Precise Peak Detector !! ideas ?

Hi,

I'm looking for a design of precise peak detector able to detect signals between 100mV and 5V at 5Mhz.

Moreover, i don't find design with not dangerous RESET (I have a differentiel Op Amp at the input of this peak detector, and whith a simple diode RF and capacitor, I can do court-cicuit the capacitor to the ground (because the output Op Amp too !!)


Regards

two stage

For these high speeds the usual method is to have a fast detector first followed by a slower one that can hold the voltage for a longer time.

This way the first detector can closely follow the signal peak because it is not driving a large capacitor. Its output will hold for a long enough time for the following detector to charge its larger capacitor.

Ok ...

but how design you the first stage (fast detector without Capacitor ????!!!)

first stage

The first stage uses a small capacitor. If your peaks are far apart it will totally discharge between them and not require a reset. The second stage will be single ended and can be reset with a mosfet across the capacitor.

The problem is that i want to measure in "real time" the level's signal which always emitted !!! So I must reset with a MOSFET when i want to take the measure.. you know ?

all digital way

Try this way. Use a high speed ADC that samples frequently enough to get at least one sample time at the top of your waveform. Put the output to a digital magnitude comparator which compares the number to one in a latch. If the new ADC value is higher, load the value into the latch. This will then find the peak. The magnitude comparator can also sense when lets say two samples following the peak are smaller. When this happens, the latch value is loaded out to some other part of your circuit and the same circuit is used to find the lowest signal level by using the less than output of the magnitude comparator. Once the dip is found and the next two samples are higher the magnitude comparator circuit is switched to the peak finding mode again.

Alternately, if you know the voltage input will go below the 100 mV when there are no pulses, you can use the adc output going below say 50 mV as the condition to clear the latch and start the peak finding again.

If you want an analog out, the other part of your circuit can be a high speed DAC.

i agree flatulent.
another,a logarithmic amplifier can fit you?

further ideas

Here is a further idea if your signals are a series of pulses with somewhat rectangular shape (steep rising and falling edges).

Use the traditional two op amp peak detector circuit with the diode replaced by an analog switch. The switch is turned on and off by an edge triggered flip flop.

Starting our description from the time the signal is at the zero volt condition, a comparator examines it and when it gets up to 75 mV it turns on the analog switch which allows the capacitor to track the pulse up to the top. Another comparator compares the input signal with the detector output. When the input declines below the held value, the flip flop is turned the other condition which allows the peak detector to hold the peak until the next pulse comes along.

You will possibly get a little glitch during the pulse rise portion but the output signal can be low pass filtered to remove the glitch but not the stair step waveform.

What I have described is a track & hold amplifier. You could buy one ready made.

If the statistics of the pulses are that the next pulse is at most n% of the previous one you could change the 75 mV threshold to a fraction of the circuit output.

If you have another signal that is synchronized with the pulses, use it to control the track & hold amplifier.

If the pulses change amplitude at a much lower rate than the pulse frequency, you can low pass filter the pulses. Use a 0.001 dB ripple elliptical filter with the cutoff being the geometric mean of the highest pulse height changing frequency and 5 MHz. Have the first notch in the stop band at 5 MHz.