Understanding then using PWM Module Signal Generation - Timing Configuration

Hi Amit,

 Yes that is what I thought. This is also what my many years of MSP430 work have instilled in me, so I was confused at my errors here.
 

I used your math (f_clk*period) from the beginning, and re-ran them this AM and recorded the results, see below and attached for result using the code above. N-Attempted listed below is used for 'ui32Period' in PWMGenPeriodSet():

Questions

• What on earth is happening here wrt N->t_meas??
• How to correct this, how to obtain T=20ms?

Measured Results

• test-results.zip
• calcs.xlsx

Justin Reina89 said:

I am having the worst time conceivable understanding and using the PWM module

Yet - you reported success w/in a related, earlier PWM post - did you not?   Might your "worst time conceivable" then be "Self-Caused?"

Changes - even those believed to be small/inconsequential - are to be avoided when entering new territory.   (superior, non MPS430 MCU)

Again - you must learn to "Build effectively" upon past successes.   (thus avoiding "worst times.")

I will break this into two separate posts:

• Post #1: To say thank you, express my gratitude and share my woeful, humorous story here (this post)
• Post #2: The working final solution (code attached)

With the unique, well-balanced and high-impact feedback above from Robert (t&m proc), Amit (fclk) and cb1 (peripheral config), you guys helped me surmount a challenge and a task that was of unquantifiable and had unachievable magnitude and presence for me, and now after today it is no big deal. THANK YOU!!! :)!!

Here are my key learning's which resulted and allowed me to come to resolution. In retrospect they are quite embarrassing in their naivety, but now that they are closed, I'll get over it :)

1. YOU ACTUALLY CAN ACCESS THE DANG Tiva registers directly!!! O... M... Sheesh!

When I first ramped into Tiva, ~2mo ago I did so with TivaWare. All of the code, documentation and examples never ever use the dang registers, they always use the cute and fancy SysCtl and TivaWare calls.

And as there was no #include <tiva.h>, like I was used to in my MSP430 work I just assumed "this was the best interface you could get. It was ARM Cortex, this is just the best that life gets. And so when I found the contents of TivaWare/inc today, for the first time ever, my world changed. Thank you so much cb1 for that final nudge to this realization!!!!

2. The PWM0 Load & Compare Registers are actually only 16-bits!!!

So previously when I tried to use enormous vals to them, thinking they were 32-bits I was completely wrong! :(

3. There is a clock division feature on the PWM Module, so I can make it work and still leave F_CLK_CPU unmodified!! :)

*But the SysCtl & Pwm library routines don't work?

4. I am a bonehead and had my T&M tool timescales dramatically wrong. My OScope is down for the count and I am using an NI DAQ right now... :(!
    

Once those four were addressed, the answer kind of just came naturally. This is thanks to you three, thank you!!! :)!

Key Solution Points:

• Use valid F_Clk and make use of PWM input clock division
• Use valid PWM0 Load Values (2^16 max)

Remaining Open Question:

The timing value for period below was empirically established. Every attempt I have taken to correctly calculate has failed, I ultimately ended up simply empirically finding them. Help please! :(...

Code (also attached in Zip):

#define XTAL_HZ		(25000000)
#define F_CPU_HZ	(XTAL_HZ / 1)											/* @hazard	must be N division of XTAL					*/

#define USEPWM_BIT		 	(1<<8)										  /* PWMCC.USEPWM. enabling adds %2, in addtn to 'N'		*/

#define PWM_CLOCK_DIV_N		(2)
#define PWM_CLOCK_DIV_BITS	(PWM_CLOCK_DIV_N<<0)
#define PWM_CLOCK_HZ		(XTAL_HZ / PWM_CLOCK_DIV_N)						//?

