Duty cycle detection

earckens · Oct 11, 2017

I need to detect a duty cycle going past a certain threshold. Here is a picture of the duty cycle in yellow; as this decreases from the current approx. 95% to say 50% (adjustable), it needs to trigger an output (TTL level). Preferably with some upper threshold too.

What is the best way to approach this?

Edit: the amplitude of the signal to be detected is constant amplitude TTL level (so, yellow trace at TTL level).

Diver300 · Oct 11, 2017

The easy way is to use an RC filter to get the average voltage and then use a comparator to detect the level.

You can also use a microcontroller.

How accurate do you need it?

earckens · Oct 11, 2017

I would like to be able to set the on-trigger at about 50% with +/-5% accuracy, and the off- trigger at about 30% with same accuracy.

Edit: the frequency remains fixed at about 2kHz; amplitudes always are TTL level.

flatfootskier · Oct 11, 2017

I would stick a capacitor on the input side & stuff it into an Arduino's A/D, where you can set the thresholds to your heart's content

earckens · Oct 11, 2017

flatfootskier said:
I would stick a capacitor on the input side & stuff it into an Arduino's A/D, where you can set the thresholds to your heart's content

That is a software solution. I prefer hardware because it makes the module I want to make independent of any software.

Ultimately I will use the output of the hardware solution in a microcontroller, but not at this stage.

earckens · Oct 11, 2017

Diver300 said:
The easy way is to use an RC filter to get the average voltage and then use a comparator to detect the level.

You can also use a microcontroller.

How accurate do you need it?

Preferably not a microcontroller; anyone with a solution using a 555 and/or comparator?

ronsimpson · Oct 11, 2017

earckens said:
anyone with a solution using a 555 and/or comparator?

A 555 can any job! I am using one as a door stop.

How fast do you want the circuit to respond? The "RC" circuit will take several cycles to respond. I can make one that will respond in one cycle.

ronsimpson · Oct 11, 2017

Input is the green trace. (V2) To the left it was 100%. Then went to a low % and then after 14 cycles went back to 100%.
The blue trace is after the RC. It averages the 0V and 5V to get a level.
Output is in red.
Right now it is set to trip at 33% and 66%. That can be changed.
Not right but something to think about.

I do not like using the RC filter method from "TTL". Because TTL does not pull up well. You should buffer the signal with a CMOS buffer so the pull up is good and predictable. CD40xx or 74Cxx

AnalogKid · Oct 11, 2017

Another approach is for the leading edge of the signal to trigger a fixed monostable circuit that times out at the 50% point. The trailing edge of the signal clocks a flipflop. The monostable output is the D input to the ff. If the signal is less than 50% DC, its trailing edge clocks a 1 into the FF, signalling a fault condition. If the signal is greater than 50% DC, it clocks a 0 into the ff because the monostable has timed out. There is a bit more to it, but that is the basic approach.

The benefit of this approach is cycle-to-cycle monitoring without the delay of a lowpass filter.

ak

ronsimpson · Oct 11, 2017

Attached is the SPICE file.

earckens · Oct 11, 2017

ronsimpson said:
A 555 can any job! I am using one as a door stop.

How fast do you want the circuit to respond? The "RC" circuit will take several cycles to respond. I can make one that will respond in one cycle.

Funny .
Speed is no requirement at all; This system has a step response of an hour!

earckens · Oct 11, 2017

ronsimpson said:
I do not like using the RC filter method from "TTL". Because TTL does not pull up well. You should buffer the signal with a CMOS buffer so the pull up is good and predictable. CD40xx or 74Cxx

Right, I agree. Can you maybe include this where needed in your schematic please?

