# Arduino Timer 1 counting in excess of 16 bits

Status
Not open for further replies.

#### dr pepper

##### Well-Known Member
I want to measure a frequency 1Mc.
I can get it to work but I get a reading 1,065535 Mc, 65.5365Kc too high.
After a lot of reading I found timer 1's overflow trigger doesnt occur at 0 its at FFFF, hence my reading.
Is there any way I can get the overflow to interrupt on 0, I tried setting Ocr1 to 0 but that doesnt work.

#### be80be

##### Well-Known Member
Why not just subtract FFFF
1,065535 Mc - 65.5365Kc

#### jpanhalt

##### Well-Known Member
It appears the Atmega has a function similar to the PIC capture/compare on Timer1:
datasheet said:
The 16-bit comparator continuously compares TCNT1 with the Output Compare Register (OCR1x). If TCNT equals OCR1x the comparator signals a match. A match will set the Output Compare Flag (TIFR1.OCFx) at the next timer clock cycle.
On a PIC, you can preset the comparator register(s) and get an interrupt/interrupt flag at any count.

Can you describe in a little more detail how the measurement is done and its math? Are there multipliers/dividers and/or pre- or post-scalers involved?

John

#### dr pepper

##### Well-Known Member
I cut the code down to this, its all messed up, subtracting and preloading just makes things worse.
I think the 'duino's operating system is messing things up, I've seen other frequency counters disable timer0, maybe thats it.
I really need to use interrupts for other things, I think I might time share, count freq for a bit then do something else.
Seems I found another 'duino not so nice, they are ok at first but you soon outgrow them, I think I'll go back to pics.

For some reason the forums code post function wont work so I'll cut & paste.

ISR(TIMER1_CAPT_vect) // PULSE DETECTED! (interrupt automatically triggered, not called by main program)
{
TCNT1 = 0; // restart timer for next revolution
masterOscCount += ICR1; // save No. of counts between pulses
}

ISR(TIMER1_OVF_vect) // counter overflow/timeout
{
masterOscCount += 65536;
}

void loop ()
{
{
//TCCR1B = B01000000; // Timer 1 no prescaler, capture on rising edge, no noise filter, ext clock rising
measure();
//TCCR1B = B01000111; // Timer 1 no prescaler, capture on rising edge, no noise filter, ext clock rising
}

}
void measure ()
{
Serial.println (masterOscCount / 1.048576);
masterOscCount = 0;
}

#### be80be

##### Well-Known Member
If it's atmega 328 you have
Code:
TOP Signalize that TCNT1 has reached maximum value.
BOTTOM Signalize that TCNT1 has reached minimum value (zero).
So you can test both ways has it hit top or bottom
One more thing you can use arduino without the cores which means you can use pure avr code.
which takes way less space.

#### dr pepper

##### Well-Known Member
Yes I found top & bottom, had a quick look to see if you can access them, top looks as though its in OCR1, not sure where they hide bottom.
I didnt know you could turn the core off, I'll look into that, for some stuff I'd be better without it, but I'm guessing all the niceties would disappear with it like serial & print.

I think I'm going to try a bodge when I get home (just finished work), it seems that the counter goes from correct reading to 65535 out and doesnt vary, I've never seen it change just by a few hz and I've been using a ocxo, so its either correct or miles out, frequency readings are not going to change by that much so I'll check each reading against the last and if its changed by a large amount dump the current reading, then if the core is causing a major error once in a while (it seems to be Ok for a minute then I'll get a few 65535 out readings, then it'll be Ok again for a while so it does look as though its beating frequency wise with some operation in the core) it will be ignored.

Last edited:

#### dr pepper

##### Well-Known Member
I dumped the idea altogther, I compiled paul stoffregan's Freqcounter library and tested that out, it works, then I tried it and my other code multitasking, it works fine.
The thing now though is the Freqcount library uses the 'duino's internal crystal for timing, I want to use an external 1pps clock pulse, so I'll have to dive into the .cpp file and see if I can make it do that, as well as posting on the Pjrc forum to see if anyone has any input of course.

Edit: just reading something I found out if your dealing with int's or long variables interrupts should be disabled as if the code is interrupted during a word or long word transfer then one or more of the bytes gets corrupted - interesting.

Last edited:

#### be80be

##### Well-Known Member
You can still us the stuff you want the arduino ide uses every thing kind of like turning off mikro c libraries

#### Ian Rogers

##### User Extraordinaire
Forum Supporter
Edit: just reading something I found out if your dealing with int's or long variables interrupts should be disabled as if the code is interrupted during a word or long word transfer then one or more of the bytes gets corrupted - interesting.
Bad compiler!! If the context is saved correctly the carry should never be missed.. However!! If you read the PIC timer doc, it mentions the bother with reading the timer in the wrong way!! You really need to stop the timer then read..