/************************************************************************************************************************************/
/**	@fcn		int main(void)
 *  @brief		generate PWM signals for GPIO pins PF2 and PF3
 *  @details	uses PWM0 Generator Module 1 (PWM0_1)
 *
 *  @section	Purpose
 *  			Help Justin figure out how to use the dang Tiva PWM module
 *
 *  @pre		any
 *
 *  @post		PWM0.1 is configured and running:
 *  			- PWM2 (PF2) running with t_period=20.3ms, t_on=1ms
 *  			- PWM3 (PF3) running with t_period=20.3ms, t_on=19s
 *
 *  @section	Operation
 *				initialize the GPIO & PWM modules, then configure PWM Generator 1 for continuous output on both output signals
 *
 *  @section	Opens
 *  			Be able to set the PWM period and duty cycles mathematically, current values are empirically established
 *
 *  @note		Periods and Pulse Durations were empirically established, there derivations are presently unknown
 */
/************************************************************************************************************************************/
int main(void) {

	//Set Clock
	SysCtlClockFreqSet((SYSCTL_XTAL_25MHZ | SYSCTL_OSC_MAIN | SYSCTL_USE_OSC), F_CPU_HZ);

	//******************************************************************************************************************************//
	//	 											    GPIO MODULE INIT															//
	//******************************************************************************************************************************//
	SysCtlPeripheralEnable(SYSCTL_PERIPH_GPIOF);

	GPIOPinConfigure(GPIO_PF2_M0PWM2);
    GPIOPinConfigure(GPIO_PF3_M0PWM3);

    GPIOPinTypePWM(GPIO_PORTF_BASE, GPIO_PIN_2);
    GPIOPinTypePWM(GPIO_PORTF_BASE, GPIO_PIN_3);


	//******************************************************************************************************************************//
	//	 											   PWM MODULE #0 INIT															//
	//******************************************************************************************************************************//
    SysCtlPeripheralEnable(SYSCTL_PERIPH_PWM0);								    /* Enable the PWM0 peripheral						*/

    while(!SysCtlPeripheralReady(SYSCTL_PERIPH_PWM0));							/* Wait for the PWM0 module to be ready				*/

    PWMGenConfigure(PWM0_BASE, PWM_GEN_1, 										/* Configure the PWM generator for count down with 	*/
    				PWM_GEN_MODE_DOWN | PWM_GEN_MODE_NO_SYNC);					/* with immediate updates to the parameters 		*/

    PWM0_CC_R = (USEPWM_BIT | PWM_CLOCK_DIV_BITS);							    /* Set the PWM clock to the system clock			*/


	//******************************************************************************************************************************//
	//	 											 PWM GENERATOR #1 INIT															//
	//******************************************************************************************************************************//
    PWMGenPeriodSet(PWM0_BASE, PWM_GEN_1, 500000);								/* Set the period of Generator Module 1 to 20ms		*/

    PWMPulseWidthSet(PWM0_BASE, PWM_OUT_2, 2000);							    /* Set the pulse width of PWM2 for a 1ms pulse		*/

    PWMPulseWidthSet(PWM0_BASE, PWM_OUT_3, 497500);								/* Set the pulse width of PWM3 for a 19ms pulse		*/


	//******************************************************************************************************************************//
	//	 										      PWM GENERATOR #1 START												        //
	//******************************************************************************************************************************//
    PWMGenEnable(PWM0_BASE, PWM_GEN_1);											/* Start the timers in Generator Module 1			*/

    PWMOutputState(PWM0_BASE, (PWM_OUT_2_BIT | PWM_OUT_3_BIT), true);			/* Enable the outputs								*/

    SysCtlSleep();																/* End in sleep, pwm0.1 on							*/

    for(;;);																	/* safety											*/
}

ref_pwm_servos-e2e-2.zip    measurement.zip

I measured the following values, using the code attached above (ref_pwm_servos-e2e-2.zip):

This table is also attached in the XLSX below. Thoughts? These values seem offensive they are so strange!

:(

measurement values.xlsx    msrPic1-4.zip   

My strong apologies. I am starting a new venture and working completely solo right now, wearing many dang  hats (mechE, compSci, biz, marketing, etc...). And when I get back into my primary and best seat, that of firmware development it is quite bumpy! ;)...

