Post by thread:[OT] counter

> > I need to be able to count to 4,000,000 (24 bits) and when complete I wish > to address serially for the result. The counter is clocked at about 0.5MHz > which is a little fast for the PIC to keep up with. Can anyone suggest a > chip that would do this for me. I do not want a parallel output as it uses > too many pins. If this is not possible then perhaps a counter and shift > register?? I'm a bit confused. While this may be a bit much for a PIC to count solely in software, using its hardware counter is an ideal application. The keep the software interaction down you can simply interrupt on rollover and have a short interrupt routine that bumps up the top two bytes. In short you're going to create a whole lot of extra hardware doing this with discretes. If you really feel you must have an extra piece of hardware, then I suggest using a second PIC whose sole purpose is to count and report the count serially to the master PIC. A 12C508 could give you the requested functionality in an 8 pin package. a three line serial interface should be sufficient for both interrupting the master PIC, and commanding and reading the counter PIC. This is the type of application where microcontrollers excell. BAJ > > TIA, > > > Peter Grey >

On Mon, 15 Feb 1999 20:28:23 +1100 Peter Grey <spam_OUTmartechTakeThisOuTOZEMAIL.COM.AU> writes: >I need to be able to count to 4,000,000 (24 bits) and when complete I >wish >to address serially for the result. The counter is clocked at about >0.5MHz >which is a little fast for the PIC to keep up with. Timer 0 will count this fast, but it's only 8 bits. It's no problem to expand the count in software though. On a 14-bit PIC, you can have teh timer cause an interrupt that increments a 16-bit high count in RAM. Or just poll the T0IF flag in software and advance the high count whenever you find it set. The polling has to be done fast enough that the timer doesn't overflow twice before it is polled again. Another method I've used is to periodically read TMR0 and add any change it has made to a large variable holding the final count: read timer (call that value N) I = N - L (L is the last timer value read) L = N C = C + I (C is the 24 bit count) Do this process as often as you want but at least fast enough that the counter won't count more than 256 counts inbetween. This process, though more complicated than using T0IF, works on 12-bit PICs as well. You could dedicate a 12C508 to the counter function and have it feed the results serially to another PIC. Can anyone suggest {Quote hidden}>a >chip that would do this for me. I do not want a parallel output as it >uses >too many pins. If this is not possible then perhaps a counter and >shift >register?? > >TIA, > > >Peter Grey > ___________________________________________________________________ You don't need to buy Internet access to use free Internet e-mail. Get completely free e-mail from Juno at http://www.juno.com/getjuno.html or call Juno at (800) 654-JUNO [654-5866]

Try this, Use the RTCC input as a counter. Have a 3 byte software counter for your 24 bits. Add the contents of the RTCC to the 24 bit counter in a contineous loop. The RTCC accumulates the count upto 255 which can be added to the 24 bit counter. After adding, clear the RTCC. This method works for me with a 12C508. Pailoor Peter Grey wrote: {Quote hidden}> I need to be able to count to 4,000,000 (24 bits) and when complete I wish > to address serially for the result. The counter is clocked at about 0.5MHz > which is a little fast for the PIC to keep up with. Can anyone suggest a > chip that would do this for me. I do not want a parallel output as it uses > too many pins. If this is not possible then perhaps a counter and shift > register?? > > TIA, > > Peter Grey Chip Technologies Microchip Design Consultant No. 70, 9th Main Road, Mathikere, Bangalore - 560 054. INDIA Tel : +91-80-3362807 Fax : +91-80-3369451 Email : .....chiptechKILLspam@spam@vsnl.com Webpage : http://business.vsnl.com/chiptech

At 08:28 PM 2/15/99 +1100, Peter Grey wrote: >I need to be able to count to 4,000,000 (24 bits) and when complete I wish >to address serially for the result. The counter is clocked at about 0.5MHz >which is a little fast for the PIC to keep up with. Can anyone suggest a >chip that would do this for me. I do not want a parallel output as it uses >too many pins. If this is not possible then perhaps a counter and shift >register?? There's a counter chip, 74HC4020 I think, that can be cascaded to do the job, there are three in that seires, 4020, 4040, 4060, one of them has all stages with outputs, that's the one you want.. Otherwise, an AVR8515 or 1200 running at 4 MHz or less will do it, and give you a uart or SPI port as well. One of my apps uses a 300kHz interrupt with some fair work to do, and I've got plenty of time remaining.

