Re: Peak detector



Tom Bruhns wrote:
50 usec is a very long time in terms of response of modern electronic
circuits. It should be possible to capture the peak using either
analog or digital techniques. How accurately do you wish to capture
it, and what amplitude is it? (What is the range of possible
amplitudes, including the peak and the value at other times?)

A digital method: let a very simple microprocessor (or even a logic
circuit) monitor the output of an ADC that's taking reading at an
appropriate rate, say 100k/second, and hold the maximum value in a
register. If a new ADC reading is higher than the value in the
register, replace the register contents with the new reading. This
could easily be equipped with a digital readout and a reset button, or
it could respond to an external electronic reset line and provide its
register value to a computer...lots of flexibility.

An analog method: there are many op amp peak detector circuits around.
I'd look on the TI and Analog Devices and National Semiconductor
websites for ideas, or look in op amp databooks. I particularly
remember some good ones in the Burr-Brown data books (now part of TI).
You can read out the value with an ADC or digital voltmeter; beware
that the design has low droop rate. It should be possible with modern
parts to get a value held to 1% accuracy for more than 10,000 times as
long as it takes to acquire the peak value. The peak is held on a
capacitor, and the ability to hold a value a long time is set by having
a high ratio between leakage currents that affect the capacitor voltage
and the current available to charge the cap to the peak value to be
held. It's capturing the held value quickly enough to avoid droop
problems in the analog circuit that make the digital solution
particularly attractive. (Analog methods are probably more appropriate
for capturing very short peaks.)

Obviously, you can make two such circuits to monitor the two inputs.
And you can arrange gating to enable each, either digital or analog,
only for a "window" of time around when you need the result.



Hmmm- a third possibility would be an analog-digital hybrid with a run-of-the-mill el-cheapo analog classic pre-conditioning the input for a quasi digital job similar to circuit shown here: http://www.elecdesign.com/Files/29/3473/Figure_01.gif .
Infinite hold times with 0% error and easily adapted to provide an output comparison of two channels.

.

Quantcast