Hi Johirul,

generally, math programs like matlab provide many useful functions but are in certain cases orders of magnitude slower than plain C programs on the same hardware. So, if you intended to use matlab during the development phase for graphically displaying the data and/or applying some transformations like FFT, digital filters, etc., I would write a C program instead, maybe with the help of the book "Numerical recipes in C", and a graphics library, and compile it using the free compiler by Bloodshed Software, which, despite its creepy name, works fine.

In the working prototype, you should avoid using several microcontrollers / microprocessors. If the ready application involves, for example, digital signal processing, the 8-bit ATmega328P (in the Arduino) is certainly too small and too slow. You could switch to a 32--bit microcontroller which supports floating point arithmetic by hardware. Examples are Atmel's SAME70 (evaluation board SAME70-XPLD), and Texas' AM335x (used on the BeagleBone boards). The latter is in several aspects more powerful, but personally, I prefer Atmels IDE. In both cases, the hardware provides powerful DMA mechanisms, so transferring the data between internal components doesn't even take CPU time (apart from the configuration).

BTW, I guess "1000 ns" = 1 microsecond < 84 microseconds is a typo.

Hope this helps.

Your not the only one struggling with the real time processing in MATLAB.

Johirul's answer is excellent!

Some concerns from another project are adrdresssed here:

https://uk.mathworks.com/matlabcentral/answers/401780-real-time-processing-with-matlab#answer_321195

If I were you, I would buffer the data first in the Atmega chip first then transfer the data chunk to the matlab. Lets say buffer every fey bytes of data.