need precision peak detector 100KHz to 1MHz

[sgugan]

my signal is 0 v t0 20Vp and 100khz to 1MHz...
I want high precision peak detector circuit... most of the op-amps not meeting required slew rate and votage rating to construct a peak detector...
any idea???

 

[chuckey]

One circuit I have come across is to feed a 0->1V signal into one input of a fast comparator and the other input is connected to a storage capacitor. While the output of the comparator is "high" a oscillator runs, whose output pulses "pump" up the storage capacitor via a diode. Once the storage capacitor has a higher voltage on it then the input voltage the oscillator stops, so the store cap's voltage stays constant, just being pumped up now and then to make up for losses. The charge transfer cap/the storage cap = step size (i.e. 1:10,000). The more pulses the store cap needs the slower the circuit. Very linear circuit, good for 30 MHz, bad for following rapidly changing voltages.
Frank

 

[DXNewcastle]

I've learned something new today, thanks to that suggestion, very succinctly explained.
Hopefully, the OP has had their request solved.

That's an excellent little circuit, but why didn't you tell me about it years ago, when I could have used to real benefit?

 

[keith1200rs]

My definition of a peak detector would be that it would track & hold a single peak. While the suggested circuit is interesting it is not what I would call a peak detector.

The normal way round the slew rate problems of opamps in peak detectors is to use a dual opamp/dual diode configuration which prevents the opamp output from spending most of it's time on the rails so it doesn't have to slew so far.

There is a technique which uses cascaded bipolar transistors which I have used up to maybe 1GHz but I would have to find it later.

Keith

 

[chuckey]

This circuit was designed by the BBC (circa 1970) to measure the amplitude of the ITS waveforms for remote measuring of video parameters. It was then used by Marconi Instruments in a commercial version of the BBC's efforts.
Frank

 

[hithesh123]

Can someone post the circuit that Keith mentioned.

 

[keith1200rs]

This is the idea (at least my interpretation of the original source material I based it on).

Keith.

 

[leo_o2]

A simpler way is just a NMOS + a capacitor. Connect NMOS's gate to the sampled voltage.
NMOS' drain is connected to supply voltage. NMOS's source is connected to the capacitor.
The capacitor will store the peaking value.
However, there is a voltage shift to the sampled result. It is Vt of the NMOS.
However，it could be added back or be cancelled in the later processing.
Its speed is high. A drawback is the input voltage range is limited at Vt above.

 

[keith1200rs]

It depends on what speed you need. The circuit I posted is very fast.

Also, if you were designing a CMOS IC then you would have different devices available.

Keith

 

[FvM]

The problem with non-feedback peak detectors, either using diodes or transistors, is their time dependant offset. In practice, they can achieve a sufficient accuracy only with single pulses of constant waveform. This is e.g. the case with the suggested MOSFET circuit. If the peak pulse width increases or the pulse is repated, the capacitor voltage rises.

For the originally requested "100 kHz to 1MHz" speed range, a standard two OPs + diodes peak detector, as mentioned by keith, should work best. Reasonable fast OPs are required, however.

P.S.: A slightly tuned version of the classical two OP peak detector can be. e.g. found here (with waveforms) https://www.analog.com/static/imported-files/data_sheets/AD8033_8034.pdf

 

[keith1200rs]

The circuit I posted was based on some paper I read and was an experiment for part of a laser rangefinder with 20ns pulses. You need to pick a circuit appropriate to your specifications which may be different. For 'normal' peak detectors I usually use a conventional dual opamp circuit.

Keith

 

[leo_o2]

I think all schemes will have accuracy problem caused by delay.
For periodic signal, a discharging and a sampling circuit should be added in MOSFET solution.

The problem with non-feedback peak detectors, either using diodes or transistors, is their time dependant offset. In practice, they can achieve a sufficient accuracy only with single pulses of constant waveform. This is e.g. the case with the suggested MOSFET circuit. If the peak pulse width increases or the pulse is repated, the capacitor voltage rises.

For the originally requested "100 kHz to 1MHz" speed range, a standard two OPs + diodes peak detector, as mentioned by keith, should work best. Reasonable fast OPs are required, however.

P.S.: A slightly tuned version of the classical two OP peak detector can be. e.g. found here (with waveforms) https://www.analog.com/static/imported-files/data_sheets/AD8033_8034.pdf