Hi David, The reason for choosing for a step increment of 2 is to limit the number of steps. If I had chosen to use a step increment of 1 then the timer has to run 256 times faster than the chosen period time in order to be able to have 256 steps in one period.. So in case of 80 Hz frequency timer has to run at around 20 kHz or an interrupt time of 50 us. That would not leave much time left for the main program so that's why I used a step increment of 2 and to reducing the timer frequency around 10 kHz. Of course you can decide to go for 20 kHz so that you can use the 8 bit value of the ADC. I assume your PIC runs at a higher frequency than 4 MHz.
Below another example using four LEDs, using timer 2. STEP_BRIGHT is called UP_BRIGHT here but you get the idea. procedure pwm_cycle is pragma interrupt if PIR1_TMR2IF & PIE1_TMR2IE then PIR1_TMR2IF = FALSE ; First check the if we need to start a new period. if (pwm_period == 0) then ; New period. Switch leds on and start a new duty cycle. pwm_period = PWM_PERIOD_TIME led_1_timer = duty_cycle_1 led_2_timer = duty_cycle_2 led_3_timer = duty_cycle_3 led_4_timer = duty_cycle_4 led_1 = LED_ON led_2 = LED_ON led_3 = LED_ON led_4 = LED_ON else pwm_period = pwm_period - 1 end if ; If the timer of the LED is 0 then the duty cycle time has passed. if led_1_timer >= UP_BRIGHT then led_1_timer = led_1_timer - UP_BRIGHT else led_1 = LED_OFF end if if led_2_timer >= UP_BRIGHT then led_2_timer = led_2_timer - UP_BRIGHT else led_2 = LED_OFF end if if led_3_timer >= UP_BRIGHT then led_3_timer = led_3_timer - UP_BRIGHT else led_3 = LED_OFF end if if led_4_timer >= UP_BRIGHT then led_4_timer = led_4_timer - UP_BRIGHT else led_4 = LED_OFF end if end if end procedure Kind regards, Rob ________________________________ Van: jallib@googlegroups.com <jallib@googlegroups.com> namens pinhe...@gmail.com <pinhead...@gmail.com> Verzonden: woensdag 21 juni 2023 11:33 Aan: jallib <jallib@googlegroups.com> Onderwerp: Re: [jallib] Software PWM @Kiste (Got it ;-) ) Clever way to do it ! I guess it took some time to figure that out, and get it running properly, specially if you further had to manage the LED multiplexing. Hats off ! I guess the PIC is running at a fairly high speed to be able to achieve all that process ? BTW: My nickname "Pinhead" is usually a term referring to a pinball freak (which I am). Furthermore, I like horror movies, and Pinhead is the name of one of those bad guys we love to hate (in the Hellraiser franchise). That's why ;-) @Rob The resolution is not critical for dimming LED's in the way I intend: 127 steps, as in your solution, are enough. But why limit it with STEP_BRIGHT = 2 ? If, for example, I would like to use the ADC value of a LDR to set the duty_cycle, I'd simply use the 8 MSB's of the ADC value, and that would give me 8 bits resolution... What was the requirement in your application to limit the resolution to 128 steps ? Also, is the PWM frequency critical ? Otherwise, maybe you could get rid of the "pwm_period = PWM_PERIOD_TIME" line, isn't it ? All in all, those two answers are food for thoughts, and I will certainly take your good advices. Thanks for that ! Le mercredi 21 juin 2023 à 07:59:57 UTC+2, Rob CJ a écrit : Hi David, Oliver, What should be the resolution of the PWM? I sometimes made it in software using one timer and controlling several LEDS. In the example below I control one LED using 128 steps (step increment of 2 (STEP_BRIGHT)) but can be extended with more LEDs. Would that be sufficient for your application? This was running on a 12F675 running at 4 MHz. The refresh rate was about 80 Hz to prevent flicker. Below this program you find some calculations. Normally I used Timer 2 because of the automatic reload but this PIC did not have that. ; Interrupt routine for timer 0, handling the PWM signal in software. ; This routine is called every 107 us (measured) and takes between 18-28 us. procedure timer_0_interrupt() is pragma interrupt if INTCON_T0IF then ; First preset the time for the next interrupt. TMR0 = TIMER_0_RELOAD INTCON_T0IF = FALSE ; Check the if we need to start a new period. if (pwm_period == 0) then ; New period. Switch LEDs on and start a new duty cycle. pwm_period = PWM_PERIOD_TIME pwm_timer = duty_cycle ; Only turn LED on if it has a duty cycle to prevent flashing. if (pwm_timer > 0) then LED = LED_ON end if else pwm_period = pwm_period - 1 end if ; If the timer of the LED is at the mininum brightness then time has passed. if (pwm_timer >= BRIGHT_MIN) then pwm_timer = pwm_timer - STEP_BRIGHT else LED = LED_OFF end if end if end procedure Some constants I used. ; Specify the values for the minimum and maximum brightness of the LEDs and ; the value to increase or decrease it. The minimum value is not zero since ; we want to know if the device is on or off so there always has to be a ; minium brightness when the device is switched on. const byte BRIGHT_MIN = 10 ; Must be >= STEP_BRIGHT. const byte BRIGHT_MAX = 254 ; Must be even and <= 254 const byte STEP_BRIGHT = 2 ; This gives 254 / 2 = 127 steps. ; Timer reload value. The clock is 4 MHz so the timer clock is 1 MHz. Let's ; use 80 Hz as PWM frequency for the LEDs. This means that the timer has to ; run at 80 * 125 steps = 10.000 Hz or a period time of 100 us. With a prescale ; of 2 (minimum) the we need a timer value of 1.000.000 / 2 / 10.000 = 50. ; Since we use the timer overflow the reload value becomes 256 - 50 = 206. ; The measured value is 9.333 Hz, most like due to processing time. const byte TIMER_0_RELOAD = 206. ; See timer init, 10.000 Hz / 125 gives 80 Hz. Measured is 74 Hz. const byte PWM_PERIOD_TIME = 125 Kind regards, Rob ________________________________ Van: 'Oliver Seitz' via jallib <jal...@googlegroups.com> Verzonden: woensdag 21 juni 2023 07:38 Aan: jal...@googlegroups.com <jal...@googlegroups.com> Onderwerp: Re: [jallib] Software PWM Salut pinhead, If I had to extend an existing controller with PWMs, I'd use e.g. PIC16F18313. It has 8 pins, I2C and 4 independent PWMs at a cost of less than 1,50€. I cannot tell however if the MCLR pin can be used for I2C. A safe choice would be 12F1572 which is even cheaper and has 4x 16bit PWM. You can use serial_software for communication, as the PWMs run happily on their own. Yes, the way you're describing you could dim LEDs, but each time something else is happening, like new I2C-input is processed, all your LEDs would slightly flicker. When I did bitbang-dim 120 multiplexed LEDs, i did it binarily: If bit 7 is set, light up the LED for 128µs. Then, if bit 6 is set, light up the LED for another 64µs and so on. The brightness sums up in the eye of the beholder. I've programmed a timer to generate an interrupt in succession after 128, then 64, then 32... µs. However, 4,2 and 1 µs I did without leaving the ISR. I did not use any "if" for that, I had an 8x8bit matrix with the dimming values, which I rotated by 90°, so I had a byte full of "bit7" of each channel, a "bit6" of each channel and so on. So, the ISR had nothing to do but move the "bit7"-byte to the output port right before the 128µs interrupt delay, the "bit6"-byte before the 64µs delay... That way I could dim 8 LEDs (all on e.g. PORTA) for the price of one. Oliver is my real name, but Kiste (=Carton(fr) =Doos(nl) =Box(en)) is my very common nickname, even in real life. Greets, Kiste Am Dienstag, 20. Juni 2023 um 22:10:07 MESZ hat pinhe...@gmail.com <pinhe...@gmail.com> Folgendes geschrieben: Hello Oliver, Indeed, PIC18F27J13 could be a candidate. I overlooked this one too quickly (I already ranted about the parametric search tool on the µchip site :-{ ) I'll check for the other characteristics. No internal OPAMP and internal VREF are a bit annoying, but if it's the only missing parts, it's ok. Anyway, what about a PIC that would be programmed as a I²C generator with several independant channels, and I²C slave. He would have nothing else to do than generating the signals. Consider one channel: - A register contains the value of its duty cycle (from 0 to 255) - In the main loop, a 8-bit counter is incremented. - At each increment, its value is compared to the duty cycle register, and if needed, the corresponding output is toggled. You could generate as many PWM signals as there are IOs available. The I²C-slave interface would be handled by the MSSP module with interrupts (yes, this could cause glitches in the PWM signals if the time needed to process the I²C processes is greater than the time it takes to increment the PWM counter by one, but if it is only to dim some LED's, I think I could live with it) The frequency could be adjusted by changing the oscillator frequency and/or adding delays in the main loop when the main counter is incremented Just for the sake of it, I think I will give it a try..... Hoping I am not overlooking a major impediment Thanks for your suggestion, Oliver (or Kiste ?) Le mardi 20 juin 2023 à 20:43:01 UTC+2, Oliver Seitz a écrit : Hi! A software PWM library isn't easy. Without a timer, it blocks everything. With a timer, it's complicated and would only add a few channels. I've done software PWM for 120 LEDs independently with a 28 pin controller, but it couldn't do much more than that at 64MHz. Have you had a look at, say, PIC18F27J13? It has 3+7 independent ccp modules (with 3 TMR2-type timers). Greets, Kiste Am Dienstag, 20. Juni 2023 um 19:49:33 MESZ hat pinhe...@gmail.com <pinhe...@gmail.com> Folgendes geschrieben: Hi, I'm falling short of PWM outputs for a project. I need 6 independant outputs, and the only way to get them is to use a 64-pin 18fxxxx, which is a bit overkill for me, or use a second PIC, bond to the first (I²C or SPI) which is also overkill but could be acceptable in the absence of another "clean" solution. It will only be used for dimming LED's, and I don't need particular features like dead-band, half or full bridges, precise or stable frequency..... To my surprise, I didn't find a software PWM library in Jallib 1.8..... Or do I miss something ? Have a nice day -- You received this message because you are subscribed to the Google Groups "jallib" group. To unsubscribe from this group and stop receiving emails from it, send an email to jallib+un...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/jallib/923bd215-4855-49d9-b7e1-17e02482c219n%40googlegroups.com<https://groups.google.com/d/msgid/jallib/923bd215-4855-49d9-b7e1-17e02482c219n%40googlegroups.com?utm_medium=email&utm_source=footer>. -- You received this message because you are subscribed to the Google Groups "jallib" group. To unsubscribe from this group and stop receiving emails from it, send an email to jallib+un...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/jallib/9aad71fb-a1f2-47f3-8cc4-82a1110327c2n%40googlegroups.com<https://groups.google.com/d/msgid/jallib/9aad71fb-a1f2-47f3-8cc4-82a1110327c2n%40googlegroups.com?utm_medium=email&utm_source=footer> . -- You received this message because you are subscribed to the Google Groups "jallib" group. To unsubscribe from this group and stop receiving emails from it, send an email to jallib+un...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/jallib/1639372286.1548147.1687325901648%40mail.yahoo.com<https://groups.google.com/d/msgid/jallib/1639372286.1548147.1687325901648%40mail.yahoo.com?utm_medium=email&utm_source=footer>. -- You received this message because you are subscribed to the Google Groups "jallib" group. To unsubscribe from this group and stop receiving emails from it, send an email to jallib+unsubscr...@googlegroups.com<mailto:jallib+unsubscr...@googlegroups.com>. To view this discussion on the web visit https://groups.google.com/d/msgid/jallib/87f34955-bd10-4bca-980b-8cce7b5dac0an%40googlegroups.com<https://groups.google.com/d/msgid/jallib/87f34955-bd10-4bca-980b-8cce7b5dac0an%40googlegroups.com?utm_medium=email&utm_source=footer>. -- You received this message because you are subscribed to the Google Groups "jallib" group. To unsubscribe from this group and stop receiving emails from it, send an email to jallib+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/jallib/GVXP195MB1637B7D84672E7A24C12D980E65DA%40GVXP195MB1637.EURP195.PROD.OUTLOOK.COM.