And what components on the 555 are "in charge" of the 33% and 66% triggers?

earckens · Oct 11, 2017

AnalogKid said:
Another approach is for the leading edge of the signal to trigger a fixed monostable circuit that times out at the 50% point. The trailing edge of the signal clocks a flipflop. The monostable output is the D input to the ff. If the signal is less than 50% DC, its trailing edge clocks a 1 into the FF, signalling a fault condition. If the signal is greater than 50% DC, it clocks a 0 into the ff because the monostable has timed out. There is a bit more to it, but that is the basic approach.

The benefit of this approach is cycle-to-cycle monitoring without the delay of a lowpass filter.

ak

Thank you for that input! Very usefull too. Do you maybe have a schematic handy for this idea?

earckens · Oct 11, 2017

ronsimpson said:
Attached is the SPICE file.

Hi Ronsimpson, please tell me how to open that file? I tried with Eagle but does not seem to work?
Edit: sorry, I found it.

ronsimpson · Oct 11, 2017

earckens said:
Can you maybe include this where needed in your schematic please

Put the CMOS gate/buffer just ahead of the RC. So the TTL signal probably is 0V and 4V the CMOS buffer 0V/5V.

earckens said:
555 are "in charge" of the 33% and 66% triggers?

I tied 2 and 6 together and injected the RC signal there.
The 1/3 and 2/3 voltage points are set by the three resistors from 8 to 1. I don't remember but I think the resistors are 10k. ????
Connect a 1k from 8 to 5 and the "TRIG" voltage will be just below 50% and "THERS" will be about 95%.
***I need to look at the data sheet to see if the inputs will work that close to VCC.***

Another option is to add some resistors to 6 and some to 2 to shift the voltage up or down.

earckens · Oct 11, 2017

ronsimpson said:
Put the CMOS gate/buffer just ahead of the RC. So the TTL signal probably is 0V and 4V the CMOS buffer 0V/5V.

I tied 2 and 6 together and injected the RC signal there.
The 1/3 and 2/3 voltage points are set by the three resistors from 8 to 1. I don't remember but I think the resistors are 10k. ????
Connect a 1k from 8 to 5 and the "TRIG" voltage will be just below 50% and "THERS" will be about 95%.
***I need to look at the data sheet to see if the inputs will work that close to VCC.***
View attachment 108565
Another option is to add some resistors to 6 and some to 2 to shift the voltage up or down.

Hi Ronsimpson, what are "2 and 6", and "from 8 to 1" and "from 8 to 5"?

I just downloaded Splice, I wonder why I have not used that before

Edit: oh I see, are these the 555 pins?

ronsimpson · Oct 11, 2017

It appears the "top" input THRES will work to VCC. The bottom input TRIG will work all the way down to GND.

I do not like this circuit but I took up the challenge.

earckens · Oct 11, 2017

ronsimpson said:
It appears the "top" input THRES will work to VCC. The bottom input TRIG will work all the way down to GND.
View attachment 108566
I do not like this circuit but I took up the challenge.

Why do you not like it?
Please feel free to propose an alternative?
Thks!

ronsimpson · Oct 11, 2017

I started out using a "74ls122" and some logic. [analogkid #9] Set one timer for min time and the other timer for max time.
Too many ICs but will give you a results in one cycle. (time dependent not duty cycle dependent)

Next I did a analog design that is just like a 555 with out the three resistors. Two pots sets the two voltage levels. If I could get at the three resistors and force the levels to 50% and x% with a POT or two, then I would feel better.
(duty cycle dependent)
Could use a dual CMOS 555 and use 1/2 as a buffer.

ronsimpson · Oct 11, 2017

Just for fun I added a 1K from pin5 to pin8. Switches at 92% and 46%.
Maybe the resistor should be closer to 820 or 750.

Duty cycle detection

Active Member

Attachments

Well-Known Member

Active Member

Member

Active Member

Active Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Attachments

Active Member

Active Member

Active Member

Active Member

Well-Known Member

Active Member

Well-Known Member

Active Member

Well-Known Member

Well-Known Member

Similar threads

Privacy & Transparency

Privacy & Transparency