baudrate

According to MSDN:Sample, Maybe you need to set a pin's signal state to indicate the data has been sent/received in your microcontroller program. More details reside in your serial communication transmission of data standard. And you should write code according to the result of WaitCommEvent(hCom, &dwEvtMask, &o); like the linked sample.

I'm assuming this is due to your code re-initializing the connection each loop. I recommend trying the following code instead --

Does the library auto-assign the slave number, or we have to define it?

With MinimalModbus you pass the slave ID through in the minimalmodbus.Instrument('/dev/ttyUSB0', 1) call (the 1 is the Slave ID). How you set the slave ID on the device itself varies (this is not covered by the Modbus over serial line spec; could be a configuration program, DIP switches, based on the serial number etc. Other libraries may take different approaches (for example defaulting to Slave ID 1).

From the datasheet, is it the register number is the address?

The header in the spec tables says "No(Address)" so "10001(0000)" means register 1, address 0 (these refer to the same thing; I recommend reading the "Modbus: When 40001 Really Means 1, or 0 Really Means 1" section in this article which explains some of the issues around addressing).

And how many decimals do I expect if there's two data sequence as a response?

Not quite sure what you mean by "two data sequence". The Modbus spec only details the sending of bits (coils) and 16 bit values (input/holding registers). From a quick look at your spec it looks like this device just uses a single registers; for instance "OUT1 Output time" has "unit: ×10㎳" so take whatever is in the register and divide by 10 to get ms.

Is the CRC16 check already included within the library as i shouldn't code it?

Any decent Modbus library will look after the protocol details (such as CRC) for you (so no you don't need to code this; MinimalModbus will calculate it for you)

The device manual isn't clear about the register start address, but the first register it mentions has the address of 1.

Similarly, the mbpoll command-line utility (not the one with GUI) isn't very clear about the start address. But its documentation mentions that the default value for -r parameter is 1.

I think it's safe to assume that both use the same addressing which starts from 1, as the command-line tool has no problems accessing the value.

But MinimalModbus API clearly mentions that its register start address is 0. So when using this library, you need to use registeraddress = 45 for accessing the temperature, not 46 or 40046.

But why won't 46 work? Normally, one would expect it to grab data starting from the next register and print some garbage, but not timeout. But we can't know how the device works internally. Maybe a request to access the temperature register actually triggers some measurement function and then returns a value. A request to access an unaligned data (with a wrong register value) can be simply rejected by the firmware.

If you still get timeouts with registeraddress = 45, your Python runtime may have some problems accessing the serial port. As I stated in my comment, I recommend using a logic analyzer to see what's going on on the wire. Without such a tool, you're doing blind-debugging.

You can't use serialPuts to send the null terminator. As with all similar functions, it will stop when the null terminator is encountered in the string. In this case I think your best option is to add a function that uses the ordinary write function that is used internally by WiringPi's own functions.

You could make a wrapper function to make it look similar to the other calls:

Your first line for the USART3 initialization is wrong :-)

RCC->APB2ENR |= 1; // enable clock for AF

must be (without a clock the USART doesn't work)

RCC->APB1ENR |= (1<<18); // enable clock for USART

And as an additional hint, do not use all these magic numbers for the bits. This is much more readable (and the needed defines are already done in the CMSIS:

RCC->APB1ENR |= RCC_APB1ENR_USART3EN; // enable clock for USART

Assuming the code you included is the complete program source code (ie. you're compiling it as a command-line app), the problem is that the program terminates right after sendString() is called, so the port has no chance to receive anything. You need to keep the program alive and running. There are multiple ways to do that, but the simplest is just to call RunLoop.main.run() as the last line of the program. That will cause the run loop to run indefinitely processing input from all sources, and will both keep the program alive and run the system machinery that ORSSerialPort relies on.

The problem lies with

There is a simple solution by changing the way in which you use your show_frame function: