Filtering wide width pulses

[amsdesign]

Hello,

I have two different type of digital pulses flowing out of a gate.

Narrow width and wide width pulse : See pics below.





How do I filter out the wide width pulses and allow only the narrow width pulses to pass through?

 

[BradtheRad]

Decide what exact length of time makes a wide pulse. When a pulse begins, wait for that length of time, to see if a pulse drops back to 0V before the time window closes.

If it is still high at the end of that time window, then you do not send a pulse. Otherwise you send a pulse.

It may be possible to configure a 555 timer IC to do the job.

 

[amsdesign]

Decide what exact length of time makes a wide pulse. When a pulse begins, wait for that length of time, to see if a pulse drops back to 0V before the time window closes.

If it is still high at the end of that time window, then you do not send a pulse. Otherwise you send a pulse.

It may be possible to configure a 555 timer IC to do the job.

For my space requirements, a 555 timer is way to big a block to solve this problem. I was thinking of just putting something like a series capacitor which would allow the short pulses to pass through (cause its almost an high frequency ac signal) and block the wide width pulse (cause it's like a DC signal) Is this method correct/feasible?

 

[FvM]

For my space requirements, a 555 timer is way to big a block to solve this problem. I was thinking of just putting something like a series capacitor which would allow the short pulses to pass through (cause its almost an high frequency ac signal) and block the wide width pulse (cause it's like a DC signal) Is this method correct/feasible?

It doesn't work this way. The edges of the wide pulse contain the same high frequent signal components as the small pulse.

Firstly your problem specification is "way too" vague. I understand that you intend a pulse by pulse filtering rather than presence detection of a fast pulse train. Pulse by pulse detection requires a time filter that actually measures pulse width. Must not necessarily use a 555 timer, but at least RC circuits with logic gates that regenerate a digital waveform.

Secondly, I suggest to elaborate your detection algorithm. Obviously, a small pulse can't be recognized before the falling edge occurs. Deciding to pass the pulse a this time isn't possible, the only way is to replicate a (delayed) small pulse when the falling edge occured.

 

[amsdesign]

It doesn't work this way. The edges of the wide pulse contain the same high frequent signal components as the small pulse.

Firstly your problem specification is "way too" vague. I understand that you intend a pulse by pulse filtering rather than presence detection of a fast pulse train. Pulse by pulse detection requires a time filter that actually measures pulse width. Must not necessarily use a 555 timer, but at least RC circuits with logic gates that regenerate a digital waveform.

Secondly, I suggest to elaborate your detection algorithm. Obviously, a small pulse can't be recognized before the falling edge occurs. Deciding to pass the pulse a this time isn't possible, the only way is to replicate a (delayed) small pulse when the falling edge occured.

There isn't a "problem specification" exactly, I have an XOR gate that is measuring the phase difference between two pulses and phase differences beyond a particular value (the wide pulses) should be a filtered and the narrow pulses should be allowed to pass.
As of now, I don't have any detection methods.

 

[FvM]

Your latest explanation actually revokes the previous specification:

I have two different type of digital pulses flowing out of a gate.

It turns out that there not two different pulse widths rather than a continuous pulse width range, I guess with a constant pulse period.

It seems to me that you are tackling the problem from the wrong end. There will be surely a signal processing step where pulse width is converted to phase angle. Why not simply applying a threshold to the final phase result?

In case you only need the digital information "phase difference above or below a given threshold", you can simply invert the criterion, filter the small pulses.

 

[crutschow]

A small microprocessor could readily perform the discrimination function for the two pulses with no added parts.

 

[bigdogguru]

I would agree with crutschow, a microcontroller should be considered as a possible option in this case.

As size and the required PCB real estate seem to be an issue, there are several manufacturers which offer low power microcontrollers which feature internal oscillators and are available in relatively small packages, like MSOP, DFN, UDFN, etc., in the 3mm x 3mm size range. More importantly they typically offer a wide range of peripheral modules and features which might be ideal for your specific task. For example, you might want to checkout the Microchip PIC12LF family of microcontrollers.


BigDog

 

[D.A.(Tony)Stewart]

A 1-shot can discriminate based on pulse width but you need a delay line or an oversampled shift-register to determine if the "mark" is gated Out or not.

Seems like a poor way to filter data.

Define the timing precisely.