Hi, Peter Grey wrote: > > I need to be able to count to 4,000,000 (24 bits) and when complete I wish > to address serially for the result. The counter is clocked at about 0.5MHz > which is a little fast for the PIC to keep up with. Can anyone suggest a > chip that would do this for me. I do not want a parallel output as it uses > too many pins. If this is not possible then perhaps a counter and shift > register?? You will need only a single resistor (abt 1K) in series with TMR0 and an extra pin shorted to it. Set the prescaler to /256. This will give you 16 bit resolution and up to 50 MHz count (limit of the prescaler). The extra 8 bits overflow can be counted via software which is only abt 800Hz. TMR0 register can be read directly but the prescaler is nearly impossible. But there is a nice technique to get the value by self clocking. | PIC | RA4 |----+--/\/\/\/\-------O count input | | RA3 |----+ | | This is done by: 1. set to TMR0 to external count, prescaler /256, advance on low to high 2. tristate pin (RA3 for example) 3. wait for a certain time while counting TMR0 overflow 4. pull pin low and remove tristate 5. save overflow value to cnt1 (cnt1 msb, cnt2, cnt3 lsb) 6. save TMR0 value to cnt2 7. pulse pin (low to high) until value of TMR0 changes and count the pulses 6. compute cnt3, cnt3 = 256 - pulses 7. goto 2 regards, Reggie

At 02:15 AM 15/02/99 -0800, you wrote: Sorry for short explanation. I have an external clock signal which is gated with data. I need to count the number of times that data is high when a clock pulse arrives. To use a PIC would involve an external clock. I cannot use the internal timers. Also as each instructiona cyclye is 0.4microsecs, I do not have the time to store this. TIA Peter {Quote hidden}>At 20:28 02/15/99 +1100, Peter Grey wrote: >>I need to be able to count to 4,000,000 (24 bits) and when complete I wish >>to address serially for the result. The counter is clocked at about 0.5MHz >>which is a little fast for the PIC to keep up with. > >how that? can't you use eg. timer 1 as a counter? > >ge > >

At 08:05 AM 15/02/99 -0500, you wrote: To translate, 0.5MHz means 0.5 microseconds. If I clocked a PIC at 10MHz then I would have time to do about 1.25 instructions during this time. I do not want to go above 10 MHz for my crystal. I am also doing a lot of other things inthe micro and wish to be informed when the counts have finished (another signal) Thanks, Peter {Quote hidden}>> >> I need to be able to count to 4,000,000 (24 bits) and when complete I wish >> to address serially for the result. The counter is clocked at about 0.5MHz >> which is a little fast for the PIC to keep up with. Can anyone suggest a >> chip that would do this for me. I do not want a parallel output as it uses >> too many pins. If this is not possible then perhaps a counter and shift >> register?? > >I'm a bit confused. While this may be a bit much for a PIC to count solely >in software, using its hardware counter is an ideal application. The keep >the software interaction down you can simply interrupt on rollover and have >a short interrupt routine that bumps up the top two bytes. > >In short you're going to create a whole lot of extra hardware doing this >with discretes. > >If you really feel you must have an extra piece of hardware, then I suggest >using a second PIC whose sole purpose is to count and report the count >serially to the master PIC. A 12C508 could give you the requested functionality >in an 8 pin package. a three line serial interface should be sufficient for >both interrupting the master PIC, and commanding and reading the counter PIC. > >This is the type of application where microcontrollers excell. > >BAJ >> >> TIA, >> >> >> Peter Grey >> > >

At 11:31 AM 15/02/99 -0500, you wrote: This is about where I have headed. I was going to use 2 of the 74HC4040 cascaded to give me 24 bits and then attached a 74HC166 (3 off) to the outputs and read the counts out serially when finished. Thanks for the help, Peter {Quote hidden}>At 08:28 PM 2/15/99 +1100, Peter Grey wrote: >>I need to be able to count to 4,000,000 (24 bits) and when complete I wish >>to address serially for the result. The counter is clocked at about 0.5MHz >>which is a little fast for the PIC to keep up with. Can anyone suggest a >>chip that would do this for me. I do not want a parallel output as it uses >>too many pins. If this is not possible then perhaps a counter and shift >>register?? > >There's a counter chip, 74HC4020 I think, that can be cascaded to do the >job, there are three in that seires, 4020, 4040, 4060, one of them has all >stages with outputs, that's the one you want.. > >Otherwise, an AVR8515 or 1200 running at 4 MHz or less will do it, and give >you a uart or SPI port as well. One of my apps uses a 300kHz interrupt >with some fair work to do, and I've got plenty of time remaining. > >

