LeafLabs Garden » Topic: Can I change the timer overflow inside the interrupt?

LeafLabs Garden » Topic: Can I change the timer overflow inside the interrupt? http://forums.leaflabs.com/topic.php?id=74337 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:14:16 +0000 http://bbpress.org/?v=1.0.2 <![CDATA[Search]]> q http://forums.leaflabs.com/search.php jappiemike on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105419 Sun, 01 Jun 2014 06:13:49 +0000 jappiemike 105419@http://forums.leaflabs.com/ I'm looking for a timer interval of at least 100hz for the longest screen write interval so the matrix looks "solid", so it's not very fast but it needs to be consistent or the illusion falls apart. The reason for the varying intervals is to reduce the load on the MPU from 256 cycles to 8 cycles through the wonders of BAM. I do intend to take out all the serial prints in the final code they are just there for debugging at the moment. Cheers, Mike ala42 on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105418 Sat, 31 May 2014 11:58:34 +0000 ala42 105418@http://forums.leaflabs.com/ As I wrote above, use timer.setOverflow(tar[bam]-1); gbulmer on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105417 Fri, 30 May 2014 11:54:26 +0000 gbulmer 105417@http://forums.leaflabs.com/ jappiemike - "... I did notice that the program runs faster when I open the serial window in the IDE to watch the values which is unexpected" That is very likely because SerialUSB.print() blocks waiting to output. If practical, remove all uses of SerialUSB from your program. "... took out the final timer.refresh(); ... seem to be getting a varying time between pulses now ... " IIRC the libmaple code configures the timers so that the overflow register is buffered (by a 'preload register'), so it isn't updated until there is another 'update event' (see 15.3.1 "Time-base unit" of RM0008 for more details). <code>timer.refresh()</code> should only wiggle some control bits in a timer register to get the timer to update its overflow register. I can't figure out if the update event is supposed to be set by the interrupt routine. If you take all of the code out of the interrupt handler, and put it in loop, and replace the handler with something simple (as I wrote in <a href="http://forums.leaflabs.com/topic.php?id=74337#post-105404)" rel="nofollow">http://forums.leaflabs.com/topic.php?id=74337#post-105404)</a> does it work correctly? What range of frequencies do you need? How accurate does it need to be? If the fastest is about a 100 microseconds or more, with some jitter, you could leave the timer interrupting at a constant rate, and do the rest of the work in loop(). jappiemike on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105416 Tue, 27 May 2014 19:03:13 +0000 jappiemike 105416@http://forums.leaflabs.com/ Hello all, so I took out the final timer.refresh(); in the ISR and it seems to be working now, although the documentation seems to indicate you need to use it. I'll do some more checking with my ChronoVu but I seem to be getting a varying time between pulses now. I also had the tar values backwards and I've reverted to 2800 as the prescaler. I did notice that the program runs faster when I open the serial window in the IDE to watch the values which is unexpected. Thanks for your help. Cheers Mike jappiemike on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105415 Tue, 27 May 2014 18:38:56 +0000 jappiemike 105415@http://forums.leaflabs.com/ Hello ala43, I changed the prescale to 1000 and changed the tar values to volatile int tar[]={32000,16000,8000,4000,2000,1000,500,250}; //{128,64,32,16,8,4,2,2}; and still didn't see a change in the frequency. Cheers Mike jappiemike on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105414 Tue, 27 May 2014 18:17:44 +0000 jappiemike 105414@http://forums.leaflabs.com/ OK Here is the full code listing such as it is. `#include <timer.h> #define LED_RATE 2800 // 2800 is best rate /* a prescale 7200 1=2.5 khz 100=49.5 hz 200=24.9 hz 300=16.6 hz 400=12.5 hz 500=10.0 hz a prescale of 2880 128=99.6 hz 64=197.7 hz 32=389.4 hz 16=756.0 hz 8=1.428 khz 4=2.570 khz 2=4.284 khz 1=6.426 khz a prescale of 720 1=25 khz 100=495 hz 200=249 hz 300=166 hz 400=125 hz 500=100 hz a prescale 288 500=250 hz 1000=125 hz 2000=62 hz 3000=41 hz 4000=31 hz 5000=25 hz */ // We'll use timer 2 HardwareTimer timer(2); volatile byte Lar[10]; volatile int tar[]={128,64,32,16,8,4,2,2}; volatile int bam; void setup() { // Set up the LED to blink pinMode(BOARD_LED_PIN, OUTPUT); pinMode(30, OUTPUT); pinMode(31, OUTPUT); pinMode(32, OUTPUT); pinMode(33, OUTPUT); pinMode(34, OUTPUT); pinMode(35, OUTPUT); pinMode(36, OUTPUT); pinMode(37, OUTPUT); //SerialUSB.begin(9600); // Pause the timer while we're configuring it timer.pause(); // Set up period timer.setPrescaleFactor(LED_RATE); // in microseconds // Set up an interrupt on channel 1 timer.setChannel1Mode(TIMER_OUTPUT_COMPARE); timer.setCompare(TIMER_CH1, 1); // Interrupt 1 count after each update timer.setOverflow(1); timer.attachCompare1Interrupt(handler_led); // Refresh the timer's count, prescale, and overflow timer.refresh(); // Start the timer counting timer.resume(); } void loop() { // delay(1000); SerialUSB.println("START UP"); for(int a=0;a<8;a++) { Lar[a]=255; // sets the LED on delay(50); SerialUSB.println(" **"); Lar[a]=0; // sets the LED off delay(50); SerialUSB.println(" ***"); } SerialUSB.println("Done Testing LEDs"); for(int a=0;a<255;a+=10) { for(int b=0;b<8;b++) { Lar[b]=a; SerialUSB.print(Lar[b]); SerialUSB.print(" "); delay(50); } SerialUSB.println(" *"); } for(int