How to detect wave crest accurately?

[X.Y.]

A normal waveform is converted to digital signal by A/D converter. And
we use a microprocessor to detect the ware crest. We have considered
the method that comparing one value and next one. However, there will
be ripple or noise around the peak. It will make a mistake if we
compare the two values simply. So, how to detect wave crest accurately
in real time?

 

[Phil Allison]

"X.Y."

A normal waveform is converted to digital signal by A/D converter. And
we use a microprocessor to detect the ware crest. We have considered
the method that comparing one value and next one. However, there will
be ripple or noise around the peak. It will make a mistake if we
compare the two values simply. So, how to detect wave crest accurately
in real time?

** That is simply an impossible task.

One cannot detect maxima in a unpredictably varying value UNTIL after it
has passed.

Any uncertainly ( ie noise) in the readings ( samples) will only add to the
delay in identifying the actual instant the rate of change reversed sign and
determining exact peak value achieved.

Identifying the exact moment of a maxima only comes after all the relevant
samples have been smoothed and re examined.



....... Phil

 

[Rich Grise]

A normal waveform is converted to digital signal by A/D converter. And
we use a microprocessor to detect the ware crest. We have considered
the method that comparing one value and next one. However, there will
be ripple or noise around the peak. It will make a mistake if we
compare the two values simply. So, how to detect wave crest accurately
in real time?

Maintain a running maximum in software, that you reset each cycle.

Good Luck!
Rich

 

[David L. Jones]

A normal waveform is converted to digital signal by A/D converter. And
we use a microprocessor to detect the ware crest. We have considered
the method that comparing one value and next one. However, there will
be ripple or noise around the peak. It will make a mistake if we
compare the two values simply. So, how to detect wave crest accurately
in real time?

You can't really do that, a crest on a waveform is only relevant over
a given time period.
Why do you want to do it?
What sort of input waveform are you measuring?
Is it known and predicable, or unknown?

For say a smoothly varying waveform like a sine wave you could use a
(small) running average and then detect the change in direction of the
average (and reset your average), but that only gives you an
indication several samples after the actual maximum.

What are you really trying to do?

Dave.

 

[Jasen Betts]

A normal waveform is converted to digital signal by A/D converter. And
we use a microprocessor to detect the ware crest. We have considered
the method that comparing one value and next one. However, there will
be ripple or noise around the peak. It will make a mistake if we
compare the two values simply. So, how to detect wave crest accurately
in real time?

figure out exactly what you mean by a wave crest as it is represented
by the samples you are taking, and then write software that detects
that.

Bye.
Jasen

 

[John Larkin]

A normal waveform is converted to digital signal by A/D converter. And
we use a microprocessor to detect the ware crest. We have considered
the method that comparing one value and next one. However, there will
be ripple or noise around the peak. It will make a mistake if we
compare the two values simply. So, how to detect wave crest accurately
in real time?

Polynomial curve fit, maybe.

John