> > At 08:05 AM 15/02/99 -0500, you wrote: > > To translate, 0.5MHz means 0.5 microseconds. If I clocked a PIC at 10MHz > then I would have time to do about 1.25 instructions during this time. I do > not want to go above 10 MHz for my crystal. I am also doing a lot of other > things inthe micro and wish to be informed when the counts have finished > (another signal) Peter, By setting the counter to overflow, you'll only see an interrupt every 128 uS by using the TMR0 counter. As I said below it's a bit much to do in software but by using the timer hardware, it because doable if not easy. If the timer isn't available, you definitely need another external part. BAJ {Quote hidden}> > Thanks, > > Peter > > >> > >> I need to be able to count to 4,000,000 (24 bits) and when complete I wish > >> to address serially for the result. The counter is clocked at about 0.5MHz > >> which is a little fast for the PIC to keep up with. Can anyone suggest a > >> chip that would do this for me. I do not want a parallel output as it uses > >> too many pins. If this is not possible then perhaps a counter and shift > >> register?? > > > >I'm a bit confused. While this may be a bit much for a PIC to count solely > >in software, using its hardware counter is an ideal application. The keep > >the software interaction down you can simply interrupt on rollover and have > >a short interrupt routine that bumps up the top two bytes. > > > >In short you're going to create a whole lot of extra hardware doing this > >with discretes. > > > >If you really feel you must have an extra piece of hardware, then I suggest > >using a second PIC whose sole purpose is to count and report the count > >serially to the master PIC. A 12C508 could give you the requested functionali ty {Quote hidden}> >in an 8 pin package. a three line serial interface should be sufficient for > >both interrupting the master PIC, and commanding and reading the counter PIC. > > > >This is the type of application where microcontrollers excell. > > > >BAJ > >> > >> TIA, > >> > >> > >> Peter Grey > >> > > > > >

> > At 02:15 AM 15/02/99 -0800, you wrote: > Sorry for short explanation. I have an external clock signal which is gated > with data. I need to count the number of times that data is high when a > clock pulse arrives. To use a PIC would involve an external clock. I cannot > use the internal timers. Also as each instructiona cyclye is 0.4microsecs, I > do not have the time to store this. More information which helps. If the internal timers are not available then you definitely need another part. I'd vote for a second PIC (for its timer and serial capabilities) with an extern and gate to gate the clock signal. So the second PIC would count whenever the data signal is high using its internal counter, and could then transmit the information to the primary PIC when requested. Would take an 8 pin and a 14 pin part which is a smaller part and pin count than the 74HC4040 and serial shift register solution I saw in the other post. Well worth taking a look at before committing to the multiple counter solution I think.... BAJ {Quote hidden}> > TIA > > Peter > >At 20:28 02/15/99 +1100, Peter Grey wrote: > >>I need to be able to count to 4,000,000 (24 bits) and when complete I wish > >>to address serially for the result. The counter is clocked at about 0.5MHz > >>which is a little fast for the PIC to keep up with. > > > >how that? can't you use eg. timer 1 as a counter? > > > >ge > > > > >

>More information which helps. If the internal timers are not available then >you definitely need another part. I'd vote for a second PIC (for its timer >and serial capabilities) with an extern and gate to gate the clock signal. So >the second PIC would count whenever the data signal is high using its internal >counter, and could then transmit the information to the primary PIC when >requested. Would take an 8 pin and a 14 pin part which is a smaller part and >pin count than the 74HC4040 and serial shift register solution I saw in the >other post. An interesting approach might be to use a single 4040, and a pic, instead of multiple 4040s and multiple buffers.

On Tue, 16 Feb 1999, Peter Grey wrote: > At 02:15 AM 15/02/99 -0800, you wrote: > Sorry for short explanation. I have an external clock signal which is gated > with data. I need to count the number of times that data is high when a > clock pulse arrives. To use a PIC would involve an external clock. I cannot > use the internal timers. Also as each instructiona cyclye is 0.4microsecs, I > do not have the time to store this. 24bits, .5us resolution? What's producing the pulse? Why does it have to be measured so accurately? It's possible to sample a signal with a resolution of 3 instruction cycles AND do other things at the same time. Dwayne and I solved this problem a while back when he had the task to accurately measure the pulse width produced by a temperature sensor. At 10Mhz, 3 instruction cycles will give you a sample resolution of 1.2uS. (If I had the code handy, I'd post it...). ----- Also, for the external counters, you're implying that there's an external clock too. Why can't you use this clock to gate TMR0? An AND gate who's inputs are the clock and data signal could control when the TMR is being clocked. Scott