a=0;a<8;a++) { Lar[a]=0; // sets the LED off } delay(100); } void handler_led(void) { digitalWrite(13,LOW); for(int s=0;s<8;s++) // work along LED columns { if(bitRead(Lar[s],bam)==1) { digitalWrite(30+s,LOW); // on //digitalWrite(13,LOW); } else { digitalWrite(30+s,HIGH); // off // digitalWrite(13,HIGH); } } if(bam++==7) { bam=0; } timer.setOverflow(tar[bam]); timer.refresh(); digitalWrite(13,HIGH); //toggleLED(); } I think the frequency is off by a factor of 2 because of toggle() only changes every 2nd pass through the interrupt. I've changed it to digitalWrites to get a more accurate idea. I've hooked my ChronoVu up to pin 13 and the frequency doesn't change from when the ISR exits until it enters it again at about 40us. I did notice that none of the timer keywords are syntax colored in my code, does this indicate a problem? As far as I know bitread is a macro according to the documentation. Cheers Mike ala42 on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105413 Tue, 27 May 2014 18:01:22 +0000 ala42 105413@http://forums.leaflabs.com/ The counter counts one step more than you think is does, as you can easily see from your measurements: 4=2.570 khz 2=4.284 khz (4+1)*2570 / (2+1) = 4283.3 So you should use timer.setOverflow(tar[bam]-1); Besides that, use a smaller prescaler and larger counter values. gbulmer on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105412 Tue, 27 May 2014 14:14:05 +0000 gbulmer 105412@http://forums.leaflabs.com/ jappiemike - In future, would you please post all the (relevant) code? There is no way we could adequately understand the code. '1' at 'a prescale of 2880' does look weird. So does '1' at 'a prescale 7200', shouldn't that be close to 4.9KHz, not 2.5KHz? However, 6.426KHz represents quite a lot of instructions, about 11,000. I'd guess that the handler code should run in 10-30 microseconds unless something is blocking or bitRead is amazingly complex. What frequency do you need for '1'? I'd suggest running a performance test/benchmark, to see how fast the code is actually able to run. Don't use the interrupt handler, instead put all the code in loop(). (You might want to disable SerialUSB, and not use it, to reduce jitter.) jappiemike on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105405 Mon, 26 May 2014 19:31:34 +0000 jappiemike 105405@http://forums.leaflabs.com/ Hello GBulmer, I have used a frequency meter on the #13 pin and got these readings <pre><code>a prescale 7200 1=2.5 khz 100=49.5 hz 200=24.9 hz 300=16.6 hz 400=12.5 hz 500=10.0 hz a prescale of 2880 128=99.6 hz 64=197.7 hz 32=389.4 hz 16=756.0 hz 8=1.428 khz 4=2.570 khz 2=4.284 khz 1=6.426 khz</code></pre> I did note the last reading was off and suspect this is where the interrupt starts to fall down and will use two 2s in the final code. The final code, if I can get it to work, will send the data out via SPI to some shift registers so this is just trying to get the routine to work in the first place. Cheers Mike gbulmer on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105404 Mon, 26 May 2014 11:16:35 +0000 gbulmer 105404@http://forums.leaflabs.com/ jappiemike - what evidence do you have that this code can work anyway? Setting the timer overflow to 1, 2, 4, 8, and 16 (maybe even 32?) seems far too short to get back out of the interrupt handler, especially with the call of <code>toggleLED();</code>. So that looks very suspicious to me. I'd probably take all the working code out of the interrupt handler, and put it into loop, then reduce the interrupt handler to something like: <code>volatile int triggered = 0;</code> <code>void handler_led() { triggered = 1; }</code> and the main loop() would have: <pre><code>void looop() { if (triggered) { triggered = 0; // assumes this code is quick enough // body of interrupt handler } }</code></pre> A couple of small points: 1. In <code>for(volatile int s=0;s<8;s++)</code>, <code>volatile</code> is not necessary. <code>s</code> is on the stack and local to the <code>for</code> loop, so it can't be accessed (validly) by concurrent activity. 2. <code>if(bam++==7) { bam=0; }</code> could be coded as <code>bam = (bam+1) & 0x7</code> jappiemike on "Can I change the timer overflow inside the interrupt?" http://forums.leaflabs.com/topic.php?id=74337#post-105403 Mon, 26 May 2014 06:38:38 +0000 jappiemike 105403@http://forums.leaflabs.com/ Hi, I'm trying to implement a BCM routine for an LED matrix and am having trouble with the changing of the interrupt frequency by changing the overflow number from inside the interrupt routine. Here is my code <pre><code>volatile int tar[]={128,64,32,16,8,4,2,1}; volatile static int bam; void handler_led(void) { for(volatile int s=0;s<8;s++) // work along LED columns { if(bitRead(Lar[s],bam)==1) { digitalWrite(30+s,LOW); // on //digitalWrite(13,LOW); } else { digitalWrite(30+s,HIGH); // off // digitalWrite(13,HIGH); } } if(bam++==7) { bam=0; } timer.setOverflow(tar[bam]); timer.refresh(); toggleLED(); }</code></pre> The tar array hold the overflow numbers and if I change them inside the loop() at compile time the interrupt works fine but I work my way along the array in the interrupt it fails. LAr[] is an array holding the LED values and bam is a variable to step along the array. I haven't put my logic analyser on it yet (it's at work) to see whats happening. I have read the forum questions but none are changing the overflow from inside the ISR but the documentation seems to indicate I can do this. Can I do this or is it a bug in the timer routines? Cheers Mike