Let's examine only the following test vector, for simplification and for elimination of prior errors... :(

Test Vector: Last posted code with the following changes:

	//******************************************************************************************************************************//
	//	 											 PWM GENERATOR #1 INIT															//
    //
	//******************************************************************************************************************************//
    PWMGenPeriodSet(PWM0_BASE, PWM_GEN_1, 40500);								/* Set the period of Generator Module 1 to 20ms		*/

    PWMPulseWidthSet(PWM0_BASE, PWM_OUT_2, 40500/2);					   		/* Set the pulse width of PWM2 for a 10ms pulse		*/

    PWMPulseWidthSet(PWM0_BASE, PWM_OUT_3, 40500/4);							/* Set the pulse width of PWM3 for a 5ms pulse		*/

My apologies again for any and for all errors that I have injected in this thread. Work is monstrous right now! :(

Assumptions:

• These are valid values

Questions:

• How does the math work for the calculation of period?
• My math shows this period should be 6.48ms, see attached XLSX

final math.xlsx

My friend - you ask that we examine (only) the following test vector - which is wildly incomplete. Yet - few more lines down - we're presented w/"finalmath.xlsx" which conflicts w/your (earlier) [test vector only] direction. That's krazy making - is it not?

A 25MHz PWM clock yields a period of 40nS - that's easily handled by a "focused" Justin - is it not? You must "zero in" on this PWM clock period - to have any chance of PWM success.

We note that your comments (w/in the "test vector") are displaced far to the right - forcing more (unwanted) effort upon your "for free" help crue! (that's unwanted & improper - surely you know that.)

It was your choice to "attempt to wear many hats" - might that have been a key source of your (now) obvious overload?   You write (uniquely), "worst time conceivable" - that's packed w/emotion and reflects far more upon your state of mind - little upon this MCU.

The decision to, "Do everything by yourself" (while) learning a powerful, new MCU - may not set records for excellence. Relaxed learning always trumps "rushed/pressure-cooker" learning - it would appear that "IF" your past MCU is "up to task" (its usage) would have proved a better decision.

You're smart enough to know all of this - and the PWM clocking & duty cycle setting is NOT that challenging.   All this suggests that your thinking has become muddied - and that cannot bode well for your project's success.

With very humble, and lightly embarrassing words I agree, you are quite correct.

I took a breather just now, stepped out and stretched my legs. And realized in reflection how asinine it is for me to debug MCU timing with no Oscilloscope and only a 48 KS/s USB DAQ. My oh my, the crazy train needs to take a quick pause here, who does that!!

I will return and post up working code once my new Oscilloscope comes in, but that is a few months out.

Thank you guys for your help, and your patience!! :)

Perhaps another option is a local school - or Tech Club (ham radio even) - or the "rental" of a scope for a 30-60 day period.   (often the rental firm will apply the bulk of that rental fee to the instrument's purchase price.)

We note that you employed at least one line of (nicely banned, at least discouraged) "DRM" code for a key/critical PWM Frequency Setting.   That's unwise - don't add complications when you've previously experienced trouble in that (or similar) area.

While you await instrumentation - nothing prevents you from the study & engaging of one of the MCU's Timers.   Then - is it not possible to route your (suspect) PWM output into that (very) Timer input pin - with the Timer "enabled" for a brief - but precise period.   You may wish to dial down the PWM Frequency during these checkouts - yet many, many here employ the MCU Timers to perform key/critical: pulse width, pulse count, and pulse frequency measurements.

We remain confused as you've reported (both) mistaken (outside of expectation) PWM frequency measurements AND lack of instrumentation.   Some way/how - those two reports fail to completely mesh...

Hi cb1,

With respect to the measurements in error I will be in response here soon. My target is two weeks upon purchase of our own oscilloscope. Thank you for the advice on a school or tech club option, we have local amateur houses which I can visit if needed (e.g. Hedron PDX)! I am quite embarrassed on my failures here, and my weaknesses this illustrates.

The timer feedback and suggestion is quite interesting, I will be considering this during upcoming work thank you.

Question:

Which line of PWM Frequency Configuration are you referencing as 'bad mojo', as DRM? DRM is 'Direct Register Modification' correct?

Here is my guess:

• PWM0_CC_R = (USEPWM_BIT | PWM_CLOCK_DIV_BITS);