At 07:48 PM 15/02/99 -0500, you wrote: VOILA!! I see what you mean. Thanks to all for the suggestions. Peter {Quote hidden}>> >> At 08:05 AM 15/02/99 -0500, you wrote: >> >> To translate, 0.5MHz means 0.5 microseconds. If I clocked a PIC at 10MHz >> then I would have time to do about 1.25 instructions during this time. I do >> not want to go above 10 MHz for my crystal. I am also doing a lot of other >> things inthe micro and wish to be informed when the counts have finished >> (another signal) > >Peter, > >By setting the counter to overflow, you'll only see an interrupt every 128 uS >by using the TMR0 counter. As I said below it's a bit much to do in software >but by using the timer hardware, it because doable if not easy. > >If the timer isn't available, you definitely need another external part. > >BAJ >> >> Thanks, >> >> Peter >> >> >> >> >> I need to be able to count to 4,000,000 (24 bits) and when complete I wish >> >> to address serially for the result. The counter is clocked at about 0.5MHz >> >> which is a little fast for the PIC to keep up with. Can anyone suggest a >> >> chip that would do this for me. I do not want a parallel output as it uses >> >> too many pins. If this is not possible then perhaps a counter and shift >> >> register?? >> > >> >I'm a bit confused. While this may be a bit much for a PIC to count solely >> >in software, using its hardware counter is an ideal application. The keep >> >the software interaction down you can simply interrupt on rollover and have >> >a short interrupt routine that bumps up the top two bytes. >> > >> >In short you're going to create a whole lot of extra hardware doing this >> >with discretes. >> > >> >If you really feel you must have an extra piece of hardware, then I suggest >> >using a second PIC whose sole purpose is to count and report the count >> >serially to the master PIC. A 12C508 could give you the requested functionality >> >in an 8 pin package. a three line serial interface should be sufficient for >> >both interrupting the master PIC, and commanding and reading the counter PIC. {Quote hidden}>> > >> >This is the type of application where microcontrollers excell. >> > >> >BAJ >> >> >> >> TIA, >> >> >> >> >> >> Peter Grey >> >> >> > >> > >> > >

At 05:06 PM 15/02/99 -0800, you wrote: Thanks for the suggestion. I was overlooking the fact that the timer0 overflows only every 128 clocks so I have plenty of time to store this data. I am using a CCD device to scan some paper for marks. The resolution is 2048 bits wide by 2048 bits long. The 500KHz clock is necessary to drive the CCD device. This is made by Sony. Thanks, Peter {Quote hidden}>On Tue, 16 Feb 1999, Peter Grey wrote: > >> At 02:15 AM 15/02/99 -0800, you wrote: >> Sorry for short explanation. I have an external clock signal which is gated >> with data. I need to count the number of times that data is high when a >> clock pulse arrives. To use a PIC would involve an external clock. I cannot >> use the internal timers. Also as each instructiona cyclye is 0.4microsecs, I >> do not have the time to store this. > >24bits, .5us resolution? What's producing the pulse? Why does it have to >be measured so accurately? > >It's possible to sample a signal with a resolution of 3 instruction cycles >AND do other things at the same time. Dwayne and I solved this problem a >while back when he had the task to accurately measure the pulse width >produced by a temperature sensor. At 10Mhz, 3 instruction cycles will give >you a sample resolution of 1.2uS. (If I had the code handy, I'd post >it...). > >----- > >Also, for the external counters, you're implying that there's an >external clock too. Why can't you use this clock to gate TMR0? An AND gate >who's inputs are the clock and data signal could control when the TMR is >being clocked. > >Scott > >

On Tue, 16 Feb 1999 10:54:59 +1100 Peter Grey <martechKILLspamOZEMAIL.COM.AU> writes: >At 02:15 AM 15/02/99 -0800, you wrote: >Sorry for short explanation. I have an external clock signal which is >gated >with data. I need to count the number of times that data is high when >a >clock pulse arrives. Use an external AND gate to combine the clock and the data. The output of the AND gate will go high only if the data is high while the clock is also high. That signal is suitable to apply to the counter input of a PIC. ___________________________________________________________________ You don't need to buy Internet access to use free Internet e-mail. Get completely free e-mail from Juno at http://www.juno.com/getjuno.html or call Juno at (800) 654-JUNO [654-5866]

