Is it a good idea to continuously reconfigure SPI form 8 bit to 16 bit or vice versa since the situation demands it.

[Pramod singh]

I have two peripheral where one works on 8-bit mode and the other on 16-bit mode. 

What could be the best possible solution for such situations?

[S.Ma]

I assume 2 peripheral means 2 SPI slave devices controlled by SPI Master MCU?

Use 8 bit mode to get something simple and robustly working first.

If your SPI has built-in FIFO, the difference between 8 and 16 bit transmission will be diminished.

(example on STM32L4 family).

Otherwise, reconfiguring for each addressed device is also possible, while adding complexity increases risks for bugs.

[Pramod singh]

Thank you for your reply.

So basically you mean to say that, I will configure my master in 16 bit mode and use it for both the 8 and 16 bit slave if my controller supports FIFO.

However the controller i'm using is stm32f4 series

[S.Ma]

if SPI doesn't have FIFO, better stay in 8 bit mode for both to get started.

Unless transmitting big amount of data, or having a low SYSCLK, it should be good enough before "optimizing" such as dynamically adjusting the 8 or 16 bit mode.

[Pramod singh]

by the way my controller have FIFO, I just checked it out. So what do you suggest now to do

[S.Ma]

if you program the SPI in 8 bit mode, a 32 bit FIFO lies between the DR and the shift register.

It means if you do a 8 bit write to DR, the fifo will store 8 bit to be "digested" into the shift register.

If you do a 16 bit DR write, it will be decomposed and queued by the FIFO as 2x8 bit to feed the shift register. you could write 3 bytes or 4 bytes at once if you'd like. With FIFO, there wouldn't be much benefits to configure the SPI in true 16 bit mode here.