If timing is critical.. Use preload as I mentioned

#### dr pepper

##### Well-Known Member
I tried pre loading with 1 and subtracting 1 and a few other ideas, but the result just changes by one.
The issue comes & goes, it'll read 1,000,000 for a 1mhz input, and then every 20 results or so it'll go 1,065,535 sometimes just the once, sometimes a few times, then it'll go Ok again, so subtracting 65,535 obviously wont work.
I think the core is robbing processing time and the timer interrupts are not being serviced when requested and the count gets too high, a bit odd though its always 65,535 too high, I've tried a 1mc signal and a 3mc signal, both come up with the same 65,535 over read.
I find it odd the error is the same, plus if you divide 1mc by 16 bits it isnt close to an integer so the 16 bit timer will not be rolling over at the measurement interrupt anyway.
My sig gen square wave o/p goes to 240kc this counts fine every time.
It can be done as the freqcount lib works ok, it undereads on 3mc, and if I then measure 1mc the underead is 1/3 less so its a crystal accuracy error not a count error, the lib for this at first looks easy but after some messing you soon get bogged down.
Reading the 16 bit timer understandably could go wrong if there was an interrupt inbetween hi & lo byte reads, I've tried disabling interrupts during counting, but the issue is exactly the same.
So not sure what to do right now, maybe a hardware solution.

Last edited:

#### dr pepper

##### Well-Known Member
It would seem google is my freind.
This is a frequency counter using 1PPS from a gps system, I'm not using gps, but I do have 1PPS.
If it does the job it saves me a lot of headache.
Interesting to see they dont use timer 1 in capture mode, an interrupt pin is used as the gate instead.
https://github.com/JChristensen/gpsFreq

#### dr pepper

##### Well-Known Member
Unbelievable, I compiled the above example program, and its the same, not exactly but near enough, the program doesnt over read by 65,535 on occasion, it under reads by that amount.
I definately should have stuck with pic's.

Edit: A bit of lateral thinking, I want measure the accuracy of 1mc and 3mc oscillators so I can plot a graph, maybe also generate a control voltage to control them in a software FLL loop.
Its not necessary to implement a 32 or 48 bit counter that counts an absolute frequency value, the osc will only be off by a max of 200hz, so I can let the 16 bit timer1 overflow and ignore it, if the oscillator is running at the correct freq the 16 bit contents will the same every time, with a 1 second gate counting 1mc with 16 bits will be 16,960 decimal, so if I subtract this from timer1 every 1 second capture I'll get deviation from the desired freq, there will be a couple other desired freq's too but I can take care of that in the code.
Dont know why I didnt think o that sooner.

Last edited:

#### dr pepper

##### Well-Known Member
Found this on a website, a method of counting frequency using timer1, it seems you need to take care if the counter rolls over:

case GO:
prev_t1_count = t1_count; // prev. counter snapshot
t1_count = ICR1; // new counter snapshot

// compensate for counter wraparound
if (t1_count < prev_t1_count) {
t1_delta = 65536 - (prev_t1_count - t1_count);
}
else {
t1_delta = t1_count - prev_t1_count;
}

You were on the right lines Burt, the bit I failed to get was the check for a wraparound before subtracting 65536.

#### dr pepper

##### Well-Known Member
This worked:

ISR(TIMER1_CAPT_vect) // PULSE DETECTED! (interrupt automatically triggered, not called by main program)
{
//TCNT1 = 0; // restart timer for next revolution
oscCount = ICR1; // save No. of counts between pulses
}

ISR(TIMER1_OVF_vect) // counter overflow/timeout
{
overFlow++;
oscCount -= 65535;
}

void loop ()
{
{
measure();
}

}
void measure ()
{
Serial.println (((oscCount - oscCountLast) + (overFlow * 65536)) / 1.048576);
overFlow = 0;
oscCountLast = oscCount;

Icr1 is the capture register, when the capture pin goes active an interrupt is generated and this register contains the snapshot value of timer1 at that time, so it doesnt matter about interrupt latency.
Pretty darn simple once you sussed it, the only change is a -65535 when the 16 bit timer rolls over, all this does is rolls the timer back to where it was +1, kinda like a combination of what you were saying, subtracting and pre-loading.
The reason for the / 1.048576 correction is my gate time isnt 1 second, its 1mc / 2 to the power 20, ie I just used 2 4040 counter ic's instead of 3 'hc390's to / by 1,000,000 to get 1hz.

Last edited:
Status
Not open for further replies.