Interrupt overflows being missed?

laborratte · Joined: 27 Jul 2005 Posts: 299 Location: Berlin

Hi,
A) You are creating a "artifical" 16Bit timer0 with your overflow counter isr and use timer1 as a pwm with a resoultion of 80 (dec) - less than 7 bit. Why not vice versa? When you use timer0 as pwm source you can have the 16bit timer1 without serving an extra overflow counter

B) When working with timer1 anyway you can use in addition the input capture mode like

teccs · Joined: 14 Oct 2006 Posts: 69

Hello laborratte.

Thanks for the reply. I did change the use of Timer0 and Timer 1 around in the second posting of my code. I have not been able to find good material to read so that I can truly understand the timers and pwm. I copied the code in the second posting of my code from the helpful suggestion given by O-Family.

I would love to find a good resource about the timers and pwm. If you have any I can read, please let me know.

albertsm · Joined: 09 Apr 2004 Posts: 5921 Location: Holland

i removed some non technical posts.
private messaging is not activated in this forum. but if some users like to exchange emails to exchange thoughts about non technical matters, let me know in an email.
this forum can however not be used for that.

O-Family · Joined: 23 May 2010 Posts: 320 Location: Japan

Hi,
The tachometer program I guided uses a 16-bit Timer1 capture.
This program also accurately measures frequency, period and rotation speed in 32 bits.
However, it may be difficult for you to understand because the articles and program comments on the site are in Japanese.

If you wish, I can translate the comments into English, although it will take some time.

laborratte · Joined: 27 Jul 2005 Posts: 299 Location: Berlin

EDC · Joined: 26 Mar 2014 Posts: 971

Hello. Take a look on this code. It is based/mixed from those codes -> https://www.mcselec.com/index2.php?option=com_forum&Itemid=59&page=viewtopic&t=14403&highlight=freq+duty.

The frequency is aqurate into 25kHz (lok attached pics) and you have opportunity to set Duty from 1 to 99% i think Very Happy

In the meaning time uC counts RPM by Timer0 Input pin. Only if it counts more than 256 then interrupt is fired only for counting overflows.
After one second those overflows are Shifted Left 8 because it is 8 bit Timer0. Then rest of Timer0 is added. It measure my generator which I set into 500Hz. I think it measure well Very Happy

This can be done in another way but maybe this will be sufficient.
Some improvements also can be made here. This is couple of minutes code only. For example Overflows can be even a Byte type cause they are copied into Word later...

SZTRAD · Joined: 30 Dec 2019 Posts: 165

Hey, EDC
I confess, I don't know how this works.
Timer0 never overflows, so the interrupt doesn't start.
Timer2 won't start at all, or am I wrong? If I remember anything in mathematics, it should apply to speeds of 500 rpm, the frequency is 8.3Hz (500/60). This means that in one second the value of 8 is loaded into Timer0. How you calculated 500 of this value is beyond me. I will also point out that this value needs to be divided by 2. The fan sends 2 signals per revolution (at least the vast majority of them).
I can't seem to keep track of new programming trends.
And I mean, honestly, I really miss how it works.

RS

EDC · Joined: 26 Mar 2014 Posts: 971

Timer2 or any Timer configured in Bascom dont need command "Start Timer". They start automatically. Command Start Timer is used only if you Stop Timer to start them again. So Timer2 is working for shure but I dont need interrupt from it. I only check its flag.
So for Timer2 I check its TimerInterruptFlagRegister TIFR2 where Timer showing Overflows or CompareA/B Event.
Then I clear the flag by writing One into it (not Zero).
So I can measure the time. 100x 10ms = 1s

Timer0 have special pin on the PORTD.4 (name T0). When Timer0 is configured as a Counter (not Timer) then it counts Falling or Rising edges on this pin. Every of this event is counted till 255 because its 8bit. Only then Overflow appear. So events can be counted without INT0 or INT1 and everytime firing Interrupt. Simply Timer count silently.
Even if overflow will not happen then Timer0 have for example 120 ticks after second and this is Hz (Revolution per Second here)

Now I saw that I name it RPM - my mistake Very Happy

And ofcourse if motor generate two ticks per revolution then simply whole result can be Shift , Right , 1 for divide by two.

Value 500 is from my square wave generator connected into T0 pin to show that uC counts.

SZTRAD · Joined: 30 Dec 2019 Posts: 165

Witam EDC
ja to przeżyłem ponownie i zrozumiałem, ale dopiero po wysłaniu i nie będę masował pytania to nie byłoby fair. Chcę tylko powiedzieć, że to rozwiązanie jest nieprzydatne dla pytającego. Przy niskiej częstotliwości (a on będzie mierzył niską częstotliwość) błąd będzie + -1Hz (co w jego przypadku oznacza błąd 420-540 obr / min). On musi mierzyć długość okresu.
Z Timerem2 masz rację. Ja tak jakoś przyzwyczaiłem się do puszczania sobie wszystkiego sam.

Embarassed

Oops I'm sorry
Hi, EDC I went over it once more and I got it, but after I sent it and I won't delete the query it wouldn't be fair. I just want to say that this solution is useless to the questioner. At a low frequency (and he will measure the low frequency) the error will be +-1Hz (which in his case means an error of 420-540rpm). He has to measure the length of the period.
You're right about Timer2. I'm kind of used to playing everything myself.
RS

EDC · Joined: 26 Mar 2014 Posts: 971

Yes. I didn't read the whole topic in depth. Maybe this is not the best uC for this job but this can be done.
What are minimum and maximum RPM?
If two impulses are for one revolution then this will be good to measure them both, then add to eachother (or even every new to previous) and then divide by two. This will average eventually disproportion betwen them.

Remember that with Arduino board you dont stuck with 16MHz. You can use Clockdivision = 2 and get 8MHz for example. Then 25kHz can be easy achieve on the 8bit timer with resolution 160 steps with Prescale1. So ICP1 from Timer1 can be used for measure.

teccs · Joined: 14 Oct 2006 Posts: 69

Thank you all again for the replies.
I am going back to the beginning of everyone's comments and working on implementing them and seeing the results.

The suggestions from MWS are where I am starting. I get a undefined variable error when I compile the code below. It is due to a reference to a register. I thought these did not need to be Dimensioned but maybe I am wrong?

EDC · Joined: 26 Mar 2014 Posts: 971

Please enable Code Explorer for runtime code checking during programming.
There are TimerInterruptFlagRegister`s TIFR`s for every timer not TIVR`s.

SZTRAD · Joined: 30 Dec 2019 Posts: 165

I took your program and adjusted it a little bit.
I don't have time to compose a real application for testing, but it should probably work. All the settings are yours,so you know what you've been doing.
The difference is you measure the pulse length between the two leading edges.
In this state it should be able to measure up to 1s. The resolution is due to timer0 overflow and amounts to 0.000016s.
If you need to soften it, set some value to timer0, but watch out for Timer0Overflows overflows.

teccs · Joined: 14 Oct 2006 Posts: 69

I could not get this to work. Results are in attached spreadsheet.
There is a non-linear relationship between the actual RPM from the oscilloscope and the RPM reported by BASCOM.

teccs · Joined: 14 Oct 2006 Posts: 69

So, from there, I tried the code EDC posted above.
It works perfectly.
It reports frequency, not RPM.
RPM = Frequency * 30
So, an error of 1 Hz gives and RPM error of 30 RPM.
That is plenty close enough for fan RPM measurement.
Thank you EDC.

I will go on to the other examples posted by others.