Post by thread:[PIC] Servos and PWM

Konstantin K. wrote : > Hi, > > I have a tank-like bot where each thread is controlled by a servor (2 > servos in total). This makes a single pic not suitable for driving > both at the same time without extensive multitasking. If you look at the problem as a *single* task controling *two* (or more) servos, you'll see that you don't need any "multitasking" at all. > The servos are modified for continuous motion. What is "continuous" here ? > I only need to be able > to go forward and backward. Speed control is not needed. As far as I know, you never directly control the "speed" of a RC servo, only the position (using the pulse width). > So to drive > the servo forward and backward I only need two different pulse widths. > > One of my options is to get a 12F685 for each servo... > My other idea was to have a 555 timer for each servo... Both those solutions seems as a major overkill and far to complicated. > Does anyone have any suggestions on what would be the best way to go > about it? Or any other ideas on how to drive 2 or 3 servos from the > same pic. Set up a timer as a basic "time base" (interrupt driven). Create a number of counters and flags to hold the "status" of each servo output. At each timer interrupt, check each servo counter and, if needed, change the output to the servos. This is normal RC servos (20 ms period 1.5 +- 0.5 ms pulse), right ? That is a very long time for a PIC... :-) Jan-Erik. ____________________________________________

Thanks a lot for your answers. I will probably take your advice and stick with the microcontroller. However, it seems as if many people are abandoning "designing circuits" in favor of microcontrollers. It seems reasonable because they are so cheap and easy to use but somehow it makes me uneasy. If you think in commercial terms (in my case), a micrcontroller costs a little over $1 where a 555 timer and a few resistors, capacitors, and transistors might cost 50c, which is about 3 times as cheap. Let me know what your thoughts are about this. On Fri, 29 Oct 2004 11:51:40 +0200 (CEST), Jan-Erik Soderholm <spam_OUTjan-erik.soderholmTakeThisOuTtelia.com> wrote: {Quote hidden}> Konstantin K. wrote : > > > Hi, > > > > I have a tank-like bot where each thread is controlled by a servor (2 > > servos in total). This makes a single pic not suitable for driving > > both at the same time without extensive multitasking. > > If you look at the problem as a *single* task controling *two* (or more) > servos, you'll see that you don't need any "multitasking" at all. > > > The servos are modified for continuous motion. > > What is "continuous" here ? > > > I only need to be able > > to go forward and backward. Speed control is not needed. > > As far as I know, you never directly control the "speed" of a RC servo, > only the position (using the pulse width). > > > So to drive > > the servo forward and backward I only need two different pulse widths. > > > > One of my options is to get a 12F685 for each servo... > > My other idea was to have a 555 timer for each servo... > > Both those solutions seems as a major overkill and far to complicated. > > > Does anyone have any suggestions on what would be the best way to go > > about it? Or any other ideas on how to drive 2 or 3 servos from the > > same pic. > > Set up a timer as a basic "time base" (interrupt driven). > Create a number of counters and flags to hold the "status" > of each servo output. > At each timer interrupt, check each servo counter and, if needed, > change the output to the servos. > > This is normal RC servos (20 ms period 1.5 +- 0.5 ms pulse), right ? > That is a very long time for a PIC... :-) > > Jan-Erik. > > > > ______________________________________________

> >If you think in commercial terms (in my case), a micrcontroller costs >a little over $1 where a 555 timer and a few resistors, capacitors, >and transistors might cost 50c, which is about 3 times as cheap. In my AVR application, I controlled eight servos (could have done 16 easily) plus two displays, switches, and a stepper motor. Each servo needs only one pin on the micro. You could control two servos easily with the Tiny-11, which costs $0.25 The processor allows you to do more than analog would, with way less effort, and to easily reconfigure it on the fly. ____________________________________________

On Fri, 29 Oct 2004, Sascha wrote: > Hi, > > If you want just full left and full right on both servos it's more than > simple to do it with the pic since you have to output a 20ms period signal > with a 1ms pulse width for full left or 2ms width for full right. You just > have to set a timer to generate a 1ms interrupt and drive the outputs from > there. Or if you end up deciding that you do need more resolution, you could implement parallel PWM outputs: http://www.dattalo.com/technical/software/pic/pwm8.asm For 8 outputs, this pwm routine takes 23 cycles per update. To generate a servo PWM, you'd probably want to implement the 20mS period interrupt in which the pwm8.asm (or a simplified version) would run as an isochronous loop. E.g.: Every 20 milliseconds: interrupt_routine: ; entry save stuff... ; start the pwms: movf pwm_dc0,w movwf pwm_cntr0 movf pwm_dc1,w movwf pwm_cntr1 clrf loop_counter ; loop_counter = 256 ; Drive the pwm outputs high: movlw (1<<bPWM0) | (1<<bPWM1) iorwf PWM_PORT,F DELAY_US(1000) ; 1 millisecond for the guard pulse pwm_loop: movlw b'11111111' ; Assume outputs stay high decfsz pwm_cntr0,f andlw ~(1<<bPWM0) ; PWM 0 is over decfsz pwm_cntr1,f andlw ~(1<<bPWM1) ; PWM 1 is over andwf PWM_PORT,F ; Update the outputs decfsz loop_counter,F goto pwm_loop ; exit interrupt routine The pwm loop takes 256 iterations * 9cycles/iteration = 2304 cycles. A midrange pic running from a 20MHz oscillator would take 460.8 uS. Giving you the shortest pulse of about 1.5 uS and longest pulse of about 460 uS. There are several ways to increase the pulse width. You can change the oscillator, you could add some delay in the loop, you can add more PWM's, you can poll other time critical stuff while in the loop, etc. Scott ____________________________________________

