> I want to measure the peak amplitude and duration of a non repetitive
> signal from a transducer. =A0Ideally as a pair of numbers on a four digit
> LCD display.
>
If it's really the digitization of a pulse that you're after, look into Wilkinson D/A converters. These are a combination of amplifier/rectifier that does the peak-hold function and a simple single-slope converter. The differential linearity is very good, and a few op amps, some analog switches and such, is all the hardware you need except for programming a PIC to do a count.
As for duration, you need to find some way to discriminate the signal first. If it isn't already digital, what IS the meaning of 'beginning' and of 'end'?
I hadn't realised that I failed to say that the input signal is a sinusoid in my initial post.
Did you mean A/D converter ?
The signal only lasts for a few cycles and occurs at random intervals. I am trying to get the peak/peak amplitude and its duration. I would like to achieve four digit resolution of both values.
Are you sure about four digits? You would need at least 13 bits A2D. PCB layouts beyond 12 bits are tricky. You need a very good noise controlled PCB for it.
If measuring a peak will do, the Wilkinson type converter is OK, but you indicated peak-peak measurement was wanted? So, you need to track/hold-peak/detect and also track/hold-negative-peak to get both an indication of the (+) and (-) excursion. That means two opamps, two rectifiers, two hold capacitors.
The 'timing' is more awkward. The time of first detected peak or trough can be taken from the 'peak detected' output (the op amp saturates (-) when a peak is detected, and the negative-peak op amp will saturate (+) when it detects the negative peak). But that means you have to combine two indicators, and it doesn't tell you the real onset time unless the waveform always appears at a known phase...
For timing, assuming that it IS a sinusoid, consider just filtering to pass the sine into a rectifier and integrator. The integrator voltage will drift slowly when there's no input (you can track that, and reset it if necessary) then undergo a change when the burst arrives, then go back to slow drifting when the burst is over. Measure the change with an ADC. A little math with the (known) peak/peak amplitude will tell you the burst duration, with error limits related to the onset and dropoff phases, which are still unknowns.
Precision rectifiers and even RMS averagers are best for the timing rectifier.
Peak-hold circuits need to be reset, of course, and you have to be sure the burst is over before you read the peaks or do the resets. This, is a small matter of programming, as is handling the numeric display output.
I can see that will get me the amplitude of the first cycle. But what about the following one ? Wouldn't I have to grab each one, calculate its value before I could get the next one ?
Yes the phase should be constant since there would be zero output from the transducer with no signal and a rising sinusoid at the onset, a peak and then a negative swing to a similar value.
A scope signal always shows a positive going starting waveform of a few cycles decaying to zero.
I'm going to have to do a bit more study.
What kind of device would you suggest ? PIC's ?
If its any help I am tempted to try use an old mobile phone LCD display as the indicator device. Particularly since they are easily come by.
If the subsequent peaks aren't larger than the 'first', the value doesn't change. I'd wait for the full duration of the disturbance before resetting the hold capacitor (and in the Wilkinson scheme, that reset IS the voltage readout; you just sink a small current from the hold capacitor until it gets to 'zero', and look at your clock to see how long that took).
If you only want amplitude, measuring max and min will suffice, but if you want ALL the peaks separately accounted for, it won't be a two-number readout later. I've done a lot of measurement where a recorded data set was fit to a multivariate model, but it's not something to take lightly. If you want to analyze the burst for ALL its characteristics, you need a digital oscilloscope and major software packages. If the word "eigenfunction" is familiar to you, you can write your own software for this :-)
PIC should be able to handle the measurement easily enough, the requirements are relatively modest. As for a display, it'd be easier to buy a display-plus-interface module, these take serial or parallel protocols and understand ASCII. The cellphone display will require you to form your own characters from bits, and understand a multiplex display-scan algorithm...
Cell phone display are often proprietary, since there is usually one customer only. Segment LCDs are easy to work with and low power. We use a 6 segments x 4 commons 3 digits LCD. So, a single PIC or AVR can drive two or three of them (tying all the commons together).
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.