如何用PIC24EP系列单片机在触发模式下使用OCX实现一次触发功能？

PIC MCU的增长领域之一是遥控器（RC）无人机（如四头飞行器）的控制模块。在这些应用中，MCU通过产生用于电子速度控制器（ESC）的控制信号来控制模型的电机。ESC由脉冲控制，脉冲的持续时间决定了ESC提供给电机的信号（节流）水平。例如，在传统的RC应用中，1毫秒脉冲对应于0%节气门和2毫秒脉冲至100%节气门。在1至2毫秒范围内改变脉冲的持续时间允许在0和100%之间的任何级别设置电机控制。由于控制速度对于RC应用是至关重要的，所以较新的ESC接受较短的脉冲——对于0%节流阀125usec，对于100%节流阀250usec，这减少了传输控制信号的延迟。ESC需要控制脉冲之间的一些延迟，所以对于较老的ESC（由1-2毫秒脉冲控制）来说，最大刷新频率大约为400-450Hz，对于较新的ESC（1-2KHz），控制RC应用中ESC的典型方法是在PWM模式下使用输出比较模块将刷新速率设置为m。当来自控制模块的电机控制的新更新时，ESC接受的轴和改变PWM的占空比以匹配所需的脉冲持续时间。由于OCxR寄存器在PWM模式下是双缓冲的，因此我们可以随时更新占空比，在下一个同步事件中占空比会发生变化。被ESC接受的级别，迫使ESC处理新的脉冲，即使没有电机控制级别的更新；由于占空比仅在同步事件上更新，随机时间生成的OCxR的更新必须平均等待PWM周期的一半，直到它可用于ESC通过改变持续时间的新脉冲，幸运的是OC模块为PWM模式提供了一个功能！根据FRM:13.3.3.14.2“单击功能”的第13.3.3.14.2节……通过设置TRIGMODE位(OCxCON1)启用“单击模式”。此事件设置TrutStAT位，计时器开始计数。当计时器滚动到0x000时，如果TrimeMod＝1，则TrutStButt将被硬件清除。这将定时器保持在Reset，直到下一个触发事件，创建一个单次定时器。他持续时间的OCXR，然后把它下来产生单脉冲所需的持续时间。当OCx定时器在OCxRS寄存器（PWM周期）中达到值时，OCx模块将清除TRIGSTAT位，并将定时器保持在RESET中，直到我们重复这个循环。不幸的是，它不是这样发生的！在将定时器放在复位之前，OCX模块检查OCXR寄存器的值，如果不是0，则会增加与模块相关联的引脚上的信号。因此，下次我们通过更新OCXR和设置TrimStAT＝1来重复周期，我们将得到脉冲持续时间，它将是OCXR的新值加上从先前触发的PWM脉冲结束的间隔的持续时间，直到触发下一个脉冲的时刻为止！这是根据FRM的部分：133.3.1边缘对齐PWM模式…边缘校准的PWM模式操作：-当同步发生时，在下个增量周期发生以下四个事件：-定时器复位为零，并恢复计数-OCx管脚设置为高（例外：如果OCxRS=0000，OCx管脚将不设置）-OCxR和OCxRS缓冲寄存器是更新从OCXR和OCXRS -中断标志，OCxIF，设置-当定时器和OCXR匹配，引脚将设置为低。显然，在模块分析TRIGMODE位并且意识到它处于触发模式并且需要在RESET中保持定时器之前，上面的顺序，包括将OCx引脚设置为高。我认为这种行为在硬件上是BUG，应该反映在Errata文档中。保持OCX引脚低，直到我们重复循环。另外，在设置TRIGSTAT=1之前更新OCxR之后，请注意6条Nop()语句——由于PWM模式下OCxR的双缓冲特性，模块需要几个周期来传播更改。小于6个（），我们可能或不可能得到一个脉冲，小于3个（），我们将永远不会得到脉冲！附上完整的MPlab-X项目，说明产生单次脉冲的完整工作程序；它已经在PIC24EP512GU810和PIC24EP512GP806上进行了测试。但是，由于PIC24EP和dsPIC33EP的整个版本都由同一个FRM所覆盖，我猜这个代码可以适用于所有这些。如果您在我的逻辑或代码中发现错误，或者有更简单的实现，我将非常感谢您的评论！谢谢您。



