me to editing of the library and found the solution actually.

Problem is according to my guess, when you use Serial.write() function, due to async nature of ESP32 it doesn't wait for the buffer to send all data on serial port. It instantly pass through other command line for disable to Transmit Enable pin.

A solution to this create exact amount of time required to send the byte frame of modbus like this at transaction function on ModbusMaster.cpp:

(For ModbusMaster library)

// flush receive buffer before transmitting request
while (_serial->read() != -1);

// ADD THIS after write
long baud = BAUD_RATE; // Your baudrate
float symbol_time = 1000000 / baud; // Create time for a single symbol
float wait_time = ( ( (1 + 8 + 1) * symbol_time ) * u8ModbusADUSize); // Create exact payload size.

Then after write of payload:
for (i = 0; i < u8ModbusADUSize; i++)
{
_serial->write(u8ModbusADU);
}
// WAIT ENOUGH TIME TO SEND DUE TO ESP32 ASYNC NATURE
delayMicroseconds(wait_time);
delayMicroseconds(1700); // Time for frame ending for modbus standart (I didn't have time to fully implement would be gratefull if someone else can does it)

Of course this is only transmit part. At the receive part also whenever you read serial buffer it starts with 0x00 byte. So I created an ugly workaround

int start = 0; // Put this here
if (_serial->available())
{
// PUT this to here, it reads first byte and do not put into modbus payload
if(start == 0) {
_serial->read();
start = 1;
continue;
}

With this workaround I could able to successfully read modbus response from payload as well.

But I think here with this write and read issues, SDK team should look and create a more understandable and robust solution.