At 08:05 AM 15/02/99 -0500, you wrote: > >To translate, 0.5MHz means 0.5 microseconds. If I clocked a PIC at 10MHz 0.5MHz has a period of 2 microseconds, not 0.5!! {Quote hidden}>then I would have time to do about 1.25 instructions during this time. I do >not want to go above 10 MHz for my crystal. I am also doing a lot of other >things inthe micro and wish to be informed when the counts have finished >(another signal) > >Thanks, > >Peter > >>> >>> I need to be able to count to 4,000,000 (24 bits) and when complete I wish >>> to address serially for the result. The counter is clocked at about 0.5MHz >>> which is a little fast for the PIC to keep up with. Can anyone suggest a >>> chip that would do this for me. I do not want a parallel output as it uses >>> too many pins. If this is not possible then perhaps a counter and shift >>> register?? >> >>I'm a bit confused. While this may be a bit much for a PIC to count solely >>in software, using its hardware counter is an ideal application. The keep >>the software interaction down you can simply interrupt on rollover and have >>a short interrupt routine that bumps up the top two bytes. >> >>In short you're going to create a whole lot of extra hardware doing this >>with discretes. >> >>If you really feel you must have an extra piece of hardware, then I suggest >>using a second PIC whose sole purpose is to count and report the count >>serially to the master PIC. A 12C508 could give you the requested functionality >>in an 8 pin package. a three line serial interface should be sufficient for >>both interrupting the master PIC, and commanding and reading the counter PIC. {Quote hidden}>> >>This is the type of application where microcontrollers excell. >> >>BAJ >>> >>> TIA, >>> >>> >>> Peter Grey >>> >> >>

On Mon, 15 Feb 1999 20:58:29 +0530 Ravi Pailoor <.....chiptechKILLspam.....VSNL.COM> writes: >Add the contents of the RTCC to the 24 bit counter in a contineous >loop. > >The RTCC accumulates the count upto 255 which can be added to the 24 >bit >counter. After adding, clear the RTCC. Don't write to RTCC while it is counting. Hardware prevents it from counting for 2 instruction cycles after a write. If a pulse comes in during that time, it may not be counted. You can read the counter at any time without disturbing the counting process. Use the method I presented involving storing the last reading and subtracting the new reading from it. It can be optimized to this: movfw oldrtcc ;Get old count subwf RTCC,w ;(TMR0 on newer PICs) W = new - old addwf oldrtcc,f ;old = old + (new - old) = new ;Now add the value in W to the 24 bit count. addwf count,f ;Add low bytes. skpnc ;Skip if no carry to high bytes incfsz count+1,f ; Carry - advance middle byte goto done ;No need to advance MSB incf count+2,f ; Advance high byte done The whole process takes exactly 8 PIC cycles. It only has to be done every 255 cycles or less (assuming the counter is counting as fast as it can, at abput 1 per instruction). So 97% of the PIC's processing time is available for something else. ___________________________________________________________________ You don't need to buy Internet access to use free Internet e-mail. Get completely free e-mail from Juno at http://www.juno.com/getjuno.html or call Juno at (800) 654-JUNO [654-5866]

At 08:51 PM 15/02/99 -0500, you wrote: Ravi, This looks very interesting. However, I only need to know the final count not any intermediate counts. Thus I should be able to count the number of times timer 0 overflows and store in a couter (3 off) and then when I get the command to stop, just read timer 0 and add this to the counts. Thanks for teh suggestion on counters Peter {Quote hidden}>On Mon, 15 Feb 1999 20:58:29 +0530 Ravi Pailoor <chiptechspam_OUTVSNL.COM> >writes: > >>Add the contents of the RTCC to the 24 bit counter in a contineous >>loop. >> >>The RTCC accumulates the count upto 255 which can be added to the 24 >>bit >>counter. After adding, clear the RTCC. > >Don't write to RTCC while it is counting. Hardware prevents it from >counting for 2 instruction cycles after a write. If a pulse comes in >during that time, it may not be counted. > >You can read the counter at any time without disturbing the counting >process. Use the method I presented involving storing the last reading >and subtracting the new reading from it. It can be optimized to this: > > movfw oldrtcc ;Get old count > subwf RTCC,w ;(TMR0 on newer PICs) W = new - old > addwf oldrtcc,f ;old = old + (new - old) = new > >;Now add the value in W to the 24 bit count. > addwf count,f ;Add low bytes. > skpnc ;Skip if no carry to high bytes > incfsz count+1,f ; Carry - advance middle byte > goto done ;No need to advance MSB > incf count+2,f ; Advance high byte >done > >The whole process takes exactly 8 PIC cycles. It only has to be done >every 255 cycles or less (assuming the counter is counting as fast as it >can, at abput 1 per instruction). So 97% of the PIC's processing time is >available for something else. > >___________________________________________________________________ >You don't need to buy Internet access to use free Internet e-mail. >Get completely free e-mail from Juno at http://www.juno.com/getjuno.html >or call Juno at (800) 654-JUNO [654-5866] > >