Transmit bitstream through serial port

Hi Matlab enthusiats,

I`m currently trying to send data (calculated from matlab) via the serial port of my computer to a board with a small microcontroller on it. The serial communication works in principle but I first want to convert the numbers from float to fixed point numbers for easier handling on the microcontroller. The transformation to fixed point works as well but as I`m trying to send the data matlab want`s a precision variable in the write command which yields to matlab interpreting the data and converting each single bit to a number of the given precision type.

So my question is there a possible way to directly transmit a bitstream (chopped down to 8 bits packages) via the serial port? (I know I can run the char array through a for loop and convert it to an integer number and transmit that but i was wondering if there would be an easier solution).

Thanks in advance!

Greetings

Benjamin

PS: My test code:

device = serialport("COM5",9600)
q = quantizer('fixed', 'nearest', 'saturate', [16 8]);
write(device,num2bin(q,pi),"uint8");

Answer 1

[Shaik]

Hi,

% Open the serial port
device = serialport("COM5", 9600);

% Convert the floating-point number to fixed point
q = quantizer('fixed', 'nearest', 'saturate', [16 8]);
fixedData = num2bin(q, pi);

% Convert the bitstream to uint8
binaryData = reshape(fixedData.', 1, []);
binaryData = binaryData - '0';  % Convert from char '0'/'1' to numeric 0/1
binaryData = uint8(binaryData);

% Transmit the bitstream
fwrite(device, binaryData);

% Close the serial port
delete(device);
clear device;

[Benjamin]

Thank you for the quick answer.

I`ve just tried it and the problem is that matlab transmits each bit individually as one byte (so 16 bytes) on the serial port. My goal would be that matlab only transmits the necessary 2 bytes.

Greetings

[Shaik]

% Open serial port connection
s = serialport("COM1", 9600);  % Replace "COM1" with the correct port and 9600 with the desired baud rate

% Data to transmit
value = int16(1234);

% Convert the value to bytes
bytes = typecast(value, 'uint8');

% Transmit the bytes over the serial port
write(s, bytes);

% Close the serial port connection
delete(s);
clear s;

[Walter Roberson]

@Shaik

The question has to do specifically with having used the Fixed Point toolbox to create a representation of the data. The example given was to represent pi in 16 bits with 8 of the bits being fraction. You cannot just int16() a floating point number and typecast() to uint8 .

In the particular case of [16 8] you might be able to use typecast(int16(VALUE*256), 'uint8') -- though it would be a good idea to check the boundary conditions to be sure that values close to the boundaries are treated the same way by this code as compared to the code I show using num2int()

Answer 2

[Walter Roberson]

bytes = typecast(swapbytes(int16(num2int(q,VALUES))), 'uint8')

Note: this has been deliberately arranged to have the high-order bytes first -- for example, [3 36] in transmission order corresponding to 3*256 + 36 = 804. The default, without swapbytes, is to use "in-memory" order, and in-memory order for all supported CPUs is currently "little-endian", low byte followed by high byte in memory. If you are transmitting to another MATLAB system with the intention of reconstructing the fixed-point number then the swapbytes adds overhead that might not be technically necessary (but in my experience, "big-endian" is notably easier to debug than "little-endian")