      以下为原文

    One of the growing area of PIC MCUs is the control modules for remote control (RC) drones like quadrocopters. In these applications MCU is being used to control model’s motors by generating control signals for Electronic Speed Controllers (ESCs). ESCs are being controlled by pulses were duration of the pulse defines the level of signal (throttle) provided by ESC to the motor. For example, in traditional RC applications 1 msec pulse corresponds to 0% throttle and 2 msec pulse – to 100% throttle. Changing duration of the pulses in the range between 1 and 2 msec allows to set motor control at any level between 0 and 100%. As the speed of control is critical for RC applications, newer ESCs accept shorter pulses – 125 usec for 0% throttle and 250 usec for 100% throttle, which reduces delay in transmitting control signal. ESCs require some delay between control pulses, so for older ESCs (controlled by 1-2 msec pulses) the maximum refresh frequency is around 400-450 Hz, and for newer – 1 to 2 KHz.
Typical way to control ESCs in RC applications is to use Output Compare module in PWM mode setting refresh rate to the maximum accepted by ESC and changing the duty cycle of PWM to match required duration of the pulse when a new update for motor control comes from the control module. As the OCxR register is double-buffered in PWM mode, we could update duty cycle at any moment of time and change in the duty cycle will happen on the next Sync event.
There are two disadvantages in using straight PWM for controlling ESCs:

• To reduce delay, we need to set PWM frequency at the highest level accepted by the ESC, forcing ESC to process new pulses even when there is no update to the level of motor control;
• As duty cycle is being updated only on Sync event, an update to OCxR generated at random time would have to wait on average half of the PWM period until it is made available to the ESC through a new pulse of changed duration.
  

Luckily OC module offers One-Shot functionality for PWM mode! According to section 13.3.3.14.2 of the FRM:

13.3.3.14.2 One-Shot Functionality
… The One-Shot mode is enabled by setting the TRIGMODE bit (OCxCON1).
In One-Shot mode, the timer remains in Reset until a trigger event occurs. This event sets the
TRIGSTAT bit and the timer begins to count. When the timer rolls over to 0x0000, the TRIGSTAT
bit would be cleared by the hardware if TRIGMODE = 1. This holds the timer in Reset until the
next trigger event, creating a one-shot timer.

Looks simple – while TRIGSTAT is 0, load OCxR with the value corresponding to the new pulse duration and set TRIGSTAT=1; OCx module will wake up raise the value on the pin associated with OCx and keep it up for the duration of OCxR, and then bring it down generating single pulse of required duration. When OCx timer reaches value in OCxRS register (PWM period), OCx module will clear the TRIGSTAT bit and hold the timer in RESET until we repeat the cycle.
Unfortunately, it does not happen this way! Prior to putting timer in RESET, OCx module checks the value of OCxR register and if it is not 0, will raise the signal on the pin associated with the module. So the next time we repeat the cycle by updating OCxR and setting TRIGSTAT=1, we will get pulse duration of which will be the new value of OCxR PLUS duration of the interval from the end of the previous triggered PWM pulse until the moment we trigger next pulse! This is in accordance with section of the FRM:


13.3.3.1 EDGE-ALIGNED PWM MODE
... Edge-Aligned PWM Mode Operation:
- When synchronization occurs, the following four events occur on the next increment cycle:
- The timer is reset to zero and resumes counting
- The OCx pin is set high (exception: if OCxRS = 0000, the OCx pin would not be set)
- The OCxR and OCxRS Buffered registers are updated from OCxR and OCxRS
- Interrupt flag, OCxIF, is set
- When the timer and OCxR match, the pin would be set low. This match does not generate
interrupts.



Apparently the sequence above, including setting OCx pin high, happens BEFORE the module analyzes TRIGMODE bit and realizes that it is in a triggered mode and need to hold timer in RESET! I think this behavior is a BUG in the hardware and should have reflected in Errata document.
Anyway, the work-around and the way to implement proper one-shot pulse is the following:
           if (0 == OC1CON2bits.TRIGSTAT)
                {
                OC1R = OC_Width;
                Nop();
                Nop();
                Nop();
                Nop();
                Nop();
                Nop();
                //--------------------------
                OC1CON2bits.TRIGSTAT = 1;
                //--------------------------
                Nop();
                OC1R = 0;
                }

The important piece here that one Nop() after setting TRIGSTAT=1, the OCxR is set to 0 so at next Sync event module will keep OCx pin low until we repeat the cycle. Also, please note 6 Nop() statements after updating OCxR before setting TRIGSTAT=1 – because of the double-buffered nature of OCxR in PWM mode, module requires several cycles to propagate the change. With less than 6 Nop() we may or may not get a pulse; with less than 3 Nop() we will never get pulse!
Attached is complete MPlab-X project illustrating fully working program generating One-shot pulses; it has been tested on PIC24EP512GU810 and PIC24EP512GP806. However, as the whole slew of PIC24EP and dsPIC33EP are covered by the same FRM, I guess this code will be applicable to all of them.
If you find a fault in my logic or in the code, or have a simpler implementation, I would appreciate your comments!
Thank you.
   Attachment(s)

GU810-20-OneShotSample.zip (15.38 KB) - downloaded 132 times