> If you think in commercial terms (in my case), a micrcontroller costs > a little over $1 where a 555 timer and a few resistors, capacitors, > and transistors might cost 50c, which is about 3 times as cheap. > > Let me know what your thoughts are about this. Component (buy) cost is not the only production cost. Placement (PP-machine-time, and labour time for placement of the reels) and PCB area must also be taken into account. I did this some time ago, and was flabbergasted by the cost contribution of simple resistors! And $1 seems to be unfair for a 555-replacement, a 10F or ATtiny11 reasonable amounts should cost much less. But of course for using a uC the cost fof programming must be taken into account. And of course the engineering (one-time, hopefully) cost will be affected. Wouter van Ooijen -- ------------------------------------------- Van Ooijen Technische Informatica: http://www.voti.nl consultancy, development, PICmicro products docent Hogeschool van Utrecht: http://www.voti.nl/hvu ____________________________________________

Don't forget the cost of stocking multiple parts (square footage) and test cost. The processor version can probably "self test" its circuit, whereas the discrete version would need to be tested on a bed of nails to ensure that all parts were present, they were the correct value, and were truly soldered down. Typically, labor costs are the biggest portion of the testing cost. Anything done to minimize test time (labor time) is a high priority. Ken .....klumiaKILLspam@spam@adelphia.net {Original Message removed}

On Fri, 29 Oct 2004, Jan-Erik Soderholm wrote: > Konstantin K. wrote : > >> Hi, >> >> I have a tank-like bot where each thread is controlled by a servor (2 >> servos in total). This makes a single pic not suitable for driving >> both at the same time without extensive multitasking. > > If you look at the problem as a *single* task controling *two* (or more) > servos, you'll see that you don't need any "multitasking" at all. > >> The servos are modified for continuous motion. > > What is "continuous" here ? The potentiometer is removed and replaced with a fixed resistor divider. Then the servo can be used as drive motors having forward - stop - reverse (it will run with variable speed within certain limits (close to the 'stop' point) forward and reverse as the servo chip implements a type of pwm when the servo error term is not zero). You have to refresh the servo at 50Hz (as closely as possible) because some servos have drastic lockouts for glitches and for control loss and won't respond outside a 50Hz refresh rate. Also some servos will return glitches conducted on the data line and can glitch pics. Last, servo start current is large enough that it will collapse your main supply line if you start more than one at the same time from a less than perfect battery. You must account for this. Peter ____________________________________________

Konstantin K. wrote (in one post) : > Thanks a lot for your answers. I will probably take your advice and > stick with the microcontroller. However, it seems as if many people > are abandoning "designing circuits" in favor of microcontrollers. It > seems reasonable because they are so cheap and easy to use but somehow > it makes me uneasy. But I read your original post as that you was planning to use, not only *one* processor, but also some more (12Fxxx) processors (or some 555's) in *addition* to the main processor, not ? > If you think in commercial terms (in my case), a micrcontroller costs > a little over $1 where a 555 timer and a few resistors, capacitors, > and transistors might cost 50c, which is about 3 times as cheap. But, the idea (at least mine) was to cut out the extra processors (and/or 555's) and not replace them with anything ! Zero extra cost. Why are you comparing costs ? You already have the PIC, don't you ? Or are you looking at a solution with no processor at all ? I must be missing something... Regards, Jan-Erik. ____________________________________________

Peter L. Peres wrote : > On Fri, 29 Oct 2004, Jan-Erik Soderholm wrote: > > > Konstantin K. wrote : > > > >> The servos are modified for continuous motion. > > > > What is "continuous" here ? > > The potentiometer is removed and replaced with a fixed > resistor divider. Then the servo can be used as drive motors > having forward - stop - reverse. Oh, I see ! But, that doesn't change it much from the PIC side of it. The PIC just feeds a 20ms/1-2ms PWM signal and don't care what the servo actualy does, not ? That is, if it is "positioning" or "free running". And that doesn't, in itself, call for some extra PICs or 555's to drive the servo. Jan-Erik. ____________________________________________

On Oct 29, 2004, at 7:47 AM, Wouter van Ooijen wrote: > and $1 seems to be unfair for a 555-replacement, a 10F or ATtiny11 > reasonable amounts should cost much less. > Not in small volumes. The cheapest 10F is more than $1 at digikey... (and in high volumes, your 555 gets lower than $0.50, too.) I was a little surprised at the 10F pricing; we'll see what happens... The Tiny11 is a better candidate, but perhaps only due to a marketing push at the moment... BillW ____________________________________________