See the section about "Use of the Up/Down Mode" in the Timer A Chapter of User's Guide.

i know this information

Yassine Akrmi said:

i know this information

yes i know this information, what i want is a code from  texas, and output this two signal, but i have another problem, how to synchronise two msp430 to output the two PWM ?

TI has code examples.

Your

MSP430F2101

will do both PWM channels simultaneously while enforcing dead time.  You won't need to synchronize 2 MSP430 devices.  Just 1 is enough for the job.

Jeff

yes i know that this mcu will do the job but in my application i will use two MCU ( to generate 4 PWM) so i need that this MCU will be synchronized

Move up the chain and look for a 430 with 2 or 3 counters then instead.

If it's not too late, you should consider a different MSP430.  You will be much better off with a Timer_A5 or Timer_B7 so you have enough CCRs to do all four PWMs with just one MCU.  The MSP430F5304 would be perfect (48 LQFP), and you would need only 1.

Code examples from TI can provide almost 100% of your PWM code if you use just 1 MSP430.  However, if you use 2 MSP430s and try to synchronize them, you'll have to come up with a lot of code on your own.

Jeff

Jeff and Tony,

The original poster will probably tell you that he knew that too. It seems that he only wants the "Texas code" for some specific requirement he did not disclose yet.

OCY

Two synchronized MCUs require two things: a synchronized system clock and a synchronized start.

You can do it by outputting MCLK/ACLK/SMCLK (whatever is highest and always on) and route it to the otehr MSPs XIN pin instead of a crystal. THen you know both MSPs 'tick' with the same clock.

If you jsut want to sanchronize the two PWM output timings, you can route the same signal that ticks teh tiemr on MSP1 to the external timer clock input of MSP2. Then the two timers are synchronized only. But then the timer of MSP2 is not synchronous to the MSP2 MCLK anymore, which may rise soem problems when you read the timer registers.

The other signal you need is the initial synchronisation signal that esnures that the two tiemrs start at the same moment. This is difficult. one MSP needs to set a port pin low and the other waits until the port pin goes high again. Then both MSPs reset their timer. However, code execution and reaction times may easily cause some asynchronity. Even if you count code cycles, you'll encounter an uncertainity of some MCLK cycles which you cannot eliminate. You can try to capture the edges of the MSP1 PWM output and compare it with the timer of MSP2, giving you the current timing idfference in timer ticks, but it is still difficult to eliminate this completely even if you knwo how many cycles the two tiemrs are apart.

It is indeed best, you handle all PWMs on the same MSP, one with a TimerA5 or better a TimerB5..7 (which even allows synchronized change of the duty cycles)

please i have another question how can i generate an external clock ?using the crystal 32 khz ?

Yassine Akrmi said:

when i simulate the circuit when only using one system clock for the two mcu, i obtained two signal synchronized, i juste set the same system clock without synchronized start

The problem is that when you power-on the MSPs, both start running on DCO, independently and on slightly different speeds. You'll need to configure the clock output form the one and the clock input from the second later. However, without an additional sync signal, the second does not know when the first has enabled the output (as the second only sees the clock when it arrived at the point to switch to the other.

I tmight be done if the second MSP switches to external clock (and then freezes as there initially is no external clock) and a safe distance later, the first MSP enables the clock poutput. THne both know that they continue form a synchronized point. Then the timers need to be reset, and it is cycle counting to make them both reset at the exact same time. After this point, the two MSPs as well as their timers run synchronized and so will the PWMs - as long as you don't do anything that may disturb the running of the timers, e.g. changing the PWM frequency (CCR0).

Great, i have know used the PIN P1.0/TACLK in order to input the two clock to be the same in the msp1 and msp2, but i don't know how to generate an external clock, or can i generate this clock using the msp1 ? and input it in the msp2?

Yassine Akrmi said:

how can i generate an external clock ?using the crystal 32 khz ?

IIRC, it is also possible to output ACLK/2 or ACLK/4.

simply i configure the pin P2.0 in my case  P2DIR |= 0x01;    P2SEL |= 0x01; and after simulation nothing happen :(

Yassine Akrmi said:

and after simulation nothing happen

Yassine Akrmi said:

and after simulation nothing happen

yes i understand what you mean, but simulation in proteus i mean, 

And how well does proteus simulate an MSP? Usually, the simulation ends at the simulation of the CPU core, sometiems some internal things like timers and maybe a clock generator. But usually, the complete hardware isn't simulated.
So it's well possible thatthe crystal input isn't simulated as proteus doesn't know how to handle this. (or rather, nobody implemeted this aspect of the MSP).
You should re-check the documentation about what is included into the simulation and what not. (hint: "CPU core" or such only refers to the simulation of assembly instructions and memory locations)

msp430 has its model in proteus so you can simulate it easly, 

i resolved the problem of the crystal, by adding an input clock, non to connect a crystal, and it works well

the simulation here give a result, and when i would like to connect to microcontroller in a real experience this result will be the same ? how can i start the two mcu at the same time ? should i insert a push button ?

Yassine Akrmi said:

the simulation here give a result, and when i would like to connect to microcontroller in a real experience this result will be the same ?

Yassine Akrmi said:

how can i start the two mcu at the same time ? should i insert a push button ?